1 Democracy-as-Fairness: Justice, Equal Chances and Lotteries Ben Saunders Jesus College Thesis submitted for the degree of D.Phil in Politics in the Department of Politics and International Relations at the University of Oxford. March 2008 (Hilary Term) Word Count: 94,210 (excluding bibliography) (82,186 excluding footnotes) 2 Abstract Democracy-as-Fairness: Justice, Equal Chances and Lotteries Ben Saunders, Jesus College, Oxford D.Phil in Politics. March 2008 This thesis challenges the close association of democracy and majority-rule. I argue that democracy requires political equality, but there is no reason to suppose that this is only realized by majority-rule. I suggest that we can think about democratic procedures contractually, and reject the claims that majority-rule will be agreed to on the grounds that it will produce better or more equal outcomes or that it will necessarily be fair to all involved. Fairness is often connected to each person having an equal chance of getting their way, but majority-rule may de facto exclude a permanent minority, who would then have no reason to accept the procedure as treating them equally. The argument is not merely negative, however. My positive suggestion is that, if we want everyone to have an equal chance of casting a decisive vote, then we can realize this goal by entering all votes into a lottery, so one is randomly chosen to determine the outcome. The thesis goes on to describe how this ‘lottery-voting’ would work in practice, offering examples of small-scale direct democracy in which it might be applicable, reflections on how it might fit into a larger democratic framework (for instance, the relation between this voting procedure and deliberation and constitutionalism) and assesses it in light of the criteria of social choice (such as decisiveness, anonymity and neutrality) and rationality. The aim is not to defend lottery-voting as the best decision-procedure for all circumstances, but to illustrate that it is a democratic possibility and therefore to stimulate new debates within democratic theory, for instance on the justification of majority-rule. 3 Table of Contents Abstract ................................................................................................................. 2 Table of Contents.................................................................................................. 3 Introduction................................................................................................................... 6 (0.1) The Importance of Political Equality ........................................................... 6 (0.2) Equal Chances .............................................................................................. 7 (0.3) What Lottery-Voting Is ................................................................................ 8 (0.4) Literature Review....................................................................................... 11 (0.5) Political Morality........................................................................................ 15 (0.6) The Logic of Democracy............................................................................ 20 (0.7) The Concept of Constituency..................................................................... 25 (0.8) Plan of the Thesis ....................................................................................... 29 1 Democracy as Freedom and Equality ...................................................................... 37 (1.1) Rule of the People ...................................................................................... 37 (1.2) The Possibility of Self-Government........................................................... 39 (1.3) Equal Relations and Respect ...................................................................... 41 (1.4) The Alleged Obviousness or Necessity of Majority-Rule ......................... 44 (1.5) Contract and Consent ................................................................................. 48 (1.6) Contracting to Majority-Rule ..................................................................... 50 (1.7) Proportionality............................................................................................ 52 (1.8) False Dichotomies and Neglected Options ................................................ 55 (1.9) Conclusion.................................................................................................. 60 2 Maximizing Arguments for Majority Rule .............................................................. 62 (2.1) Introduction ................................................................................................ 62 (2.2) Procedural Justice....................................................................................... 64 (2.3) Perfect Procedural Conceptions of Democracy ......................................... 67 (2.4) Problems with Utilitarian Outcomes .......................................................... 69 (2.5) Can Intensities be Accommodated? ........................................................... 70 (2.6) Imperfect Procedural Conceptions of Democracy ..................................... 87 (2.7) Minor Problems with the Condorcetian Paradigm..................................... 89 (2.8) Indeterminacy............................................................................................. 91 (2.9) Conclusion.................................................................................................. 95 3 Using Lotteries to Adjudicate between People........................................................ 97 (3.1) Introduction ................................................................................................ 97 (3.2) The Uses and Abuses of Lotteries.............................................................. 98 (3.3) Justifications of Lotteries ......................................................................... 101 (3.4) The Numbers Debate................................................................................ 105 (3.5) Taurek’s Argument for Equal Chances.................................................... 107 (3.6) Scanlon’s Objection to Equal Chances .................................................... 109 (3.7) Scanlon’s Argument for Saving the Greater Number .............................. 111 (3.8) The Weighted Lottery: Pooling Chances ................................................. 113 (3.9) Re-constructing the Fairness of a Weighted-Lottery ............................... 115 (3.10) Counting Individuals, Again .................................................................. 119 (3.11) Scanlon’s Argument Against Weighted Lotteries.................................. 121 (3.12) Prior Randomization versus Fixed Majorities........................................ 122 (3.13) Proportional Chances versus Proportional Outcomes............................ 124 (3.14) Conclusion: Towards Political Application ........................................... 127 4 Lottery-Voting Described ...................................................................................... 130 (4.1) Introduction .............................................................................................. 130 4 (4.2) The History of Lot.................................................................................... 130 (4.3) Modern Proposals..................................................................................... 133 (4.4) From Representation to Direct Decisions ................................................ 137 (4.5) Lottery-Voting Described ........................................................................ 138 (4.6) The Book Buying Example ...................................................................... 141 (4.7) The Reading Group Example................................................................... 147 (4.8) Vote Counting .......................................................................................... 152 (4.9) When do Losers’ Votes Count? ............................................................... 154 (4.10) Contingent Outcomes............................................................................. 158 (4.11) Predictability .......................................................................................... 161 (4.12) Randomness in Democracy.................................................................... 162 (4.13) Conclusion.............................................................................................. 165 5 Practicalities........................................................................................................... 166 (5.1) Introduction .............................................................................................. 166 (5.2) Lottery-Voting as a Part of the Decision Process .................................... 167 (5.3) Minority Motions ..................................................................................... 168 (5.4) Compromise and Collusion...................................................................... 172 (5.5) Liberalism and Constitutionalism ............................................................ 174 (5.6) Incentives for Deliberation....................................................................... 178 (5.7) The Nature of Deliberation and Limits of Reason-Giving....................... 182 (5.8) Openness and Compliance ....................................................................... 183 (5.9) Scrutiny .................................................................................................... 187 (5.10) No Repeats ............................................................................................. 188 (5.11) Dividing Decisions................................................................................. 190 (5.12) Conclusion.............................................................................................. 194 6 Minimal Conditions of Social Choice.................................................................... 196 (6.1 Introduction) .............................................................................................. 196 (6.2) May’s Conditions ..................................................................................... 197 (6.3) Decisiveness ............................................................................................. 198 (6.4) Anonymity................................................................................................ 199 (6.5) Neutrality.................................................................................................. 201 (6.6) Positive Responsiveness........................................................................... 203 (6.7) Arrow’s Conditions .................................................................................. 206 (6.8) Universal Domain .................................................................................... 208 (6.9) Pareto........................................................................................................ 212 (6.10) Independence of Irrelevant Alternatives ................................................ 217 (6.11) Non-Dictatorship.................................................................................... 227 (6.12) Arrovian Conditions Concluded............................................................. 232 (6.13) Non-Manipulability................................................................................ 234 (6.14) Weighted-Voting.................................................................................... 238 (6.15) Conclusion.............................................................................................. 241 7 Rationality.............................................................................................................. 243 (7.1) Introduction .............................................................................................. 243 (7.2) The Nature of Rationality......................................................................... 244 (7.3) Maximization ........................................................................................... 249 (7.4) Consistency between Decisions ............................................................... 251 (7.5) Consistency over Single Decisions .......................................................... 254 (7.6) Individual Rationality............................................................................... 258 (7.7) Collective Rationality............................................................................... 262 (7.8) Rational Use of Decision-Mechanisms .................................................... 265 5 (7.9) Path Dependence ...................................................................................... 269 (7.10) Conclusion.............................................................................................. 271 Conclusion ................................................................................................................ 273 (8.1) Summary of the Argument....................................................................... 273 (8.2) The Importance of Thought Experiments ................................................ 274 (8.3) Practical Possibilities................................................................................ 277 (8.4) A Utopian Example.................................................................................. 280 (8.5) The Place of Democracy .......................................................................... 283 Bibliography ............................................................................................................. 287 (B.1) Books and Articles................................................................................... 287 (B.2) Other Acknowledgements ....................................................................... 306 6 Introduction “If it is not controversial, it is not about democracy”1 “[I]t has been said that democracy is the worst form of Government, except all those other forms that have been tried from time to time”2 (0.1) The Importance of Political Equality It is often assumed that democracy means, or at least requires, some form of political equality. For instance: “[E]quality is the fundamental value underlying democracy”3 “One characteristic that most persons regard as essential to democracy is political equality. A familiar way of describing this trait is ‘one man, one vote’… the right of each citizen to count as one in the decisionmaking process of the community”4 “[T]he democrat wants us all to contribute equally to delivering an answer”5 I will not here investigate these claims. I will simply take it for granted that we do indeed want all votes6 to count equally, as is often claimed: “The principle of political equality… is that every man counts for one vote, and that one man’s vote is the equivalent of the next man’s”7 “[Elections] constitute an important arena for ensuring political equality between citizens, both in access to public office and in the value of their votes… [E]ach vote should have the same weight or value, regardless of where people happen to live or which party they vote for”8 Woodruff (2005) p.ix [not emphasized in original]. Churchill, speech of 11/11/47, in Churchill (1974) p.7566 [not emphasized in original]. 3 Christiano (1996) p.17. 4 Ranney and Kendall (1956) p.16. 5 Swift (2006 [2001]) p.183. 6 I discuss the possibility of weighted voting in chapter 2.5 and 6.14. Even if we accept that some people should have more weight than others, however, we would want their votes to be equal, so giving someone double weight can be achieved by giving them two votes. 7 Sartori (1965) p.335. 8 Beetham and Boyle (1995) pp.31, and 49. 2 1 7 “The statutory right to an equally powerful vote reflects the widely shared view that democratic institutions should provide an equal opportunity to influence political decisions”9 “[F]airness requires that each person receive an equal share or an equal chance”10 If it turns out that democracy does not require that votes be equal, as is sometimes suggested11, then my thesis is not about all democratic systems, but only those that do count all votes equally. (0.2) Equal Chances Assuming we want votes to count equally in some form, it remains to be seen what this means. Numerous interpretations could be put forward, but I focus on one – namely the often-stated ideal that each vote should have an equal chance of determining the outcome of the election. “Majority rule is a genuinely egalitarian rule because it gives each person the same chance as every other to affect the outcome”12 “[A] representation system satisfies procedural equality if it accords to every voter an equal a priori probability of influencing any particular legislative choice”13 “[I]f we are concerned about political equality, it seems reasonable to require not only that each person’s vote count the same as that of each other person, but also that each person, regarded as someone with particular proposals to advance, have an equal chance to have his proposals adopted”14 9 Guinier (1994) p.72. Dahl (1985) p.58. 11 For example, Vernon (2001) p.44 and Rehfeld (2006) pp.23 and 195 argue that because the chance of a vote deciding the outcome of an election is so small, we need not worry that the chances are unequal. This seems to contradict the idea that equality is often supposed to matter more the less we have, but this is because the chance is practically nothing – it is like distributing not slices but crumbs of cake, in which case we may not care whether one person has three crumbs and another two (I owe this example to Jerry Cohen). 12 Christiano (1996) p.55. 13 Beitz (1983) p.72. 14 Nelson (1980) p.19. 10 8 “The purpose is not to guarantee “equal legislative outcomes”; [but] equal opportunity to influence legislative outcomes”15 If all votes should have an equal chance of determining the outcome, an easy way to achieve this is to put them all into a metaphorical ‘hat’ or tombola and randomly select one to decide the outcome. This is roughly what lottery-voting amounts to, although a fuller description is needed. (0.3) What Lottery-Voting Is The idea of randomly-selecting a single vote to determine the outcome of an election is not a new one. It is sometimes appealed to by those working in social choice. It has also been put forward as a possibility – if only to be rejected – by a number of political theorists, as discussed in the next section. The only serious advocate I know of is Akhil Amar, who proposed what he called ‘lottery voting’ in the Yale Law Journal over twenty years ago16. Although the name comes from this article, lottery-voting is proposed here for direct decision-making, rather than electing representatives. The added complexity of large-scale representative democracy is bracketed by focusing on small bodies, such as clubs or associations, making their own decisions (see the examples in chapter 4.6-7). As radical as the idea may seem, it is not itself original, as revealed in the next section. What is distinctive about the present argument is that, instead of starting from such a proposal and seeking to justify it by appeal to its consequences, I intend to argue for such a system, starting from basic ideas of democracy and fairness in adjudicating between persons. As such, the argument begins with general considerations about democracy and political equality, leading to criticism of 15 16 Guinier (1994) p.14. Amar (1984). 9 majority-rule. Lottery-voting is not fully developed as an alternative until chapter four, though it lurks in the background throughout the earlier criticisms of majorityrule. Lottery-voting is given a brief introduction here, so that the reader can keep this eventual aim in mind through the earlier material. A fuller statement must, however, await until after we have diagnosed potential problems with majoritarian systems. In essence, the proposal is that individuals cast votes17 for their most favoured option, as in a standard ballot under plurality-rule. Instead of the votes being counted up and the side with the most votes winning, however, one vote is drawn at random to determine the result. This is sometimes known as a ‘random dictator’ method, for an arbitrarily-chosen person is taken to decide the outcome, regardless of everyone else’s votes. It does, however, mean that each person’s vote has an equal chance of affecting (in fact, deciding) the outcome – because every vote has an equal probability of being drawn. Overall, the effect will be that side A in the contest has a probability of winning the vote equal to the proportion of people who vote for A. That is, if one side wins 60% of the votes, then they have a 60% chance of winning – because there is a 60% chance of one of these votes being randomly drawn. If the other side took the other 40% of the votes, however, they still have a 40% chance of winning. Thus proportionality is preserved up until this stage, though once one vote is drawn the ‘winner takes all’ – i.e. the decision goes wholly their way – and there is no further compromise when it comes to the actual policy implemented. All of these details require argument, and that will be provided throughout the following thesis. One might wonder, for example, whether it is fair that one side has a 60% chance and the other only 40%. Of course, this means that some people will have a far greater expectation of getting their way, but it follows from treating each 17 I avoid the term ‘election’ as I am concerned with direct decision-making. There is danger of ambiguity though, because ‘votes’ may mean election or an individual’s expression of preferences. I hope the meaning is clear to the careful reader. 10 person equally – what groups get is proportional to their size. This proportionality is certainly more equal than majority-rule, where the 60% are guaranteed to get all their way. The other alternative – giving every group an equal chance of getting its way, regardless of its size – may seem just, but is not democratic, for it wholly ignores numbers – it would be as good, on this proposal, to have one vote as to have all bar that one vote18! Even once we accept proportionality, however, it may be objected that proportional chances introduce it in the wrong place – for once the lottery has taken place, some people have all their own way and others nothing, so there is still no retrospective equality. It is not my aim to defend retrospective equality, however. The problem with majority-rule is a procedural one – there is no point in members of a permanent minority continuing to play the democratic game, for it is already predetermined that they will lose. Maybe sometimes a compromise on the final outcome is the best way to respect all equally, but sometimes such will be impossible or result only in a ‘fudge’ that pleases no one. Lottery-voting does not rule out a compromise that all can agree to, for instance after deliberation, but where such an accommodation cannot be reached it takes proportional chances to be the fair way to decide between conflicting interests. This will hopefully lead to all getting their way sometimes, but this result is merely a likely consequence of a fair procedure, not its main motivation. Moreover, it may be hoped that lottery-voting will encourage mutual understanding and a spirit of tolerance – because everyone knows they are not guaranteed to get their way, so no one will want to oppress others, since they know those others may win next time. 18 A similar argument is made by Jones (1983) p.175. 11 These issues, and more, are addressed at greater length in what follows, so they cannot be settled decisively here – for now, I only wish to outline what lottery-voting is, to aid in the understanding of the argument for it which follows. (0.4) Literature Review If we take seriously the idea that each vote should have an equal chance of determining the election, then one way in which this can be realised is by randomlyselecting a single vote and taking that to decide the outcome. A number of theorists briefly discuss such a proposal, usually only as a starting point or reductio ad absurdum. For instance: Robert Paul Wolff considers and rejects such a proposal, observing that “legislation by lot would offer some chance to the minority, unlike rule by the majority, but it would not offer to each citizen an equal chance that his preference be enacted”19. This is to assume a different account of fairness, which I shall argue in chapter 3.10 is undemocratic – we cannot give each individual an equal chance of getting their way and respect votes. In any case, it is unclear why Wolff cares whether chances are equal, since his strong anarchism would presumably reject any procedure that involved a decision going against one’s conscience or autonomy, even if one had had an equal or greater chance of success20. Bruce Ackerman notes that “The more familiar way of making a political outcome responsive to a constituency is to add up the judgments of individual citizens to form an overall total representing the view of society… [But] there is a second way of relating citizen views to political outcomes. Under this “probablistic” 19 20 Wolff (1976) p.45. Hyland (1995) pp.141-4. 12 approach, every citizen is given a finite chance of deciding the political outcome” 21. He concludes that there is no reasoned basis to prefer majority-rule to a responsive lottery. Peter Jones, having earlier criticized the view that political equality requires majority rule22, goes on to describe a lottery-based alternative: “Suppose a group of individuals agreed to make decisions which applied to them collectively by way of a lottery. For every issue there would be a lottery in which each individual held one and only one ticket. The individual who held the winning ticket for a particular issue would have the right to decide on that issue on behalf of the whole group. Thus, for any issue, each individual qua individual would have an equal chance of being the decider. The odds in favour of any particular proposal being adopted would be proportionate to the number of individuals who favoured that alternative in the total group”23 David Estlund also considers the proposal, which he calls ‘Queen for a Day’, on the basis that it amounts to a random dictator for each issue, and admits: “I know of no strong moral argument against it as compared with ordinary voting. Insofar as it is distasteful, bear in mind that none of the approaches to democratic legitimacy canvassed in this essay has any reason to reject it. It is fair, and it can take place after individual views are shaped by public deliberation”24 Even so, the idea is often given short shrift. The only sustained defences I am aware of are two articles by the aforementioned Akhil Amar25. Even where others admit they have no reasoned basis on which to reject such a proposal, they put it aside remarkably quickly. This is true even of those willing to embrace a significant amount of randomness, for example: Having spent several pages on its advantages, Jon Elster dismisses it in less than page, remarking that: 21 22 Ackerman (1980) p.288. Jones (1983) p.160. 23 Jones (1983) pp.170-1. 24 Estlund (1997) p.193. 25 Amar (1984) and (1995). 13 “[L]ottery voting has several disadvantages which explain why it has never been adopted and suggest that it never will be. Most obviously, the lack of continuity among the representatives counts against the proposal… Furthermore, the predictable rise of numerous small parties would make the Fourth French Republic a paradigm of stability by comparison… Finally, the risk of some lunatic fringe coming into power would not be acceptable, even if the chance were very small”26 It is not clear that all of these criticisms are successful. The point of lottery-voting is to break up the two-party hegemony that typifies countries with a first past the post electoral system in order to represent minorities. This will lead to the rise of numerous small parties, but there is no reason why this must obviously be more unstable than any other system of proportional representation. Since proportional representation is stable in many countries, using examples like the French Fourth Republic, or Weimar Germany, to discredit such systems is unfair – presumably the electoral system was no more than part of the problem in those cases. Nor is it obvious that lottery-voting would create a lack of continuity amongst representatives, for popular incumbents would still be likely to retain their seats and change would still only take place in elections every few years. The final worry, about extreme minorities, may be more substantial, and it is one I address in chapter five. Neil Duxbury argues “were such a system of lottery voting to lead to the emergence of numerous small parties, formation of governments might prove difficult if not, on occasions, impossible to achieve” and concludes “Lottery voting does not appear to constitute a particularly promising combination of decision-making devices”27. Barbara Goodwin provides a literature review covering numerous proposals for using lotteries28. She argues that many, including Amar’s lottery-voting, only give the lottery a very limited role – by, for example, combining it with voting or restricting it to the election of representatives not determination of policy – and that 26 27 Elster (1989) pp.89-90. Duxbury (1999) p.150. 28 Goodwin (2005) pp.181-90. 14 these limits show a lack of faith in the lottery and/or equi-competence of the people29. She goes on to point out that the rationale for lottery-voting would logically lead to the abolition of representatives altogether30. Amar proposed lottery-voting not for making decisions, or even individual appointments, such as the president, but for electing representative chambers. The random element could therefore be seen as a sampling mechanism, approximating proportional representation. Most of what discussion there is, therefore, focuses on the potential use of lottery-voting in such contexts. My concern here is not, however, with representative democracies (though I have some remarks on sortition-based proposals in chapter 4.3). Representative democracy is left aside because it involves further complexities, for instance whether lottery-voting should be used in electing representatives, their decision-making, or both, and why it would be preferable to a non-random form of proportional representation. Electing a representative is merely one decision that a group of individuals can make and, if it is reasonable to employ lottery-voting for such decisions, then it may also be reasonable to use it in making other decisions. My focus is on cases of collective decision-making in which all are supposed to participate as equals31. I therefore ask whether Amar’s lottery-voting could be applied to other cases of direct decision-making; something that has received even less attention in the existing literature. Whether lottery-voting – or any other democratic method – will be appropriate in a given circumstance depends on the specifics of that case, but some examples are given in chapter 4.6-7 of cases where lottery-voting may be, all things considered, at least as good as any other decisionmechanism. 29 30 Goodwin (2005) p.190. Goodwin (2005) p.191. 31 Note that these need not be democratic as such, because the group deciding as equals may be a small oligarchy. 15 While lottery-voting may indeed face problems, these will be addressed in the substantive parts of the following thesis. For present purposes, I merely want to motivate that further consideration. Having seen how briefly lottery-voting is considered, if at all, in most of the mainstream literature, I want to suggest its potential interest by showing its relevance to three recent books, all of which neglect the possibility. These books are Richard Vernon’s Political Morality32, Anthony McGann’s The Logic of Democracy33 and Andrew Rehfeld’s The Concept of Constituency34. I will now briefly summarize the arguments of each, showing how lottery-voting might fit in. (0.5) Political Morality Richard Vernon’s Political Morality: A Theory of Liberal Democracy35 attempts to resolve perceived tensions between liberalism and democracy. For present purposes, however, let us focus on Vernon’s argument for majority-rule, which I will argue actually better supports lottery-voting. It seems that Vernon either has to accept lottery-voting or provide a better justification for majority-rule, which rules out lottery-voting. Vernon begins with a historical narrative: since the decline of non-political – for example, religious – authority, societies are forced to negotiate arrangements for collective decision-making and establishment of authority36. Adopting an appropriately contractualist version of democracy, therefore, Vernon suggests we would adopt a limited majority rule. He supposes a liberal default injunction against 32 33 Vernon (2001). McGann (2006). 34 Rehfeld (2005). 35 Vernon (2001). 36 Vernon (2001) p.19. 16 coercion, which is why we need to justify democratic decision-making that imposes on everyone. “If the use of compulsion were not something requiring justification, there would be no need to give one; so the claim that majorities have this entitlement implies a general background prohibition against compulsorily substituting one’s own judgement for another’s”37. Vernon appeals to a Lockean idea of compensation to limit the majority38. The ‘political morality’ he refers to in his title is not some moral constraints on officials, but rather restraint on behalf of the majority39 that safeguards the minority from ‘uncompensable’ defeats. If one suffers material loss as a result of the ruler’s decree, one can in principle be compensated; whereas there can be no compensation for loss in spiritual matters, such as eternal damnation. The point is not that the majority should actually compensate the defeated minority; but that the minority can themselves come to terms with a compensable loss. “The majority has an obligation to leave minorities a space in which they can adjust their lives to a public environment that they did not choose”40. Vernon illustrates what this might mean in the case of the abortion debate. Being forced to have an unwanted child is an uncompensable life-changing experience41; whereas the pro-lifer whose proposed ban on abortion is ruled out can go about limiting abortion in other ways – e.g. reforms of adoption, sex education, promoting contraception and so on42. Thus liberal democracy means not that the democratic impulse of majority-rule is limited by 37 38 Vernon (2001) p.104. Vernon (2001) pp.75-6, and 149. 39 Vernon (2001) p.90. 40 Vernon (2001) p.148. 41 A claim contradicting Tooley (1972) pp.52-3. 42 Vernon (2001) p.82. 17 something else (liberal rights), but rather that, beyond a certain point, the rationale for majority-rule itself runs out43. Proper liberal democracy is self-limiting in its scope. Insofar as it can be considered simply about either, Vernon’s Political Morality seems more a theory of democracy than liberalism. Liberalism, he assumes, offers us certain areas of freedom in which we are protected in the exercise of our judgements. His ‘private sphere’ seems narrow – confined only to what would be ‘uncompensable’ – but in keeping with traditional pictures of liberalism. While this is largely assumed as a background, however, Vernon offers his own quite original arguments for why we should value the democratic process – starting out by rejecting, on grounds of incommensurable value conflict, the utilitarian alternative that majority-rule will be conducive to the greatest good of the greatest number44. After discussing both outcome- and procedure-based defences of democracy, which he finds in Rousseau and Habermas respectively, Vernon outlines a combination view, which he calls ‘processual’. What is valuable is an outcome, but not the electoral outcome which could have been brought about by other means; rather it is the result of the deliberative process itself – i.e. the fact that arguments are advanced, reasons considered and interests generalized. In this way, there is an improvement in public reason, and everyone’s capacity as a moral reasoner is equally respected. Once this process has taken place, the actual result of the subsequent vote is immaterial. It is explicitly not assumed that more information will lead to either consensus or better electoral outcomes45. The main reason for individuals to cast informed, reflective votes is not backward-regarding, but rather to ensure the continuation of the reason-giving deliberative process next time. In this way, they 43 44 Vernon (2001) p.72. Vernon (2001) pp.43, 77, and 120. 45 Vernon (2001) pp.48-9, 64-9, 77, 123, 139, and 148. 18 ensure that politicians continue to have incentives to provide generalized reasons and cater for the interests of all46. There is, however, a flaw in Vernon’s case for majority-rule, which is insufficiently considered. Having abandoned the claim that a majority judgement will even necessarily best reflect the outcome of the preceding discussion47, Vernon denies any special privilege to either a relative majority or even a 50%+1 threshold48. His case for majority-rule rests simply on the claim that “majority rule has the virtue of encouraging the generalization of political argument, which is an important service to public reason; and it also has the virtue that because the counting of votes records a fact, it can yield the unequivocal answer that we need in order to have a political authority. Both of these justifications have weight even for the outvoted”49. That is, majority-rule gives politicians reasons to appeal to as wide a portion of the electorate as possible, and produces a determinate answer. If, however, these are the only virtues of majority-rule50, then picking the winning vote at random, or lottery-voting, seems to better meet Vernon’s standards. If the majority verdict is no better in itself, then one might as well simply pull a vote out of the hat, as it were, and take that as authoritative. This yields just as determinative answer as the fact provided by counting votes. Indeed, a lottery has long been viewed as a fair way to decide between ties or incommensurable claims51. As for giving parties an incentive to ‘generalise reasons’, it seems that lottery voting fares even better than majority-rule, because lottery-voting gives parties reason to appeal to as many voters as possible since 50%+1 is not enough to guarantee victory. 46 47 Vernon (2001) pp.70, 79, and 142. Vernon (2001) p.69. 48 Vernon (2001) pp.50-1, and 69. 49 Vernon (2001) p.142. 50 One other Vernon (2001) p.70 considers, in the case of small groups, is needing the goodwill of a majority for action to proceed, but he rejects this in larger political communities. 51 E.g. Greely (1977), Sher (1980), Neurath (1983) and the examples in chapter 3.2. 19 Vernon rejects supermajority requirements because, if a bill needs 67% support to pass, one only needs to appeal to the sectional interests of 34% to block it. Requiring majority support means that one has incentives to offer reasons to at least half the people, and maybe more on the grounds not all will necessarily be persuaded. If one only needs a majority, however, then there is no need to appeal to everyone. If 20% of society seemed unlikely to accept your arguments, it would not matter and one would have no reason to appeal to them, as the other 80% are more than enough to ensure victory. Picking a random vote, however, means that electoral success is never guaranteed (short of unanimous persuasion, in which case the vote as well as the lottery is superfluous). When this probabilistic element is introduced, it is no longer enough to have – say – 60% support; rather the more votes you win, the greater your chances – and thus there really is an incentive to appeal to everyone. Such a scheme epitomizes the requirement that electoral outcomes not be predictable in advance52. Vernon might seem hostile to such an idea, judging from several earlier remarks about lotteries, but in fact none of these criticisms tell against the idea of lotteryvoting. What Vernon has in mind, and rejects, is simply deciding what to do randomly, by a simple lottery in which each option has equal chances53. Even assuming that we can independently identify discrete options, such a lottery might be fair in one sense, but it would be undemocratic, because it ignores people’s preferences54. By giving those on each side equal chances of satisfaction, it may reflect a certain conception of justice, but it totally ignores numbers – so one vote is as good as 99% of them, and each option would get a 1/n chance. It is such a proposal that Vernon rightly dismisses because of this lack of responsiveness. As Vernon puts 52 53 Vernon (2001) p.39; c.f. Przeworski (1991), discussed in chapter 4.10. A possibility suggested in Estlund (2007) e.g. p.80. 54 This will be argued more fully in chapter 3.10. 20 it, “One way in which the majority principle differs from a coin-toss or lottery is that it gives voters an opportunity to influence the outcome in their favour, not just to expect favourable outcomes with a certain probability”55, to which he adds that majority-rule also provides citizens with reason to acquire information, because their vote matters56. Lottery-voting shares both these advantages, since citizens can increase the chances of a given option winning by voting for it and since every vote matters – because it may win – they also have reasons to acquire information57. So, to recap, Vernon suggests that we want a procedure that produces determinate outcomes, and gives politicians reasons to appeal to as many people as possible. He accepts majority-rule, because it is such a procedure, but in fact lotteryvoting provides just as determinate an outcome and probably a greater incentive to generalize one’s appeal, since a simple majority is not necessarily enough for victory58. As such – despite his hostility to simple lotteries – Vernon should either accept lottery-voting or needs to offer some further argument against it, if he is to defend majority-rule. (0.6) The Logic of Democracy Anthony McGann’s The Logic of Democracy: Reconciling Equality, Deliberation and Minority Protection59 is a very worthwhile read, intelligently combining normative political philosophy, social choice and empirical studies of his favoured ‘consensual’ democracies of Western Europe (his favoured examples being 55 56 Vernon (2001) p.43. Vernon (2001) p.46. 57 Vernon (2001) p.52 [ch.2 fn.5, ref. to p.43], cites Estlund (1997)’s ‘Queen for a day’ proposal, but doesn’t consider how it differs from the simple lottery he rejects. 58 One may argue there is less need to persuade as many people, since 20% may be enough – but this misses the point that one always has an incentive to persuade more if one can. 59 McGann (2006). 21 Denmark, the Netherlands, Norway and Sweden60). Though it was only published late in the development of this thesis, many of his arguments chimed in with and influenced the final expression of much of what was said in chapters 2 and 6. McGann convincingly shows how the goal of political equality can produce determinate institutional recommendations, even in modern representative democracies; specifically, proportional representation for electing legislative representatives and then majority rule at the decision-making stage. He argues that such procedures not only manifest political equality and citizen sovereignty, but also foster deliberation and actually protect minorities better than supermajoritarian constitutional limits. One strand of the argument is to show that Riker’s pessimism about democracy is ungrounded. The cycling results of social choice do not undermine any possibility of meaningful democracy, all that they do is undermine any idea of a single ‘will of the people’ or correct decision, i.e. what Riker calls ‘populism’. It does not follow, however, that we are forced to adopt a minimalist or ‘liberal’ defence of democracy – indeed, McGann points out that Riker’s arguments may undermine even such a position. Instead, he argues that democracy should be seen as a pure or quasi-pure distributive procedure – a position in keeping with my argument through chapters 13. We can endorse democracy on the grounds that it is a fair procedure for determining who gets what, without having to make claims about it expressing some mythical ‘will of the people’. Most of what McGann has to say is very sensible, but his defence of majorityrule is over-hasty and, as we shall see, relies on a crucial empirical premise. Majorityrule is defended as satisfying political equality, specifically May’s anonymity and 60 McGann (2006) p.177. 22 neutrality requirements. In his original argument, however, McGann is careful to state that he is considering only determinate procedures: “May’s (1952) theorem shows that the only determinate procedure for choosing between two alternatives that satisfies political equality is majority rule. This eliminates all the commonly used alternatives to majority rule, as any other nonrandom binary procedure privileges either some voters or some alternative”61 Later, when recapping and building on the argument, these qualifications are dropped, leading McGann to make a series of stronger claims: “May (1952) shows that majority rule is the only positively responsive, decisive, binary voting rule that satisfies anonymity (voters are treated the same regardless of their names) and neutrality (alternatives are not discriminated between on the basis of their names)”62; “Majority rule is the only decision rule that is fair in the sense of treating all voters and all alternatives equally”63; “The only decision rule that is fair in the sense of treating everyone equally is majority rule”64; “As has been shown previously, the only social decision rule that treats all people equally is majority rule”65. As argued in chapter 6.4-5, lottery-voting respects political equality, defined in terms of anonymity and neutrality. Of course, one may want to reject lottery-voting on other grounds, but this requires argument, whereas McGann simply refused a priori or neglected to consider it. We can see, however, why this might be a mistake if we consider a case where there is a permanent majority. McGann argues that, far from being a fatal problem, cycling is actually important to the operation of majority rule, because it implies there is no permanent majority or minority. This means not only that any given group in society know they could win next time, but they know that they could win this time. McGann assumes 61 62 McGann (2006) p.80 [emphasis added]. McGann (2006) p.89. 63 McGann (2006) p.134. 64 McGann (2006) p.168. 65 McGann (2006) p.174. 23 majorities must be coalitions of minorities, and so – as in the classic ‘divide the dollar’ game – are always liable to be split if another group can make a better offer. This, he argues, will mean that it is not in the interests of any winning coalition to oppress a losing minority, because if they push too hard that minority might strike a deal with some of the others, ‘selling’ their support very cheap to join a new winning majority66. McGann’s empirical studies seem to bear this out. As he points out, the consensual democracies operate largely by majority-rule, with what vetoes there are generally collective. He observes that, in these countries, winning majorities are always coalitions of minorities, but do not oppress minorities, perhaps for the very reason he suggests. However, while McGann is interested in these countries precisely because they match his institutional prescriptions (proportional representation and majority-rule), this means that he looks only at the ‘easy cases’ – ones more or less fitting Dahl’s description of a polyarchal society where there is no single majority, and so majority-rule operates reasonably well by giving all a chance to participate in winning coalitions (see chapter 3.11-12, below). However, even if society is in fact made up of many different minorities, it does not follow that all are equally likely to enter a winning coalition – e.g. if there are a number of different religious denominations, all of whom are somewhat distrustful of each other, but prefer any other religion to atheists67. Then there is the problem that some societies do have permanent majorities, rather than cross-cutting cleavages. McGann admits that lack of cycling is a problem for this very reason, but has nothing to say about what to do in such cases. 66 67 McGann (2006) p.109. For criticisms of Dahl, along these lines, see Lively (1975) pp.20-4 and Hyland (1995) pp.89-90. 24 In these situations, lottery-voting offers a potential solution. Pace-McGann’s incautious comments, it does respect anonymity and neutrality (i.e. political equality) and citizen sovereignty. Of course, once this is on the table, there is room for further debate about its merits vis-à-vis majority-rule. Like Vernon, McGann argues it is majority-rule that best promotes discussion, by requiring people to maximize those persuaded to their view68. Lottery-voting may be better, however, because it always gives all parties – majorities and minorities – an incentive to try to persuade more people to their view. Admittedly, a given group may always decide to break-off negotiation and simply hope their chance comes up, but McGann says nothing about the case of an intransigent permanent majority (because he simply assumes such does not exist). Further, McGann argues that majority-rule, even with cycling, will not produce any outcome, but likely be confined to a reasonable, uncovered set69. Lottery-voting may allow for a wider variety of outcomes, but they will be limited to the extent that they are always someone’s first preference (which means that they are extremely unlikely to spiral as arbitrarily far as predicted by some chaos results). These remarks do not amount to a case against McGann and all of these points are open to further argument, but they should be argued. McGann’s defence of majorityrule seems to apply only in favourable empirical circumstances, and he does nothing to engage with the alternative of lottery-voting, which may solve some fundamental problems in cases where there is a permanent minority. 68 69 McGann (2006) p.135. McGann (2006) pp.61-70. 25 (0.7) The Concept of Constituency Andrew Rehfeld’s The Concept of Constituency: Political Representation, Democratic Legitimacy, and Institutional Design70 focuses on representation, rather than direct decision-making. Rehfeld, however, employs lotteries at a different stage, in allocating voters to constituencies, and then assumes majority-rule in elections and voting. He advocates abandoning territorial representation altogether, arguing that large constituencies do not represent local communities anyway, but rather produce legislators more concerned with pushing ‘local pork’ than promoting the common good71. As an alternative, he proposes random constituencies: on coming of age, every American will be assigned to one of 435 constituencies, which will be theirs for life. Consequently, each representative will truly stand for a cross-section of the nation, providing them “self-regarding incentives to act as if they cared about the common good”72. Rehfeld is ready to endorse randomization – at one level – to address a perceived problem of contemporary democracy. His critique of territoriality – developed with lengthy reference to its historical justification and development (his chapters 3-6) – seems sound. He is right to point out that territorial representation – even if deeply embedded in the US by the federal system – is not necessary and that the slogan ‘all politics is local’ does not justify such constituencies because it is true in consequence of the system73, and we may instead favour representation by occupation or ethnicity74. The deeper problem is that, whether based on locality or other natural interest groups, competing partisan interests – the ‘pluralist’ model – 70 71 Rehfeld (2005). Rehfeld (2005) pp.8, 21, and 152. 72 Rehfeld (2005) p.xiv. 73 Rehfeld (2005) pp.8, and 152. 74 Rehfeld (2005) pp.36-9, 146, and 159. 26 results in a legislature divided over ‘zero sum’ issues, rather than united in pursuit of a truly common good. Rehfeld’s solution is to make each constituency a random sample of the whole nation, so each representative is accountable to a microcosm of the whole electorate. The result is that if blacks make up 15% of the whole population, they will be 15% of each constituency75. This seems to deny them the special protection of gerrymandered districts that effectively ensure black representation; but Rehfeld suggests that some minorities may be better served by having a voice in all districts, rather than controlling a few but being anonymous in most76. This is an empirical question, but many would be unhappy with the consequence he envisages and accepts – that a 51% majority of the nation, by becoming a 51% majority of each constituency, could win 100% of the seats77. In commenting on PR and group rights, Rehfeld says: “If constituencies are defined by their members’ similarity of voice (if African American representatives, for example, come from predominantly African American districts), then we promote diversity of voice within a representative body by denying it within the constituency. The demand that representative bodies should be diverse thus subordinates the deliberative diversity within a constituency to that of the legislature. Yet, if good and proper deliberation requires that all voices are heard, then it would appear that we have to choose between diversity within the legislature and diversity among their electoral constituents. Or, in terms of exclusion, the question becomes, do we exclude “voice” from the representative body itself, or from the constituent groups who select their representatives?”78 He may be right that a diverse legislature is often ensured by creating homogeneous electoral groups, and further that such groups (exposed only to their own viewpoints, not others) may radicalize, making legislative compromise harder. 75 76 Rehfeld (2005) p.214. Rehfeld (2005) p.11. 77 Rehfeld (2005) pp.227, 231, and 244. 78 Rehfeld (2005) p.27. 27 Nonetheless Rehfeld’s preference for diverse constituencies seems to rest on an unjustified assumption – empirical or normative – that democratic deliberation has to take place with fellow constituents79. Of course, this depends on whether political discussion focuses around general issues – such as the importance of the environment – or particulars, like specific candidates. If Rehfeld is envisaging the latter, then it is more likely that one will discuss the merits of particular candidates with others who face the same electoral choices. Until recently, he suggests, the need for discussion between constituents constrained us to local constituencies; non-territorial constituencies such as he recommends only became possible with mass media and particularly the internet – through which he imagines most debate and campaigning taking place80. While it is true that the internet has allowed for much democratic debate, such as political blogs81, it seems unlikely to me that citizens would deliberately seek out others with whom they had little more than a randomly-assigned constituency in common – it is far more likely they will use new technology to converse with those local to them and/or sharing similar interests or views. Leaving aside these problems for now, and assuming that Rehfeld’s goal of diverse constituencies is normatively desirable and that his proposal is practically feasible, the fact that a single group – even a majority – could seize the whole legislature82 still seems problematic. Even when individual members of that majority are exposed to competing minority viewpoints in their constituency, one has to wonder how much check that will really be. It seems that legislators will only have to 79 80 Rehfeld (2005) pp.51, 118, 129, and 215-7. Rehfeld (2005) pp.162, 174-6, 217, and 243-4. 81 E.g. a few quick links take me to (amongst others): http://publicreason.net/, http://virtualstoa.net/, http://crookedtimber.org/, http://considerphlebas.blogspot.com/, http://bensaunders.blogspot.com/, http://www.makemyvotecount.org.uk/blog/… 82 Rehfeld (2005) pp.227, 231, and 244. 28 appeal to the national majority – e.g. whites83 – to be sure of a majority in their sample constituency. While Rehfeld seems happy to accept majority-rule in his random constituencies, this is neither necessary to his project nor conducive to a common good that includes everyone. In his defence though, it should be pointed out that Rehfeld is at pains to stress that defining districts is something independent from, and prior to, determining voting procedures84. Rehfeld’s commitment to majority-rule can therefore be regarded as only provisional for, once a district is randomly-constituted, it is still an open question whether to adopt majority-rule85. Lottery-voting could be an attractive mechanism to complement random constituencies, because it would allow us to combine them with a diverse legislature. Rehfeld criticizes existing methods of proportional representation for choosing a diverse legislature at the cost of homogeneous constituencies86, but he recognizes that we might want diversity in both constituencies and legislatures87. His solution is a quota system, guaranteeing a certain number of, e.g., black and female representatives88. This creates one kind of diversity, but they will still all be black or female Republicans – it is not obvious this will result in much diversity of political viewpoints. I think the groups that matter for representation are not those based on ascriptive features such as sex or ethnicity, but those people identify with by their voting. While random constituencies and majority-voting produce a legislature dominated by the majority, it is quite possible to adopt randomly-assigned 83 84 In his futuristic utopia, Rehfeld assumes whites will be a minority (p.241). Rehfeld (2005) e.g. pp.7, and 21. 85 Rehfeld (2005) p.7: “Maybe they would use majority rule or plurality rule. Maybe they would select a representative by lottery. Our concern here is not, then, with voting rules or the questions of singlemember or multimember representation. It concerns the prior question of how constituent groupings themselves affect the legitimacy of a political regime.” 86 Rehfeld (2005) pp.13, and 26. 87 Rehfeld (2005) p.235. 88 Rehfeld (2005) pp.337-8. 29 constituencies (a la Rehfeld) and lottery-voting (a la Amar’s original proposal89), and thereby produce diverse constituencies that mirror the whole nation, and a legislature that also includes members of all these groups. Of course, since both procedures rely on randomization, neither result is logically guaranteed – but the large numbers concerned make these generalizations pretty much absolute. In pursuing heterogeneous constituencies, Rehfeld neglects the importance of different political views (as opposed to group representation) within the legislature itself. If the aim is to give representatives an incentive to pursue a common good, then any electoral mechanism that only requires them to appeal to a majority seems to cast doubt on this. However, if Rehfeld’s radical proposal of random constituencies is combined with another level of randomization – in the election of representatives from these constituencies – then we can get diverse constituencies and legislatures. Again, lottery-voting was not something considered, but here it helps address what seems a grave problem, and thereby better serves Rehfeld’s ultimate goal. (0.8) Plan of the Thesis Having explained what it is I want to argue, and why it is important, I turn now to how the argument will proceed. Apart from the above outline, lottery-voting itself is barely introduced until chapter 4, in what begins the second part of the thesis, because it is something I wish to argue to, rather than from. Thus it will be helpful to set out here the stages through which the argument proceeds, so that its overall structure can be appreciated in advance – and readers will hopefully be reassured that many questions they have now will be addressed later. 89 Amar (1984). 30 Chapter 1 introduces the notion of ‘rule by the people’. Coordination can often make all better off, but since one pattern of coordination may benefit some more than others, there may be a conflict of interests over which system of cooperation should be imposed. Equal treatment of each person’s agency amounts to democracy, the form of collective decision. However, while many have simply assumed that democracy and political equality straightforwardly imply majority-rule, this is not so. There is no ‘natural necessity’ to majority-rule and, even if the greatest force would be likely to win out, e.g. if the decision came down to violence, we cannot reach a normative justification from an empirical fact. Having accepted the principle of political equality, we still need to know how it is to be institutionalized. Majority-rule is popularly supposed to realize political equality, but it is not obvious that it is the only way to do so, so we still need an argument to justify it. The second chapter takes up the first family of arguments for majority-rule, namely broadly ‘utilitarian’ arguments that it will best serve the general interest. Majority-rule and utilitarianism are often associated, because both seem to aim at the ‘greatest happiness of the greatest number’. In fact, however, I argue that the two can diverge because they understand this differently; utilitarianism emphasizing the great happiness and majority-rule the greatest number. Consequently, majority-rule cannot guarantee the utilitarian outcome properly understood, since it neglects the intensity of preferences and interests that are not voted for, which means the outcome supported by the majority need not be best after all. Further, even if utilitarian outcomes are reached, this is an inadequate conception of fairness, because while it treats each person’s satisfaction as equally important it does not guarantee any equality of outcome – if there is a permanent majority, they may be satisfied each time, at the expense of the minority. 31 There is, however, an alternative consequentialist-based argument, offered by epistemic democrats. Those writers argue that we can identify ideal outcomes – which may be simply utilitarian or include considerations of justice – independently of people’s votes. They argue that, if voters seek to realize such outcomes, rather than merely furthering their personal interests, and we assume the majority are more likely to have successfully identified them, then majority-rule is justified. There are a number of problems with this line of reasoning, however. Requiring voters to identify and vote for impartially best outcomes places great epistemic and moral demands on them and it is far from clear that a majority are necessarily more likely to be correct. Most fundamentally, however, I have argued that democracy is a matter of adjudicating between competing interests, so the ideal of an ‘impartial social best’ will be indeterminate – while all agree on the need for coordination, they disagree over which scheme of such they prefer. This means we need to look not simply at the outcomes produced by majority-rule, but whether it is fair to all involved. The idea of procedural fairness is taken up in chapter 3. It is often argued that majority-rule is fair because it gives each person an equal chance of satisfaction. This is only true though if each person has an approximately equal chance of being in the majority – that is, if the composition of majorities and minorities is fluid and effectively random. If there is a permanent majority/minority divide, then the same people will win or lose each time, and the minority do not enjoy even prospective equality. In such cases, it is hard to accept that majority-rule really does treat everyone equally. If we want to give everyone an equal chance then we can employ a lottery. While it is obviously fair to toss a coin to decide between two people’s competing claims, however, it is unclear whether it is still fair to give equal chances to groups 32 that differ in size. John Taurek defends coin-tossing in such cases on the grounds that it gives each person an equal chance of satisfaction90 and Estlund also points out that random decisions treat each person equally91. This seems to realize one form of equal treatment, by giving each a 1/n chance of satisfaction, where n stands for the number of available options (rather than voters). Nonetheless, such a lottery seems less than democratic because it neglects the preferences of voters, meaning it makes no difference whether votes are split 50/50 or 90/10 between two options – either way, each is given an equal chance. I think this defect results from regarding voters merely as patients, rather than as agents. I argue that the most appropriate way to fairly adjudicate between unequally-sized groups is a weighted-lottery that proportions each group’s chances to size – so one-third of voters would have a one-third chance of victory. This means that each person counts equally, in contributing their 1/m chance of victory to whichever option they vote for, where m stands for the number of voters rather than options. In this way, each person’s preferences have some impact on the process, however others have already voted. Following this abstract account of equality, chapter 4 turns to how it may be implemented in democratic practice. Here I develop the brief account already given of lottery-voting, noting that it draws both from ancient Athenian use of lotteries and modern ideas of voting. Everyone casts a vote but, instead of the option with most votes winning, a single randomly-selected vote determines the outcome. This means that, even when the distribution of preferences is predictable beforehand, each individual still has an equal chance of being decisive. This differs from the simple lottery rejected in the previous chapter, since each person’s chances of satisfaction depend on the amount of support their favoured alternative enjoys. Such a scheme 90 91 Taurek (1977) p.303. Estlund (2007) pp.6, and 82. 33 therefore realizes proportional rather than equal chances. Moreover, I argue that the contingency of outcomes it creates satisfies certain understandings of both democracy and fairness. This account is developed by suggesting instances of small-scale decisionmaking in which such a procedure seems particularly appropriate. For example, if a small group are considering purchasing a set of books, and have to make decisions over whether to do so, how much to spend and what to buy, there is no obvious way to make all of these decisions. If they set an amount and then vote whether to spend that much or not they may get a different answer from if they vote whether or not to spend something and then try to decide an amount. Moreover, it will be hard for voters to decide whether or not to back one decision without knowing the outcome of the other – it may make perfect sense, for example, to support high spending but think low spending literally worse than nothing. Lottery-voting offers one way of treating everyone equally – we simply have each person declare their preferred solution (how much to spend and how) and have one of those outcomes selected by lottery. This can be considered a ‘random dictatorship’, but it treats everyone equally, allows each to express their preferred choice and means that popular preferences have a greater chance of victory (even if no one expresses exactly the same combination, those with similar tastes are more likely to get something close to their ideal). Chapter 5 further considers the practicalities of such a proposal. I do not so much advocate particular solutions as explore the possibilities of lottery-voting, e.g. whether it is combined with open or secret voting and whether vote counts are announced or merely the winner. These matters will, of course, impact on how lottery-voting fits into wider democratic practices, such as deliberation and collusion. I point out that lottery-voting has certain advantages, for example it may encourage 34 deliberation because no group short of everyone is ever guaranteed victory so all – both majorities and minorities – have incentives to try to convert others to their cause, thereby increasing their chances of winning. Further, I consider the worry that extreme minorities may win, much to everyone else’s dissatisfaction. Again, there are various ways this can be addressed – for example, we might make constitutionally-protected rights immune to any form of democratic decision-making or exclude small minorities of votes by imposing a threshold before any option receives a chance. On the other hand, we may simply accept such results as democratic. We cannot be committed to giving all voters equal chances of success and complain that the outcomes are not to our liking when people vote in ways we disagree with – if people vote for extremists, then extremists may win. This approach need not be as dangerous as might be thought, because a lot of extremist voting, e.g. for the BNP, may be no more than protest voting. It is important to remember that how people vote may be an endogenous consequence of the voting system. Where only a majority will ever win, it is easy to cast a vote for a small party with little consideration. Lottery-voting, however, requires each person to consider seriously that whatever option they vote for may actually win, and thus show more responsibility in their voting – which means lottery-voting seems to better-fit defences of democracy based on its moral and educative effect on citizens. Having described and developed the proposal at some length, the final two chapters turn to evaluating lottery-voting against what seem reasonable normative criteria to demand of any collective decision-procedures. Chapter 6 focuses on the formal axioms outlined by May and Arrow92, namely decisiveness, neutrality (no option is favoured), anonymity (no voter is privileged), positive responsiveness, 92 May (1952), Arrow (1963 [1951]). 35 universal domain (admissibility of all possible individual preference orderings), Pareto (the social decision should respect unanimous improvements), independence of irrelevant alternatives (a decision between two alternatives should depend only on those options and not their ranking against non-feasible possibilities) and nondictatorship. I argue that the most normatively-compelling are those straightforwardly connected with equality, such as anonymity and neutrality, and these are satisfied by lottery-voting. The most problematic is independence of irrelevant alternatives, though this is partly due to confusion about what Arrow’s condition actually requires and difficulty applying it to a random decision-making method. Arrow’s conditions are designed to be applied to deterministic, rather than random, processes, because he is really concerned with defining a social preference based on what individuals prefer, rather than an actual electoral rule. If lottery-voting violates his conditions, it is because it operates within a different – pure procedural – conception of democracy, according to which outcomes are legitimated because they come from the procedure, rather than because they conform to some prior standard. It is no embarrassment for lottery-voting that the outcomes could have been different, for this is what gives the procedure its fairness. Finally, I conclude this chapter with two further results of interest to those working in social choice – firstly, lottery-voting is strategy-proof, because it is never in a voter’s interests to misrepresent their preferences, and secondly it facilitates weighted-voting, should that be thought desirable. My final chapter addresses the worry that deciding by lottery is simply irrational. Arrow assumed that society should act like a rational individual, and therefore be able to order all possible alternatives and choose the highest-ranked, maximizing social welfare. Chapters 1 and 2, however, developed an alternative conception of democracy, tying the importance of political equality to fairly resolving 36 conflicts of interests and, in such a context, it is not obvious that we can talk of certain options being ‘socially preferred’ – the reality is that option x may be better for some and option y better for others. When we are dealing with competing interests, I assumed it is fair to decide between them by lottery, but this does not mean that the chosen outcome is better – indeed, its legitimacy comes from the fact that either outcome could be chosen. I argue that no decision-mechanism is inherently rational or irrational, what matters is whether it is rational for us to adopt a given rule for our present purposes. It can be rational to use a lottery, for instance to break a tie or resolve conflicting interests and, if my preceding argument is successful, it can be rational to adopt lottery-voting for making group decisions in certain contexts. Further, this should not be seen as a complete abdication of reason because each voter has to make an intelligent and responsible choice before the lottery comes into play. 37 1 Democracy as Freedom and Equality “[W]e are so deeply imbued with the ethic of majoritarianism that it possesses for us the deceptive quality of self-evidence”93 “Majority rule is justified only as a means of achieving political equality”94 (1.1) Rule of the People ‘Democracy’ comes from the Greek demokratia, meaning rule by the people. Maybe this etymology no longer tells us much about the term in modern society, however. Familiar forms of government in the early 21st century are far-removed from what the Athenians would have recognized – being in some respects intuitively more democratic (for instance, the much wider franchise), while in others perhaps less (for example, restricting meaningful participation for most to electing representatives). Nonetheless, I take it that this does reveal one crucial feature of democracy, namely that it is rule by the people. Many still associate democracy with the three conditions famously presented by Lincoln – “government of the people, by the people, for the people”95 – but only the second appears distinctively democratic. Government of the people is ambiguous, but means either simply rule by the people (as in the ‘rule of Henry VIII’ or ‘dictatorship of the proletariat’) or rule over the people, in which case it applies to all forms of government. Government for the people, meanwhile, refers to its ends rather than means. While communist dictatorships have sometimes been portrayed as democratic in this sense96, it does not seem sufficient for democracy, since even Plato’s Wolff (1976) p.42 [not emphasized in original]. Dahl (2006) p.16 [not emphasized in the original]. 95 Lincoln (1991) p.104 [‘Gettysberg Address’, 19/11/1863]; c.f. Churchill (1974) p.7565 and Sartori (1987) pp.34-5. 96 E.g. Macpherson (1966) pp.12-22. This claim is criticized implicitly by Ross (1952) pp.75-6 and 91, who insists democracy is a matter of procedure, how decisions are taken not what they are, and explicitly by Lively (1975) pp.33-5, Holden (1988) p.83, and Sartori (1987) p.35. 94 93 38 Guardians may be described as rule for the people. The problem with any benevolent dictatorship is that it effectively claims to tell the people what they really do or should want. If we believe, however, that each person is the best judge of their own interests, then rule by the people will realize rule for the people. Moreover, if we take Lincoln’s three conditions as each necessary and jointly sufficient for democracy then, even if a benevolent dictatorship really does rule for the people, while it may have much to commend it, it is not democratic. Note that, on this reading, it may also be that rule by the people can fail to be sufficient for democracy – perhaps, for instance, if the people are uninformed or pursue their own interests rather than the general interest. I have little to say on these other issues or the wider definition of democracy. I do not address who should be enfranchised; my focus here is simply on the requirements of government by the people, that is how a given group of people may decide among themselves as equals. I take decision-making by equals to be a defining feature of democracy, though it is not sufficient for or restricted to such. The question that occupies this thesis, therefore, is how the people are to decide. There is no difficulty when a single monarch or dictator makes a decision for a group97. The problem addressed here is one of collective or group decision-making. Except in the easy – and usually unlikely – case of unanimous consensus, some way must be found of turning differing individual opinions into a group decision. Note again that this is not specifically a democratic problem; whatever group makes a decision needs some decision-rule. While a democratic context is concerned, equality between decision-makers could also be a property of an oligarchy. Whether or not all are involved in making group decisions, the problems of arriving at a single decision from any (non-unanimous) group are well-documented: Can we really say those who 97 Unless, of course, they want their decision to reflect people’s preferences. See Arrow (1984) pp.55- 6. 39 do not get their way are self-governing? And on what basis, if any, can we hold these losers bound by a decision they do not agree with? (1.2) The Possibility of Self-Government If our aim is for the people to be free, then the typical liberal solution is for decentralized decision-making. Those who take this to its logical extreme typically argue there is little, if any, role for the state. However, if there is no enforcement then it seems the only freedom realized is that of the strong to oppress the weak. If freedom is important, then it is something we want to distribute equally. Moreover, there are some decisions that simply cannot be disaggregated or that, if they are, are likely to result in a worse situation for everyone. We cannot leave each to determine their own foreign policy or legal system, for example – these decisions are necessarily collective matters and must be decided collectively. Thus it seems reasonable to assume that certain decisions need to be made collectively. We need collective coordination because it makes all better-off. It makes little difference, prior to the adoption of a shared norm, for example, whether we drive on the right or left, but we want everyone to drive on the same side of the road, whichever it is. This is a pure coordination problem that can be represented in game form as follows: Fig. 1.1 Driving Co-ordination You drive on left I drive on left I drive on right 5,5 0,0 You drive on right 0,0 5,5 This is clearly a case where we want centralized decision-making, and do not really care what the decision is. While there is no particular need for this decision to be 40 democratic – if some illegitimate dictator had issued the directive that all should drive on the right then we should all be morally and prudentially obliged to obey it, provided we believed sufficient others would also – we cannot simply assume that order will emerge spontaneously. Were the British government to disappear, we may continue to drive on the left, simply because that is the more salient solution to our problem, but this would not apply to cases where coordination had yet to be established and there was no agreed focal point – either because no solution was at all salient or each of us had different preferences. Suppose we want to go for a meal together, but have different preferences over food – specifically, I like Indian while you prefer Chinese. Assume it is more important to each of us that we eat together than that we have our preferred meal, thus this is still a coordination problem, except that now it is no longer a matter of indifference which option we coordinate on. This can be represented as follows: Fig. 1.2 Restaurant Co-ordination You go to Indian I go to Indian I go to Chinese 6,4 0,0 You go to Chinese 2,2 4,6 As we can see, both of us would agree that we should go to the same restaurant, but we disagree which that should be98. If you think I will go to the Indian, then you should go to the Indian, but you would rather we both go to the Chinese, and think I am more likely to go to the Chinese if I expect you to go there. This is a case where, although all benefit from coordination, there is a conflict of interests over which pattern of coordination is to be chosen. Both of us want coordination, but we each prefer a different pattern of such. In this case, assuming we can decide together, it 98 Pettit (2000) makes this point when he observes “Matters of common, recognizable interests can often be advanced in different ways, where one way is more costly for this group, a second more costly for that, and where the different groups will prefer different approaches” (p.118). 41 seems plausible that the fairest solution would be for us to commit to going somewhere together and then to toss a coin to decide which restaurant to go to, giving us equal chances of getting our preferred option99. This problem is easily solved because of its artificial simplicity. If there are two parties with equal but competing interests, then tossing a coin seems a fair way to resolve their differences. The more interesting problems are when we deal with larger groups, where numbers on each side are unequal. Suppose, for example, a group have regular meals together, and five of them favour Indian and three Chinese – how now should they coordinate? In this case, decentralization may be more attractive (five going to the Indian, the other three for Chinese), but assume coordination is still more important. Is it still fair to toss a coin, giving each person an equal likelihood of getting what they want? 100 Or is it fair to let the majority decide? And, if so, does this mean the group should always go for Indian, to the displeasure of the permanent minority? These are among the issues that the following thesis will address. (1.3) Equal Relations and Respect I take it that ‘rule by the people’ means rule by all of the people – for it is the wider franchise that distinguishes democracy from oligarchy. This is all very well if there is unanimity, but what can rule by all the people mean in cases of disagreement? If people disagree, as in these examples, they cannot all have all their own way. Even though all get what is most important to them – coordination – the question which coordinated solution is to be imposed on all, e.g. whether the group goes to the Indian or Chinese restaurant in this case, still raises issues of fairness to each person. In such 99 This assumes a one-off decision. If we have a series of meals together, it would be fair to alternate between Indian and Chinese. I also exclude the possibility of compromise, e.g. Italian. 100 This is the argument of Taurek (1977), which is considered in chapter 3.5. 42 cases, the most reasonable solution seems to be equality: all should have, in some way, equal impact in decision-making or chances of being decisive (see the quotations in my introduction, sections 0.1-2). Democracy, as Dahl defines it, is a system of political equality and popular sovereignty101. The latter condition is necessary because a system where people have no power is still one where they are equal. Here, this is taken for granted, so the focus is simply on political equality, which I think is best defended if we conceive of decision-making as a distributive process, determining who gets their way over what. Whatever value democracy serves – for instance, liberty, power, pursuit of interests or self-development – is one that we assume all have interests in. If all have interests at stake in collective decisions, then it is generally supposed that all should have a say in them102. Thus, it is no surprise that democracy seemed to evolve, in ancient Greece, from the idea of isonomia (equality before the law)103. Moreover, while the spread of democracy has certainly not been a simple, teleological progression, equality has continued to figure prominently in calls for democracy – as in Rainborough’s remark that “the poorest he that is in England hath a life to live, as the greatest he”104 – and still occupies a central place in modern discussions of democracy. Ranney and Kendall, for instance, observe that: “One characteristic that most persons regard as essential to democracy is political equality. A familiar way of describing this trait is ‘one man, one vote,’ which we take to mean that in a democracy political power must be equally shared by all its citizens, and no man should have a larger share than any other man”105 While Dahl and Lindblom write that, in democracies: 101 102 Dahl (1956) pp.34, and 37ff, c.f. Christiano (1996) p.3, McGann (2006) pp.5, and 203. For the ‘all affected’ principle, see Whelan (1983) p.19, Goodin (2007) pp.50-55, and Arrhenius (2007) pp.12-4. Note, this does not assume all have equal interests at stake; merely that all equally have interests at stake. 103 Holland (2005) p.134. 104 Woodhouse (1938) p.53. 105 Ranney and Kendall (1956) p.16; c.f. Still (1981) p.376. 43 “Control over governmental decisions is shared so that the preferences of no one citizen are weighted more heavily than the preferences of any other one citizen… In elections the vote of each member has about the same weight”106. While justifications for and interpretations of equality differ, it is generally assumed that we are all egalitarians now107, and this is certainly the case when it comes to political equality – almost everyone claims to be (correctly understood) a democrat108. Given this point in our development, I do not feel the need to defend the premises of democracy, popular sovereignty or political equality. Nothing here is intended as a complete argument for democracy, though this chapter has sought to ground it in finding a socially optimal pattern of coordination that respects each individual’s differing preferences equally. This thesis is concerned not with justifying political equality but its correct interpretation and institutional requirements. It is putting abstract principles like equality into practice that usually leads to disagreement. For example, does equal concern for each person mean that each person’s utility should count equally in some maximizing system? Or that each should be ensured equal welfare? Or equal opportunity for welfare, or equal resources, or simply equal liberty to pursue their own interests?109 Similarly, there are varying interpretations in the political context – for instance, does political equality mean equality only of votes or of all influence over decisions?110 Should representation be proportional or majoritarian?111 What does equality require of 106 107 Dahl and Lindblom (1976 [1953]) p.41, and p.277; c.f. Still (1981) p.375. For the so-called ‘egalitarian plateau’, see Dworkin (1977) pp.179-83, Kymlicka (2002 [1990]) pp.3-4, and Sen (1992) pp.12-16. 108 Hyland (1995) p.36, Woodruff (2005) p.6. 109 These possibilities are intended to roughly characterize the positions of classical utilitarians, welfare egalitarians, Arneson, Dworkin and libertarians, respectively. 110 Dworkin (2000) pp.191-8. 111 Hyland (1995) pp.94-100, McGann (2006) pp.35-59. 44 electoral districts?112 Can it require weighted voting, to respect inequalities?113 And, the issue here addressed, how do we best treat all votes equally? (1.4) The Alleged Obviousness or Necessity of Majority-Rule Many have thought it obvious that if each vote is to count equally, then the majority must hold sway114. This move is clearly made by Robert Dahl; following the quotation from the previous section, Dahl and Lindblom go on to add: “To say that the votes of the greater number should not prevail is to say that political equality is impossible, or that it is undesirable, or both… [U]nless government policy responds to the preferences of the greater number, the preferences of some individuals (the lesser number) must be weighted more heavily than the preferences of some other individuals (the greater number)”115. Abraham Lincoln, whose characterization of democracy served as our starting point, also thought majority-rule a practical necessity, declaring: “Unanimity is impossible. The rule of a minority, as a permanent arrangement, is wholly inadmissible; so that, rejecting the majority principle, anarchy or despotism in some form is all that is left”116. Similar thoughts are often expressed, particularly in popular thinking117. Some have even gone so far as to define democracy in terms of majority rule, for example, Carritt says: “Democracy is government of the whole people by a majority and it may be carried out either for the whole people or merely for the majority”118. This is an illegitimate jump. Democracy should be defined in terms of political equality, which may turn out to require majority-rule (this appears to be Lincoln’s reasoning), but there may be other ways of treating everyone equally. While it seems natural to conclude that, if all are to count equally, then more people must somehow count for 112 113 Balinski and Young (1982), Still (1981). Brighouse and Fleurbaey (2006), Heyd and Segal (2006). 114 C.f. Waldron (1999) pp.129-30. 115 Dahl and Lindblom (1976 [1953]) p.44. 116 Lincoln (1991) p. 58 [‘First Inaugural Address’, 04/03/1861]; c.f. Holden (1988) pp.39-40. 117 Hyland (1995) pp.88-100, Berg (1965) p.140. 118 Carritt (1947) p.150. 45 more119, it is a further and unwarranted step to conclude from this that the majority must get all their own way120. If we accept this inference, then it is not clear we can distinguish democracy from majority tyranny121. Perhaps one reason for thinking it ‘natural’, or even ‘necessary’, that the majority get their way is because of the greater physical force of greater numbers. It is plausible to imagine that majority-rule might have arisen out of need for peaceful conflict resolution. While Clausewitz famously said “war is nothing but a continuation of political intercourse”122, disagreements were resolved by physical conflict long before they were resolved by what we would regard as civilized politics, and the more likely truth is that politics (in the form of civilized debate and voting) is a continuation of war by other means123. Although it is somewhat fanciful, it is easy to imagine that generals of opposing sides were often aware who was better-placed to win the battle and, where the outcome was predictable, it made sense for the likely losers to defer without battle – to surrender without bloodshed, rather than fight to defeat or death124. Hobbes, for example, argued: “[I]f the Representative consist of many men, the voyce of the greater number, must be considered as the voyce of them all. For if the lesser number pronounce (for example) in the Affirmative, and the greater in the Negative, there will be Negatives more than enough to destroy the Affirmatives; and thereby the excesse of Negatives, standing uncontradicted, are the onely voyce the Representative hath”125 This interpretation was certainly endorsed by Henry Thoreau, who claimed: 119 120 Parfit (1978) p.301. We could, for example, believe in compromise – see, e.g., Hyland (1995) pp.96-8 and Lijphart (1977) pp.38-41. 121 De Tocqueville (1994 [1835]) pp.254-62 [Democracy in America I.15], Emerson (1998) p.1. 122 Von Clausewitz (1997 [1832]) pp.357 [On War VIII.vi.B]; c.f. pp.22-3 [I.i.24, 26]. 123 This point is also made by Sartori (1987) p.42. 124 Przeworski (1999) p.48 suggests voting is a proxy for war. Sartori (1965) p.335 and (1987) p.343 quotes the expression ‘it is better to count heads than to break them’, which also appears in Riker (1982) p.243. Ross (1952) pp.96-7 considers the suggestion that democracy is a peaceful means to resolve conflict, but points out there’s no need for all to agree to democracy to resolve conflict peacefully. 125 Hobbes (1985 [1651]) p.221 [original pp.82-3, ch.16]. 46 “[T]he practical reason why, when the power is once in the hands of the people, a majority are permitted, and for a long period continue, to rule is not because they are most likely to be in the right, nor because this seems fairest to the minority, but because they are physically the strongest”126 This conjecture would also explain why the extent of the franchise was often somehow related to military service – for instance, Athenian democracy is often attributed to the need for manpower in their navy and, throughout history, the franchise was generally restricted to men, or even the rich (who were likely better-fed and -equipped), all of which could be justified by their military use127. Of course, the relationship between weight of numbers and military victory is only contingent – the larger army does not always win, especially if their opponents are better trained, organized, equipped or positioned. The problem, however, is not that this reasoning does not always favour the majority, for we can concede that other things being equal a larger army is more likely to win. Rather, the objection to this line of argument is that it does not offer much of a normative justification at all. If majority-rule originated simply as a proxy for ‘who is most likely to win, if it comes to fighting’ then it seems no more than an appeal to ‘might makes right’, which violates Hume’s 126 127 Thoreau (1993 [1849]) p.2. The link between Athenian naval power and democracy is drawn in Ober and Hedrick (eds.) (1996) p.9, Holland (2005) pp.166 and 217, and Woodruff (2005) p.25. Those who fought, even slaves, were often granted citizenship rights. See Aristophanes (1919) pp.104-5, especially note to line 694, and Aristotle (1959) p.282 [Ath Pol XIV, ch.40] and (1992) p.308 [Pol V.4 (1304a21-4)]. The association of political rights and military service in England dates at least to the civil war. Ramsborough’s demands in the Putney debates, for instance, seemed connected not only to the fact that each was subject to the laws but that men had fought for them. Woodhouse (1938) pp.56, and 61. C.f. Crawford (2001) p.197: “The category ‘citizens’ sometimes excluded women because it could refer to men who bore arms for the defence of their country”. In 1816, William Cobbett called for the franchise to be extended to all who pay tax, (1944) pp.214-5 (more precisely, to those who pay direct taxation, but it is clear this limit is a matter of practicality, not principle), but he connected this to military service, saying, “As it is the labour of those who toil which makes a country abound in resources, so it is the same class of men, who must, by their arms, secure its safety and uphold its fame”, p.207. He approved Samuel Bamford’s proposal to use militia lists for voting registers, Bamford (1967) p.19. Again, practicality may have figured, but Colley’s account, (1992) p.318, explicitly connects the reasoning to fighting; “if all adult men were worthy to fight for Great Britain, then surely they had the right to take part in its politics as well?” I thank Robert Poole for leading me to these references. Even the relatively recent extension of the franchise to women could be attributed to a recognition of their potential role, either in supplying the next generation of soldiers or – in modern ‘total war’ – taking the men’s places in factories. 47 dictum that we cannot derive evaluative conclusions from purely factual premises128. That one side would likely win is no reason to conclude that they should129, just as we do not think it legitimate of me to impose my preference for an Indian restaurant on you because I am stronger and would win in a fight. One possible solution would be to recast the idea of ‘force’ in terms that don’t refer to violence, as when one side wins due to the force of their argument. Locke, for instance, appeals to the natural laws of physics, claiming that: “[W]hen any number of men have, by the consent of every individual, made a community, they have thereby made that community one body, with a power to act as one body, which is only by the will and determination of the majority: for… it being necessary to that which is one body to move one way; it is necessary the body should move that way whither the greater force carries it, which is the consent of the majority: or else it is impossible it should act or continue one body… [Thus] the act of the majority passes for the act of the whole, and of course determines, as having, by the law and nature of reason, the power of the whole”130 This argument suffers multiple flaws. Firstly, if it means that majority-rule is a natural necessity, then there should be no need to argue for it – but clearly a body of people need not always move in the way suggested by the majority. If, however, it is proposed that we should follow the majority, because this is how physical bodies act, then it is again guilty of deriving normative conclusions from empirical premises, and arguably irrelevant ones at that. Moreover, the crucial analogy is flawed; firstly because it is not obvious that the majority-will is the strongest force – an intense minority could be stronger, and there is no reason to associate mere numbers with 128 129 Hume (1978 [1739-40]) p.469 [Treatise 3.1.1]. Rawls (1999 [1971]) p.116: “it is to avoid the appeal to force and cunning that the principles of right and justice are accepted. Thus I assume that to each according to his threat advantage is not a conception of justice”. 130 Locke (1980 [1689]) p.52 [Second Treatise §96]. Note he does not here say they consent to majority-rule; he says they consent to form one society, which makes majority-rule necessary. Waldron (1999) pp.130-50 dubs this the ‘physics of consent’; c.f. Risse (2004) p.47. 48 greater force131 – and, secondly, because the ‘physics’ of the argument is wrong in any case132. Bodies do not move only under the influence of the greatest force exerted on them, but their final motion vector is the result of all forces; if there is a great force pushing left and a smaller one pushing up, the body will move left, but also up. If the natural motion of physical bodies under forces supports anything, it is more likely some form of proportionality. Certainly it does not seem that majority-rule can be justified on grounds of physical necessity. (1.5) Contract and Consent Even if there is nothing necessary about majority-rule, as Locke assumed, it may be that parties would agree to such. Decision-rules are not unalterable natural necessities, but tools we adopt for specific purposes (a point that becomes important in chapter 7), so it seems that rule chosen should itself be a matter of agreement. The idea of social contract theory is to model people as free and equal. If we want to design social institutions that respect people as such, then one way to do this is to ask what institutions they would agree to. This model can be applied to either distributive questions or the democratic organization of society. So, if we want a decision procedure that will treat all equally, one way of reaching such will be to ask people – who do not know who they will be in the final society – what rules they would accept133. The application of contract theory to the design of democratic decision making procedures has not been entirely neglected. Madison, in the discussions founding the 131 Risse (2004) p.49 points out ‘no more has been said’ when the same argument is given by more people. 132 The latter point is made by Risse (2004) pp.55-7. 133 Note that, though I term this ‘contractualist’, I am not assuming that what is agreed to constitutes fairness. The contract here legitimizes certain decision rules, and it is perfectly coherent for people to reject certain rules because they are unfair. 49 United States, uses what Elster describes as a veil-of-ignorance argument, defending the Senate on the basis of what “A people, deliberating in a temperate moment, and with the experience of other nations before them, on the plan of Govt” would choose134. Contracts played an important part in founding government for Hobbes, Locke and Rousseau, and the tradition has been extended by modern thinkers such as Rawls135, Scanlon136 and Gauthier137. Wolff points out that many have supposed all would agree to majority-rule at the stage of the original contract – though he insists that this is not legitimating, simply consenting to the surrender of their autonomy138. The next section addresses whether those who contract to form a society would necessarily consent to majority-rule. While it seems there are powerful reasons to agree to some shared decision-rule, since we have seen that majority-rule is not naturally privileged, there seems no a priori reason why it should be agreed to rather than some other procedure, such as a lottery139. Nonetheless, the contract is a useful device for thinking about constitutional design, and such reasoning will be appealed to later, particularly in chapters 3 and 7. Of course, any choice will involve a loss, for while each person’s ideal decision-rule may be their personal dictatorship, equality requires compromise with others. Thus, we may all have to accept what we do not want sometimes in return for getting what we do want on other occasions. If coordination is more important than the particular pattern of coordination, however, then this compromise may allow everyone to get more of what they truly want (see 1.2 above). 134 135 Elster (1993) pp.197-8; for the original see Farrand (1966 [1911]) pp.421-2 [26th June 1787]. Rawls (1999 [1971]) esp. p.10. 136 Scanlon (1998) e.g. pp.5-6. 137 Gauthier (1986) e.g. p.10. 138 Wolff (1976) p.41. 139 Simmons (1993) p.94. 50 (1.6) Contracting to Majority-Rule If we accept the idea that contracts illustrate what is fair, the crucial question is whether people would accept – or contract to – majority-rule; but this is a different claim from the one (refuted in section 1.4, above) that majority-rule is simply the ‘obvious’ or ‘natural’ solution. The two are often blended together, however, perhaps because it is assumed people must (rationally) consent to that which is necessary. Hobbes, for instance, said: “[H]e that dissented must now consent with the rest; that is, be contented to avow all the actions he shall do, or else be justly destroyed by the rest. For if he voluntarily entered into the Congregation of them that were assembled, he sufficiently declared thereby his will (and therefore tacitely covenanted) to stand to what the major part should ordayne”140 There is, so far as I can see, no argument that those consenting to join a society thereby tacitly covenant to accept majority-rule. One possibility is that Hobbes was persuaded by the thought that the greater number are likely more powerful (see 1.4 above), though this may conflict with his preference for monarchy rather than democracy141. Similarly, Locke adds to his ‘physics’ the idea that we would consent to our society behaving in this way. He adds that society should follow “the consent of the majority: or else it is impossible it should act or continue one body, one community, which the consent of every individual that united into it, agreed that it should; and so everyone is bound by that consent to be concluded by the majority”142. The idea is simply that we all want the community to act as one, so we must thereby commit ourselves to acting in accordance with the majority will. He goes on, in the next three sections, to say: 140 141 Hobbes (1985 [1651]) p.231 [original p.90, ch.18] [underlining added]. Hobbes (1985 [1651]) pp.241-8 [original pp.95-99, ch.19]. 142 Locke (1980 [1689]) p.52 [ch.8,§96] [underlining added]. 51 “[E]very man, by consenting with others to make one body politic under one government, puts himself under an obligation, to every one of that society, to submit to the determination of the majority, and to be concluded by it”143 “For if the consent of the majority shall not, in reason, be received as the act of the whole, and conclude every individual; nothing but the consent of every individual can make any thing to be the act of the whole: but such a consent is next to impossible”144 and “Whosoever therefore out of a state of nature unite into a community, must be understood to give up all the power, necessary to the end for which they unite into society, to the majority of the community, unless they expresly agreed in any number greater than the majority”145 Sidgwick also assumes that it is reasonable to consent to majority-rule, pointing out that: “If the majority of a nation are able to modify, in an orderly and regular way, their laws and the action of their government, a minority desirous of change will, ordinarily, be only tempted to resort to physical force when it is hopeless of becoming a majority”146 Tellingly, however, the last two quotations from Locke actually make reference to other possibilities. Like Lincoln, he rejects unanimity as impossible, but it does not actually follow that we have to accept a simple majority, for he allows that we could expressly adopt some other (super-majority) rule. We can agree to act as one, but still choose from many different possible decision-rules, so consenting to act as one only binds us to act on the majority will if either i) that is the only way of deciding (which was rejected in section 1.4) or ii) if that is in fact how we do consent for the whole to act. Thus, nothing in these arguments from consent requires, or even favours, majority-rule. We cannot simply assume people will agree to majority-rule, unless they do so because they already think it is a fair way of resolving their conflicts. As A. John Simmons puts it: 143 144 Locke (1980 [1689]) p.52 [ch.8,§97] [underlining added]. Locke (1980 [1689]) p.53 [ch.8,§98]. 145 Locke (1980 [1689]) p.53 [ch.8,§99] [underlining added]. 146 Sidgwick (1908 [1891]) p.616; also quoted in Kuflik (1977) p.327, fn.6. 52 “Only if majority rule were obviously fairer and more authoritative than lottery, weighted lottery, votes adjusted for intensity, plural votes for the qualified, and the like, would we be obliged to interpret a commitment to political membership as a commitment to majority rule”147 The aim of this thesis is to defend one of these alternatives – viz., the weighted lottery or, as it is called here, lottery-voting – rather than majority-rule. This is not to say it is never rational to choose majority-rule, but only that, at least in some choice contexts, it could be at least as rational for a group of fair-minded, disinterested individuals to adopt lottery-voting to determine their collective decisions. The next two chapters explore two reasons that people might have, in an original contract position, to agree to majority-rule. Chapter 2 addresses claims that majority-rule produces better outcomes, in some vaguely utilitarian fashion, and that people will accept it for this reason. Such outcome-based reasons to accept majority-rule are rejected, because these outcomes are not necessarily better overall and – even if they are – they need not be better for each individual. We saw, above, that democracy offers a way of resolving conflict between socially optimal patterns of coordination, which suggests that democratic procedures must be justified in terms of fairness. Chapter 3 takes up the claim that majority-rule is a fair way of adjudicating between competing claims, arguing that, while majority-rule may be fair in certain cases (namely, where the composition of the majority is effectively random), it is not fair if there is a permanent majority, and what equality really requires is a form of proportionality. (1.7) Proportionality While Locke assumes that the majority speak and decide for the whole, others have thought it is unjust to exclude the minority. Sartori, for instance, stresses that 147 Simmons (1993) p.94. 53 “the people consist, overall, of the majority plus the minority” so we cannot allow absolute, unrestrained majority-rule148. Democracy is ideally about the rule of the whole demos, not simply a majority of them. It is often objected that if the minority got their way, each of them would have to be counted for more than one. For instance, Dahl and Lindblom claim “unless government policy responds to the preferences of the greater number, the preferences of some individuals (the lesser number) must be weighted more heavily than the preferences of some other individuals (the greater number)”149. However, this seems to neglect the fact that, if the majority get all their own way, then they also seem to be counted for more than one. Consider a 60/40 split in a typical ‘winner takes all’ system: Fig. 1.3 Proportionality % votes Majority Minority 60 40 outcome 100 0 The majority get 60% of the vote, but 100% of the outcome150. This means that each vote seems to count not for 1% but 1.67%. Also, it is not only that the majority count for more than the minority – as it might be if the outcome was split, say, 80/20 – but that the minority seem not to count at all – they get nothing for their 40% of the votes. It certainly does not seem that each vote is post facto being treated equally. Of course, post facto equality may be impossible in cases of disagreement, since some people must get their way151. If the outcome is necessarily zero-sum and winner takes all, then it may be least objectionable to let the majority have all their way, for the 148 149 Sartori (1987) p.32. Dahl and Lindblom (1976 [1953]) p.44. 150 Of course, 100% of the outcome may not be all their own way – they may, for example, be voting between proposals that are already compromises, e.g. 70/30 or 30/70. 151 Lively (1975) pp.17, and 24 notes only unanimity ensures complete retrospective equality. 54 alternative – satisfying the minority – is a greater deviation from equality (in this case, each 1% of the vote would count for 2.5% of the outcome). My concern here is not with retrospectively equal outcomes but ensuring that the procedure really does give all a fair chance – and thus with ensuring that no one faces near-certain defeat. The only defence of majority-rule is the claim that each person had an equal chance of being in the majority, so it was not predictable in advance whose vote was to be significant. Chapter 3 will take up such themes, but for now note that the equality of all is not obviously served by majority-rule. As Guinier puts it: “The proportionality principle delivers what majority rule proponents assume but do not produce: decisional rules that promote reciprocity and accountability without straying too far from the efficiency and stability norms… Proportionality seeks a reason for implementing a decision that legitimates the decision in the eyes of all voters, even those who may lose. It asks whether the process provides all voters an equal opportunity to be part of the winning coalition”152 If we want each vote to have the same weight, then our aim seems better-served by some kind of proportional compromise, e.g.: Fig. 1.4 Compromise % votes Majority Minority 60 40 outcome 60 40 I think that most issues are actually amenable to some sort of compromise. To take one example where this is not obvious, while there is apparently a binary choice between allowing and not allowing abortion, there are various ways in which this can be made more or less palatable to either side – for example, rather than making abortion freely available on demand, we could insist that it be performed in the first 20 weeks of pregnancy and require approval of two doctors, or, instead of complete 152 Guinier (1994) p.92. 55 prohibition, we could allow it only for cases of rape and medical necessity. Similarly, the decision whether or not to go to war is an apparent binary, but there are various possibilities such as ‘hot war’, ‘cold war’, simply building up defensive forces, negotiating treaties with other parties, etc. The problem is not generally that fair compromises do not exist. Rather, they require a degree of judgement or sensitivity. A compromise that gives everyone some satisfaction cannot normally be achieved like cutting a cake – for example, allowing every other request for an abortion would seem outrageous. I shall argue, in section 3.13, that compromises are not always ‘mechanically’ possible or desirable on any given issue – even if we can ‘split the difference’ it may not please anyone. There I shall argue that we should seek compromise not over individual decisions, but over either (if we can) the whole series of political decisions or, alternatively, the procedure that decides them – that is, we should ensure that even minorities sometimes get their way, so that no one is permanently excluded. The procedural solution leads to proportionality not of outcomes but of chances. Whether or not this later argument stands, the conclusion so far is that winner takes all majoritarianism does not respect each equally. If our aim is to include all equally, then we cannot complacently assume that majority-rule will achieve this. (1.8) False Dichotomies and Neglected Options I believe the complacent, almost naïve, acceptance of majority-rule comes from a failure to consider all available options153. The possibility of proportionalitypreserving solutions is often neglected. It is regularly assumed that the choice is simply between majority-rule and minority-rule, in which case the former looks 153 C.f. Guinier (1994) p.2. 56 obviously democratic, while the latter appears only to preserve oligarchy. To return to the example from the previous section, if one side are to get everything, then it is obviously closer to proportionality to give the whole outcome (100%) to the 60% rather than the 40% of voters. The fault seems to originate in a tendency to neglect certain possibilities, including lotteries or other proportional arrangements, and simply contrast majorityand minority-rule as if they were the only alternatives. This dichotomy can be traced back as far as Aristotle’s opposition between the rule of the many and rule of the few154 – though in fact he recognized the intermediate possibility of mixed constitutions. More recently, this has been forgotten and, since calls for democracy often began as protests against elite privilege, it has been assumed that the many must rule. This confuses two separate issues: i) the boundary of the demos and ii) the decision rule used within that demos. Democracy may well require the ‘rule of the many’, understood as a wide franchise, but it should not be identified with majorityrule amongst those decision-makers. Firstly, the majoritarian decision-rule is not unique to democracy but may be employed by an oligarchy155. Secondly, there may be other ways in which the demos can make decisions respecting all equally – as shall be argued below. Nonetheless, this confusion has continued to exert a profound influence on thinking about democracy, particularly in popular discourse. Recall Lincoln’s remark, quoted in section 1.4: “Unanimity is impossible. The rule of a minority, as a permanent arrangement, is wholly inadmissible; so that, rejecting the majority principle, anarchy or despotism in some form is all that is left”156. Even granting the premise that unanimity, even after deliberation, is impossible, two issues have been 154 155 Aristotle (1988) p.61 [Pol III.7 (1279a27-8)] Aristotle (1988) pp.85, and 94 [Pol IV.4 (1290a30-3), IV.8 (1294a12-4)], Lively (1975) p.13. 156 Lincoln (1991) p. 58 [‘First Inaugural Address’, 04/03/1861]. 57 confused here. We can ask: i) should rule be the permanent possession of one group or shift between groups, and ii) if one group rules permanently, should it be a majority or a minority. If one group is to have permanent rule, the tyranny of the majority seems preferable to that of a minority, but we can question the initial supposition that any group – even a majority – should be permanently in charge. Permanent government in the hands of any one group – majority or minority – seems to be a licence for tyranny and undemocratic insofar as it violates equality understood as proportionality157. Lincoln is clear that he actually favours shifting rule158, and this seems obviously desirable. Hence there have been long traditions emphasising, for example, rotation in political office159. Once we realize that we want rotation, however, rather than permanent rule of any given group, it is less obvious that majority-rule always delivers. Defences of majority-rule often emphasize that it is not some fixed group – ‘the majority’ – who exercise permanent rule160. Rather, the case for majority-rule is strongest when that majority is a loose alliance, made up of shifting individuals or groups, such that each member of society is sometimes in the majority and sometimes out of it – so while no-one always gets their way, everyone sometimes gets their way. This makes certain empirical assumptions about the nature of society, however, which need not always hold. Aristotle assumed democracy would mean the many poor ruling over the rich few, and in modern societies divisions may form along other lines, such as race, religion, geography, etc161. This is true not only at a national level, 157 158 Madison regarded the majority as simply another faction, van Mill (2006) p.142. Lincoln (1991) p. 58 [‘First Inaugural Address’, 04/03/1861]: “A majority held in restraint by constitutional checks and limitations, and always changing easily with deliberate changes of popular opinions and sentiments” [emphasis added]. 159 E.g. Aristotle (1988) p.144 [Pol VI.2 (1317b1-4, and 23-4)], Carritt (1947) p.152, Berg (1965) pp.150-4. 160 E.g. Sartori (1987) p.33, and Downs (1985 [1957]) p.57. C.f. criticisms by Guinier (1994) pp.4, 9, 17 and 77. 161 Guinier (1994) pp.9-16; c.f. Carter (1994) p.xv. 58 but even within small groups of the sort discussed here – for instance, chapter 4.6-7 uses two examples drawn from student life, and in each case there could be a clear distinction, e.g. between doctoral and masters students, or those that live to the east and west of town (these divisions are not, of course, absolutely ‘permanent’ or unchangeable, but sufficient to lead to a conflict of interests over a whole series of decisions). If we assume that there is no significant movement between majority and minority, then it is no longer clear that majority-rule – to the permanent exclusion of some minorities – is equal or democratic162. Moreover, even if society is broken down into, say, ten equal groups, any eight of which could unite into a changeable majority, that is no consolation to the other two who might still be permanently excluded163. While some theorists have still argued for majority-rule, on the grounds that at least only a minority will have their rights invaded164, it is unclear that the ‘tyranny of the majority’ is much better than the tyranny of a single individual, simply because more people are tyrants. After all, a single tyrant can hardly be blind to the plight of his people, and knows he will be deeply unpopular – and hence at risk – if he impoverishes them too much. Most people can, however, support a lavish prince, since the costs of one person’s extravagant life are spread over the very many poor. When it is a majority imposing on a minority, however, the total costs of giving all the majority a good life are much higher, and they are imposed on a much smaller segment of the population. Those in the minority may become practically ‘invisible’ to the majority, who see only their fellows and, even if the plight of the minority is 162 163 Sartori (1987) pp.32-4, Lively (1975) pp.25-7, Guinier (1994) p.78. Note, therefore, that the problem arises whenever there is a permanently excluded minority, even if the winners change. For criticisms of Dahl, along these lines, see Lively (1975) pp.20-4 and Hyland (1995) pp.89-90. 164 Van Mill (2006) p.141: “These arguments can be used to support majority rule; if sovereignty poses a threat to rights, it is better that it rests with the majority because only a minority of people can possibly have their rights invaded… instead of trying to control power, or divide it up to make it safe, the best solution is simply to give it back to the people and trust them”. 59 noticed, each individual member of the majority is likely to feel very little personal responsibility for what the majority as a whole does. Suppose we know, when designing our democratic institutions, that the society in question will involve a permanent division between a 70% ‘red’ majority and a 30% ‘blue’ minority, and these group identifications will affect the way people vote on a significant range of issues165. If we were committed to majority-rule, then we would not find this problematic, but simply accept that the blues would never get their way. If we were in an ‘Original Position’, not knowing whether we would be a blue or a red, however, it is not clear that we could accept such a possibility166. At least, if Rawls is right that we would not be willing to gamble, and so would reject utilitarianism, then it seems that for the same reason we would reject such ‘loaded’ majority-rule and insist on a principle that would guarantee us some say, even if we ended up in the blue minority167. Possible solutions would be those that ensure compromise on each issue – as may result from a unanimity rule168 or consociational elite bargaining169 – or artificially induce some rotation in office, as produced by lottery-voting. This is not, of course, to prove that one of these methods would be chosen all things considered, but lottery-voting is not ruled out. People may be willing to accept a lottery, even if it lowers their chances of getting their way, to ensure they are never excluded a priori simply in virtue of being in a minority170. The This case is discussed again in chapter 4.9, below. I do not mean to imply democratic procedures are to be chosen from Rawls’ Original Position. In fact, the choice between them can be made only with some knowledge of society – as assumed here – and so is a matter for the ‘constitutional convention’, Rawls (1999 [1971]) pp.172-4. My point is that lottery-voting remains an open possibility. 167 Harsanyi (1955) p.316 reasons to utilitarian conclusions from a similar original position, but he assumes we thereby maximize our expected utility, because we have equal chances of being anywhere – or anyone – in society. Hurley (2004) p.117ff. criticizes the idea that the natural lottery should be thought of as a proper lottery, because there is no identity before the ‘randomness’. I return to these themes in chapter 3.3 and 3.12. 168 Buchanan and Tullock (1962) pp.81-92. 169 Lijphart (1977) pp.49-54. 170 C.f. Timmermann (2004) p.112, quoted in chapter 3.14, and Guinier (1994) p.1. 166 165 60 reason that such possibilities have been neglected is, in part, I conjecture, that people have confused the fact that ‘the many’ must be the ones making the decision with the decision-rule that says the many are to be decisive in cases of disagreement between decision-makers. Now that we are aware of alternative possibilities, the next two chapters will examine justifications of majority-rule; with chapter 3 developing an alternative account of fairness that leads, in chapter 4, to me developing lottery-voting as an alternative to majority-rule. (1.9) Conclusion This chapter has shown how the need for democratic decision-making arises out of the desire for coordination between free and equal persons in conditions of pluralism. I have suggested that a contractualist approach may have considerable potential for framing decision rules – we should regard as fair those rules that all would agree to, for example, if reasonably motivated or placed behind a veil of ignorance and so unaware of their eventual position in society171. The idea of a social contract to form society is not, of course, new, but a number of historical theorists have been guilty of simply assuming majority-rule will be agreed to or a consequence of the contract, rather than showing that it would in fact be agreed. One possible reason for this failing was diagnosed, namely a confusion between the question ‘who makes decisions?’ (where democracy does indeed call for the enfranchisement of the majority) and the secondary question ‘how are decisions to be made?’ (which need not be by majority). This confusion has, regrettably, led to an over-simplistic contrast between majority- and minority-rule; a false dichotomy that neglects other possible 171 Note that the contract is used to reach democratic procedures. I do not assume a contract specifies justice, so it remains open for me to say that contractors agree to certain democratic procedures because they are just. 61 democratic procedures, such as lottery-voting, the recognition of which leaves us needing reasons to accept majority-rule. The next chapter turns to broadly utilitarian arguments, which purport to justify the choice of majority-rule on the grounds that it produces better outcomes. I argue that this is not so, because there is no necessary connection between majority-rule and better outcomes. Moreover, even if majority-rule does maximize utility, this is not an adequate understanding of treating people equally, since it need not distribute this satisfaction fairly. Chapter 3 therefore takes up outcome-independent arguments of fairness, considering what procedures contractors would agree to for resolving competing claims. I will argue that, if we think of decision rules that would be agreed to by free and equal persons, then majority-rule will not necessarily be agreed to; depending on the social situation, contractors will have reason to at least consider other possibilities that are neglected by the contrast between majority- and minorityrule, including lottery-voting. 62 2 Maximizing Arguments for Majority Rule “[D]emocracy is not founded in a concern for maximizing social utility; instead it is based on the ideal of giving citizens equal control over their social world”172 “[T]he use of the majority principle cannot bring about any significant approximation to ideally egalitarian compromises”173 (2.1) Introduction The previous chapter argued that we should not simply accept majority rule as the ‘natural’ or ‘inevitable’ decision procedure. Maybe if democracy did evolve as a means of peaceful conflict resolution between otherwise warring factions, it was natural for the numerically superior army to be awarded victory, but once we have moved beyond such a primitive state we no longer believe that might makes right, or even legitimates – as Rawls observes, “it is to avoid the appeal to force and cunning that the principles of right and justice are accepted. Thus I assume that to each according to his threat advantage is not a conception of justice”174. We want principles that all can accept on the basis of reasons, not simply a modus vivendi adopted out of pragmatic necessity175. After all, in the modern world a well-trained and -equipped elite force would often be capable of defeating a much more numerous enemy army, but we do not feel the former are entitled to over-rule the latter in a vote simply because they could beat them in a fight. If we are to accept majority rule, it must be because it is something all could rationally agree to in our hypothetical contract position. 172 173 Christiano (1996) p.96 [not emphasized in the original]. Berg (1965) p.147 [not emphasized in the original]. 174 Rawls (1999 [1971]) p.116. 175 C.f. Guinier (1994) p.1, and Mouffe (2000) p.94. 63 The present chapter takes up one strand of argument for majority-rule: broadly consequentialist claims that such a procedure maximizes some good176. Two versions of such are considered, perfect and imperfect procedural views177. The former assumes that, if everyone votes according to their interests, then majority-rule will automatically bring about the greatest happiness of the greatest number. This argument is rejected because, even if this ideal is a worthy one, votes cannot guarantee utilitarian outcomes. The imperfect procedural view, in contrast, takes the majority verdict as the most reliable indicator of ideal outcomes more broadly construed (this is generally known as an epistemic conception of democracy). This has some advantages, for instance it can accommodate impersonal ideals such as justice, but it is unlikely to succeed because of the epistemic and moral demands it places on voters. More fundamentally, I suggest that the argument that voting conduces to an independently-specified ideal outcome is undermined by indeterminacy over that ideal. If one outcome is better for me and another better for you, then we may have no way of saying that either is better overall. Moreover, wellknown problems of vote aggregation threaten the meaningfulness of any supposed majority verdict. These considerations led Riker to reject what he called ‘populism’ in favour of a more modest ‘liberal’ conception of democracy, the main virtue of which was removing potential tyrants (almost randomly) from office178. What we need is a different way of thinking about democracy. If there is no such thing as the ‘popular will’, then we should let each person vote for their preferred outcome and employ a fair procedure to decide between any conflicting interests. This may be termed the 176 I say broadly utilitarian because I am not committed to any particular conception of well-being (e.g. mental state or desire-satisfaction), but I stop short of consequentialism, because at least at first I restrict consideration to personal good – see sections 2.6-7 for a relaxation of this assumption. 177 These seem to correspond to what Waldron (1990) calls ‘Benthamite’ and ‘Rousseauian’ views. Although these labels are intended to describe a familiar interpretation of said authors, I do not believe they accurately portray their views. Perfect and imperfect procedures are defined by Rawls. 178 Riker (1982) pp.242-6. 64 ‘pure procedural’ conception of democracy, which is developed further in the next chapter179. (2.2) Procedural Justice Rawls distinguishes three types of procedural justice: perfect, imperfect and pure180. In fact, this is an oversimplification, neglecting an ambiguity between two forms of imperfect procedure181. To briefly state these four possibilities: Perfect: We know the independently just outcome and have an infallible means of reaching it. Imperfect (a): We know the independently just outcome, e.g. equal shares, but do not have a sure way of reaching it. Imperfect (b): There is an independently just outcome, but we do not know what it is and the procedure does not guarantee reaching it182. Pure: There is no independently just outcome, only what results from the procedure. Rawls uses a cake-cutting example to illustrate a perfect procedure, assuming firstly that equal slices are independently just and secondly that a ‘you cut, I pick’ rule will achieve this. In fact, both points are disputable. Even if equal slices are just, such a rule is not guaranteed to produce equality as the cutter may misjudge, so it is – at best – imperfect. If you cut the cake 55/45, then I can pick the bigger piece. Moreover, one could argue that this is just because it follows from the procedure, 179 Although I was already developing such a line of argument, I have been influenced here by McGann (2006). 180 Rawls (1999 [1971]) pp.74-5. 181 I owe this observation to Magnus Jedenheim. Some of the following also draws on remarks by Pavlos Eleftheriadis. 182 Note that it could be claimed we do know what the just outcome is in the abstract, e.g. in a jury trial we know the just outcome is to punish the guilty, it is simply that we do not know who is guilty. The line between these two forms of imperfect procedure is not clear-cut. 65 which suggests that this cake-cutting may be better understood as a pure procedure. This interpretation could be strengthened if we assume the cake is not homogenous but, for example, has a cherry on the top. In this case, you may deliberately include the cherry on the smaller piece, such that you judge the two slices (55% or 45%+cherry) equal, but I need not place the same value on the cherry. Alternatively, if you are not so hungry, you may deliberately cut the cake unevenly, expecting me to take the larger slice. Since ‘you cut, I pick’ is intuitively a fair procedure, whichever outcome eventuates can be regarded as just, regardless of whether or not it tends to independently specified equality and, in any case, an independently specified standard of equality cannot be guaranteed by such an inherently fallible procedure. If we want a better illustration of a perfect procedure, we could appeal to something like measurement to ensure that two shares were equal. Suppose, for example, that we were dividing a pile of beans between us. One way to reach an equal outcome, assuming an even number, is to take beans alternately (one for me, one for you, one for me, etc) until we had split the pile half and half. Assuming that equal shares is an independently just outcome, then this is a perfect procedure. Rawls’ example of an imperfect procedure is a jury trial, which illustrates the second type of imperfect procedure I have identified. In a trial, there is a guilty person, but we do not know who it is, and the trial is supposed to be most conducive to identifying this person. The first type of imperfect procedure, however, is one where we know the just outcome, e.g. equal shares, but do not have an infallible way of reaching it – for instance, when we split the cake only roughly in half or perhaps if we were to divide a large number of goods between us by tossing a coin for each one. One thing to notice about imperfect procedures, of either sort, is that they can be more or less ‘perfect’ or reliable as means to reach the outcome. For instance, one 66 way of trying to find the guilty party would be to randomly select a name from the phonebook. Assuming the guilty person is in the book, this has some slight chance of reaching the right answer, but obviously a jury trial is better because it is more likely to do so. Where there are independently just standards, the ‘you cut, I pick’ rule is at the other extreme, being an ‘almost perfect’ way of reaching equal outcomes, assuming this is what both parties want and that they are competent cutters/choosers. As such, the difference between perfect and imperfect procedures is not really one of kind but degree. A procedure with a 1% chance of reaching the just outcome is imperfect, as is one with a 99% chance of doing so, while one with a 100% chance is perfect. Pure procedural justice is different in kind. In cases of pure procedural justice, there is no just outcome independently of the procedure. Rawls illustrates such cases with the example of fair bets – if we agree to stake £5 on a horse race, then justice says nothing about which of us should have the money independently of the outcome of that race. Note, therefore, that this is a historical notion of justice, in that it relies essentially on the procedure actually being followed. While procedures may have independent value, when there is an independently just outcome, we may be tempted to bypass the procedure if we can better realize substantively just outcomes. For instance, we would probably give up the ‘you cut, I pick’ rule if some external authority could better cut the cake in half. Where there is no independently just outcome, however, we cannot bypass the procedure. If the horses do not race, for instance, there is no justification for an external authority to transfer my £5 to you or vice versa. 67 (2.3) Perfect Procedural Conceptions of Democracy Utilitarians assume that there is an independently specifiable ideal for government to aim at. As Bentham puts it, in the section of his Constitutional Code titled ‘Ends Aimed At’: “I recognise, as the all-comprehensive, and only right and proper end of Government, the greatest happiness of the greatest number of the members of the community: of all without exception, in so far as possible: of the greatest number, on every occasion on which the nature of the case renders it impossible by rendering it matter of necessity, to make sacrifice of a portion of the happiness of a few, to the greater happiness of the rest”183 Both utilitarianism and democracy aim to take equal account of each person’s interests – as Bentham also said, “every individual in the country tells for one; no individual for more than one”184 – so it is no coincidence that there are many parallels between them; for instance individual rights are often portrayed as a check on both utilitarianism and majority-rule185, defending individuals against majorities. Nagel also draws this connection, observing that, “The moral equality of utilitarianism is a kind of majority rule: each person’s interests count once, but some may be outweighed by others… Persons are equal in the sense that each of them is given a ‘vote’ weighted in proportion to the magnitude of his interests… the basic idea is majoritarian because each individual is accorded the same (variable) weight and the outcome is determined by the largest total”186 If this is our goal, then majority-rule seems to offer a potentially perfect way of reaching it. Suppose the contested decision represents a conflict of interests: those who get their way can be represented as having a utility of +1 while those who lose out have -1. Imagine that five people vote for policy X and three people for policy Y. 183 184 Bentham (1983 [1822-32]) p.136 [Const Code Ch.VII.2] [underlining added]. Bentham (1843 [1827]) p.334. Mis-quoted by Mill (1998 [1861]a) p.199 as “everybody to count for one, nobody for more than one”. 185 E.g. Waldron (1990) passim, Dworkin (1977) pp.90-6, Elster (1993) pp.178-9, Nagel (1979) pp.113-4, Freeden (1991) pp.83-100. 186 Nagel (1979) p.112. I return to this weighted voting in section 2.5, below. 68 If we choose policy X then total utility will be 1+1+1+1+1+(-1)+(-1)+(-1)=2, while if we choose policy Y then it will be 1+1+1+(-1)+(-1)+(-1)+(-1)+(-1)=-2. If we can satisfy either five or three people, many find it obvious ceteris paribus that it is better to satisfy the larger number. It is often said that utilitarians treat the community as a single, organic ‘super-entity’, ignoring the fact that gains to one person do not compensate losses to another187. If I offered you the choice between £5 and £3, it would be rational for you to maximize your pay-offs, choosing £5, but utilitarians extend this rationale to society as a whole, assuming that social benefits should be maximized while ignoring distributive concerns. Leaving aside, for now, the criticism that this ignores the separateness of persons, the issue is whether majoritarian voting procedures really will realize the greatest social benefit. The aim is for the voting procedure to be a simple, mechanical measurement of preferences, and so to arrive automatically at ‘the greatest good of the greatest number’. If we assume that each person successfully votes for what is good for them, then the option with most votes is by definition good for more people than any other. While the optimal outcome could in principle be identified independently of the vote, voting is supposed to provide an infallible means to achieve it via the ‘invisible hand’ of majority-rule. There are, however, a number of problems with trying to identify optimal outcomes in this way. I shall begin by pointing out that utilitarians cannot ignore intensity of preferences, so must acknowledge that the greatest good can come apart from the greatest number of votes, and then argue that this end cannot be guaranteed by any majoritarian-style voting procedure. 187 E.g. Gauthier (1962) pp.125-7, Rawls (1999 [1971]) pp.21-7, Parfit (1984) pp.329-47. 69 (2.4) Problems with Utilitarian Outcomes Even if all accept the greatest happiness as the goal to be achieved, putting aside any other problems such as justice, one difficulty is that this need not go hand in hand with the satisfaction of the greatest number. The above example, in section 2.3, assumed that what was at stake for each individual was equal, and thus satisfying the greatest number was the means to maximizing overall happiness. Maybe this is what Bentham meant by his above remark that each is to count for one, but if so it seems an implausible doctrine about interpersonal comparisons: to hold that any benefit for individual A is equal to any benefit for individual B. While we may not have an exact scale on which to make interpersonal comparisons – we cannot always tell, for instance, whether Janet or John would more enjoy a film – we often feel we can make rough comparisons, for example the loss of John’s arm is a bigger hurt than the breaking of Janet’s fingernail. To represent ‘loss of an arm’ as ‘John (-1)’ and ‘breaking a fingernail’ as ‘Janet (-1)’ is deeply counter-intuitive – we know that we would much rather break a fingernail than lose an arm, and it is generally felt that Janet may be required to accept the loss of a fingernail to save John’s arm188. Moreover, the objection doesn’t simply appeal to our intuitive responses; counting each loss the same threatens incoherence because of intransitivity – if John’s arm is treated as equivalent to Janet’s fingernail or Janet’s arm (also (-1)), which is in tension with the judgement that it is worse for Janet to lose her arm rather than her fingernail. An adequate theory of interpersonal comparison will have to take into account that there may be differences between people. These comparisons may of course be inexact – we may not be able to say whether it is worse for John that he lose an arm 188 Taurek (1977) pp.301-2. 70 than that it is worse for Janet for her to lose her arm – but we must be able to say it is worse for John to lose his arm than for Janet to break her fingernail. Once this is admitted, however, it is no longer clear that benefiting the greater number produces most happiness. The particularly problematic examples concern cases where a significant harm to a few is outweighed by smaller benefits received by many individuals. The consistent utilitarian then has to drop the phrase ‘of the greatest number’ and simply recommend whatever distribution will maximize aggregate happiness. This not only clearly separates the utilitarian criterion from any majoritarian procedure but also raises further problems of its own – for instance, if there was a single ‘utility monster’ (person who is exceptionally good at converting resources into utility)189 then utilitarians seem committed to giving him all resources, even at the cost of others starving. I do not believe that classical utilitarianism is a plausible moral theory and touch again on some problems in section 2.8, below, but here my critique is intended to show only that its aims cannot be perfectly realized by majority-rule. (2.5) Can Intensities be Accommodated? As we have seen, the utilitarian aim of maximizing aggregate happiness only justifies majority-rule if majority-rule produces maximum happiness, but that does not seem likely when people have equal votes to represent unequal interests. This section explores whether various loosely majoritarian procedures can reflect differences of interest and therefore better realize utilitarian ideals. The problem here is not simply that utilitarians cannot account for differing preference intensities between two people – the measurement and comparison of 189 Nozick (1974) p.41. 71 which is itself a problem, that I do not go into here – but that it is hard to see how these intensities (even assuming they are meaningful) can be represented by votes. When individuals are only allowed to vote for or against a proposal, their vote effectively records a +1 or -1, but no more. We cannot tell, simply from the fact Janet and John both voted against proposals, that the former did so because it threatened a loss equivalent to a broken fingernail while the latter was threatened with losing his arm. The problem is that voting only compares ordinal preferences, yet what we seem to need is a cardinal comparison – one that tells us how much more John wants X than Janet wants Y. Before giving up on the aim of reaching utilitarian outcomes, however, I want to discuss four suggested routes by which voting procedures may take some account of intensity of preference. Note, however, that while all of these may serve to produce somewhat better outcomes, they certainly do not guarantee optimal outcomes in the automatic way envisaged above (section 2.3). All of these adjustments make the system somewhat imperfect – which is why, after dismissing them, I turn (in section 2.6) to other imperfect procedures. (a) Borda Counts The Marquis de Borda proposed a rank-order method in 1770. According to this procedure, each elector orders all the n options. First preference is given n-1 points, second preference n-2 and so on down to last (least preferred), which gets no points. The winner is the option with most points. This reflects the difference between an option being ‘first preference’ and ‘second preference’ and attempts to most please everyone (collectively), rather than fully-pleasing most (as a simple plurality/majority system). This would seem to promote utilitarian outcomes; for example, it would elect someone who was widely popular, even if second choice, over someone who 72 had a narrow majority of first preferences but was otherwise widely detested. For example: Fig. 2.1 Borda Rankings Preference 1 2 3 … 24 25 26 51% A B C … X Y Z 49% B C D … Y Z A 51% (i.e. a narrow majority) rank option A top and the other 25 in order, B-Z. The other 49% rank A as the worst (26th) option, but otherwise B-Z in order, 1st-25th. Here majority-rule chooses A, which gets 51% of the vote to B’s 49%. It is unlikely, however, that A maximizes the total amount of satisfaction, because although 51% of people get their most-preferred option, the other 49% get their least-preferred190. Probably B would be better – first choice of 49% and second choice of the rest. This is the case in favour of the Borda count, which takes second-, third-, and other subsequent-preferences into account, as a proxy for intensity. The problem with this method (assuming it is trying to maximize utility, as described) is that it uses an ordinal ranking as proxy for intensity; what is really needed is some cardinal measure, which reflects the fact that, for a given voter, there may be a larger gap between A and B than B and C. We can see the problem with another example. Suppose an election consists of one conservative (right-wing), a communist (extreme left) and an extreme socialist (not quite as far left as the 190 In a vote between A and B, this may represent the case where an almost indifferent majority (the 51% that barely prefer A to B) out-vote what is presumably an intense minority. C.f. Mackie (2003) p.133. 73 communist)191. Here, there is very little difference between the communist and extreme socialist – many of those who vote for either might be almost indifferent between the two. Simply representing the voter’s rankings as (2, 1, 0) does not reflect how much he prefers one to another, when on a cardinal scale it might be (10, 9, 1) if he was left-wing or (17, 2, 1) if he was a conservative192. Even in the previous example, it is possible that the 51% really strongly prefer A, and detest the other candidates almost equally, while the 49% do not really care among any of them193 – regarding even A (their lowest choice) as not much worse than B (their first preference) – so in this case majority-rule might produce more utility than the Borda count. The more significant problem, however, is that the Borda count is manipulable because it is affected by the number of options available. In the above example, for instance, B beats A only because of the presence of C-Z; if the other 24 candidates were removed then A would win. The rank-ordering of the candidates is being used as a proxy for intensity, but in fact someone’s preference for A over B does not – ordinarily – depend on whether there is another option between them194. Condorcet criticized this as making irrelevant comparisons bound to lead to error195. The problem is not simply, however, that it may inaccurately reflect true intensities of preference. There is a further problem that voters can deliberately manipulate their preferences – e.g. putting a serious rival candidate lower down their list, giving more points to those who are unlikely to win. By distorting the expression of their 191 192 I assume a simple left-right economic dimension and single-peaked preferences. Borda appeals to something like the principle of insufficient reason, saying “because of the supposed equality between all the electors, each place given by one of the electors must be deemed to be of the same value and to assign the same degree of merit as the corresponding position given to another candidate, or to the same candidate by any other elector whatsoever” [quoted Black p.157]. This may be true in general, but seems clearly false in specific cases, such as this one. C.f. Black p.178 193 To put it in terms of ‘approval voting’, the 51% might approve only A, while the 49% would approve all 26 candidates. 194 And note that, if there are only two options, Borda collapses into simple majority-rule. 195 McLean and Urken (1987) p.34. 74 preferences, they can secure outcomes more likely to be in their interests, but this runs counter to the aim of producing the greatest social utility. In light of this problem, Borda was forced to admit “My scheme is only intended for honest men”196, for otherwise it is unlikely to realize the greatest utility. There may be contexts where a Borda count is an acceptable procedure – in particular, situations where the options are fixed independently and it is likely that differences between them are approximately equal. However, we can have no general confidence that Borda scores accurately reflect cardinal utility. As such, this method cannot be widely used to approximate utilitarian outcomes. (b) Weighted Voting The Borda method makes first preferences count for more than second preferences, but as we have seen the problem is that the difference between them need not be constant. An alternative is to simply make some votes count for more, weighting them in proportion to what each has at stake. While such weighted voting has historically had anti-democratic associations, for example in Aristotle197 and J. S. Mill198, some recent proposals have advocated it as some form of ‘proportional equality’199. It is often claimed that it is unjust to treat unequals the same; for example, Brighouse and Fleurbaey point out that one of a group of friends who is not going to dinner with the rest should have no say over the restaurant they go to, but claim that this is merely the limiting case of a proportionality principle that also 196 197 Quoted Black (1998 [1958]) p.182. Aristotle (1988) pp.145-6 [Politics VI.iii (1318a11-18b5)] suggests 1,000 poor people might have the same power as 500 rich ones – taking wealth as a sign of merit (axia), or perhaps what each has at stake in the state (polis). He doesn’t spell the details out precisely, but it seems he thinks if the two groups each own as much property, the two groups should be considered equal, despite differing in numbers. Thus what is counted equally is property (like share owners) rather than people per se, and this is equivalent to weighting each person’s vote by their property (or number of shares). 198 Mill (1998 [1861b]) pp.334-41 suggests the educated should have more votes. C.f. Berger (1984) pp.192-4. 199 I focus on Heyd and Segal (2006), and Brighouse and Fleurbaey (2006). 75 dictates that one only joining them for dessert should have less say over the restaurant200. Different justifications can be given for such schemes, but Brighouse and Fleurbaey focus on procedural fairness and good consequences201. The present chapter is concerned with the latter of these considerations, while procedural fairness is addressed in the next chapter. Brighouse and Fleurbaey propose that “Power in any decision-making process should be proportional to individual stakes”202. This serves to make democracy more conformable to liberalism for, as in the previous paragraph, it excludes those who have nothing at stake, resulting in the conclusion that only the individual should decide over self-regarding issues, and justice, for it gives those with stronger claims, according to some unspecified theory of justice, more weight. The idea of giving each person a vote weighted in proportion to what they have at stake seems a promising way to realize better outcomes, because it reflects the fact that John’s arm matters more to him (say, 5 units) than Janet or Jane’s fingernails matter to them (say, 0.1 unit each). While simple majority rule would allow Janet and Jane to out-vote John (a case of a relatively apathetic majority defeating an intense minority), a properly proportional scheme would allow John to outvote as many as 49 people who each stood only to lose a fingernail. There are some, of course, who insist this is not enough – for instance, those who believe that we should look only to pairwise comparisons between individuals would hold that John’s claim should defeat any number of (significantly) smaller claims203. Nonetheless, while such a system may not perfectly realize optimal outcomes, it offers a much better approximation 200 201 Brighouse and Fleurbaey (2006) p.7. Brighouse and Fleurbaey (2006) p.2. 202 Brighouse and Fleurbaey (2006) p.3 [emphasis in original]. 203 See Scanlon (1998) p.238 and Nagel (1979) pp.124-5. 76 than simple, egalitarian voting in which all claims are given the same weight204. Moreover, it should be noted that one thing differential weights can reflect is priority to the worst off205, and Brighouse and Fleurbaey are at pains to point out that their proposal is not therefore intimately tied to narrowly-utilitarian reasoning – it can, for example, measure stakes by resources206. As intuitively promising as such proposals seem, however, they certainly cannot guarantee better outcomes, and face a number of serious problems. Firstly, there is the need to decide what each has at stake. Brighouse and Fleurbaey are clear that this is determined by a theory of justice, not strength of subjective preferences. They suggest that this may be less controversial than the substantive issue in question207, but that is not necessarily so – there is considerable debate between rival theories of justice; they themselves have seemingly ruled out one substantive position (utilitarianism) and, as Heyd and Segal observe, it may be reasonable to think either that women should have more say on abortion issues (because their bodies are so intimately involved) or that they should have less (because they are too involved to consider the foetus’ interests)208. Moreover, it is not clear that all affected interests should have influence over the decision. For example, in the restaurant-choosing case, surely the owners and staff of nearby restaurants have their interests affected by the decision, but it is not obvious that they should have any say. Brighouse and Fleurbaey are, however, explicit that what counts as a (properly) affected interest depends on one’s theory of justice, and may therefore exclude external preferences and the like209. They may therefore reply that the worry of over-inclusiveness is addressed by 204 205 Brighouse and Fleurbaey (2006) pp.16-8. Brighouse and Fleurbaey (2006) p.18. 206 Brighouse and Fleurbaey (2006) p.17. 207 Brighouse and Fleurbaey (2006) pp.3-4. 208 Heyd and Segall (2006) p.106. 209 Brighouse and Fleurbaey (2006) pp.4-5, 9 and 17-9, Dworkin (1977) pp.234-9, Goodin (2007) p.51. 77 the close connection between justice and democracy in their theory, which resolves who should have a say and tends to outcomes that are both democratic, so defined, and substantively just. In any case, we still need some practical way of assigning weights. While Brighouse and Fleurbaey bracket this issue, Heyd and Segal propose that weights can be assigned by a prior voting stage – first, everyone assigns weights to each group, the average of which may, for example, tell us that women should count twice as much as men in the subsequent vote on sexual harassment, and then the vote is conducted with these weights – in which a proposal endorsed by 35 women and 10 men will beat one supported by 15 women and 40 men210. Heyd and Segal explicitly exclude concerns about strategic voting in the first stage, assuming that true utilities are known211; but this raises the question why they employ the first vote at all, rather than simply appealing to these utilities for their weighting. Also their proposal that one’s vote depends on “the average weight members of society are willing to give to each other’s preferences”212 seems to contradict Dworkin’s claim that it is “unfair [to count external preferences] because these preferences, like racial prejudice, make the success of the personal preferences of an applicant depend on the esteem and approval, rather than the competing personal preferences, of others”213, which implies that it should not matter what others think of you. Nor is it clear why there should be only two stages of voting: why not have a prior stage, where everyone can decide how much weight each group’s opinion is to have in assigning weights for the substantive debate, thus allowing me to give a lower weight to the Nazi’s lower 210 211 The weighted votes being 80 versus 70, as opposed to the simple numbers 45 versus 55. Heyd and Segall (2006) pp.108, 115. 212 Heyd and Segall (2006) p.105. 213 Dworkin (1977) p.238. Contradiction may be avoided because the former depends on the weight given to a group defined by ascriptive characteristics, e.g. lay men, while the latter seems to depend on the content of one’s view. I do not, in any case, intend to explore Dworkin’s critique of ‘external preferences’ here; merely raise it as a problem to be dealt with. 78 weighting of Jews? And if we cannot stop the regress at two stages, it threatens to be infinite. In practice, these difficulties need not be so severe in small groups of the sort considered here. Even if not marked by close affective ties, it is generally easier for each person to assess what others have at stake, simply because of the smaller numbers of people involved and fact that they can be observed in ‘face to face’ interaction. Many such groups operate by discussion and consensus, rather than any formal voting procedure. In such cases it is plausible to suggest that majorities may sometimes defer to minorities out of solidarity and a sense of fairness and that they implicitly do approximate something like Brighouse and Fleurbaey’s proportionality principle, recognizing that some do have more at stake in certain decisions. When it comes to larger, impersonal democratic units, however, where people do not know each other as well and are less prepared to surrender their own interests for others, then such proposals seem likely to fail on the difficulty of measuring and comparing stakes. This would be a significant challenge to any attempt to implement weighted voting at the nation state level, but such groups are not my concern here – I assume that democracy is most easily achieved in the small groups where face to face contact and the fact that everyone may know everyone makes it more likely that people can effectively gauge each other’s strength of preferences (see chapter 4.6-7). Measuring individual ‘stakes’ is not, however, the end of the difficulty. It has been well documented that voting power need not be proportional to the number of votes. For example, if votes are split 3, 3, 3, and 1 (with 6 needed for a motion to pass) the fourth person has no voting power – they are never vital to any passing motion, as it always requires at least two of the others, who are sufficient 79 themselves214. Infamously the weighted voting employed by the six-member EEC in 1958 gave Luxembourg no power whatsoever215! Moreover, for any given electoral system, different proposed measures of voting power can give very different answers216. Since Brighouse and Fleurbaey want to make power proportional to individual stakes, they need to define not only those stakes (which requires a theory of justice) but voting power. Maybe these problems are not fatal to weighted voting schemes – I cannot offer a complete evaluation of all such proposals here. Nonetheless, I think it suffices to make two observations: Firstly, while giving those who have more at stake more voting power will better conduce to optimal outcomes, they do not guarantee such since voters may still be mistaken about their own interests and, though weighted voting addresses the problem of differential intensities, it is still subject to the other objections raised in this chapter. Secondly, nothing in my positive proposal of lotteryvoting is inherently opposed to weighting voting – indeed, as pointed out in chapter 6, it is actually quite conducive to such, should we wish to weight votes for any reason. For now, however, it is clear that weighted voting faces significant obstacles if it is intended to lead to utilitarian outcomes. (c) Cumulative Voting, Log-Rolling and Vote Trading 214 If the numbers are relatively arbitrary, as in the EEC case, then this is a problem. It’s less obviously so if they reflect, say, number of constituents represented, as then the two with three votes each would represent, say, 60,000 of the 100,000 people. However, we should remember those constituents are unlikely to be homogeneous in their preferences. Even if the representatives represent a majority of their 30,000 constituents, that could be as few as 15,001 each. It would seem fairer to give them three representatives (one per 10,000) rather than grouping their votes into a more powerful bloc. 215 In 1958, the EEC employed weighted voting to reflect differences in size and economic power between its member countries – France (4), Germany (4), Italy (4), Belgium (2), Netherlands (2) and Luxembourg (1). Twelve votes were needed to pass a motion; requiring either the three large countries (France, Germany and Italy), or two of those plus Belgium and the Netherlands. Luxembourg was never pivotal to a winning coalition; because all the others had even numbers of votes, they could never combine to give eleven, so any winning coalition including Luxembourg would still be winning without. See Felsenthal and Machover (1998) p.4 and A. D. Taylor (1995) pp.45-6 and 71-5. 216 A. D. Taylor (1995) pp.217-24. 80 It may be practically impossible or undesirable to make some votes count for more, but perhaps better outcomes can be realized if people are given more freedom to use their votes as they like. This section discusses three such ways people can use their votes to bring about better outcomes: cumulative voting217 (in which people are given one vote for each of a number of issues, but can use more of those votes on the issues they feel strongly about having no say on others), log-rolling (where I agree to vote your way on issue A, in exchange for you voting my way on issue B) and vote trading (where people are allowed to buy or sell votes). One option is to give people one vote per issue, but let them divide those votes as they wish over a number of different issues218. Thus, if we face votes on policies A, B, C, D and E then I get five votes – if I care very strongly about D, marginally about B and not much about the other three, then I may cast four of my votes for D, one against B and effectively have no vote on A, C or E. This gives some crude measure of intensity – my strength of feeling for D is shown by the fact that I am willing to give up my vote on A to have more influence on D. This goes some way to solving the problem of determining the intensity of people’s preferences, but it is only very crude. Two individuals may not care equally strongly about the five policies on offer at once – one person may care very strongly about only one of them, while someone else may care strongly about two or three of them. This is not to say that one person may care more over all issues than another. Though we may find it hard to know what that means, we can imagine one person of a very sensitive disposition who cares a lot about all sorts of things, while another person is more Stoical or apathetic, and 217 I use the name cumulative voting, although this differs from the usual understanding of that term, in which voters can split their votes between different candidates in one (multi-seat) election, it shares the fundamental feature that voters can choose to split or concentrate their votes. 218 This is suggested by Guinier (1994) pp.14-6, 94-5, 107-8, and 117-23. 81 not much bothered by anything – but it is not obvious that the former should have more votes219. The more serious problem for such proposals is that the result of the votes depends on what issues are currently on the table together. The earlier vote between issues A-E would be affected by adding F to those decided at the same time, and whether or not I cared strongly about F. This violates something like the Independence of Irrelevant Alternatives – though it is not simply that the choice between coffee and tea is affected by the presence of hot chocolate on the menu, it is rather that the choice is affected by whether we are also to decide something else unconnected at the same time220. Moreover, this also makes such a system open to manipulation by strategic voters. For instance, I may actually care more about issue A than D, but think that A is effectively a foregone conclusion while the vote on D will be marginal – thus, rather than using my votes on the issue that matters most to me, A, I may use them on D, where I hope to have more influence and therefore my votes produce a greater expected benefit. By doing so, I misrepresent what I really care about and therefore undermine the objective of achieving the maximum social utility to better serve my own interests. Finally, such cases of manipulation serve to illustrate the fact that any such scheme, which lets voters allocate their votes as they wish, risks giving those who are better able to work the system more effective power. The second option is the perhaps more familiar log-rolling, whereby those who feel more strongly about A than B make an arrangement with those whose priorities are the other way round that each will support the other on the issue that matters to them – e.g. I am relatively indifferent on B, but agree to vote for it with you, if you in Compare the ‘utility monsters’ of Nozick (1974) p.41 and Scanlon (1975) pp.659-60’s observation that one’s claim on us depends on the objective, rather than subjective, importance of her interest. 220 Although of course two seemingly different issues might be complementary, e.g. what biscuits we have with our hot beverage. 219 82 turn vote against A with me. This allows two groups to coordinate in order to increase their chances of getting their way on the issues they feel strongly about by sacrificing their influence on issues they care less about. Of course, this carries overtones of elite collusion and the strategic nature of such bargaining may run counter to strict equality. It may be a problem that some groups will have better bargaining positions or more potential partners than others. Log-rolling certainly does not guarantee maximum utility, since two not-so concerned minorities can collude to form an apathetic majority. Nonetheless, it is a possibility that may be worth preserving (see chapter 5), if it is likely to lead to better outcomes, because mutually acceptable compromise can be better reached through such personal bargaining than through impersonal mechanisms such as the Borda count. Thirdly, one might suggest that an open market in buying and selling votes would promote more efficient outcomes221. If I care strongly about issue A, then I could pay you to vote against it, and ex hypothesi the payment you receive makes you better off; thus utility is served by mutually beneficial trades. While advocates of such proposals sometimes suggest that straightforwardly buying votes in this way is not so different from – and more honest than – ‘buying’ them through advertising or offering electoral ‘bribes’ (e.g. tax cuts), many think it objectionable that money should have so much influence in politics. These objections might persist even if we had initial economic equality, for example amongst those who think different goods are to be distributed according to different logics, and that money should simply not be relevant to politics any more than to healthcare or education222. One need not think the influence of money wholly inappropriate, however, to recognize that, at least as things currently stand, people are in unequal bargaining positions. Just as poverty 221 222 Buchanan and Tullock (1962) pp.257-61. Walzer (1983) pp.10-3 and passim. 83 forces some people into exploitative jobs, or perhaps worse – e.g. selling their bodies for medical experiments, prostitution or organ donation – a scheme where votes can be bought and sold may simply see the poor forced to sell their votes to the rich, and the resultant power of the rich may allow them to perpetuate – or even increase – the existing inequality. While perfect markets may create efficiency, whether or not they produce fair outcomes depends on the initial situation223, and even efficiency depends on idealizing assumptions that may not hold in a world of imperfect information and competition. Thus it is certainly not obvious that a market in votes would produce better utilitarian outcomes and, even if it did, we may well have other reasons to reject such a proposal. The three suggestions in this section have all been premised on the claim that people will be better able to use their votes to further their interests – and so utilitarian outcomes – if they are given more freedom to use their votes, e.g. by distributing them as they wish over issues or trading them for other votes or money. While there may still be much to be said for proposals that let people use their votes in these ways, principally in terms of liberty, they may also be undesirable for other reasons, e.g. unfairness. Whatever the other merits and demerits, however, it seems clear that none of them can guarantee utilitarian outcomes. Indeed, it is possible that none of them make utilitarian outcomes any more likely, but that does not have to be proved here – so far, I am only arguing against perfect procedural justifications of majority-rule. There is, however, one further related possibility, which will be discussed separately – those who feel strongly may simply try to persuade others to vote their way. (d) Pressure Group Campaigning 223 C.f. Dworkin (2000) p.68, Cohen (1989) p.933. 84 Dahl suggests that, in polyarchies, there is no monolithic majority that governs224. This would be potentially mixed news for anyone advocating lotteryvoting. It suggests that we do already have a rotation of minorities in government, so even if such is a normatively attractive goal we do not need lottery-voting to achieve it. For present purposes, Dahl’s interesting claim is that intense pressure groups can get their way because they can convince others to vote by campaigning. He supposes that the more intensely a group feel about a specific issue, the more energy they are likely to put into campaigning and the more success they will have winning over others. If they feel sufficiently strongly, then they are likely to win over a majority to their view. We should distinguish two different ways in which persuasion might operate: firstly, campaigners might convince others that they are right – bringing the others round to share their view – or, secondly, while the others may retain their original opinions they might be persuaded to vote otherwise than they would, simply in recognition of the fact that those people feel strongly and should get their way. These forms of persuasion may be more applicable to different types of decision, e.g. the former but not the latter is more plausible when discussing matters of justice, but both can operate in a single case, for instance if we were deciding between an Indian or Italian restaurant, A may seek to change B’s preferences (e.g. by pointing out that the Italian does pizza as well as pasta or has a good range of vegetarian options) or simply to make the case that they strongly prefer Italian so that B, while still preferring Indian from a self-interested point of view, accedes to going to the Italian out of regard for A’s wishes. The former perhaps represents the faith of deliberative democrats that, if people can talk and reason long enough, they can eventually come 224 E.g. Dahl (1956) p.132. 85 round to rational consensus. Certainly it seems desirable for people to engage each other in debate, and this may improve decision-making if some can convince others that they are right (see chapter 5.6-7). It appears unlikely, however, that unanimous agreement can be reached – it is not simply that people will be biased to their own interests, but there will be reasonable indeterminacy between potentially competing values, where there is no one ‘right answer’ but a plurality of reasonable trade offs. While this cannot be conclusively proven here, if I am right then each group may have a reasonable position, neither of which is actually better than the other, which makes it unlikely that one can persuade all others that they are right. In any case, if one group are able to persuade, in the first sense, the other that they are right, then that is good reason to implement the now agreed policy, because there is no longer a conflict of interests – everyone is able to get what they want. Perhaps what Dahl requires is for others to be sufficiently moved by the strength of feeling shown by campaigners to be willing to allow them to get their way, even though they are not convinced to share this view225. But how realistic is this? Maybe where some are apathetic but see that others feel strongly they can be motivated to vote out of sympathy for them – a solidaristic community may even develop informal ‘log-rolling’ conventions – but that is assuming that the people in question neither vote for their own interests or simply abstain (which would seem quite likely, if they are apathetic). In some cases where others feel strongly I may think there is a utilitarian reason to satisfy them, but in others I may judge otherwise, either because I worry about being held hostage to deliberately cultivated intensity or because I feel the intensity unreasonable given the objective importance of the issue at stake226. 225 Rehfeld (2005) p.233 proposes minorities should get their way only when they persuade the majority to vote for them. One problem is that everyone in the majority may think it’s just for the minority to get their way on some decisions, but not be able to coordinate which. 226 Scanlon (1975) pp.659-60. 86 Dahl makes a number of questionable assumptions. It is not clear that, just because a group feels particularly strongly, that this will result in campaigning activity. Some groups are better able to organize than others, and also the incentive to campaign may depend on how likely it seems to be successful – if 40% of people feel reasonably strongly, it may well be worth them campaigning, but a group of just 10% – no matter how intense they feel – may never bother. One advantage of lotteryvoting, as pointed out in my introduction (section 0.5) and chapter 4.9, is that it always gives groups reason to win over as many others as possible, regardless of whether they already have a majority or would move only from a tiny to a small minority. Further, it is assumed that this campaigning will be effective in winning over other voters. The success with which one is able to persuade others need not relate to one’s own intensity of feeling at all – I may very easily be able to persuade others of something I do not care strongly about at all, but not of things I do. This problem is exacerbated by the fact that intensity is likely to correlate to extremist or ‘non-negotiable’ positions, such as matters of religious faith, where believers feel strongly about their preferences and are generally less open to rational persuasion. Moreover, not all groups will be equally able to persuade others to join them. While, if Dahl is right, rule might be by shifting coalitions of different minorities, it does not follow that all minority groups are equally included – some might be in many coalitions while it is still possible that a certain group might be regarded as ‘untouchables’ or pariahs and never in fact likely to be included227. Dahl’s empirical assumptions at least mitigate the dangers of majority-rule, but they do not necessarily result in fair access to power for all, and his claims that intensities can be accommodated by persuasion are unsubstantiated. The idea that 227 For criticisms of Dahl, along these lines, see Lively (1975) pp.20-4 and Hyland (1995) pp.89-90. 87 others can be persuaded to vote for an outcome just because they see a certain group feel strongly about it, however, has affinities to the explicitly imperfect approach in which all vote for the social rather than individual good, to which I now turn. (2.6) Imperfect Procedural Conceptions of Democracy So far, we have assumed a utilitarian notion of the good, and seen that this cannot be perfectly achieved through any practical voting system. One could suggest that voting could be an imperfect procedure in the first sense identified above (section 2.2). If this is so, then an alternative way of reaching it would indeed be to model voting on a jury trial. Let us suppose that each person forms a judgement of what conduces to this good, and we take the majority verdict as most likely to be right. Having postulated an independently-existing ideal, the question is how we are to reach it228. Given what has recently been said, the procedure must be imperfect, but we want it to be better than a ‘stab in the dark’, like taking a random name from the phonebook. David Estlund proposes “Democratic legitimacy requires that the procedure is procedurally fair and can be held, in terms acceptable to all reasonable citizens, to be epistemically the best among those that are better than random”229 and claims an epistemic argument “promises to explain, as fairness alone cannot, why majority rule is preferable to empowering randomly chosen citizens: under the right conditions majority rule is vastly more likely than the average individual to get the morally correct answer”230. But what basis do we have to assume that majority rule is 228 Actually, in light of the two versions of imperfect procedure identified above, the question could be how to identify (and reach) the ideal or simply how to achieve it once known, but this makes little difference for now. 229 Estlund (1997) p.174. 230 Estlund (1997) p.185. 88 more likely right than wrong? One possibility (rejected by Estlund231) is to appeal to Condorcet’s Jury Theorem. The jury theorem states that: “[I]f each member of a jury is more likely to be right than wrong, then the majority of the jury, too, is more likely to be right than wrong; and the probability that the right outcome is supported by a majority of the jury is a (swiftly) increasing function of the size of the jury, converging to 1 as the size of the jury tends to infinity”232 For instance, in a group of 399 people, each of whom had a probability of 0.55 of getting the right answer, the probability of a majority getting the right answer is 0.98. For a benevolent dictator to do better, she would need a greater than 0.98 probability of getting the right answer, but we can assume no one is this wise233. Though it may appear complicated, the basic idea behind the jury theorem can be explained quite intuitively234. Consider tossing a fair coin – since neither heads nor tails is more likely, we would expect them to occur roughly 50/50, and to more closely approximate this the more times we tossed the coin (six heads out of ten wouldn’t surprise us, 60 out of 100 might, and 600 out of 1,000 really would). Since our expectation is even, there is no greater probability of 1,000 tosses producing a majority of heads than ten tosses. But now consider a chance event where one outcome is more likely – say, rolling 3+ on a six-sided die. We would expect to do this around two-thirds of the time on each roll (2/3=67%). If we roll the die three times, the chances that at least two of those will be 3+ are 20/27 (=74%). If we roll the die 100 times, we would be absolutely astounded if the majority of results were not 3+ (given that ordinarily we would expect two-thirds of them to be so). If the jury argument is successful, then it may justify a wide franchise and majority-rule, even over a particularly competent elite – as Przeworski notes 231 232 Estlund (1997) p.185ff and Estlund (2007) pp.223-236. List and Goodin (2001) p283; c.f. Przeworski (1999) pp.26-29. 233 Przeworski (1999) p.27. 234 C.f. Estlund (2007) p.15. 89 “collective competence may increase with the size of the assembly even if increasing the size lowers the average individual competence”235. However, there are a number of obvious problems with the argument. Firstly, it assumes that individuals are reasonably competent – at least that average competence is above 50%. This assumption may be unjustified, but it is arguable that if it does not hold then we should give up on democracy and submit to expert rule236. Perhaps more problematic is the assumption that there is a single right answer, neglecting the significance of conflicts of interests – this is something I touched on in section 1.2, above, but will return to in section 2.8, below. First, however, I turn to some smaller difficulties. (2.7) Minor Problems with the Condorcetian Paradigm Even if we assume that this imperfect procedural picture is possible – in the sense that there is a good that all citizens can aim at and grant that they could identify it – this seems to place significant burdens on citizens. Firstly, there is considerable epistemic difficulty in working out what is the best overall outcome for everyone, including impersonal goods, future generations and so on, compared to merely evaluating what is good for oneself. One defence of liberal-democracy relies on the idea that citizens are best placed to know their own good, but it does not follow that they are best-placed to know everyone else’s – maybe government officials can better make these judgements. Calculation is made yet more difficult by the pervasive effects of cognitive bias. Even when citizens are sincerely aiming for a collective good, they are always liable to favour their own group interests unconsciously, simply 235 236 Przeworksi (1999) p.27; c.f. Aristotle (1988) p.66 [Politics III.11 (1281a40-b15)]. Plato (1992) p.262 [Rep 590c-d], Schumpeter (1967) pp.173-7. 90 because they find these easier to identify with237. And it will be even harder to know what other people want if we cannot take their votes as indicative of their assessment of their own interests (but rather, their assessment of everyone’s interests). The burden is not only epistemic, however. Supposing that one has genuinely identified this good, it need not coincide with one’s own interests; so this procedure relies on people being able to set aside their own immediate good in order to take into account, possibly distant, others. In this case, it may be even harder than often appreciated to explain why people vote. Rational choice theorists often see relatively high levels of voting as a paradox, given the small probability of personal benefit; but on the present theory there may in fact be negative personal benefit and, even if there is a personal benefit or one is motivated by social benefit, the probability of one’s vote making a difference – given that everyone is supposed to be aiming at the same thing – is likely to be even smaller. Given these problems, it is unlikely that everyone will successfully identify and vote for the greatest collective good. This is exactly the situation the jury theorem is designed for, of course – provided mean competence remains sufficiently high, then we are best taking the majority as the more reliable guide to the good. However, we have to assume not only that people are more likely than not to identify the right answer but that they are more likely than not to actually vote for it. Moreover, there is a yet more radical challenge to be made: I now wish to question whether there is any determinate greatest social good at all. 237 Lord et al (1979) passim, Mele (2004) pp.246-50. 91 (2.8) Indeterminacy Up until now, I have assumed that the goal to be reached – through either perfect or imperfect procedure – is the simple utilitarian one of the greatest happiness, where ‘happiness’ can be understood more widely than in hedonic terms, but is essentially a factor only of personal well-beings. I have argued that it is very hard to construct any notion of maximum happiness from voting preferences, since equal votes do not reflect unequal intensities of interest, while any proposal that gives some more voting power need not accurately reflect the objective importance of their interests. Section 2.6 suggested voters could arrive at their own judgement what maximizes social good and then vote to realize this. Once we adopt this imperfect approach, we are not confined to a person-affecting notion of good. Individuals’ judgements of what is best can incorporate impersonal good-making factors, such as equality or the environment, which, insofar as they do not affect any individual in particular, are not considered by self-interested voting and, therefore, cannot plausibly be guaranteed by a perfect procedure. Our overall assessment of the state of affairs need not be solely dependent on the individual utilities in that state but can incorporate other values such as equality. This allows us to say, for example, that (6,6) is better than (7,5) or even perhaps (8,5). We can thus arrive at a ranking of social states that reflects whatever we think really good, personally or impersonally, that is immune to various objections to utilitarianism, because it can consider all possible values, including justice. We might say that our ranking of states of affairs constitutes a form of ‘representative consequentialism’238. If such an approach gave us a complete impersonal ordering of social states, then we may accept the consequentialist injunction to ‘bring about the 238 Scanlon (2001) p.39. 92 best outcome’, while differing from classical utilitarians in not understanding this in terms of individual utilities. The prospects of reaching such a complete and determinate ordering, however, appear slim239. Some are sceptical that it makes any sense to speak of an impersonal point of view and argue that we should focus only on individuals240. Taken to its extreme, this opinion leads us to conclude that only Pareto improvements can be regarded as simply better, and that any cases of interpersonal conflict must be concluded ‘on a par’ or strictly incomparable. Taurek, for example, suggests such an approach when he claims that “I cannot imagine that I could give David any reason why he should think it better that these five strangers should continue to live than that he should”241. However, Taurek does not actually deny that it can sometimes be right for one person to make a sacrifice for another. Though he thinks A would not be required to make a certain sacrifice to spare B a slightly larger loss, he thinks there comes a certain point where A would be required to do so – for instance, A might be required to suffer a broken arm to save B’s life242. While Taurek might be reluctant to express this conclusion in terms of the impersonal betterness of these outcomes, preferring to focus only on rightness, there is no reason to reject representative consequentialism as a façon de parler according to which we judge the broken arm better because it is what should be brought about. On this account, we can say that the social state (8,8) is unambiguously better than (4,4) because everyone is better off. Moreover, if we accept the Pareto criterion, we can say (8,8) is better than (8,7) because one person is better off and no one is Sen (1992) pp.46-9, 131-5 and 143-4, Shapiro (2003a) pp.39-44 and 49-50, Taurek (1977) pp.3045. C.f. Guinier (1994) p.103: “Where preferences are dispersed, decisions should not be made according to any single conception of public good”. 240 Nagel (1991) pp.64-9, Scanlon (1998) pp.229-30. 241 Taurek (1977) p.300. 242 Taurek (1977) pp.301-2. 239 93 worse off. The problem comes when interpersonal comparison is necessary, however. This is not simply a problem of measurement – though there may be such difficulties – but one of making a trade-off between two distinct people. Suppose we have to choose between (10,8) and (8,9). Even if we accept that the former has a higher level of aggregate utility, it does not follow that it is the outcome to be brought about, for this is what ignores the separateness of persons – the extra two units of utility for the first person in no way compensate the second for his loss. Since what is at stake for each person is ‘on a par’ or roughly equal, the two outcomes must be considered socially on a par. The first person justly prefers (10,8), the second (8,9), and we cannot say either is ‘socially better’243 in the way we might had the choices been (20,8) or (8,9)244. Thus, even if we can compare what I have to gain to what you have to gain, the fact that one of us has more at stake is not enough to determine what should be done. In a choice between (10,8) and (8,9) both parties have opposed interests at stake and it seems reasonable to toss a coin, offering each an equal chance of satisfaction, since the slightly greater benefit to the first person is no compensation to the second. Though these cases are not exactly zero-sum, the conflict cannot be resolved by an appeal to aggregate utility. Recall that in section 1.2, above, I argued that a co-ordinated outcome may be best for all, even if it is not the best pattern of co-ordination for given individuals (perhaps for any individual). This means we can agree that some co-ordination is socially better, but we cannot use social betterness to resolve which pattern of coordination should be adopted, because that is where there is conflict. To repeat the example used there, suppose I am better off with a ‘drive on left’ rule and you are 243 ‘Better’ here includes a value judgement that goes beyond personal utility – in the same way that (6,6) is better than (7,5). 244 This is supposed to represent something like the earlier broken arm case, where what is at stake for one person is so great as to morally outweigh the other’s interest. 94 better off with a ‘drive on right’ rule. The greatest social good is clearly attained by both of us following the same rule, rather than our individual preferences and though the resulting satisfaction can be represented either (11,9) or (9,11) – whichever of us gets our way will be marginally better off – from the impersonal point of view, neither is better, so no one else has any reason to prefer one or the other on grounds of the greatest collective good. The problem is effectively a zero-sum conflict of interests. This problem of indeterminacy is more widespread than some have appreciated when we consider interpersonal distributions. Perhaps we can all agree on maximizing efficiency – by which I understand achieving the most good possible or, in economic terms, operating on the ‘production possibility frontier’ – but we will still have to decide on who gets what. To give a more concrete example, we can agree that society should produce or realize 100 units of good rather than 80, but then the conflict over their distribution is still zero-sum, between 60/40, 40/60, etc. It might be thought that 50/50 is an obvious default, but this assumes lack of disagreement over justice – whereas in reality the issue will be complicated by conflicting claims of desert, entitlement, need, capabilities, etc. What is needed is an agreement to respect Pareto improvements, which enjoy unanimous support and ensure efficiency. We need to supplement this, however, with a fair procedure for resolving distributive questions. We cannot suppose there will be a unique ‘greatest good’ that will save us from having to make these decisions245. Thus, while a theory of democracy should indeed take into account the likely good or bad consequences it will produce, it 245 For a more radical pluralist attack on the idea of rational consensus, see Mouffe (2000) pp.90-105. 95 cannot be justified solely by such: we need an account of its fairness246. The next chapter addresses this problem, but first I recap the conclusions so far. (2.9) Conclusion Chapter 1 argued that we need justification for majority-rule. The present chapter has surveyed one possible strand of justification – consequentialist arguments that claim that we all would or should accept majority-rule because it leads to better outcomes. There is some ambiguity, however, as to whether this means better for everyone or simply the majority. To claim that majority-rule is better for everyone requires us to assume that everyone has a chance of being in the majority and so victorious, but this relies on the assumption that majority-rule is fair – something taken up in the next chapter. If we do not make this assumption, then it is unclear why those that lose out from majority-rule should accept it simply because it produces benefits for others. This utilitarian line of thinking ignores the separateness of persons, so this simple maximizing argument can be rejected as unjust. Whatever conception of the good we adopt, especially if it is an ideal one that incorporates considerations of justice, any decision procedure intended to maximize it will necessarily be imperfect. This does not mean, of course, that we should neglect the likely outcomes of our procedures, but we should be wary of seeking to justify them simply by reference to such. Even if our procedures do produce more good, there is likely to be zero-sum conflict, because of indeterminacy about what is ‘best’ or distributive questions, so we still need a procedure that is fair to everyone. The following chapter considers whether majority-rule is fair. It will also be seen that I 246 This is something Estlund acknowledges, but he assumes that majority-rule is fair and, therefore, that any epistemic advantage is decisive. I have questioned this epistemic advantage, but the decisive point is that majority-rule need not be fair if it results in perennial losers. 96 favour a procedure in which individuals can vote for their own private interests. This avoids many of the epistemic and moral burdens identified in section 2.7, above. Rather than requiring individuals to vote for what is the best overall compromise, I propose a system that – as far as possible – has them vote for what they want, and then produces compromise. This is consistent with Rawls’ ‘basic structure’ approach to justice247 – where the market and government taxation/welfare policy are set up to realize justice-as-fairness given individuals who are self-interested market maximizers. This approach has been criticized by those who think people ought to consider the demands of justice in their daily lives. I do not propose here to defend this whole approach, but I think it clear that it would be preferable if institutions could realize justice without making such demands on individuals. 247 Rawls (1999 [1971]) pp.76-7, and 242-51. It is not, however, how Rawls thinks of voting; see Rawls (1999 [1971]) pp.313-8. Note, though, that he qualifies his imperfect account by admitting that voting can be seen as ‘quasi-pure’ within an acceptable range, which is really what I am talking about. 97 3 Using Lotteries to Adjudicate between People “[I]t is precisely because of majority rule that political pluralism fails… cultural pluralism calls for another kind of democracy”248 “[W]hile democratic procedures may indeed be fair, the epitome of fairness among people who have different preferences over two alternatives is to flip a coin”249 (3.1) Introduction The previous chapter rejected consequentialist justifications of democracy. Whether conceived of either as a perfect preference-aggregation or an imperfect epistemic attempt to promote social good, majoritarian procedures are unlikely to maximize utility or any conception of social good – indeed, I argued the very idea of such a maximum was likely to be indeterminate. Even if we concede that there may be a unique ‘social maximum’ in some cases, we should not expect democracy to necessarily realize such, because it is government by the people, not simply government for the people250. If we valued democracy only to the extent that is was instrumental to some ideally good outcomes then in fact we should be willing to embrace a benevolent dictatorship, such as Plato’s Guardians, if such arrangements better served our goal251. I distinguish the realization of such ideals as an aim of good government, whether democratic or not, from what makes government democratic – which is simply responsiveness to the wishes of the people252. While a good democracy is one that produces good outcomes, democracy itself is simply a matter of treating everyone equally253. 248 249 Leca (1994) p.62 [not emphasized in the original]. Estlund (1997) p.176 [not emphasized in the original]. It is not clear why he assumes democracy and fairness must come apart, rather than exploring ways they can be reconciled. 250 Ranney and Kendall (1956) p.16, Hospers (1961) p.383. 251 Plato (1992). 252 Woodruff (2005) p.30. 253 This leaves open the possibility that democracy is only justified where it proves to also be good government. 98 This chapter explores what it is to treat people equally and argues that majorityrule is not necessarily fair under certain conditions, such as when there is a permanent majority-minority split254. Instead, an alternative account of fairness between competing claims is developed, starting from tossing a coin between two people and culminating in weighted lotteries between unevenly-sized groups. Thus, where we have a conflict between four and two people, rather than the four automatically winning, the fairest solution is to give them a two-thirds chance of getting their way. This does not give each individual an equal chance of satisfaction, but it does give each an equal chance of being decisive, and mean that each person is considered equally in the decision procedure. The next chapter develops how the fairness of weighted lotteries can be institutionalized within a democratic framework; for now, the focus is more abstract. (3.2) The Uses and Abuses of Lotteries Lotteries have a long history of use in conflict resolution, going back at least to the ancient Greeks255 and also being used in Biblical times256. While sometimes their use seemed to rely on superstition – divining the will of God, or at least taking the judgement out of human hands – they can also be justified on grounds of fairness to each participant – giving each an equal chance and removing any bias. In recent times, lotteries seem to have earned a bad reputation, both in public discourse and to some extent in academic discussion. In both cases, however, this seems to be largely due to misunderstanding. Newspapers, for example, frequently 254 C.f. Guinier (1994) pp.1-9, which points out that rules that are fair in the abstract may turn out to be exclusionary in practice – for instance, in racially divided communities. 255 Goodwin (2005 [1992]) p.53-4 256 Acts 1:15 and Jonah 1 (the latter is found in Goodwin (2005 [1992]) p.52). For other cases, see Elster (1989) pp.50 (Proverbs 16:33), 52 (Num 26:52-6 and Josh 7), 64 (Acts 1:26 and Num 26:52-6 and 33:54), 66 (Lev 16:7-10 and Jonah 1:7), and 69-70 (St John 19:23-4). 99 condemn inequalities between different local jurisdictions, such as healthcare trusts, as a ‘postcode lottery’257. Similarly, the use of penalty shootouts to decide football matches is often branded a lottery – the claim being that the winner becomes a matter of luck – usually with negative connotations258. Rawls’ attack on the ‘natural lottery’ shows a similar strand of thinking in academic debate259. However, what is wrong is that none of these examples are actually lotteries260. The fact is that postcodes are not allocated by lottery, so nor is access to any good or burden that is allocated by or dependent on where one lives. Not only does it seem unjust that those living on one side of an arbitrary boundary may have access to a drug or treatment denied to those on the other side261, but there is also the problem that the wealthy may be able to buy houses where they have better access to medical provisions, in good school catchment areas, and so on. Even the so-called ‘natural lottery’ is not a real lottery – since it determines people’s identities, so there is no continuous person, before and after the draw, who enters and may receive benefit262. 257 For example: O’Neill, Gibb and Brooke (2005) [criminal conviction rates], Anonymous (2006) [selecting doctors], Malkin (2006) [asylum applications]. Typing “postcode lottery” into Google.com produces about 346,000 results (as of 15/04/07), including an actual (partly) post-code based lottery (http://www.postcodelottery.co.uk/). 258 This claim is often made, e.g. Wilson (2007), http://blogs.guardian.co.uk/sport/2007/03/17/better_the_bore_draw_than_the.html (last accessed 17/04/07), and http://news.bbc.co.uk/1/hi/euro2000/sportstalk/812927.stm (last accessed 17/04/07); although those in the know do deny it, e.g. http://www.soccerphile.com/soccerphile/news/penaltyshootout.html (last accessed 17/04/07) and http://www.smsc.org.uk/resources/penalties.htm (last accessed 17/04/07). 259 Rawls (1999 [1971]) e.g. p.64. 260 This claim depends on the nature of lotteries. I believe we can use unpredictable natural events as lotteries, e.g. if an expectant couple agree that the father will name the baby if a boy and mother if a girl, and appropriately call this a lottery. Postcodes are not lotteries because where one lives depends on choice and, often, money. Penalty shootouts might be considered a lottery on this expansive definition, but I do not think they are because they are part of the game. Settling a football match by tossing a coin would be a lottery. Settling it by, say, a chess match would also be a lottery, because it arbitrarily uses something irrelevant to the game in question. Penalties are not a lottery because they are part of the game and the specified way of breaking ties in specified circumstances. 261 This does to some extent neglect the fact that as different authorities distribute their budgets differently the results are likely to be Pareto-noncomparable. Also the worry is rarely extended to international boundaries. 262 Hurley (2003) ch.4, esp. pp.117-21, and Heyd (2000) p.65. 100 Actual lotteries in fact have been used for a number of these purposes, such as allocating scarce medical resources263 or deciding football matches264. They have also been used or recommended from everything from the extremely trivial265, to tickets for events including Wimbledon and the Oscars266, school or university places267, broadcasting licences268, land allocation269, immigration visas270, the military draft271, and spaces on lifeboats272. There have also been many actual or suggested political uses, to which I turn in the next chapter273. While some of these uses have been criticized, the widespread practice also suggests that many have found them fair or an intuitive solution to conflicts where neither side has a greater claim than the other to the good in question. 263 Duxbury (1999) pp.45, 49 and 151, Goodwin (2005 [1992]) p.211, Elster (1989) pp.68, 70 and 734, Broome (1984b) p.39. 264 Elster (1989) p.63 notes some peripheral uses, such as which team plays first or choice of new players. More interesting is the use of lotteries to settle ties: in the European Cup, before the days of penalty shootouts, quarter finals between Liverpool and Cologne (1965) and Celtic and Benfica (1969) both went to coin tosses. For the former, see Anderson with Done (2004 [2002]) p.80; for the latter see McColl (1998 [1995]) p.96 and http://news.bbc.co.uk/sport1/hi/football/teams/c/celtic/6100458.stm (last accessed 10/04/07). In both cases, the method was criticized. C.f. Duxbury (1999) pp.43-4 fn.3. 265 The website http://www.teaoclock.co.uk/ will randomly select an office member whose turn it is to make a cup of tea for the office (last accessed 10/04/07). This does seem to confuse a lottery (random selection) with turn taking, but strict turn-taking may be impossible where the group is not constant (see the similar problems discussed in chapter 4.6), so a lottery seems a fair solution. 266 Duxbury (1999) p.45. 267 A (weighted) lottery is already used for medical school in the Netherlands, see Duxbury (1999) p.45 and Elster (1989) pp.47-8. Lotteries have been proposed for university places, e.g. Brighouse (2000), c.f. the discussion at http://crookedtimber.org/2005/09/29/lotteries-in-admissions-to-academies/ (last accessed 15/04/07), Ryan (2000) and (2007), and Schwartz (2007). The introduction of a lottery for school places in Brighton in February-March 2007 caused much controversy; see Andalo (2007), Laville and Smithers (2007), Paton (2007), Oberman (2007), and Garner (2007). I think the negative response may partly have been due to the general bad press of lotteries (see the previous paragraph), but also worries about how catchment areas would be drawn (including self-interested bias of those – predominantly rich – parents living near the good schools). 268 Duxbury (1999) pp.45 and 151-2. 269 Goodwin (2005 [1992]) p.54 [discussing Biblical, Athenian and Roman uses], Duxbury (1999) p.44. 270 Elster (1989) p.56, fn.63. 271 Greely (1977) p.115, Elster (1989) p.64, Broome (1984b) pp.38-9, Duxbury (1999) pp.65-7, 100, 131 fn.195 and 154-5. 272 Broome (1984b) p.38, Goodwin (2005 [1992]) p.53, Elster (1989) pp.64-6 and 75-6. 273 Chapter 4.2-3. 101 (3.3) Justifications of Lotteries I cannot, here, give a full justification for the use of lotteries – to some extent, their fairness is taken as a given – but I will begin with a few remarks on their justification, before turning to how lotteries can be used between competing unequally-sized groups. Broome regards lotteries as a ‘second best’ or providing “a surrogate equality in satisfaction”274. Rawls also suggests that, in cases of conflict: “[A]ll shall be satisfied equally, if that is possible… [or] an impartially arbitrary method of choosing those to be satisfied shall be adopted… Imagine a good of such a nature that it is impractical or impossible to divide it, and yet each of a number of persons has an equally strong claim on its possession or exercise. In such a case we would be directed to select one claim as meriting satisfaction by an impartially arbitrary method, e.g., by seeing who draws the highest card… [This] is impartial because prior to the drawing of the cards each person has an equal chance to acquire in his person the characteristic arbitrarily taken to be relevant”275 The ideal solution is an equal satisfaction of equal claims. Sometimes this will be possible, when interests do not conflict, and all can be satisfied. Where two people have equal claims to a non-divisible good, however, the only way to respect absolute equality is for neither of them to have the good276, but this is clearly sub-optimal – we may regard this as levelling down277 and assume neither would want to accept it. If we prefer inequality at a higher level, as recommended by the difference principle278, then we should want to give the good to one of the individuals in question, rather than waste it. Given that ex hypothesi each has an equal claim to it, the contentious issue is which of them to give it to. Again, equal chances of satisfaction are used as a surrogate for equal satisfaction itself – each person’s claim is still respected and treated equally, though only one will in fact get the good. 274 Broome (1998) p.956; c.f. Broome (1984a) p.628, Broome (1984b) pp.40 and 45-6 and Broome (1999) p.119. 275 Rawls (1951) p.193. 276 Broome (1998) p.956. 277 Parfit (2000 [1991]). 278 Rawls (1999 [1971]) e.g. pp.69-72 and 135-7. 102 This has led to criticism from some that there is no reason for a lottery because parties do not have a claim to a chance, but simply to the good in question279. Nonetheless, we cannot always distribute the ultimate objects of concern – even those who advocate equality of (opportunity for) welfare must, as a practical measure, focus on redistribution of tangible resources to achieve their end, for example. Moreover, what people have a claim on others for, as a matter of justice, need not be what they actually want – as in Scanlon’s famous example of the religious fanatic who would rather have help building his temple than food280. The mere fact that what someone wants is the good in question, rather than a chance for it, does not show that, where we have no other option – beyond giving it to neither281 – there is anything wrong with allocating chances. It could be regarded as controversial whether those given a chance that did not come up benefit, but I can remain agnostic on this provided it is accepted that they would rather have the chance of benefiting, even if they did not in the end do so, than not. One way in which people can be given equal opportunity is if the good is attached to some position of merit that each of them is equally able to attain. This is how jobs – with their attendant status and salary – are ideally distributed, to the most deserving candidate. Sometimes, however, no criterion of merit is clearly applicable, or use of some such standard, e.g. educational qualification, does not provide equal opportunity because parties were unequally situated when they came to compete. An alternative form of equality is provided by a lottery. Here the good is attached to some arbitrary criterion – such as heads on a coin, or a number drawn from a hat – that each party has an equal opportunity of satisfying282. If two parties have equal but 279 280 Hirose (2007) pp.54-5. Scanlon (1975) pp.659-60. 281 Broome (1998) p.956. 282 Rawls (1951) p.193. 103 conflicting claims, it is likely that they will agree to something like tossing a coin to resolve them. Actual agreement, however, is not necessary – that may not be forthcoming for various reasons, such as unequal bargaining positions leading one person to think she can seize the good. What matters is that reasonable people would agree to such a procedure, so it should be taken as fair. This shows that there is a difference between the ‘natural lottery’ and an actual allocative lottery, because the former does not give concrete individuals equal chances. It has been claimed that some forms of contractualism cannot justify employing an actual lottery283. Harsanyi, for instance, assumes that our ‘ethical preferences’ over social states are those that we would adopt if we imagined we had an equal chance of being anyone in that state284. This is effectively to adopt the stance of an ‘impartial spectator’ and so produces utilitarian conclusions in which the gains to one person can be added to, and weighed against, losses to another. It is unsurprising that such an account has little space for fairness. If we count each party’s satisfaction or dissatisfaction equally, it seems a matter of indifference which of them gets a given good, so there is no need for a lottery – such would be justified only as a low cost way to resolve the matter (seeing the problem as like that of Buridan’s ass, rather than resolving conflicting claims) or if one or both parties preferred to win or lose through such a process. In principle, however, the ‘natural lottery’ could be employed to determine the distribution – for instance, the good could be allocated to the tallest person, on the grounds that all would accept this if they regarded the tallest person’s preferences equally with their own or assumed that, from some pre-birth position, they had an equal chance of being the tallest person. 283 284 Stone (2007) pp.290-1. Harsanyi (1953) p.435 and (1955) pp.315-6. Hurley (2003) pp.263-7 criticizes ‘equal chance’ procedures for allowing cognitive bias. 104 This illustrates one potential problem with ‘veil of ignorance’ or ‘original position’-type reasoning, as employed by John Rawls285. What seems fair behind the veil – to one in ignorance of their eventual position in society – need not still seem so once it is lifted and they are aware of their actual personal attributes. Parties in the original position might be rationally indifferent between settling a dispute by tossing a coin or by an arm wrestling contest – since neither knows who will be stronger, neither has any greater or lesser expectation from the latter procedure. The reason that tossing a coin is preferable to arm wrestling is that it still seems fair once the parties are situated and aware of their identities, whereas the outcome of an arm wrestling contest may seem predictable and hence no longer fair. If Mike Tyson suggests an arm wrestling contest to Steven Hawking, for example, then the fact that both may have accepted such from an original position no longer seems relevant, because it is clear which of them, as actually situated, will win. This difference seems captured by the stress Rawls places on the ‘strains of commitment’286, though we can by-pass the need for such by simply appealing, like Scanlon, to reasonable agreement from the start. My argument will be that, whatever form of contractualism we endorse, majority-rule faces an analogous problem to arm-wrestling. While it may seem fair to all from some a priori original position where no one knows whether or not they will be in the majority or minority, it may not seem so to particularly situated individuals, who find themselves in a permanent minority and can reasonably foresee defeat287. If we take it for granted that the way to adjudicate between two persons’ competing claims is to toss a coin, or hold some other equal chance lottery between 285 Rawls (1999 [1971]) pp.118-23. Of course, I do not mean to imply that Rawls would use such a procedure to decide this case of conflict between two individuals – his contractualism is designed to govern the basic structure of society. 286 Rawls (1999 [1971]) pp.153-4, and 475. 287 Again, Rawls’ response would presumably be to say that the design of democratic institutions is a matter for the constitutional convention – Rawls (1999 [1971]) pp.172-4, 203, and 311-2 – where such facts about society are known. 105 them, then this also seems fair when there are two larger equal-sized groups, for instance deciding between two groups of five. The more interesting question is what to do when there are unequal numbers on either side. Here I turn to the possibilities suggested in the course of another debate, between consequentialists and deontologists. (3.4) The Numbers Debate We need some account of what it is to treat two different groups equally. Surprisingly, this question seems under-addressed in the democratic literature – although a few do suggest the need for some form of proportional compromise, there are fewer institutional specifications for how this can be brought about. There has, however, been much written about the analogous ‘saving the greater number’ debate in recent moral philosophy. Here, the debate is about what a rescuer should do if they could save either of two differently-sized groups from some loss, typically death. While the utilitarian answer is obvious – all else being equal, save the greater number – many deontologists have opposed aggregation of benefits and harms across distinct persons on the grounds that it can yield intuitively unappealing consequences, such as that enough people’s mild discomfort can outweigh one person’s severe pain or death. These anti-aggregationists have typically invoked the separateness of persons to argue that benefits and pains to distinct individuals are not felt by any social super-organism and cannot be summed. The question then becomes whether, given this individualist restriction, they can still deliver intuitively acceptable solutions to cases where harms 106 are equal and numbers differ – or, as one journal article title put it, ‘Can NonConsequentialists Count Lives?’288. This debate offers what is needed here, because it concerns the fairest way of adjudicating claims between competing groups while putting aside utilitarian considerations. Scanlon, for instance, tries to defend saving the greater number, based not on the alleged betterness of doing so, but on the fairness to each of deciding in this way. If his argument succeeds, then it seems one could give a closely analogous argument for the fairness of majority-rule when two groups have conflicting preferences about social arrangements. However, Scanlon’s arguments have been widely criticized and it will be argued here that they do not succeed or, at least, rely on doubtful assumptions. There have been two other common positions taken in the debate: a simple lottery (equal chances to each group) and a weighted lottery (in which the chances for each group depend on its size). I will argue that the latter best respects each person, and in the next chapter I develop a theory of democracy from this starting point. First, a few remarks on the appropriateness of the analogy. Taken in the abstract, the debate is how to treat the opposing claims of, say, one person and five others fairly. This is exactly what is at issue when we have conflicting interests in a vote. Note that overall utilitarian reasons are set aside. Taurek, for example, denies the legitimacy of any impersonal judgement. He argues that one outcome is better for the one, and the other better for the five, but no more can be said about which is ‘impersonally better’. These assumptions may be hard to accept in general, but they are actually more reasonable in our democratic context. The previous chapter showed that votes do not necessarily reveal social good, because all concerned are given equal 288 Wasserman and Strudler (2003). 107 votes, even though preferences may differ. If the issue at vote is a zero-sum distributional one, for instance, then ex hypothesi neither option is objectively better – the winners collectively gain only what the losers lose. Indeed, in such a case, because the same total loss is shared amongst fewer people, each of the losers loses more than each of the winners wins. In this case, a pairwise comparison principle, such as Scanlon’s, that compares only individuals gains or losses without regard to the numbers on each side, would actually favour minority rule. Moreover, even if Taurek’s scepticism seems exaggerated, because we are sometimes able to make uncontroversial judgements between two social states, it was argued above (in chapter 2.8) that these comparisons would generally be vague or indeterminate – thereby admitting cases that cannot be resolved by appeal to social good. Further, note that while the ‘saving the greater number’ debate has frequently proceeded in terms of who to save from death, nothing special is supposed to follow from the fact that lives are at stake. It applies in principle to any losses borne by distinct individuals, e.g. the loss of arms or distribution of ice cream289. Moreover, I am not the first to consider the possibility of extending this discussion to conflict of votes rather than objective interests. Several in the debate suggest looking at preferences, rather than assuming everyone wants to be saved290 – but I am, I believe, the first to explore it in more detail. (3.5) Taurek’s Argument for Equal Chances John Taurek’s central claim is that there is no obligation to save the greater number in cases of conflict. He argues that five people dying does not involve anyone 289 290 Taurek (1977) pp.301-2, Kamm (1985) p.188, fn.9. Taurek (1977) pp.310-4, especially p.314, Kamm (1985) p.181, Wasserman and Strudler (2003) p.81, fn.18 and p.92, fn. 33. The only reference to the ‘numbers debate’ that I am aware of in the democratic literature is Risse (2004) p.50, fn.22. 108 suffering anything worse than a death – we cannot add these separate harms and claim that anyone suffers more bad, because no individual suffers anything worse than death. Indeed, Taurek suggests that – at least in conflict cases – we cannot make some impersonal judgement about which state of affairs is better291. Where our alternatives are to save one person, whom he calls David, or five others, then all we can say is that the first is better for David while the second would be better for the five; we cannot say anything beyond these individual viewpoints. It is not clear how far Taurek’s scepticism goes about the ‘view from nowhere’. He could still make his arguments while accepting that, if in a non-conflict case, it could be judged simply worse for someone to die, since that situation is worse for one and better for no one. Further, he does accept that there are in fact cases where one person should suffer a loss to save others – for instance, David may be required to bear a hangnail in order to save five (or just one) other lives – though it is unclear whether he would want to describe these outcomes as better. Though Taurek would judge David blameworthy, or at least somehow defective, for preferring to avoid his own mild discomfort in such cases, it is because neither of the life-saving situations is all-things-considered better that David can permissibly prefer his own life be saved292. The general argument has certain libertarian features, for instance Taurek places great emphasis on the absence of any prior contract, the right of the rescuer to do what he wants, and the question whether anyone is wronged293. His conclusion is that it would be quite permissible for the rescuer to save David, based merely on a mild preference – David being someone he ‘knows and likes’, which seems to presuppose 291 292 Taurek (1977) pp.299-300, and 304. Taurek (1977) pp.302, and 305. 293 Kavka (1979) p.288ff. C.f. Taurek (1977) pp.297, and 305. Note Taurek sets aside – but does not reject – the claim that someone should be saved (p.293 fn), whereas a Nozickian would presumably accept the conclusions Kavka derives, e.g. that the rescuer can eat the drug himself because it tastes nice, Kavka (1979) pp.289-90. 109 that the rescuer can do what he likes with his drug subject to not violating anyone’s negative rights. The more interesting claim, for our present purposes, however, is that if the rescuer wants to be fair, he should toss a coin294. Taurek has already argued that the numbers do not count for anything, and neither outcome is objectively better. Thus, the situation in the five-against-one conflict is in no important way different from the one-against-one conflict: tossing a coin gives each person the same thing, a 50% chance of survival. (3.6) Scanlon’s Objection to Equal Chances Although several criticisms were made of Taurek’s argument, it was Scanlon who largely revived the debate on aggregation and numbers, by trying to justify saving the greater number given the individualist restriction of his contractualism295. Scanlon accepts the fairness of tossing a coin between two people’s competing claims296: that is something that, in the absence of other means of resolution, no reasonable person could reject. He argues, however, that something changes when more people are involved – that it is no longer fair to toss a coin because it does not consider all people’s claims. Let us illustrate with Taurek’s David, and call the other five A, B, C, etc. In the one-against-one case (David against A), tossing a coin gives each an equal chance of being saved. Now add B to A, so we have a two-against-one case. Tossing a coin still gives each individual a 50% chance of being saved but, Scanlon claims, B can complain that her claim is not duly considered because it has made no difference to how the matter is decided – that is, to still toss a coin is 294 295 Taurek (1977) p.306. Scanlon (1998) 229-30ff. 296 Scanlon (1998) p.232. 110 effectively to ignore her presence297. This, he claims, allows B to reasonably reject tossing a coin as a way of resolving the two-against-one conflict. This objection does not seem convincing. Firstly, it has been criticized for being implicitly aggregative, because B’s claim only outweighs David’s if taken together with A’s298. If this is so, then Scanlon seems to have neglected his individualist restriction. More importantly, however, it is not clear whether B can justifiably complain that she is not counted. After all, as Taurek would point out, her interests are taken into account in exactly the same way as those of David and A, i.e. she is given a 50% chance of survival299. Moreover, one could say that the procedure is no longer exactly the same as it was: instead of tossing a coin between David and A, the rescuer is now tossing a coin between saving David or saving A and B. To assess B’s complaint, we need to know what difference B’s claim should make. One interpretation of making a difference is that the presence of B’s claim must lead us to change the basic decision procedure, e.g. from coin-tossing to saving the greater number. This, fairly clearly, cannot be what is meant, as there are only finite possibilities and, once we are already committed to saving A and B, the presence of C alongside them cannot make any further difference – unless it were perversely to change us back to coin-tossing or a weighted lottery! Presumably what Scanlon means is that each extra claim should be taken into account within the procedure. This is most clearly seen in the case of weighted lotteries that proportion the chances of each group to their numbers300. 297 298 Scanlon (1998) pp.232-3. Otsuka (2000) p.291, Timmermann (2004) p.110. 299 A point made by Otsuka (2000) p.291 and Otsuka (2006) p.115. 300 Technically the proportionately weighted lottery is only a special case of the more generally weighted lottery – for giving the five people a chance between 51% and 99% would be to accord them greater weight. In what follows, I assume proportional weights unless specified otherwise. 111 (3.7) Scanlon’s Argument for Saving the Greater Number Scanlon thinks that, if each person must make a difference, this is enough to reject coin-tossing. He also thinks he can reject weighted lotteries, which he admits do respect each equally301, on independent grounds (see 3.10, below). His argument for saving the greater number must therefore have two parts. Firstly, he has to show that each person does ‘make a difference’, in the way that he requires and that cointossing fails to satisfy, on the saving the greater number policy. Secondly, he must have some other reason to reject weighted lotteries. Scanlon rejects coin-tossing because the presence of B makes no difference. It is easy to see how B does make a difference if the rescuer saves the greater number – because by changing the situation to one where there is a greater number, rather than a tie, B now obligates the rescuer to abandon David and save the two. But does this generalize? Suppose we now add person C, alongside A and B. Intuitively, C’s presence makes no difference – one was already required to rescue A and B, and now one is still required to save A and B, as before, but also C. Of course, one may point out that it would be unwise of C to press this complaint – after all, as things stand, he is to be rescued anyway, even if only on account of being with A and B. Were C to demand a procedure that took account of his life, the result might be to adopt a weighted lottery, which in this case would give him only a 75% chance of rescue rather than certainty of such. Admittedly, then, it would not be prudentially wise for C to complain, but Scanlon is committed to the position that what we are entitled to as a matter of right is not always what we would subjectively prefer – for example, we are entitled to food rather than help building a temple for our god, even if we 301 Scanlon (1998) p.234. 112 would go without food to build that temple302. Therefore, if all C is entitled to is to be counted, then the fact that he would prefer to be rescued does not matter – after all, everyone would prefer that, which is what generates the conflict to begin with. At least in C’s case, being saved would seem to compensate for not being counted. Now suppose, however, that as we are about to rescue A, B and C, we notice another person – D – alongside David. Now surely his claim is ignored. If we still proceed to save the larger group, it does not matter whether the case was three-against-one or three-against-two. It seems the only way in which Scanlon can say these people, C and D, do count is hypothetically. For instance, had B not been with A, then C’s presence would have been enough to break the tie with David – and, in that case, D would then have reestablished the tie against A and C. So it is not that each person must actually make a difference, because they might already be out-weighed by others, but had the numbers been otherwise they could have made or broken a tie. But if this is all Scanlon means by making a difference, it is not obvious that he can use the claim to reject the equal chances procedure. Return, for example, to his original case of adding B to the A against David conflict. Scanlon thinks, if we still toss a coin, B can complain of making no difference. But the defender of coin-tossing can now say to him ‘Of course, with the numbers as they are, you did not actually make a difference, but don’t you see how you could have made a difference? Had A not been there then, instead of simply saving David, we would have been obligated once more to toss a coin. That is, you would have created a tie’. Scanlon’s criticism of equal chances fails, on an ad hominem basis, because he cannot find a sense of ‘making a difference’ on which saving the greater number does 302 Scanlon (1975) pp.659-60. 113 allow each to make a difference but equal chances do not. But this does not mean that equal chances are vindicated from the original charge. It might rather be that saving the greater number is equally condemned, because individuals only make a difference when the numbers are already close – extra people added to either side of an already very uneven contest are practically irrelevant. If this is so, then we should turn to the remaining policy – the weighted lottery – to see why Scanlon rejects that (see 3.10, below). We must begin, however, with the positive case for such a procedure. (3.8) The Weighted Lottery: Pooling Chances Technically a weighted lottery is any random procedure that gives the larger group a greater chance – for instance, in a three-against-one conflict that could range from giving the three a 51% to 99% chance. Here, however, I shall use the term to mean more specifically a proportionately-weighted lottery, that is one that assigns chances to each group in proportion to numbers, so in the four cases discussed earlier303: the chances alter each time from 50/50, to 33/67, to 25/75, to 40/60. It is easy to see how each person counts equally, contributing a 1/n chance to their group. Admittedly, each extra person added reduces the significance of every person – by the end, each person has only a 20% individual claim – but this is simply the result of total chances being shared between more people, just as dividing a cake between more people results in smaller slices. The basic idea is that each person has some individual baseline chance (one over n), coupled with the intuition that it is permissible for them to pool these chances – so in the three-against-one case the three have a 75% chance of all being saved, as opposed to each having just a 25% chance. Timmermann represents this with a wheel of fortune, supposing each individual 303 Respectively: 1 David vs. A; 2 David vs. A&B; 3 David vs. A&B&C; 4 David&D vs. A&B&C. 114 occupies an equally-sized slice, and the would-be rescuer spins the wheel to decide which one person they should save – only to then later acquire secondary obligations to save those others they can, once the situation is no longer one of conflict304. This clearly does respect Scanlon’s individualist restriction, and each individual is given an equal chance of being the one chosen, even if they have de facto different chances of actually being saved. It is this difference in actual prospects that has motivated some criticisms of pooling chances305. Timmermann seems to suppose the wheel is divided into four quarters, the first assigned to David and the other three each to A, B and C respectively. But why, one could ask, must these chances be exclusive? Having put David and A in the first two quarters, when it comes to placing B – giving her a 25% chance – why can one not do that by putting her in the same quarter as A? Since, if that second quarter comes up, both will be saved, this seems to respect both persons’ 25% chances. Then, the same reasoning would have it, we can do the same with C. The result is the first quarter of our wheel tells us to save David, the second quarter to save A and B and C and the third and fourth quarters are unallocated. The result is equal chances between the two unequal groups, plus some wasted chance. Since we presumably want greatest equal chances, we should now abolish this wasted section, and increase the first two quarters to halves, which effectively takes us back to tossing a coin. I think there are two answers to this. The first takes us down a slightly longer and more complex route, which I will develop in the next section. This is not, however, essential to my overall project. The second is simply to accept equal chances as one possible fair procedure, but claim that we can reject it on democratic 304 305 Timmermann (2004) p.110. Hirose (2007) pp.49-50. 115 grounds because individuals do not make any difference and, in the political context, it would be prone to manipulation. I return to this in section 3.9, below; but first, the longer response. (3.9) Re-constructing the Fairness of a Weighted-Lottery As well as criticizing the weighted lottery for giving individuals unequal de facto chances of being saved, Hirose also points out that – as described by Timmermann – it could actually lead to an inverse lottery if the rescuer responds to his dilemma by randomizing who not to save and then reasoning similarly306. That is, if they draw A as not-to-be-saved, they realize that they cannot now save B – for to go that way will obligate them to save A too – and so they end up saving David with a two-thirds chance. However, his first criticism was that the de facto unequal chances of being saved only arise because of the assumption that we must save A and B together – whereas to give each a true one-thirds chance might require that we, for example, saved A but left B. If we don’t assume that A and B must be saved or left together from the start, though, then – contrary to the second objection – we don’t get an inverse-lottery. Instead, having picked A as not-to-be-saved, we still face a choice between David and B. Then we must flip a coin between David-alone and B-alone. On this view, we can save B-alone, as A has already been considered, and lost in the lottery. I shall come back to whether we should save B-alone, when we could also save A. For now, the two-stage lottery can be represented by a decision-tree: Fig. 3.1 Partial Decision Tree 306 Hirose (2007) p.51. 116 David not saved 1/3 Who not to save? David, A or B 1/3 B not saved If we draw A, and then B, as the unlucky victims not to be saved (the route marked in bold), then we should save David. The likelihood of this is one-sixth (i.e. one-third multiplied by one half). If we draw A, and then David, not to be saved, the outcome is that we save B-alone, again with one-sixth probability. There are six possible combinations – 3 and 4 being those completely shown above. Fig. 3.2 Results of Two Draws Possible combination 1 2 3 4 5 6 David David A A B B Loser of first draw Loser of second draw A B David B David A B A B David A David Who is saved 1/3 A not saved 1/2 B not saved 1/2 David not saved Each these six is equally likely, with probability one-sixth (i.e. one-third times onehalf). As each person can be saved by two combinations (e.g. David is saved by notA, then not-B or not-B, then not-A), each has a one-in-three chance of being saved somehow. This is the result of them not-losing the first draw (two-thirds) and then, if they got this far, not losing the second (one-half). Thus, by abandoning the requirement that A and B are saved or left together, we get back to the situation suggested in the first objection, whereby each of the three gets a one-third chance. This may be wasteful, as it can tell us to save A without B, and vice versa, but it is fair – all again have equal chances. 117 We can refine this outcome slightly, however. Suppose the first draw results in David being selected as not to be saved. Now there is no longer a conflict, since it is possible to save A and B together, so a second draw is unnecessary. Thus there is a one-third probability of saving A-and-B together. If A is drawn in the first lottery (one-third probability), however, we still must choose between David or B-alone (each one-sixth probability); while if B is drawn in the first lottery, we must choose David or A-alone (again one-sixth probability). The result is a two-sixths (or onethird) chance of saving David, a one-third chance of saving A-and-B, and one-sixth chances of saving each of A and B alone. By this reasoning, A and B each have a onehalf chance of being saved, though not necessarily together. There is also a one-third chance of waste, i.e. of saving only one of A or B. Fig. 3.3 Second Draw Not Needed Chance 1/3 1/6 1/6 1/6 1/6 Draw 1 David (not saved) A (not saved) A (not saved) B (not saved) B (not saved) Draw 2 (not needed) David (not saved) B (not saved) David (not saved) A (not saved) Outcome (saved) A-and-B B-alone David A-alone David This, I maintain, is perfectly fair. Although A and B have larger de facto chances of survival, David has no ground for complaint: he was merely unlucky that he was in conflict with both A and B and drawn as the one not to survive. Each is respected equally. To reduce A and B’s chances would not benefit David, but would simply be a form of ‘levelling down’ – reducing their chances without improving anyone else’s, just for the sake of equality307. Therefore, fairness is satisfied by a lottery over the four alternatives: saving David (probability one-third), saving A-and-B (probability 307 Parfit (2000 [1991]). 118 one-third), saving A-alone (probability one-sixth) and saving B-alone (probability one-sixth). It can still be objected that this fair procedure is wasteful. There is a one-inthree chance of saving either A or B alone, when we could do better by saving both. In these cases, however, I think it would be permissible to save the extra person, even though the lottery had selected them as ‘not to be saved’. Fairness alone does not fully determine how we distribute goods – we must also pay attention to outcomes. When it comes to dividing a good between two competing claimants, fairness demands only that they be treated equally, and this can be achieved either by dividing the good between them or a lottery to see who gets it. Sometimes division seems more appropriate, as when the good can be divided without losing value and each is likely to prefer half of the good to a 50% chance of all of it – for instance, when we cut a cake in two. Other goods are not so easily divisible, at least not without sacrificing what is valuable about them – for instance, a needed transplant organ or a child – and in these cases we prefer a lottery to division because it achieves better results. In this example we can make a Pareto improvement and this seems reasonable grounds on which to depart from strict equality308. Fairness does not require us to level down, where we can benefit one person without harming anyone else. The lottery was only introduced because we faced the tough decision of having to abandon one or more. If it is now possible to save an extra person, whom we were previously forced to abandon, it would be fetishistic to leave them anyway, simply because we had selected that course of action when we could not save them too. If David were to suddenly teleport to join A and B, we would not abandon him anyway. 308 At least on a ‘leximin’ interpretation of Rawls’ difference principle. 119 Likewise, although we were prepared to abandon A, we should not do so gratuitously, if we can now save her at little cost, while rescuing B. Thus, if our lottery indicates that we should save A or B alone, then we should in fact save them both, so we are effectively back to the original weighted-lottery. The re-constructed procedure is extensionally equivalent: there is a two-thirds chance of saving A-and-B, and onethird of saving David. David cannot complain about our Pareto improving deviation from strict equality, as his one-third chance is not affected, whether we save A-alone or A-and-B. It appears that the first objection to weighted-lotteries has defeated the second; and that we have re-constructed a case for them, as a practical equivalent of two-stage lottery draws. Thus, we have a principled defence of weighted lotteries. (3.10) Counting Individuals, Again Regardless of what one thinks of the argument of the previous section, there is one compelling reason to prefer weighted lotteries to equal chances for each group when it comes to resolving conflicts of interest democratically: proportional chances do allow each individual to make a difference, whereas equal chances effectively ignore the distribution of preferences. Here we may assume that there are two things people have interests in – one is that their view prevails, but they also have an interest in its prevailing because it is their view, i.e. in their interest being counted and effective. This is why it is not simply enough that people could move to a country with laws they approve of, or for benevolent rulers to impose laws people happen to agree with. Democracy requires that the people’s preferences be efficacious. Equal chances may give people fair chances at getting what they want, but it does not respect each person’s preferences because the numbers make no difference. 120 If we insist on equal chances, then it makes no difference whether ten voters are split 5/5 or 9/1, since we will be tossing a coin either way. That effectively means that, once one person has voted for each side, the other eight are ignored. Equal chances is fair in that it gives each person an equal likelihood of satisfaction, even after votes are cast, but it does not give each voter an equal chance of determining the issue. Moreover, it would have perverse effects within a wider democratic system – for instance, there would be no incentive to try to win over other voters, and very little reason for the majority of people to vote, as long as they were sure one person of their view would. Further, attempts to realize equal chances may break down to something closely approximating proportional chances anyway, provided that voters can co-ordinate their actions. Because it is hard to delineate separate options that deserve separate and equal chances, voters may vote for different proposals that are close enough to their ideals to satisfy them, but claim they deserve another chance. For example, suppose on some spending decision three voters want £10,000 spent. If all of them vote for this figure, they will have only an equal chance with any other option; yet if two of them vote strategically – for figures of £9,999.99 and £10,000.01 let us say – then they can claim an equal chance for each of these three figures. Since the difference between these options is negligible, the result is practically indistinguishable from giving £10,000 three chances. Since I have argued that proportional chances are fair, I think this is what should be done anyway. The fact that equal chances may in practice result in such – or, worse, better reward only co-ordinated groups – seems further reason to favour proportional chances anyway. 121 (3.11) Scanlon’s Argument Against Weighted Lotteries We have already seen arguments for and against the weighted lottery, but now it is time to return to Scanlon’s objection. Recall that Scanlon accepts that weighted lotteries do count each equally – indeed, I have argued they do so considerably better than saving the greater number (see 3.7 above). Given this, it is somewhat obscure why Scanlon rejects the possibility of weighted lotteries – he simply asserts that saving the greater number is a better procedure309. Although he recognizes that minorities will have prudential grounds to call for any sort of lottery, he thinks this is irrelevant, for he claims that it would be unreasonable of them to do so. As he puts it, “There is no reason, at this point, to reshuffle the moral deck, by holding a weighted lottery, or an unweighted one”310. I shall now attempt to reconstruct what seems to motivate Scanlon’s objection, before going to argue that it depends on the contingencies of the particular circumstances and so cannot rule out weighted lotteries where those conditions are not met. I believe Scanlon’s ‘reshuffling’ metaphor is revealing. A lottery is only reshuffling if the deck has already been shuffled to begin with. It seems, therefore, that Scanlon is implicitly assuming that it is already a matter of chance who has ended up in which group, and thus who is in the greater number and who is in the minority311. This is, however, effectively to adopt a Harsanyi-style original position, in which we effectively assume an equal chance of being anyone and so are no longer worried by inequalities, and so it is unsurprising that it delivers utilitarian conclusions. We can 309 310 Scanlon (1998) p.324. Scanlon (1998) p.324. 311 I will turn, shortly, to political implications, but note the similar assumption underlying majorityrule; c.f. Guinier (1994) “Simple majority rule assumes that the majority and minority are fungible, meaning that the outcome of voting procedures depends solely on the shape of the distribution of the preferences and not on which voters hold certain preferences” (p.77), but this breaks down in the case of racial (or other deep) divisions, because “When majorities are fixed, the minority lacks any mechanism for holding the majority to account… The permanent majority simply has its way, without reaching out to or convincing anyone else” (p.9). 122 see how important this assumption is if we make this prior randomization explicit, and compare it to cases where there is no such randomness. (3.12) Prior Randomization versus Fixed Majorities Let us make explicit the assumption of prior randomization. Suppose that our David is on a ship with A, B, C and D when it breaks up in a storm. Four of them manage to get to a lifeboat, while the remaining one is left floating on a piece of wreckage. It is a matter of chance who ends up in each position – i.e. any of them had a roughly equal probability of being stranded on his own, as in fact happened to David. In this case, we have four who will be saved and one who will not, but ‘the greater number’ is an abstract category or non-rigid designator312 – it does not pick out definite people a priori. Each individual had a four-in-five chance of ending up in this larger group and thus each had an equal (four-fifths) chance of being saved. In this case, to now hold a weighted lottery would not only be to “reshuffle the moral deck”, but it would be to diminish everyone’s chances – their chances would now be just 68%313 instead of 80%. Thus, while it would still be fair – in that it gives the same chance to each – it violates our criterion of greatest equal chances – it is suboptimal given that we could guarantee saving more people, which increases everyone’s ex ante chance of being saved, without being unfair to anyone. This seems to be why Scanlon favours saving the greater number, because he assumes some such randomization has already taken place. But now consider the situation if this is not so: suppose David goes out to sea alone in a small fishing boat, 312 I owe this phrase to Jerry Cohen, who so describes ‘the worst off’ in Rawls’ difference principle – whoever happens to be worst off, not any definite prior group. So it is with the majority – they are not a group picked out by anything beforehand. 313 This is (chance of being in the four) x (chance of the four being rescued) + (chance of being the one) x (chance of the one being rescued), or 0.8x0.8+0.2x0.2. On this point, see Otsuka (2006) p.124. 123 while A, B, C and D are in the same waters in their larger ship. I am assuming that this a regular occurrence, so that who goes to sea on their own is not plausibly a matter of prior randomization – suppose, for example, that David is a fisherman who regularly undertakes such trips alone, while the other four are pleasure-trippers who only do so in groups. Now if there is a storm and all get into trouble, it is clear that it will be A-D who are the larger group and David will be on his own. There is no randomization in this case – David knows, when he goes to sea, that if both groups were to get into trouble he would be the one not saved. In this case, would it be unreasonable of David to object to the policy of saving the greater number? There is no real sense in which he – as a concrete individual – had the chance to be in the greater number, and so saved. He had a chance only if we appeal to something like the original position, but I have already argued that this does not justify all arrangements that no longer seem fair once people are actually situated (see 3.3 above). This difference exactly mirrors that between a permanent, fixed majority and a fluid society in which each individual has an equal chance of being in the eventual majority. It is no surprise that those who defend majority-rule generally assume or require that the majorities are or should be changing ones, and thus that each person does have a chance of being in the majority – if not this time, maybe next. The fairness of majority-rule therefore seems to be premised on the essential ‘randomness’ of who is in the majority. In such situations, majority-rule would indeed be fair – and, indeed, probably the best fair procedure, since more people get their way. Lottery-voting, like the weighted lottery, however, is motivated by cases where it does not seem that all have had this prior equal chance – cases where there is a single significant cleavage dividing society into a permanent majority and 124 minority314. In these cases, the minority have no chance to have their say on anything under majority-rule315, so it seems that they could reasonably reject it in favour of lottery-voting, which gives them their proportionate chance. Where this is not the case, I would still hold that lottery-voting is both fair and democratic, but that it is not all-things-considered best, because majority rule is both fair and democratic and ensures that more people get their way on each decision, thereby better meeting the ‘greatest equal chances’ condition. (3.13) Proportional Chances versus Proportional Outcomes I have accepted that majority-rule can be fair in certain circumstances, viz. those where everyone has an approximately equal chance of being satisfied by being in the majority. Where these conditions are not met, however, equality calls for proportionality to be realized in some other way. Kamm accepts both of these points, but as well as considering the proposal discussed here, ‘chances proportional to numbers’, she considers another possibility – ‘results proportional to numbers’ – according to which there should be a compromise that partially satisfies everyone. This seems less obviously applicable to life-or-death saving cases, but Kamm still thinks there may be fanciful instances where it is appropriate. I accept that if we can engineer a compromise that gives everyone some of what they want then that is often more appealing than a winner-takes-all lottery, because it also achieves more retrospective equality. Thus, if one-third of a group want to spend their group budget on X, and the other two-thirds want to spend it on Y, for example, then it may be perfectly reasonable to split the budget proportionally between X and Y. Kamm 314 This can be over a series of decisions or in a one-off decision where it is clear there is an obvious majority. 315 C.f. Guinier (1994) pp.1-9, and passim. 125 points out that such procedures are often used “in coalition governments when a compromise policy reflects the position of members in proportion to their numbers”316. This may also be the result of consociational arrangements or log-rolling between groups. Such proportional compromises have been defended in democratic literature for best respecting retrospective equality between all parties317. However, not all decisions are susceptible to compromise. This is not because there are no intermediate policy positions. While it is often dichotomies that seem most salient (e.g. war or peace, is abortion legal or not?), this masks the fact that there are a range of different policy options, e.g. peace with disarmament or peace with increased defence spending, abortion on demand or with restrictive controls. Nor need the problem be that people are ‘unreasonable’ or absolutist in their demands – as perhaps the case between anti-abortionists and pro-choicers. It may simply be that compromise is impossible in the particular situation – e.g. splitting the budget between X and Y cannot give either group what they want – or too difficult to come to318. While it may be that discussion and deliberation naturally engenders compromise, with each side taking on some of the others’ position, it may be that it is an impossible goal, even in the small group, face-to-face democracy focused on here. I certainly do not want to give up on compromise where it is possible, but it will be hard – if not impossible – to institutionalize. Lottery-voting is based on compromise at a higher level that a single policy outcome. While it may be a good thing if decisions ‘even out’ over the long run, so each group get their way on a 316 317 Kamm (1985) p.191. E.g. Jones (1983) pp.168 and 175, and Hyland (1995) pp.94-8. 318 For instance, the range of ways in which abortion could be restricted include permitting it only up to a certain stage of pregnancy, requiring ‘medical grounds’, doctors’ approval, consent from male partners/parents, a limited number of operations per person, a financial charge (as opposed to state funding) and so on. Even supposing those pro- and anti-abortion were to agree that they should compromise in principle, working out a particular compromise arrangement between them would be a very complicated matter. 126 roughly proportional number of issues, that is not part of the case for lottery-voting because a chance-based procedure can never guarantee this and it would require further checks to ensure, for example, that each issue voted on was of approximately equal importance. Rather than the minority getting partial consideration in each case, they are likely to get all of their way on a few. This may be a more attractive outcome for all concerned, but it is not what motivates lottery-voting – though it will be a consideration in deciding whether it is the best decision-rule all-things-considered. Lottery-voting is instead a compromise at the procedural level, designed to ensure that no group’s opinion is ever defeated and discounted from the start. This seems to treat everyone’s interests fairly and, moreover, strengthens the minority’s bargaining position in any negotiations aimed at compromise. If the majority know that they can have all their own way, or most of it, then they have little reason to make concessions. If all sides know that, if no agreement is reached, the matter will go to a lottery, then they may have powerful incentives to reach a compromise that all can agree on, in order to protect what they see as their most vital interests. Certainly the ideal of compromise is attractive enough to be worth serious consideration, I simply have doubts that it can be easily achieved through institutional processes. In any case, it is only one possibility, and Kamm thinks deciding individual cases by assigning ‘chances proportional to numbers’ (i.e. weightedlotteries) is another fair procedure, and this is what is explored here. While, where it can be achieved, compromise may (like majority-rule) be a better fair procedure, allthings-considered, that does not detract from the fact that lottery-voting is both fair and widely applicable and therefore a reasonable solution where compromise on the particular outcome looks unlikely. 127 (3.14) Conclusion: Towards Political Application While parallels have been noted, most of the argument in this chapter is not explicitly concerned with democracy. Nonetheless, it is applicable because it is framed in terms of more general notions of fairness or equality when it comes to deciding between competing group claims. For example, Taurek’s aim is to show ‘equal concern and respect’ to each person, which is a general precept of liberal democracy319. The parallel – or extension – of these arguments to voting procedures or public policy has been widely noticed within the debate320, and in particular Scanlon’s insistence that each person counts mirrors the basic premise that each vote should make an equal and positive difference in favour of the option for which it is cast. The relevance can be increased if we simply replace the phrase ‘saving the greater number’ throughout with ‘satisfying the greater number’, which we can generally assume will come to the same thing in practice. Indeed, Kamm identifies the former with ‘majority-rule’ and speaks not only of saving people but also of satisfaction and preferences321. In the cases we are concerned with, it seems that numbers should make some difference (pace Taurek). If we adopted an analogue of Taurek’s coin-tossing proposal it would mean each option or proposal being given an equal chance – whether favoured by just one voter, or 99% of them322. I have already pointed out that this would be profoundly undemocratic because it does not count numbers at all (see 3.9, above). The question, however, is what difference numbers should make. If the 319 E.g. Dworkin (1977) pp.180-3, Dworkin (2000) esp. pp.184-5, Ackerman (1980) pp.289 and 301, Sartori (1987) p.343. 320 E.g. Taurek (1977) pp.310-4, especially p.314, Kamm (1985) p.181, Wasserman and Strudler (2003) p.81, fn.18 and p.92, fn. 33. Scanlon (1998) p.397 fn.38 notes his argument is restricted to the case in question and “in an electoral district with a permanent minority it might be argued that a weighted lottery in which everyone votes and the decision is made by choosing one of the votes at random is fairer than majority rule as a method for making political choices”. 321 Kamm (1985) e.g. p.191. 322 C.f. Estlund (2007) pp.6 and 82-4. 128 fair thing to do, when faced with two equal groups, is to toss a coin then how should we be affected by a single extra person on one side? Must they really determine the outcome? This may seem reasonable if we move from one-against-one to twoagainst-one, because two is significantly more than one. It has, however, seemed less obvious to many when we are dealing with larger numbers, such as 1,001-against1,000. In this latter case, tossing a coin may seem fairer, and that is – I suggest – because it approximates the ideally fair solution of proportional chances. Unlike simple coin-tossing or equal chances, lottery-voting, because it proportions chances to votes, does respect numbers. It means that if 60% of people vote a certain way, they have a 60% chance of victory. The lottery allows each group to preserve their chance of victory, after the votes have been cast, in recognition of the fact that vote-casting is not an entirely random or unpredictable matter, and that where there are permanent majority/minority divides some would not have any prospects of success under majority rule. I have argued that, if we are sure that ‘prior randomization’ has already effectively taken place in the distribution of votes, then all should be willing to accept saving the greater number or majority rule, because this maximizes their chances without discriminating against anyone (see 3.11 above). The problem is when it is predictable how the numbers will fall. As Kamm points out: “Majority rule does not allow the single individual any chance to win, but CPN [chances proportional to numbers] implies that it would be unfair to deprive him of his chance”323 And Timmermann, speaking of the rescue cases, adds: “It is rational for the members of a society not to choose to maximize the probability of being saved. A somewhat lower overall probability is the price they would be willing to pay for their claim’s never being discounted right at the beginning”324 323 324 Kamm (1985) p.187. Timmermann (2004) p.112 (though I would say reasonable, not necessarily rational). 129 So far, I have argued that proportional lotteries are a fair way of adjudicating between differently-sized groups. This is what lottery-voting offers: one individual is randomly-selected to decide the issue, and all those (ideologically) close to them also benefit by getting their way, which exactly mirrors Timmermann’s ‘individualist lottery’. Nonetheless, more needs to be said about how this account of fairness can be incorporated into democratic institutions in practice. The remainder of this thesis attempts to describe and assess lottery-voting as a way of realizing political equality in the necessary conditions. 130 4 Lottery-Voting Described “[T]he defects in majority rule… oblige us to look with the utmost skepticism on the claim that democracy necessarily requires majority rule”325 “[I]t would appear that democracy is almost a random choice process”326 (4.1) Introduction So far, I have rejected majority-rule as necessarily fair in all contexts and argued that democratic equality requires a form of proportionality, and the previous chapter argued for proportionality of chances. The present chapter describes how this can be achieved, by combining voting with a lottery over the votes – something I call lottery-voting. I begin by pointing to the history of sortition (random selection) in democratic practice, and observe that it retains a place in the theory of representation. What I want, however, is to apply lotteries not to selecting representatives, but direct decision-making, illustrated with two hypothetical examples. I go on to argue that this achieves the desirable goals of counting each vote equally and making outcomes contingent, which ensures there are always incentives for participation and avoids the problem of excluded minorities. The next chapter further explores whether lotteryvoting meets the reasonable minimal criteria of a democracy, by comparing it to the axiomatic conditions of May and Arrow. (4.2) The History of Lot The previous chapter drew attention to use of lotteries in various non-political distributive issues. Lotteries, or random selection devices, were also a key feature of 325 326 Dahl (1989) p.162 [not emphasized in the original]. Tullock (1967) p.33 [not emphasized in original]. 131 democracy, at least in its Athenian origins. Greek democrats believed all citizens327 were equally competent to exercise the duties of the assembly (ekklesia). While certain important posts, like generals, were elected, this was seen as inherently undemocratic – rather it was a means to finding the best man for the job, and thus aristocratic in the original sense of ‘rule by the best’ (aristos). Thus, as Aristotle put it, “the filling of offices… by lot is regarded as democratic”328. Whether participation in the council (boule) was seen as a privilege or burdensome civic duty, it was seen as something that all citizens were equally qualified for and entitled to, much like modern jury service. The obvious way to appoint some of these equals, therefore, was to use a device known as the kleroterian – essentially a lottery machine that selected random citizens to fill the roles. The result was a rotation of office-holding, in which all citizens had a reasonable chance of serving at some point in their lives. This is not to say, of course, that the Greeks trusted wholly to the lottery. All office-holders were scrutinised both before and after taking up their position. Thus the equal opportunity of the lottery was combined with a form of quality control. Such ‘checks’ are common in combination with random selection; few would want to trust themselves entirely to chance329. Interestingly, a similar idea is operative in Philip K. Dick’s science fiction novel The Solar Lottery, where a ruler is appointed by a game of chance called ‘the bottle’, but a challenge in the form of an officiallyappointed assassin is in place to quickly eliminate weak rulers330. In lottery-voting, it is the vote that provides a check on the suitability of options – an option only gets a 327 328 A term whose extension was considerably narrower than the modern franchise, of course. Aristotle (1988) p.95 [Pol IV.9 (1294b8)]. Though note he later appears to reject this en doxa, suggesting what really matters is whether the appointment is made by and from some or all, rather than whether it is by election or lot, pp.107-8 [V.15 (1300a20-b4)]. 329 Those who do accept unchecked lotteries to appoint representatives, for example Callenbach and Phillips, have the safeguard of a larger representative chamber in which they can be reasonably sure of chances evening out – producing something akin to proportional representation in a chamber, where a majority continue to get their way. This does not prevent a few crackpots being selected, but is effectively a check against extreme minorities wielding any ultimate power. 330 Dick (1955) esp. pp.38, and 75. 132 chance if someone has been persuaded to vote for it, and perhaps a significant number of people (I return to the possibility of imposing minimum thresholds in chapter 5.3). The use of lotteries was not some quirk unique to the ancient Greeks and science fiction, however. Complex lotteries continued to be used in other contexts, for example to select between candidates for office in Italian city states through the middle ages and renaissance. That they fell out of use in modern times was probably due to a distrust of such a radically egalitarian – indeed, democratic – method. The Founding Fathers of the US, for example, were primarily republicans rather than democrats, as shown by the way they embraced institutions such as the Electoral College, with the aim of promoting wise rather than simply popular decisions. They were not primarily concerned to make government responsive to the will of the people, but to ensure checks and balances between all competing interests, with the result that government is predominantly liberal – leaving people to lead their own lives – rather than simply democratic. Similar concerns were operative on both sides of the Atlantic, for while J. S. Mill called for extension of the franchise he coupled this democratic drive with a distrust of the masses thereby empowered, which led him to call for plural voting for the intellectual elite and for government to confine itself to preventing harm, rather than trying to make people happy. Nonetheless, lotteries remained part of both democratic practice and theory. Montesquieu and Rousseau both followed Aristotle’s analysis by associating democracy with random selection, remarking respectively that: “Voting by lot is in the nature of democracy; voting by choice is in the nature of aristocracy. The casting of lots is a way of electing that distresses no one; it leaves to each citizen a reasonable expectation of serving his country”331 331 Montesquieu (1989 [1748]) p.13 [Spirit of the Laws II.2]. Note that he refers to lot as voting and election. 133 and: “[E]lection by lot is more natural to democracy… the holding of office is not a benefit, but an onerous duty, which cannot justly be imposed upon one citizen rather than another. Only the law can place this duty on him to whom it falls by lot… There would be few disadvantages to appointment by lot in a true democracy… When elections and drawing lots are both employed, the former should be used to fill posts requiring particular abilities, such as military positions; the latter suits those in which common sense, equity and integrity are sufficient”332 This is a trend that has survived in some recent, radical democratic theory. (4.3) Modern Proposals In 1998 the British think-tank Demos published a pamphlet titled The Athenian Option333, which recommended reforming the House of Lords by creating ‘people’s peers’, randomly-selected from the electoral register in the same way as juries. This was not an isolated or wholly idiosyncratic suggestion, but built not only on the previously described history of sortition but a revival of theoretical proposals inspired by Athenian practice in the 1980s. Some of these are extremely radical, for example Burnheim proposes abolishing the state apparatus, in favour of anarchist communities governed by randomly-selected functional groups334. Callenbach and Phillips put forward something more similar to the Demos proposal for the US in their 1985 book, The Citizen Legislature335. They go further, however, by suggesting replacing not the upper but the lower house of Congress – that is, making the House of Representatives a genuine Representative Chamber, drawing a true cross-section of the population by randomly-selecting 435 ordinary citizens to fill office on overlapping three-year terms. While they appeal to the fact the Senate will still be able to exert a moderating 332 333 Rousseau (1994 [1762]) pp.139-140 [Social Contract IV.3]. Barnett and Carty (1998); c.f. Van Mill (2006) p.173. 334 Burnheim (1985). Another anarchist, Wolff (1976), also considers but rejects something close to lottery-voting, pp.44-7. 335 Callenbach and Phillips (1985). 134 oversight, this proposal actually places power of legislative initiative in the hands of ordinary people. Similar proposals for randomly-selected representatives have also been made by Carson and Martin336 and Manin337, drawing on the fact it results in a form of proportional representation. Other proposals have involved using random selection in setting constituencies for representation338, but these are not my concern here. Randomly selecting ordinary citizens to participate in some form of legislative assembly has obvious attractions – for instance, it gives all an equal chance of involvement, spreading participation much wider than a narrow political elite, and tends to realize the ideal of ‘mirror representation’ since the random sample should – by the law of large numbers – include members of all major social groups. However, there are also many equally obvious objections. Many ordinary people may, for various reasons, simply not be well-qualified for the job, or unwilling to do it. Forcing people to serve would not only be coercive, but also probably counterproductive, since those people would probably do a worse job than if they were to participate voluntarily. On the other hand, if the random selection was limited to those who volunteered, then equal representation would be extremely unlikely, since volunteers would almost certainly not be a representative sample of the population and, if it was primarily the rich and educated from whom representatives were drawn, then the result would probably not be so different from what we have today. One way of avoiding the main objections to simple sortition is to combine it with an element of choice. That is, we can allow those who are selected to choose who will represent them – if they wish, they can take their seat in the assembly but if, for whatever reason they do not wish to do it themselves then they can send a 336 337 Carson and Martin (1999). Manin (1997), and Przeworski, et al (eds.) (1999). 338 Rehfeld (2005). 135 substitute to represent them339. This moves away from the idea of mirror representation, but arguably does at least as good a job of representing people’s interests (since even if individuals best know their own interests, there is no guarantee they are best placed to speak on their behalf – and, of course, if they feel they can only be represented by other women, blacks, etc then they can appoint a representative of the same ascriptive group) while allowing people greater freedom. If an ordinary member of the public does not wish to represent their interests in person, but is happy to appoint a replacement who is willing to do so, on her behalf, then this is a Pareto-improving division of labour, which suggests that we should allow such arrangements. Of course, we do not always allow people to transfer positions they are entitled to – for instance, the successful job applicant is not allowed to pass the offered post to someone else if he no longer wishes to take it, and under current electoral practices I would feel rightly aggrieved if, having voted for an MP, she then transferred her place in parliament to another. However, it is important to see what I am describing is not like this latter case: when I vote for an MP, I do so on the understanding they have put themselves forward to represent me, and others. What I am in effect doing is authorizing them to speak for me in the legislature, and this is extensionally equivalent to substitution in the lottery case – only rather than it being the majority who get to appoint some collective representative, it is the single selected person who can appoint the representative they want (which may or may not coincide with the wishes of the majority)340. 339 We can be neutral between mandate or trustee models of representation here. The idea of sending a substitute in one’s stead if selected is controversial, particularly if one pays them to take one’s place, but has been used in some military drafts, see Elster (1989) p.78 and Duxbury (1999) pp.150-8. 340 Interestingly, Charles Dodgson [Lewis Carroll] proposed allowing candidates to transfer surplus votes they received to other candidates – see McLean and Urken (1995) p.54. 136 Although he does not argue his case in this way, Akhil Amar has proposed effectively this scheme341. The proposal just described was one where all individuals’ names are placed in the hat, with those selected then entitled to either represent themselves or appoint a replacement, but one could see voting as a way of naming such a representative in advance. While each citizen is entitled to have his or her own name in the hat, those who are particularly keen to take on the office may announce that they are ‘standing’ for such, and then others are given the option to confer their vote – and thus their chance – on those running. In effect, when I vote for Jones, I replace my own chance with another chance for him – which is exactly what would happen were I to have my own name in the hat but, if drawn, abdicate in his favour: Jones gets two chances (his and mine), while I no longer have an independent chance of being chosen myself. Logically, everyone could decide that they want to represent themselves – thereby effectively voting for or nominating themselves, with the result being like the simple sortition method we began with342 – however, it is also possible that almost everyone would prefer to nominate one of a few self-selected candidates in their stead. In this case, we effectively have Amar’s proposal: votes are cast as usual, and a single vote is randomly-chosen to determine the outcome of the election. He calls this ‘lottery voting’, and it is from his proposal that I draw inspiration. Nonetheless, the lottery-voting I here advocate is considerably different from that he describes: most notably, Amar’s proposal is outlined in the context of appointing a representative assembly, whereas I intend to apply a similar procedure for making direct decisions, albeit at a smaller scale. 341 342 Amar (1984) and (1995). Amar (1984) p.1304 actually considers this ‘pure lottery’ a danger to be avoided, arguing candidates should have to appeal to a certain number of other citizens. In the 2005 British general election, Caroline Taylor-Dawson polled just one vote standing in Cardiff North for the Vote For Yourself Rainbow Dream Ticket, though it was not apparently her own. Source: http://news.bbc.co.uk/1/hi/uk_politics/vote_2005/wales/4523583.stm (last accessed 15/04/07). 137 (4.4) From Representation to Direct Decisions Lotteries have long been used to make appointments. The Athenian council (boule) and juries were so appointed, though both made decisions by majority. Similarly, the Italians may have used lotteries to select office-holders, but once appointed they made the decisions. The theoretical proposals of the previous section were also based around representatives, rather than actual decision making. It is easy to see why lotteries have been associated with representation, for the same reason that surveys use random samples of the population – given large numbers, the resulting sample will tend to be a cross-section reflecting the underlying opinions, attitudes and interests of the whole people. Nonetheless, there is no reason why proportionality should be restricted to matters of representation. I have argued that equality requires proportionality in decision making, which can be achieved either by compromise or proportional chances. Moreover, I have argued that there are often reasons to favour proportioning chances: A compromise over any given decision may be hard to engineer, since there will usually be various issues that may be involved, and subject to various trade-offs. Further, on some issues the intermediate result may please no one, so it may be preferable to give each group a chance of getting its way, with the result that each will be able to point to certain features of the social environment and say ‘at least that went our way’. The decision of a group to appoint a particular representative can be seen as a special case of their making a direct decision – after all, until they have a representative it is the people themselves that must make the decision. One can imagine a small city-state, along the lines of ancient Athens or Rousseau’s Geneva, in 138 which the citizens directly decided matters of internal government and appointed a representative to some larger federal body. In such a situation, depending on the power of the larger body, that appointment might be a particularly important decision, but it is still just another direct decision of the populace. If it is reasonable to employ lottery-voting to appoint a representative, then it might be reasonable to employ it in all decisions343. (4.5) Lottery-Voting Described I take the name lottery-voting from Amar, but abstract from his representative use. The idea is not unique, it has been suggested by – among others – Bruce Ackerman, David Estlund, Peter Jones and Robert Paul Wolff344. The proposal is that each voter expresses his or her first preference on the ballot paper but, instead of the winner being whichever option has most votes, a single vote is randomly selected to determine the outcome. Whereas traditional electoral methods involve simply counting up the votes – perhaps after some form of redistribution (e.g. STV) – in lottery-voting, the outcome is not determined solely by numbers, though it is influenced by such. Rather, once all of the votes are cast, they are placed in a tombola of some sort, and a single vote is selected at random. This vote determines the outcome, so whichever option it is cast for wins. If the vote is spoilt, a new draw is made, until a valid vote selects an eligible winner. The random selection of any vote from all those cast obviously satisfies the requirement that we took as a premise, namely that each vote should have an equal chance of determining the outcome of the 343 Of course, there may be reasons not to employ it. Amar (1984) pp.1283 fn.2 and 1303 recommends it only for choosing an assembly, not appointing a president. The reasonability of employing it for particular decisions may therefore depend on whether we see them as one-off decisions or part of a larger series of decisions. 344 Ackerman (1980) p.288, Estlund (1997) p.193, Jones (1983) pp.170-1, Wolff (1976) p.45. 139 election345. Note, however, that from the fact that each vote is equally likely to be decisive – or, equivalently, that each individual is equally likely to be the one whose vote determines the outcome – it does not follow that each individual is equally likely to be satisfied. The chances of any particular alternative winning the election depend on the number of votes it attracts. Thus the proportionality of chances, defended in the previous chapter, is maintained. If A is amongst 60% of people voting for option X, while B is one of the remaining 40% who all vote for option Y then, though they each have an equal (1/n) chance of their vote deciding the outcome, A is more likely to get the result she wants – because there will be a 60% chance of a vote for X being selected. To briefly recap the reasons to favour a proportional lottery over equal chances, what we want is for each vote to count equally and, when the votes are not divided equally, this departs from equal chances of satisfaction. There might be some cases where tossing a coin between the two opposing groups would be fair, since this would give each individual a 50% expectation of satisfaction. But if we were to give each option an equal chance of success, this would mean one that attracted a single vote would have as great a chance of victory as one that attracted all the rest of the votes, and this would hardly be to count all those extra votes equally – indeed, it would not count them at all. Further, such cases may break down to something closely approximating proportional chances if voters can co-ordinate. For example, suppose on some spending decision three voters want £10,000 spent. If all of them vote for this figure, they will have only an equal chance with any other option; yet if two of them vote strategically – for figures of £9,999.99 and £10,000.01 let us say – then they can claim an equal chance for each of these three figures. Since the difference 345 This is expressed by, e.g., Christiano (1996) p.55. See my introduction, sections 0.1 and 0.2. 140 between these options is negligible, the result is practically indistinguishable from giving £10,000 three chances. So far, I have simply assumed that we want each vote to count equally, and shown that this is satisfied by lottery-voting. This is not yet to complete the argument, though the previous chapter’s defence of proportional chances goes a long way to establishing its fairness. It remains to be shown that lottery-voting meets certain requirements of a reasonable decision procedure – in chapter 6, I assess it by conditions taken as axioms in social choice and, in chapter 7, I consider its rationality (or alleged lack thereof). The remainder of this chapter discusses lottery-voting in the light of some more intuitively democratic characteristics, showing that it is desirable because it counts all votes equally, renders outcomes uncertain and incentivizes all to participate in the process. But first, I give two practical examples of real-life, smallscale decision making that could be resolved by lottery-voting, to illustrate how it might work in practice and some advantages and disadvantages. Note that both of these examples are small-scale in two senses – they involve only small groups and only small decisions. The first of these is a matter of principle – I believe that democracy operates best in such groups and I am deliberately excluding representative institutions. The second of these is merely contingent. I believe that lottery-voting could be used by a small group making major decisions, e.g. whether to go to war. I do not, however, focus on such for two reasons: firstly, it is less obvious that we want such decisions to be made democratically and, secondly, lottery-voting is less likely to be accepted all-things-considered until we have seen how it operates in practice, so it would be more strategically sensible to follow a principle of caution and introduce it, at first, for relatively trivial decisions. Recall also that my claim is not that lottery-voting is uniquely democratic or even the best 141 democratic procedure. Rather, lottery-voting is a potential voting rule that stands at the same level as, for instance, the Single Transferable Vote rule346. Which of the various potential democratic procedures we want to use to make any particular decision will have to depend on the context – and, as such, some of the reasons given below to favour lottery-voting in these specific contexts are appropriately ad hoc considerations. (4.6) The Book Buying Example My first example of lottery-voting in practice is drawn from a debate in my Graduate Common Room (GCR)347, which nicely illustrates the arbitrariness or path dependence of any other method: It was suggested that the GCR should buy some reference materials – dictionary, thesaurus and atlas – for the general use of GCR members. This divided opinion in several ways. Some thought any such purchases would be undesirable, for example because they might change the Common Room from an informal social venue to a place where people worked. Also, however, it was pointed out that we were in close proximity to both the college and Bodleian libraries, and that the Common Room already had a computer with online access to OED, 346 Shepsle and Bonchek (1997) pp.167-72 offer six examples of voting rules: the simple plurality, plurality run-off, sequential run-off, Borda count, pairwise comparison (Condorcet) and approval voting. Lottery-voting may be on a different level from these, as they are arguably all – with the possible exceptions of Borda and approval voting – attempted implementations of majority-rule, where majority-rule is assumed to treat all equally. If we question that assumption, however, then lotteryvoting can be seen as just another attempt to realize the common aim of political equality. 347 Lengthy debate was involved in the first meeting of Trinity Term 2004 before a rather vague motion was withdrawn: https://gcr.jesus.ox.ac.uk/meetings/meeting_minutes/minutes_05_05_2004.html (last accessed 19/11/06). A refined motion was subsequently passed in the second meeting of term: (last accessed https://gcr.jesus.ox.ac.uk/meetings/meeting_minutes/minutes_01_06_2004.html 19/11/06). This discussion also draws on my communication with the GCR president, Timothy Burns, particularly an email of 28th May 2004. Oxford colleges have proved a fertile ground for study and advancement of social choice, since Dodgson in Christ Church – see McLean and Urken (1995) pp.52-5. Mueller (2003) p.589 also points out the difficulty of a hypothetical choice involving number and kind of books. Note that on 05/03/08 the GCR was renamed MCR (Middle Common Room), which will probably soon affect these links. 142 Wikipedia, etc. As such, if the books were merely to be used for crosswords or general information – rather than academic purposes – some also felt it was not worth buying ‘top of the range’ reference materials348 (a worry backed up by the lack of security). Thus we could see a number of different positions: some thought that no books should be bought, while those that wanted to buy them were divided between those that wanted to spend a lot for top of the range materials (e.g. OED) and those that thought cheaper materials would suffice. This does not immediately look threatening to the prospects of collective decision, but there were two further complications. Firstly, there are not simply three options. Expenditure is a continuous variable, so it is not as if the choice were nothing, £50 or £150. It would be possible, for example, for someone to propose we buy a decent thesaurus and atlas, but only a cheap dictionary because we had access to the OED online anyway. Thus a variety of different figures were possible, as well as different views over how we should spend any given intermediate amount. Secondly, preferences were not necessarily singlepeaked349. It would have been easier if everyone had had some ideal figure in mind and preferred any number closer to that ideal to one further away – so someone who wanted us to spend £50 would prefer £20 to nothing and £120 to £150. However, this did not hold. While I cannot reconstruct anyone’s actual preferences, it would have been entirely possible for someone to hold, for example, that ideally we should not buy any books but if we were going to then we had to go for top of the range ones, assuming cheap ones are a waste of money and literally ‘worse than nothing’. 348 A variety of cheaper reference materials are available and naturally these vary in quality. I simply assume a correlation between price and quality in what follows, so it is supposed that the OED is both more expensive and better than any other dictionary, but there are a variety of cheaper and worse alternatives. 349 See Black (1998 [1958]) pp.14-45. 143 The usual procedure for deciding on motions in the GCR is simple majority rule, with the option to amend motions prior to the final vote350. Since motions are put forward and those present merely vote for or against, the likely consequence of any particular proposal is rejection by a coalition who favour other options. It is quite possible nothing will be done, by default, but repeated motions or amendments could finally lead to a compromise being passed. This nicely illustrates the difficulty such traditional decision mechanisms have dealing with multi-dimensional issues where preferences are not single-peaked. Trying to come up with an alternative procedure reveals just how general this problem is. If we begin by voting on an expenditure figure, perhaps between several set options (e.g. £50, £100, £150) and then vote on whether or not to go ahead (i.e. whether to spend this agreed sum or not) then there are mixed incentives – those who would rather not buy the books at all have reason to vote for the amount of money they believe most likely to fail at the final hurdle. Conversely, if we begin by deciding whether or not to buy the books, then those who vote yes are effectively being asked to write a blank cheque – they do not know (yet) whether they are voting for buying top-quality or cheap books, though this could reasonably affect whether or not they want to make the purchase at all, and in light of this uncertainty it might be reasonable of them to say no to buying anything. In either case, the outcome will depend not just on the preferences of the individuals, but how the decision-procedure treats them – just as a Condorcet cycle is often resolved (unnoticed) by the order in which votes are taken351. The worst case scenario is that an ingenious and The GCR Constitution can be found here: https://gcr.jesus.ox.ac.uk/constitution/GCR_Constitution.pdf (last accessed 19/11/06), with procedures for meetings in section VI.C. 351 Arrow (1997) p.5, Riker (1982) e.g. pp.91, 140-2, and 152-6, Wolff (1976) p.62. Note that the empirical likelihood of cycles has been questioned, e.g. by Mackie (2003) pp.88-90, and Regenwetter, et al (2006) pp.32-6. 350 144 unscrupulous agenda-setter may be able to manipulate the process to favour his preferred option. The best is that no great thought goes into the procedure, so the result is essentially random – but even this is precisely the kind of “arbitrary dependence of the final social choice on the choice path” that Arrow wished to avoid352. Maybe the difficulty with these options is that they treat the case as two separate problems: whether to buy books at all and how much to spend. While this might be a natural way of breaking the issue down into two component decisions, it will inevitably require us to decide which should be taken first, since this may determine the final outcome353. The alternative is to treat not buying the books as simply another option, on a par with (rather than qualitatively distinct from) the others. For example, instead of the choice between £50, £100 and £150, to be followed or preceded by a simple yes/no vote, we could present the case as a single decision between four options, one being £0. This case still faces the difficulty of setting the available options, for example, why those given above and not simply £0, £75 and £150 – or, for that matter, £0, £149 and £150354? And, assuming we settle on any options other than just £0 or £150 – which is effectively to disregard the ‘how much?’ question – we need some procedure that is capable of choosing between three or more options. In these situations, there are numerous possibilities, such as Single Transferable Vote, 352 The quotation is from Chapman (1998) p.297, c.f. his pp.293 and 303, and Arrow (1963 [1951]) e.g. pp.28-30. 353 If we were to take the two decisions separately, I think it would be quite reasonable to toss a coin to decide which should be taken first – but once it is realised that this may be what effectively decides between two different possible outcomes, we have no reason to reject more randomness. Miller (1992) suggests that, in some cases, deliberation may uncover what is at stake for each side, and suggest compromise or a reasonable way of resolving the decision. I accept this may be true of some cases, but I do not see an obvious solution in this instance – as suggested in the previous paragraph, one might want a decision whether to buy books at all before how much to spend or vice versa. Some difficulties are discussed by Chapman (1998) pp.304-12. 354 This last looks obviously unreasonable, but on what principled basis can we exclude it? And on what basis can we insist that £0 has to be an option but no other figure is similarly privileged? Or must we then consider 151 different alternatives? 145 Plurality Run-Off or Plurality Rule, but since these may produce different outcomes from the same underlying preferences355 our result still depends on the procedure. We could go for some more radical proposal, for example, having each voter write their ideal expenditure figure (which may be £0 and up to £150) on their ballot paper, and simply taking an average. This would produce a compromise outcome, in line with what Kamm called ‘results proportional to numbers’356. It does have some plausibility, because it gives each person an equal weight. For instance, if two people are split such that one favours £0 and the other £150, then it gives a result equidistant between them (£75). If there is a third person – favouring £150 – then the average is now £100: the two have more weight, so pull the outcome closer their way, without wholly determining it357. This averaging seems problematic, however, in that it can make majorities hostage to a few extremists. For instance, suppose there were thirty people, twenty-eight of whom wanted nothing spent and just two (the proposer and seconder) wanted to spend £150. In this case, the average is £10. The case is even worse when one realises the incentives to vote strategically358. Suppose twenty people favoured £50 and ten favoured £100. In this case, the sincere average is £66.67. If the ten, however, were all to act strategically and declare their preferences as £150, then the average of the preferences given would be (20x£50+10x£150=) £83.33 – closer to their actual ideal. Further, while one case for such averaging might be that it would minimize aggregate dissatisfaction, if measured by how far each voter’s ideal is from the final outcome, it is fairly clear that this fails if preferences are not single peaked. Return to our earlier example, where two voters are divided between the extremes (£0 and £150). The average, £75, may satisfy no one – indeed, it is quite possible both 355 356 Shepsle and Bonchek (1997) pp.167-72. Kamm (1985) p.191. 357 Compare Jones (1983) p.175. 358 I owe this point to Toby Ord. One of the advantages of lottery-voting, discussed in the chapter 6.13, is that it encourages true preference reporting. 146 would actually prefer the other to get their way than settle for this compromise (e.g. if both thought cheap reference materials so near to useless as to be just a waste of money). It is possible, therefore, that both would prefer to toss a coin between their two extremes, and risk the other getting their way, than settle for the middle position that pleases neither of them. We can see already how close we are to lottery-voting. If the way to adjudicate fairly between two people of opposed preferences is to toss a coin, then what about the case of two against one? One option is the weighted compromise suggested in the previous paragraph, but again this might satisfy no one. Another alternative is still to toss a coin, regardless of the different numbers, but, as I argued in chapter 3.10, while this may be fair (in that it gives each an equal chance of satisfaction), it is not plausibly democratic because it gives the third person’s preferences no weight whatsoever, and therefore makes votes irrelevant. Lottery-voting achieves a compromise by assigning chances proportionally to numbers; giving each individual an equal chance to affect the outcome. Of course, this may produce a different result from any other decision procedure, but that cannot be a reason for choosing between any two procedures. If we simply have every individual write down their ideal – an amount and how it will be allocated – and then randomly select one of those votes to become policy, then this gives each an equal chance to determine what it is to be. There is an obvious sense in which the outcome is arbitrary, but at least it is only an arbitrarily chosen preference: it will be someone’s preference and although anyone may be required to live with another’s chosen policy, which they dislike, no one is required to compromise their own view. For any given set of preferences, there may be an alternative procedure that performs better. For instance, if voters are evenly split between £50 and £100, but all 147 would prefer a £75 compromise to tossing a coin between the two options – because they find the middle figure greatly superior to getting the other option – then it would indeed be reasonable to split the difference359. As observed above, however, it may equally be that all would prefer to toss a coin rather than accept what pleases no one. The best democratic procedure to adopt for a given decision always depends on the context, including the type of decision being made and structure of people’s preferences. While it may be possible, in specific cases, to do better than lotteryvoting, its fairness is not impinged by this fact (i.e. it would still be just as fair and democratic – though suboptimal – to toss a coin, even if everyone would rather compromise) and no procedure can guarantee to do better in the abstract. Thus, lottery-voting seems a perfectly reasonable possibility for resolving such an issue. (4.7) The Reading Group Example The second example is more informal, but also shows how lottery-voting might be used by a small group seeking to make decisions. A political theory graduate students’ reading group meets each week in a different pub around Oxford360. Obviously the choice of pub depends on various factors – somewhere quiet, in terms of clientele and music, is appreciated, with the range and price of drinks also an important consideration – which not all agree on. Perhaps the most important, or at least contentious, issue though is location. With few suitable central venues, the group often meets in places like The Old School (west), The Old Tom or Head Of The River (south), Royal Oak (north), or Angel And Greyhound or The Half Moon 359 Note that lottery-voting is compatible with two groups agreeing to such a compromise, and voting for a middle position, or with their agreeing to support each other on different issues (log-rolling), though either of these solutions may require public voting. 360 Again, this is loosely based on personal experience, though at present the group has no formal decision procedure. 148 (east), yet obviously moving too far in any direction imposes costs on those living further away. The group has no formal decision mechanism – usually the decision where to meet next is reached at the end of the night. Typically, one person suggests somewhere and that is either settled on or rejected. If there are serious objections (e.g. ‘that’s too noisy/crowded/far/etc’) then there is often a counter-proposal and some informal discussion, often leading to consensus or compromise. This system works reasonably well; although it does sometimes involve relatively high decision costs everyone gets some influence. Suppose, however, it was agreed to adopt a more formal procedure. One option might be for everyone to take turns choosing a pub, so that everyone got his or her choice every six weeks or so, for example. The problem with such is that group membership is not perfectly stable – while some people come (nearly) every week, there are others who participate as the timing, location, reading or inclination suits them. In this case, it is not clear whether those not present on a given night should take a turn, or whether the choice should go to the person present whose turn is next. In the latter case, attendance will likely be affected by whose turn is next – i.e. people will be more likely to come if they think they are next in line to choose the venue – but it will likely become very hard to keep track of whose turn it is, if everyone who has or might come needs a place in the rota, but many of those are skipped through absence. Instead of rotating around people, another option is to rotate around venues – to regularly move in a north, east, south, west pattern, for example. This would mean that everyone should be equally satisfied – for example, those living in the east will find the venue near them a quarter of the time. However, this does not solve all the problems – one still has to define the relevant areas (Why settle for four compass 149 points? Why not treat St Clements, Cowley Road and Iffley Road as separate ‘options’, for example?) and choose a pub within them, which could be more or less central (e.g. The Angel And Greyhound or Port Mahon). Another potential problem is that such a procedure seems undemocratic. While not all decisions necessarily have to be made democratically, it would seem pointless for the group spend a quarter of their time in north Oxford if no one lives in north Oxford. Moreover, if half of the regular attendees live in east Oxford, it may well be fairer (based on a principle of proportionality) to meet there more often – otherwise the interests of those extra people are effectively ignored. These difficulties mirror those of giving equal chances for each group – discussed in chapter 3.10, above – and, thus, the same solution suggests itself. Instead of rotating round individuals present, if we want each to have an equal say, we can give each person present an equal chance of determining the venue for the next meeting by implementing a lottery (‘random dictator’). This will produce a form of proportionality: for instance, if half the group live down Cowley Road (east Oxford), and therefore favour meetings in that direction, then the group will tend to meet there around half the time. There is, however, a serious danger of extremism. If the decision simply represents the will of one person, there is no need or incentive for them to moderate their demand – to give an example, why stop at The Angel And Greyhound if Port Mahon is closer to home? There is a danger that one person getting all of what they want can impose greater costs on others. Moreover, the group may become in a sense self-selecting – if the group meets far enough out, then it is likely that only those from east Oxford will attend, and in that case they will be the ones who decide where to meet next and it will be east Oxford again – the group has been ‘captured’. 150 This illustrates a danger with lottery-voting. While one problem with compromise is that it may satisfy no one (see 3.13 above), allowing a ‘random dictator’ to have their way may result in a decision that almost everyone else finds greatly unsatisfactory. I will return to some ways of dealing with this, including constitutions and thresholds, in chapter 5, where I discuss the practical logistics of lottery-voting. For now, one thing to say is that such a proposal may require some restraint on behalf of the voting demos. That one could have all one’s own way does not mean it is reasonable to exclude others totally. In particular, since no one is guaranteed victory, each may be more inclined to consider others – that is, those in east Oxford may not propose anywhere too far out because they know that then those from other directions will not only be able to say ‘how would you like it?’ but actually be able to retaliate by doing likewise when they get chance to decide. Obviously, this self-restraint may only be effective in solidaristic groups, but that is not unrealistic when I am talking about small groups who meet face-to-face. Perhaps the successfulness of democracy depends on each side having some sympathy with their opponents, and thus limiting what they demand when they win, to ensure they can bear the costs when they lose361. If each side is willing to press its demands to the fullest, then it will be no surprise if the group breaks up – the result, in the case I am describing, might be two separate groups, meeting in north and east Oxford – the small-scale parallel to secession. This would be unlikely in this case, not only because each individual limits their demands out of a sense of reasonableness, but also because the success of the group depends on participation. If each person wants both to visit a variety of venues and enjoy discussion with many other people, then they 361 Sartori (1987) pp.31-3 and Vernon (2001) p.72-91 both emphasize the importance of looking at losers rather than winners. 151 are forced to compromise their own preferences as to the location to realize these other goods. Note that, while I do appeal to compromise to produce better outcomes, it is left to each individual to make the compromise. There is no mechanical way in which compromise can be struck between two different people (for example, if one proposes Gardener’s Arms and another Port Mahon, we cannot simply look for somewhere equidistant between the two – even if there is a ‘middle ground’ it may be a pub that, for other reasons, pleases no one). What we can do is rely on each individual to make compromises themselves, and demand only what is reasonable – for instance, moderating their respective demands to the more central Royal Oak and Angel And Greyhound. Thus, rather than each demanding the ‘whole cake’, we rely on each to propose something they think a fair division. Because each compromise is made by a single individual, we can assume it to be consistent and to satisfy at least that one person – it is what they want, given the demands to consider what others want. We recognize, however, that they are unlikely to agree unanimously on some ideal that will please everyone. While each is willing to move to a more central venue, out of consideration for others, there is still a divide between those who favour somewhere north-central and those who favour somewhere east-central362. In this situation, the ‘random dictator’ method would be one way for the group to decide where to meet on an equal basis, i.e. democratically. Of course, this does not mean it is necessarily the best decision-method all-things-considered: the above discussion pointed to some dangers of such a method, which show that democracy is not the only value when it comes to group decision-making. If we want to encourage 362 I think the ideal of some deliberative democrats – that unanimous consensus be reached – is unrealistic because, even if all are sincerely motivated to search for such there may be multiple possibilities that are equal or incomparable with respect to justice and reasonability. Moreover, aside from the fundamental point that there may not be ‘one right answer out there’, people’s conception of what is just may be affected by cognitive biases about which they can do nothing. 152 independently just or ‘middle ground’ positions, that is another matter, which may not be best achieved through a democratic machinery, or at least will influence our choice between democratic methods. Nonetheless, lottery-voting respects the equality of all in that each person will have an equal chance, on each occasion, to be the one who decides where to go next time. Again, this does not give each person an equal chance of satisfaction – if there are more people from east Oxford, then there will be more chance of them all being satisfied – but this is for the reasons given earlier (see chapter 3.10 and 4.5) to favour proportional rather than equal chances. Moreover, this practice will ensure no one is excluded – it is not like majority rule (which might lead the group to always be in east Oxford), or turn-taking (which distorts incentives to participate) – even if the decision has not gone one’s way, there is always an incentive for each person to turn up, because they know they might be the decisive person next time. Having offered these two examples, I turn now to generalizing the merits of lottery-voting. (4.8) Vote Counting I said that lottery-voting treats each person equally, but one might object to the fact that lottery-voting does not in fact count votes at all. Whereas in traditional majority-rule procedures, outcomes are determined by numbers of votes cast, so counting is all there is to determining the winner, lottery-voting is simply chancebased. Since we never need to know the chances of any option, we can hold the lottery without ever counting votes. If votes are not even counted at all, one may think this necessarily violates the requirement that all votes be counted equally. This objection, however, rests on a confusion between a vote’s being counted and its 153 simply counting. The first sense given to the verb count in the Oxford English Dictionary is: “1. a. To tell over one by one, to assign to (individual objects in a collection) the numerals one, two, three, etc. so as to ascertain their number; to number, enumerate; to reckon, reckon up, calculate; also, merely to repeat the numerals in order up to a specified number, as to count ten” This is a familiar sense in which votes, and many other things, can be counted – or, as sense 10.b puts it, ‘reckoned numerically’. It is how we might know how many people voted altogether, or for each option. Formerly, the word ‘tell’ was used, as in Bentham’s ‘each to tell for one’363 – seemingly reconciling the requirements of both democracy and utilitarianism. However, simple enumeration is not all there is to counting. Other senses given by the OED include: “2. a. To include in the reckoning; to reckon in” “13. To be reckoned or accounted” “15 count out c. To leave out of count or consideration; to reckon as not to be counted or depended upon; to exclude. colloq. (orig. U.S.)” In these senses, it is not so clear whether all votes do really count under majority-rule. It is clearly not enough that votes simply are counted. After all, a dictator might hold mock elections in which voters are given total freedom to vote for what they want, and all votes are duly counted and honest totals recorded, but then totally ignore the result and impose his wishes regardless. In this case, votes have been counted, but they do not count for anything. That votes are enumerated is of no intrinsic importance; what matters is that they can count for something – that they can affect a result. Similar cases can arise under majority rule, however. Suppose 101 voters vote on a given issue, between options X and Y. Now imagine we have counted 80 of them, and so far 53 are in favour of X, while 27 support Y. There are still 21 votes 363 Bentham (1843 [1827]) p.334. Mis-quoted by Mill (1998 [1861]a) p.199 as “everybody to count for one, nobody for more than one”. 154 left to be counted, but even if they are all for option Y it will be defeated 53 to 48. Whether or not we actually count these votes is immaterial, since they no longer count for anything. I am not claiming that this fact in itself amounts to a decisive objection against majority-rule. After all, retrospective equality is almost impossible to preserve in cases of disagreement – at the end of the day, some people will get their way and others will not. That these particular votes will not count for anything seems to depend only on how others have already voted and the order in which the votes have been counted. If we assume that the latter is effectively random, then we may say that – since each vote had an equal chance of being the first one counted – each had an equal chance to count. Further, since each remaining vote then has an equal chance of being the second counted, the third counted, and so on – all the way up until at least the 51st vote – one could argue more votes will be counted this way. Under lotteryvoting, it could be objected, only the first vote counts at all, and all other votes do not count for anything. The point at present is simply that votes do not have to be actually counted in order to count, at least in a hypothetical sense. The next step is to ask at which stage votes should count in this manner. The problem arises, I will argue, when people know in advance that they will lose. In this case, their votes may as well have been excluded a priori. (4.9) When do Losers’ Votes Count? Under lottery-voting, all votes count in the sense that they increase the chances of the option they are cast for winning. Once one has been randomly selected, however, there is a very real sense in which only this one counts for anything. The rest are as immaterial as those cast in any other election, once the winner has already 155 been determined. The crucial issue then, is at what stage votes must count – i.e. make some kind of difference – in order to be meaningful. If their equality of impact must be preserved at all points, then it seems that equality can only be retained – if at all – by some form of proportionate compromise over final policy364. Yet if making a difference need only be hypothetical, then it is enough that one’s vote could have counted, had others voted differently365. There is no objection to the fact that, once a majority has been established and a winner therefore determined, the votes of those who happen to be in the minority no longer actually count for anything. It seems likely that equality will have to be given up at some point. After all, except in the extremely unlikely situation of voters being unanimous or evenly divided, we cannot maintain the equal impact and satisfaction of each. This is no doubt why the requirement of equality is often spoken of in terms of each voter’s chance of influencing the outcome. The question is at what stage a voter must have this equal chance. It is no more reasonable, in general, for a member of an out-voted minority to complain that they no longer have an equal influence than for one who as it happened was not randomly selected by the lottery to complain his vote no longer has an equal chance. There mere fact that anyone could have been in a majority from a hypothetical original position is not, however, enough, any more than the fact that Steven Hawking could have been able to beat Mike Tyson in an arm wrestling contest. Given the ‘strains of commitment’, we need a procedure that actually situated people can still accept as fair. Where there is a permanent minority, this means we need a procedure that still gives them something – either partly their way (compromise) or a chance of victory. 364 365 Kamm (1985) p.191’s ‘results proportional to numbers’. Note that this parallels the counting issue from chapter 3.7. 156 Suppose one lives in a deeply divided society, say one with a 70% red majority and a 30% blue minority. Each of these groups has relatively homogeneous preferences over a wide range of political issues, but there are great differences in culture and preferences between the groups. Thus we can imagine that a whole series of decisions, each put to majority vote, have all gone the way of the reds. They may have their way about policies concerning fiscal matters, industry and commerce, religion, official languages, schooling, infrastructure and more. They may even encroach on the rights of the blues, but we need not assume this – let us suppose that all constitutional protections are in place, and in particular that the blues’ formal rights to participation – understood as an equal vote – are in place. Now suppose some new decision is to be made, concerning, perhaps, healthcare spending. Let us assume there are again predictable divisions between the two groups366. Take a moment to put yourself in the position of a blue voter considering going to the polls. You know that, if you vote, your vote will not arbitrarily be declared void, and will indeed be counted according to proper procedures. You even know that, if there are more blue votes than red votes, your preferences will indeed hold sway and be enacted as policy. However, all of this formal equality is tempered by the knowledge that this is very unlikely – blues are out-numbered by more than two to one in the society. It takes just 43% of the reds to turnout to ensure their victory, even if you and every other blue person were to vote. Further, the consequent unlikelihood of winning tends to discourage many of your fellow blues from even voting. Given that defeat seems a foregone conclusion, many have concluded their votes count for nothing, and it is not worth casting them. This merely results in a vicious circle, because the lower 366 These could arise because, for example, genetic predispositions to various diseases differ between the two groups. They could perhaps also arise from some cultural phenomenon, e.g. a religion that forbids certain medical practices in one group. 157 turnout amongst blues further reduces their chances of electoral success, consequently discouraging other would-be blue voters367. In such circumstances, formal a priori equality is not enough. While the example may be exaggerated, there are examples of deep and permanent majority/minority divisions, be they based on tribe, culture, language, or even sex, geography (such as urban/rural or north/south) or party368. An ideological split, such as between conservatives and socialists or economizers and environmentalists, may appear less serious if we assume that these groups are open to rational persuasion and some may shift their positions. However, it is not at all clear that this is the case and, where there is a permanent split, it may even be more problematic, because the only thing that differentiates these groups is their political preference, whereas we might imagine other groups will not actually have homogeneous political preferences. It is hard to draw a sharp line in theory, but what I am concerned with is votes that seem a foregone conclusion, such that voters rightly feel, before even casting their votes, that they will not actually count for anything. This requires more than an abstract, hypothetical chance of affecting the outcome369, but conversely recognizes that at the end of the day some will get their way and others will not – it only demands that the winner not be clear in advance. Of course, this does not mandate lottery-voting – it may be that the majority is unpredictable or be possible to achieve proportionality by compromise370 – but the case of permanent division does tell against the equality of majority rule. 367 368 This process is described by Guinier (1994) pp.2-3, and 76-7. ‘Red’ and ‘blue’, for example, could stand as proxies for a racial white/black division, or for political parties, e.g. Labour and Conservative. 369 This seems no more ‘real’ than the chance Steven Hawking had of winning his arm-wrestling match against Mike Tyson in chapter 3.3 – he could have been born someone stronger, and one’s view could theoretically have been in the majority, but that is no consolation when one knows how one is situated. 370 C.f. Kamm (1985) p.191, discussed in chapter 3.13, and Hyland (1995) pp.93-100. 158 (4.10) Contingent Outcomes I have assumed it is desirable that outcomes are not known in advance, but we live in a society where media polls try to call every election in advance and – particularly in small groups of the sort I am focusing on – public debate may give each voter a fairly reliable indication of the way the vote is likely to go. Is the idea that outcomes are unknown in advance – contingent and inherently ‘open’ – either empirically realistic or normatively desirable? Przeworski describes democratic systems as “organized uncertainty”371. This is not an appeal to the total chaos predicted by McKelvey and Schofield372. Przeworski does not say anything could in fact happen, outcomes are not actually decided randomly373, rather he stresses that outcomes appear uncertain374. As he puts it, “there is no group whose preferences and resources can predict outcomes with near certainty”375. This is what is wrong with the above example of the reds and blues – since reds clearly outnumber blues, we can predict in advance that their preferences will hold sway. But what exactly is wrong with this predictability? Przeworski’s answer is that it is uncertainty that draws competing groups into participation376. This is not premised on the assumption that more participation is always intrinsically good, but the pragmatic grounds that a certain minimum level is necessary to a functioning democracy. The danger is that perennial losers, like our blues, will see no reason to continue to comply with a game that seems rigged against them. Thus, Przeworski argues, democratic institutions “must give all the relevant political forces a chance to 371 372 Przeworski (1991) p.13. Riker (1988 [1982]) p.186-8, Mackie (2003) pp.173-96, Mueller (2003) pp.123-6 and 237-8. 373 Przeworski (1991) p.42. This does not actually rule out lottery-voting. 374 Przeworski (1991) e.g. pp.12, 40, and 49. 375 Przeworski (1991) p.47; c.f. Vernon (2001) pp.39-40. 376 Przeworski (1991) p.13; c.f. McGann (2006) p.25, and Guinier (1994) e.g. pp.1 and 9. 159 win from time to time in the competition of interests and values”377. The possibility of winning next time makes it rational for losers to continue to play within the democratic system, in spite of certain and immediate losses378. Admittedly, a group need not win at all if they think democracy brings gains for everyone – e.g. economic development – or that even continued defeat in such a system is better than the uncertainty of revolution, in which defeat could be more total379. Nonetheless a group who face consistent defeat, but are kept loyal only by fear that any radical change could be even worse, seem in no better a position than those forced to accept an absolute Hobbesian sovereign simply because it is a less frightening prospect than the war of all against all. Such compliance may be prudentially rational, but it hardly vindicates the system as fair. The same point can be made in terms of Rawls’ ‘strains of commitment’. Suppose, as we suggested in chapter one, that members of society contract to find a decision-making mechanism agreeable – and fair – to all. If someone signed up, behind the veil of ignorance, to a procedure that it turned out gave them no chance of victory, it would be very hard to live with. As Rawls says of the parties in the original position “They cannot enter into agreements that may have consequences they cannot accept… when we enter an agreement we must be able to honor it even should the worst possibilities prove to be the case”380. Now, it is not clear whether this requires, or is even compatible with, lottery-voting – after all, the worst possible outcome of such is that some fringe lunatic may get to decide policy. But, on the other hand, it is quite possible that this excludes majority-rule. After all, the parties do not know particular facts about their society. It could be that it is as deeply divided as our 377 378 Przeworski (1991) p.33. Przeworski (1991) pp.19, 24, and 29. 379 Przeworski (1991) pp.31-3. 380 Rawls (1999 [1971]) p.153. 160 earlier red and blue example and, in this case, it is not clear that one who took seriously the prospect that he might be a blue and have no chance of victory under majority-rule could accept such a procedure for all public decisions either. This is why the final decision on matters like the voting rule can only be made with some knowledge of context, including whether there are such divisions in society – that is, in what Rawls calls the ‘constitutional convention’381. It is possible that, if contractors find themselves in a divided society, they will prefer lottery-voting to majority-rule because it ensures that no group are guaranteed to lose. Thus it seems there are respectable reasons for wanting democratic decisions not to be foregone conclusions – that is, for wanting the outcome to seem undecided as voters go to the polls382. We can further support this ideal if we appeal to the intuitive ideal of electoral competition. A one-sided competition is hardly a competition. If we know in advance who is certain to win an election, there is little reason for either candidates or voters to care. We see similar reactions in sports contests – if the winner is already decided (e.g. in the final few matches of a league season), then the excitement and interest drops markedly. So it is with votes where the winner is already known, or confidently predicted. The ritual of voting seems an empty formality of little importance. Apathy is hardly surprising. Lottery-voting, however, ensures that the outcome is always contingent until the last possible moment. Even if one knows how everyone else has voted, one’s own vote is never worthless – it can always affect the chances of one’s desired option winning383. 381 382 Rawls (1999 [1971]) pp.172-4, and 311-2. There may be a problem announcing results in some districts before polls have closed in others. 383 Of course, in a large electorate, this chance would be small, but it may be significant in the small groups I am principally considering. 161 (4.11) Predictability I have suggested that the very unpredictability of lottery-voting is an advantage, by arguing that a foregone conclusion is a bad thing, but it remains to be seen at what stage predictability can be surrendered. After all, tossing a coin is often seen as fair because it is unpredictable, and gives each side an equal chance. Not all fair methods of adjudication need be so, however. Consider an impartial referee, for example, in a sporting contest. The referee’s decisions are not unpredictable384, in the sense that if a foul has taken place then all who saw it should be able to predict at least approximately how the referee will apply the rules. At what stage, then, is predictability a problem? The referee example is very different from the coin-tossing case, however. The coin is used to resolve a dispute between two equal claimants, whereas the referee’s decision is predictable precisely because the two parties involved are no longer equal – if one player has committed a foul, then the decision (free kick, penalty, etc) should go in favour of the victim. Further, what is predictable is only a particular decision and we cannot say, from the moment of kick-off, that either side is more likely to win a free kick or penalty. This is not so problematic. Recall that, in chapter 3.12, I accepted that majority-rule could be fair. If an issue arose in which there were no obvious prior divisions, but after debate almost all tended to one side, then it is reasonable that they should win. This is rather like the impartial referee recognizing reason to penalize one side. What is problematic is if it is predictable before deliberation has even begun – this is more like a bribed referee who has given every decision the way of one team. 384 I owe this example to Bas van der Vossen. 162 There are in fact two things going on here, although they are related. The first issue is whether the predictability we are worried about concerns a particular decision or the general series, and the second is whether it is predictable prior to or only after deliberation. The predictability of a single, given issue need not be a problem – because there would be some extreme proposals that almost everyone should want to reject, and other reasonable proposals that all might accept (note that because lotteryvoting respects unanimity, these outcomes will be predictable). It need be no serious objection that people with an unpopular view on a given issue know they are likely to lose, as that is simply the consequence of counting each person’s preference equally. What is problematic is when a given group know that they have no chance of winning anything ever. This is when they are likely to be alienated from the democratic process itself – when they see no point in participating, and may well resort to undemocratic means to get their point across. This was the case in the previous section, where blues have no incentive to even bother, since defeat is a foregone conclusion. A procedure that gives each group a chance, however, motivates all to participate. What is important is not the individual decision, but the process as a whole; however introducing chance in each decision makes the whole unpredictable – thus all groups will be aware that, even if their chance of victory on any given issue is small, they may win next – or this – time. (4.12) Randomness in Democracy Przeworski goes as far as saying that “Democracy appears to be a system in which everyone does what he or she expects is for the best and then dice are thrown 163 to see what the outcomes are”385, but – even if this is not an exaggeration – it is only describing appearances. He recognizes, of course, that “democracies – at least modern ones – have no institutions that function as randomizing devices”386. This is certainly true – even if the results are essentially random, for instance because of a majority cycle, this is unintended – they are not explicitly and deliberately randomized in the way that lottery-voting is. The immediate issue is not, however, whether or not our current practice is essentially random. What may be at question is whether or not a democratic decision could be random – how is this really rule of the people, as opposed to ‘lottocracy’ or ‘klerotocracy’387? Again, however, this objection is based on a misconception. The lottery does not decide what we are to do, independently of the people’s will(s) – as if we were to write all the options on a wheel of fortune, and simply spin it to decide what to do. This is the kind of lottery Vernon objects to, as giving people no influence over decisions388. Perhaps there may be occasions where such is not a bad decision mechanism, but it is not democratic for the very reason that it pays no regard to preferences389. Moreover, such a lottery would require that all options could be identified in advance and then randomized over. A slightly more responsive lottery could take votes to define eligible options, and then randomize over only those alternatives that attracted at least one vote. This would be slightly less perverse, but would still mean an option with a single vote gets an equal chance to one with all bar 385 386 Przeworski (1991) p.12. Przeworski (1991) p.42. 387 The latter term is derived from the kleroterian, the random selection machine employed in ancient Greece. I owe this to Conall Boyle. See http://www.conallboyle.com/Aboutme.html (last accessed 05/11/06) 388 Vernon (2001) p.46. 389 See the remarks on equal chance devices, chapter 3.10 and 4.4, above. 164 that one of the votes390. This hardly treats votes equally, even if it may be an interpretation of fairness to each voter (giving them an equal chance of satisfaction). The lottery employed in lottery-voting is not a straightforward lottery over options, but a lottery over votes. It is the means by which the winner is determined, from all the votes people cast, thus it is a way of treating all these inputs equally, rather than ignoring them. It is not the lottery that rules or decides – it is merely the device for choosing between the opposing votes. Thus it is no more accurate to say that this ignores the will of the people in favour of rule by lottery than to say majority rule ignores the will of the people in favour of rule by numbers or by counting. It is not simply counting that determines the outcome in the latter case, but counting of votes. So it is in lottery voting: it is not a simple lottery (over options), but a lottery of votes. So a random lottery does not detract from the rule of the people, because it is simply a device by which their votes are to be treated equally, and not an independent source of decisions. But one might still think this outlandish enough that it hardly counts as democracy, if by democracy one understands rule of the majority. I have already tackled the assumption that democracy naturally implies such in my first chapter, however. Further, there is no intrinsic reason why random selection should not be part of a democratic mechanism – as described in above (4.2-4), sortition has had a long history in democratic practice and theory. If this is unconvincing though, then my simple response is ‘what’s in a name?’ Lottery-voting need not be given the epithet ‘democratic’ to share the value of being based on political equality391. If this is what is good about democracy (understood as majority-rule or in any other way you 390 This might also be open to the objection that ‘artificial’ options could be created to increase the effective chances of a numerous group, as described in chapter 3.10 and 4.4, above. 391 Indeed, I noted in chapter 1 that, insofar as political equality is not sufficient for democracy, lotteryvoting is not really about democracy as such, merely a way of making decisions. 165 like), then it is also good about lottery-voting, which also realizes equality between voters. (4.13) Conclusion Chapters 1 to 3 established that democracy is based on equality, and that equality may be better satisfied by a weighted lottery rather than majority rule. The present chapter has suggested how this might be implemented, through a procedure known as lottery-voting. I have given two examples (sections 4.6-7), to illustrate how it might work in practice. The next chapter further explores the practicality of lotteryvoting, in particular addressing the problem of victories by extreme minorities, and in the process how lottery-voting fits into the wider democratic context, including the place of constitutionalism and deliberation. After that, I turn to assessing lotteryvoting, beginning in chapter six by comparing it to formal axioms of social choice and concluding in chapter seven with a discussion of rationality. 166 5 Practicalities “[W]e cannot understand a political theory or use its principles to evaluate existing practices until we engage in the process of formulating its principles, translating them into practices and judging the practices against our convictions”392 “[P]olitical forms, of themselves, can accomplish nothing; their value depends upon the spirit which energises them”393 (5.1) Introduction It is important to remember that the primary purpose of this thesis is theoretical – to show that democracy does not conceptually require majority-rule and to offer a potential alternative. It is not the purpose of this thesis to argue that we should actually be adopting lottery-voting in general practice. Though there may be cases where lottery-voting recommends itself as an appropriate practical decisionprocedure, its all-things-considered desirability depends on various background conditions, the extent to which we value democracy above other goods and on the transition costs that would be incurred in switching from familiar democratic systems to such a proposal. Given this limitation, it may seem strange to devote a whole chapter to the practicalities of lottery-voting. It is, however, important that democracy is a procedure for making practical decisions (see chapter 7.2a). Even though not necessarily recommending that we actually adopt lottery-voting in practice, it would be no good to propose a decision procedure that was unworkable. I need to show how, in appropriate conditions, lottery-voting could be practical. Chapter 4 described how lotteries could be used in a democratic procedure, offering two detailed examples where lottery-voting could plausibly be employed in practice. There remain a number of practical questions to address before we have a 392 393 Gutmann (1999) p.18 [not emphasized in original]. Laski (1933) p.149 [not emphasized in original]. 167 clear idea of exactly how it would operate, however. The present chapter therefore supplements the previous one by dealing with how lottery-voting will be implemented and tackling a number of practical difficulties. Perhaps the most important section, 5.3, deals with the issue of extremist minorities, but the chapter goes on to discuss the place of constitutionalism, compromise and deliberation in light of lottery-voting. Then discusses the openness and scrutiny of the process, before turning to the potential problems of ‘parcelling’ decisions and repeated motions. As stated above, however, I do not here consider the problems of transition, which are addressed briefly in the conclusion. The aim of this chapter is merely to show that lottery-voting could be practical and how it might be implemented. (5.2) Lottery-Voting as a Part of the Decision Process It must be remembered that lottery-voting is intended as an alternative to majority-rule, but majority-rule – or, indeed, voting in general – is not all there is to democracy. In practice, some way of resolving disagreement will be near-essential, as there will be few interesting cases of deliberation to unanimous consensus, but this always occurs within a wider context. That the primary concern here is with the ultimate decision stage, rather than for example the prior rules of agenda-setting and deliberation, does not mean that these other matters are unimportant. Lottery-voting is not in itself a complete theory of democracy. Indeed, presumably the primary concern of democracy should be the franchise rather than the decision rule, for either majority-rule or lottery-voting could be used by oligarchic groups seeking to make decisions as equals. This is not something explored further here, but the present chapter considers how lottery-voting could interact with wider democratic practices. 168 (5.3) Minority Motions The chief objection constantly raised to lottery-voting in discussion is that, because it gives all groups a proportionate chance of victory, it will sometimes allow minorities to win. If this is merely an affirmation of majority-rule and the assertion that it is never justified or democratic for minorities to get their way, then its fairness has already been argued for (see chapter 3). There is nothing democratic about a 60% bloc getting 100% of their way 100% of the time and the others being totally unsatisfied. Lottery-voting treats all equally by enshrining the principle of proportionality, interpreted as giving a chance to each group. One may attack proportional chances, believing compromise over individual decisions more reasonable – a matter addressed in chapter 3.13 and 5.4, below – but if one believes in equality, then one should not be willing to deny the minority any satisfaction simply because they are the minority. (Of course, as pointed out in chapter 3.11-12, it is usually implicitly assumed that the composition of the majority and minority is fluid, so everyone is sometimes satisfied – on those occasions in which they happen to end up in the majority. I am concerned with cases where this is not so). The objection cannot be simply that it is wrong for a minority to get their way, merely because they are a minority, for this begs the question and would need to be supported by an argument for majority-rule – the most well-known justifications for which were criticized in chapters 2 and 3. Those who wish to push the worry usually seek to make the problem more dramatic, however, by illustrating it with a minority holding what they take to be objectionable views, e.g. what if some extremists, such as the BNP, were to have a 2% chance and win? The danger of giving everyone a chance is that this may include not only reasonable opinions, supported by a majority or sizeable minority, but a few crazy crackpots whose favoured policies are either 169 unjust or greatly at odds with everyone else’s preferences. There are four main responses to this challenge. (a) Constitutional limits One traditional bar against unjust minorities has been to exclude certain important issues from the democratic agenda. Many present democratic systems, from large states to small groups and organizations, embed certain principles in a constitution to protect them from majorities. While some hold that such checks are themselves democratic, I hold that even when they are democratic in effect, they are still limits on the democratic process itself. Democracy is rule by the people. If judges override the wishes of the people, even to preserve democracy in the long-run (which is what makes their action democratic in effect), then that intervention is not itself an exercise of democracy. The judges are like the dictator who decides to democratize a previously autocratic regime. The decision creates democracy but is not itself a democratic decision, since it is made by the dictator rather than the people394. (b) Exclusion of small minorities Further, if we are worried that objectionable views may always be able to persuade a few people, then we can compromise our principle of proportionality by excluding very small groups. Many countries with proportional representation impose some minimum threshold, such as 5% of the votes, before a party is eligible for any representation. We could similarly say that an option with less than a certain threshold of votes will not get any chance. Although, in small groups, one vote may be enough to meet a threshold like 5%, we could add an absolute requirement – e.g. the need for at least two votes, so no one gets a chance of being decisive unless they can persuade at least one other person to share their view. Again, this compromises 394 Conversely, if the people vote to, for example, exclude certain groups from the vote, that should be considered a democratic decision, based on how it was made, even if undemocratic in effect. The people can democratically decide to abolish democracy, if they wish. 170 democracy, understood simply as political equality, for the sake of other values, concerning good outcomes. That may be sensible, all-things-considered, but cannot be defended on grounds of democracy itself. (c) ‘It’s still democratic’ Both constitutional safeguards and thresholds are undemocratic measures because they deviate from political equality (which lottery-voting is one way of institutionalizing) in order to reduce the risk of bad outcomes. If we are unwilling to compromise our democratic values, however, then we can simply accept bad outcomes. After all, democracy is not all good things, and if people vote for unjust outcomes then they will, at least sometimes, get unjust outcomes. A simple majoritarian system does not guarantee that the majority will not vote for unjust outcomes, so nor can lottery-voting. The problem remains, however, that minority groups may continue to hold unjust views even if they cannot win over a majority. The commitment to democracy does not mean that we are blind to other concerns, but they need not be considered when we are focused only on democracy itself. (d) Endogenous voting Finally, there is reason to believe that such worries about extremist views may be exaggerated. Far-right or -left-wing groups have been known to poll relatively strongly in majoritarian systems, but we cannot be sure that these votes are genuine first preferences. Mayer’s studies of the Front National in France suggest many of their votes are actually protest votes from disillusioned voters395, and when Le Pen did reach the second stage of the presidential election in 2002 he was defeated heavily396. Under majoritarian electoral systems significant numbers may be willing 395 396 Mayer and Perrineau (1992) pp.132-7. To give another recent example, de Quetteville (2006) quotes a Fatah supporter who votes Hamas in protest, saying “I voted Hamas so that my own Fatah Party would be shocked and change its ways… I thought Hamas would come second. But this is a game that went too far. Nobody thought Hamas 171 to cast protest votes (‘none of the above’), because they know that the extremists they vote for have no realistic chance of winning. Under a system of lottery-voting, they would have to take seriously the possibility that a vote for the extremists gives them a chance of winning, and then they might not vote that way. In this way, lottery-voting would encourage serious reflection and responsibility among voters. Of course, it may be that more people actually vote for extremists once they have realistic chances of winning, but in this case it reflects their genuine preferences which are denied by a majoritarian system – so we have to turn back to saying that this is democratic, though blocking it if need be by one of the above routes. Finally, while there may be legitimate worries about the chance that ‘undesirable’ minorities will get their way, there is the counter-possibility that enlightened minorities will also sometimes get their way. What is more, the consequence of these groups sometimes having their way will be that all get to see the implications of such policies in practice. Hopefully it will be seen which groups are wise and worth listening to, and which are undesirable397, and these lessons will be incorporated in future rounds of debate and voting – that is, even if people are not immediately put off voting for extremists, they may be so after they have seen such policies in action. would win – even them. I know lots of people who voted Hamas, who regret it now. If I could vote again, I would vote for Fatah”. Lottery-voting will end both such irresponsible protest-voting and the need for it. 397 Of course, it will always be open for these groups to complain that their policies were unsuccessful because only implemented in a few cases. They can argue that the success of their platform would depend on its consistent, wide-ranging implementation – but this is an empirical claim the truth of which we can leave voters to judge. 172 (5.4) Compromise and Collusion While people may have more reason to vote responsibly, that does not however mean that they will necessarily vote for ‘middle ground’ policies. I later claim it an advantage of lottery-voting that it gives each person incentive to vote for their true first preference, rather than strategically (see chapter 6.13), but this same fact seems to threaten the possibility of compromise. There is seemingly no reason for someone who is genuinely an extremist to support a middle ground compromise, when their vote would give them a chance of getting their ideal398. We can illustrate the problem by considering two people dividing a cake between them. We may think that the ideal result is to split the cake 50/50, but lottery-voting instead divides the chances – if each person votes that they get the whole cake, then there is a 50/50 chance of either of them getting it all. Since this still treats each equally, it is fair, but likely suboptimal, given the plausible story about diminishing marginal utility would have us favour splitting the cake equally399. The problem is that lottery-voting seems to prevent the realization of this better compromise because, even if parties agree they would prefer to split the cake 50/50, when it comes to the vote each has incentive to defect and vote for the whole thing. If one group know that, no matter how unreasonable their demands appear to others, they still have a chance of success, there appears little reason for them to moderate their demands. I do not see this as necessarily a great problem. While a middle ground compromise may best reflect some kind of average of what everyone wants, it may also be fudge that satisfies no one. Recall the book-buying example (from chapter 398 399 I owe this objection to Clare Chambers. Compare chapter 3.12, above, where I point out that if majority-rule is fair it is probably all-thingsconsidered better than lottery-voting because conducive to better outcomes. 173 4.6); it may be quite reasonable for people to prefer either extreme – e.g. high spending or none at all – to a middle ground position. Faced with a whole series of decisions, people may prefer the prospect of getting wholly their way on some to a guarantee of getting partly their way on all. While any particular outcome will necessarily deviate from retrospective equality when there is a difference of opinion and no possibility of proportionate compromise, each decision will satisfy prospective or forward-looking equality400. Not all goods are divisible, like cake, and it may be perfectly reasonable for people to prefer equal chances of a whole good to equal shares of a divided good. Nonetheless, the worry remains that it does seem better to divide other goods that are like cake. The objection is that lottery-voting will prevent such compromise where it is possible and desirable. Although nothing in lottery-voting itself prevents mutual agreement or people voting for such compromises, it gives all involved an incentive to defect from these agreements. For instance, suppose you and I agree it would be better to split the cake 50/50. When I enter the polling booth, my dominant strategy is to vote for me to have all of it – which will give me a 50% chance of getting all of it, regardless of how you vote (if you stick to the agreement, I will still have a 50% chance of getting my half share; while if you also defect I am no worse off). Lottery-voting does, however, require all to face up to the real possibility of losing. Under a system of majority-rule, it is all too easy for majorities to ride roughshod over the interests of minorities, either deliberately or simply through oversight (if the minority are not publicly noticeable). I have argued that a proper democracy is one where outcomes are unpredictable so, whether they employ lottery-voting or majority-rule, everyone knows that they might lose. Because everyone knows they 400 Lively (2007) p.24. Note I do not rely on the claim that the overall series of outcomes will likely be divided proportionately. 174 might lose, they have reason to care how losers fare401, and this might give them reason to reject any collectively-binding decision, where it is possible instead to ‘live and let live’. This liberal option may seem similarly unstable to the dividing the cake solution, but there is less incentive to deviate if we assume people are more concerned to ensure they can live as they like than to ensure that they can impose their way of life on others. There are a number of ways that we can make such compromises more stable, which I will be returning to later: for instance, we can constitutionally guarantee certain liberties, make voting public so defection is obvious or rely on solidaristic bonds and a sense of fairness to keep people honest. Moreover, once collusion becomes possible, lottery-voting should be compatible with processes like log-rolling or other group bargaining and can be defended as giving each the appropriate baseline position to negotiate from – since each can fall back on their own proportionate chance, they need only make agreements where it seems beneficial. While more may be said about such informal cooperation and bargaining, and its likely form against the backdrop of lottery-voting, it is likely to be conducive to better outcomes than lottery-voting alone. (5.5) Liberalism and Constitutionalism As stated above, in sections 5.3 and 5.4, one way of limiting the damage that can be done by extremist minorities (and majorities) and ensure compromises are stable is to place constitutional limits on the democratic process. If we think free speech, for example, is a matter of justice, and we are unwilling to see this right put at risk, we may exclude either the passing whims of the people or randomness of the 401 On the importance of how defeated minorities fare, see Sartori (1987) pp.31-3 and Vernon (2001) pp.72-95 on uncompensable defeats. 175 lottery by enshrining it within a written constitution. In other words, not everything need be up for grabs in a democracy, and certain rights can be guaranteed and nonnegotiable at the voting stage. Some argue that constitutional protections can be part of democracy, broadly understood, for they preserve the equality necessary to the system402. Certainly it is true that certain rights, such as that to vote or participate more generally in government, are essential to an on-going democratic process, but these rights can still be protected by non-democratic means, even within what is generally a demcoracy. If democracy is conceived as a process, not an outcome, then whether a given measure is democratic depends on how it comes about, not its effects. Thus, if an absolute monarch chooses to resign power in favour of elected government, this decision – though democratic in effect – is not itself democratic403. Conversely, if the people vote to abolish democracy, their decision was still democratically-taken, even if its consequence is to end democracy. If the people vote to, say, deny blacks the vote, that is also a democratic decision that is undemocratic in effect, and a court that strikes down such a move is an undemocratic way of guaranteeing democratic outcomes (which include, of course, the possibility of future decisions being made democratically). To regard such a court as simply democratic is to mistake democracy for rule in the interests of the people, rather than rule by the people – but on this view even Plato’s Guardians would count as democratic404. 402 403 E.g. Dworkin (2000) pp.184-210. C.f. Hyland (1995) pp.61-3 and 154-5, and Berg (1965) p.117. 404 I say this because Plato insists that their purpose is to make the whole city happy, Plato (1992) p.95 [Rep 420b]. Of course, I intend it as a reductio. Plato describes his ideal city as a kingdom or aristocracy, Plato (1992) p.121 [445d], and is in contrast unfavourable in his description of democracy, Plato (1992) p.227 [557a]. Macpherson (1965) considers rule in the interests of the people a form of democracy, e.g. pp.5, 12-22, but this is rightly criticized by Lively (2007) pp.35-6, Sartori (1987) pp.376-9, and Holden (1988) p.83. Of course, the court may be democratically authorized, but (though I cannot fully argue this here) I do not think representatives are really democratic. 176 That a constitution is itself undemocratic does not mean that we need be even prima facie suspicious of judicial review. Democracy, understood as merely procedural, is only one value amongst many. We may have good reasons to limit democracy for the sake of other values, such as just or stable outcomes. If the people vote, for example, to criminalize homosexuality and judges overturn this decision, then I would welcome it, but this does not mean that we should confuse the value of personal freedom, which is defended, with democracy, which is compromised. Once we are clear in making these distinctions, absolute or unchecked democracy may appear all-things-considered undesirable, because of the other things we value. This is true whether democracy is implemented by majority-rule or lottery-voting at the voting stage, although lottery-voting perhaps involves a greater risk of unjust outcomes, since they can be brought about by only a small minority winning. The recognition of need for some kind of limit – constitutional or otherwise – only reminds us why, for so long, opponents of democracy regarded it as dangerous. As I argued, in the previous section, lottery-voting may encourage liberal compromises because it requires each group to face up to the possibility of losing. Majority-rule may do nothing to stop a majority group imposing their religion on minorities. If the potential democratic solution is a lottery between me imposing my religion on you or you imposing yours on me, however, then we may both agree to ‘live and let live’ rather than risk losing. Thus the very fact that lottery-voting is a potentially dangerous form of democracy also gives all groups all the more reason to agree to checks on that democratic process. As such, the constitution can come about as the product of a higher-order overlapping consensus between groups more concerned to protect their own rights than impose their way of life on others. 177 If we assume the presence of a constitution, however, then we need some provision for constitutional change, and if this is decided by lottery-voting then it may prove too easily modified and be threatened by the unreasonable minorities it is supposed to stop. Given the point of constitutional protections, they must be harder to change than non-constitutional laws and decisions. This could be achieved involving lotteries, e.g. we could say that any proposed changes be put to a vote and then two votes be randomly drawn and both must be in favour of change. If there was only 40% in favour of the change, then the chances of both random votes coming from this minority would be only 16%405. While such a system may seem more in keeping with lottery-voting, because it preserves the use of lotteries, there is no reason not to accept more traditional super-majority requirements for constitutional change, e.g. twothirds or three-quarters. That such is a deviation from lottery-voting need be no more embarrassing than the fact that it is a deviation from simple majority-rule is for those who favour such systems. Moreover, we can justify such by offering principled reasons why such a supermajority is likely to be less dangerous than a simple majority: when it comes to voting on constitutional provisions, people are more likely to be motivated by considerations of justice than self-interest; when it comes to justice, deliberation and persuasion may have greater parts to play than in the competition of interests; if there is any truth in Condorcet’s Jury Theorem, then a supermajority has much greater chances of identifying just outcomes; a larger group is less likely to be ‘captured’ by shared, factional interests; and finally constitutional protections are likely to require widespread support and compliance to be effective anyway. 405 Given that the group numbers less than one hundred, drawing the first vote from this minority would further reduce the odds of the second coming from the minority – e.g. 8 out of 20 then 7 out of 19 (=14.7%). 178 (5.6) Incentives for Deliberation In face of the risk for losing, parties might be able to agree to and uphold a reasonable compromise (perhaps through a constitution), rather than risk an ‘all or nothing’ vote. This would require deliberation and agreement between the various groups, however, which one might worry is excluded by lottery-voting. Remember though that lottery-voting is merely intended to replace majority-rule at the ultimate decision stage; it does not follow that it has to accept unreflective preferences as given and it is quite compatible with preferences being (re-)shaped by deliberation prior to the actual voting. The effects of deliberation will be discussed in the next section, but first it will be argued that lottery-voting provides incentives for deliberation. Firstly, deliberation can simply mean individual consideration of the issues in hand. Obviously, insofar as people want to cast informed votes for what they think, on reflection, the best policies, then they still have reason to deliberate about how to vote. Indeed, while lottery-voting may not solve the problem of rational ignorance, it does mean that anyone’s vote could be the decisive one, and this means that responsible individuals may give their vote more thought (see 5.3d, above) – whereas under majority-rule an individual vote often makes no difference and can therefore be cast with little thought. Those who call themselves deliberative democrats406, however, are not usually interested merely in encouraging individual reflection on issues, but in collective group deliberation407. The relevant question, then, is how lottery-voting interacts with such discursive group practices. As emphasized at the beginning of this chapter (5.2), lottery-voting is simply an alternative to majority-rule at the decisive voting stage, so it has no implications 406 407 E.g. Gutmann and Thompson (1996). Hence Dryzek’s ‘discursive democracy’: Dryzek (1990) pp.40-6 and Dryzek (2000) p.3. 179 about what happens before that. If it is desirable for people to discuss and deliberate together – which I think is often the case, but has been disputed408 – then lotteryvoting is quite consistent with this. People can be brought together both out of a commitment to finding reasoned agreement justified to all and because lottery-voting offers strategic incentives to try to persuade others. Of course, unless unanimity can be reached, the outcome of deliberation is never guaranteed to be implemented, but that is part of the reason there remains an incentive to try to persuade the others and, so long as there remain a significant number unpersuaded, why should we ignore their views? The previous chapter described how it is that all votes count (4.8-9), by influencing the chances of the option they are cast for, and pointed out that contingency and unpredictability of outcome gives people incentive to participate (4. 10-11). If the outcome seems a foregone conclusion, because there is already a sizeable majority, then there is little reason for members of the minority either to vote or to try to persuade others – they face a seemingly hopeless task trying to get their way. Because lottery-voting means that every vote could be decisive, and influences the chances of whichever option it is cast for, then it means there is always reason to vote. For example, even if one knows one is in a small minority, it is better that they have an 11/51 chance than a 10/50 chance. While, in large groups, the 1/n chance of one’s vote being decisive will still be tiny, it is more likely than being the one to make or break an exact tie (which is what needs to happen for a vote to be decisive under majority-rule)409. Moreover, aside from this rational incentive, the very message that ‘every vote counts’ may boost turnout, and campaigners are likely to do 408 409 E.g. Rousseau (1994 [1762]) pp.134-6 [Social Contract IV.1]. Although, of course, in this case all those votes are decisive (for if any one of them had not been cast, the outcome would be different). Blais (2000) pp.63-7 estimates the probability in Canadian constituencies as between 1 in 10,000 and 1 in 50,000. 180 more to mobilize support knowing that a) they do not need to win over a majority to their cause and b) even having a majority does not guarantee victory. Because lottery-voting gives proportional chances, it means that it is always better to have more support, since more votes imply more likelihood of winning. This means that not only is there an incentive for people to vote and encourage turnout from their like-minded fellows, but there is also a strategic reason to try to persuade opponents to join one’s cause. A minority may have no realistic hope of winning over enough opponents to become a majority, at least in the short-term, but if they can persuade a few to join them then they can at least shift the odds in their favour, e.g. from 70/30 to 65/35. Similarly, the majority cannot rest on their laurels – only unanimity can guarantee victory, so they also have an incentive to try to win over members of the currently unpersuaded minority. Thus lottery-voting increases the incentive for both existing majorities and minorities to try to persuade others410, since, whatever option one favours, one always wants as many votes on one’s side as possible. One may object that, if voters start off evenly divided (50/50) between two alternatives, and deliberation results in a 70/30 split, then that is grounds to believe the first alternative better supported by reasons, so we should go with this as the outcome of deliberation411. However, this ignores the 30% who were still convinced of the opposing view, having heard reasons of the other side – and presumably having reasons of their own. It is true that this is no longer simply a conflict of interests, however – the claim is that the first group should get their way not simply because 410 This does not mean one will necessarily phrase one’s argument to appeal to as many as possible. It may be better to make a persuasive case to a few (e.g. an argument with an 80% chance of convincing 40% of the people – expected to win over 32%) than a weaker case to a larger number (e.g. an argument that has a 50% chance of winning over 60% of the population – expects to win just 30% of the vote). I owe this point to Tom Porter. 411 I owe this objection to David Miller. 181 there are more of them but because they were able to offer more persuasive arguments, winning over some of their opponents. This highlights one important feature of the argument, however – it does not actually support majority-rule. The suggestion, reasonable enough in itself, is that we should take arguments that persuade people as better – this does not show they are right, but there are democratic grounds to acknowledge the outcome of people’s reason. However, majority-rule simply favours the larger group, which is not necessarily the one with the more persuasive argument. Suppose the initial split was 80/20, but the majority is actually ignorant and the enlightened minority are able to offer more persuasive reasons, resulting in a 60/40 split by the end of deliberation412. Majority-rule still favours the former group, yet a rule responsive to the outcome of deliberation would favour the latter group. I know of no rule that is responsive only to persuasion in the way we would want – the obvious problem with taking two votes before and after deliberation is that people may strategically express the opposite of their true opinion in the first, so it would look like they were persuaded, and count for more, when they reveal their real view after deliberation413. Lottery-voting does, however, go some way to recognizing the outcomes of deliberation, because each side gets a greater chance the more people it can persuade. If deliberation moves us from 50/50 to 70/30 then the former group, though not guaranteed victory, will be more likely to win. Meanwhile, if an enlightened minority are able to change a few people’s opinions – going from 80/20 to 60/40 – then they double their chances of winning. Again, lottery-voting gives each 412 There is no reason to expect good arguments will be able to persuade everyone, e.g. if the majority are affected by cognitive bias. This may be the situation faced by those campaigning against slavery or apartheid, say (though they did eventually win majorities). 413 Though, of course, the risk of this strategy is that if they really were persuaded to the other side in debate, their vote would count for nothing as it would look like they held it all along. 182 group incentive to try to persuade others to share their point of view, or to agree to a compromise (see 5.4, above). (5.7) The Nature of Deliberation and Limits of Reason-Giving One objective of deliberation is to reach a reasoned consensus, justified to all. This is supposed to be different from a compromise or modus vivendi414 reached through purely self-interested bargaining. It may be objected that the incentive structure described in the previous section gives parties only a strategic rather than moral reason to care about persuading others. Parties to deliberation are supposed to be concerned with what is even-handed and just, and reaching an agreement that no one could reasonably reject, rather than striving to maximize their self-interest415. While the two are theoretically distinct, the contrast is less-marked in practice. Often some parties will be genuinely concerned to reach a reasonable accommodation of all parties’ interests, while others will simply be out to get as much as they can for themselves. More than this, however, parties may be doing a bit of both – since justice is often indeterminate or imprecise, they want to advance their own interests as far as allowed by justice, without unjustly harming others. Democratic politics can therefore be seen as a competition between partly opposing interests within the limits of justice – this is what differentiates it from all-out war. Even when parties are acting in good faith, they can simply find it hard to appreciate opposing viewpoints or interests they do not share. Lottery-voting encourages mutual consideration because all parties really do have to consider the possibility of losing. This is not simply a point about respective bargaining positions (although I do think proportional chances 414 415 Rawls (1996 [1993]) e.g. p.147. This formulation brings out how deliberation is more like Scanlonian contractualism than Gauthier’s contractarianism. 183 make these fairer), but encouraging the sympathy for others that is essential to reasoned justification. One worry is that, by giving all groups a guaranteed chance of victory, we will actually discourage attempts to persuade others. Those groups who realize that they have little chance of persuading others, perhaps because they cannot phrase their comprehensive conception of the good in terms of ‘public reasons’416, may simply choose to withdraw from debate and rely on the fact they will still have at least a chance of victory. We can see these groups as choosing to ‘dig their heels in’, and accept the chance they already have of victory, without attempting to engage and persuade others. Since I am not committed to claims about the value of deliberation, this need not be problematic for me. Nonetheless, others may worry that this is simply one example of groups choosing to take an uncompromising hard-line, and having no reason to budge given they have a chance of success (getting all their way). No democratic system can guarantee harmony if people take such an attitude, but I would contend majority-rule is probably worse since it can consistently satisfy one group to the exclusion of minorities. Lottery-voting is fully compatible with reason-giving and deliberation and it need not accept people’s current preferences as given. Where reasons run out, however, it still gives each opposing group their fair chance of satisfaction. (5.8) Openness and Compliance Public deliberation would give people some idea of the prevailing opinion in the group. Further, section 5.4 suggested that there might be a case for open, public 416 Vernon (2001) p.151 gives a good example. But note I think ‘we want this, for religious reasons’ may be acceptable in public debate, since all can reasonably accept the importance of living by important conscientious convictions. C.f. Scanlon’s point that having one’s preferences satisfied is sometimes a matter of objective importance, Scanlon (1975) p.659. 184 voting under lottery-voting. It is a more general question, whether we adopt open or secret voting, whether we want to count and reveal the number of votes, or whether it is enough to simply determine the winner. One may think that it is best for the system to be as transparent as possible, but there is a worry that if people knew they were in a majority and lost because of the random element that they might either refuse to accept the legitimacy of the decision or attempt a re-vote on the same issue (this latter possibility is explored further in section 5.10 below). While the random selection method of the decisive vote should presumably be an open process – as described in the next section – there is a question as to whether the vote counts should be announced or not. Since the counting of any significant number of votes is a costly and time-consuming process, that both delays results and imposes administrative burdens, it would be advantageous if it could be avoided. Under a simple form of lottery-voting, all of this can be dispensed with by simply drawing out a vote, and announcing the winner, without ever having to count the other ballots. No one need know whether the winning vote in fact came from a majority or minority. This may have appeal sometimes, but it should not prove decisive – indeed, it should be borne in mind that we may have other reasons to count votes, for instance to monitor turnout or to impose minimum thresholds (see 5.3b, above). These might be less burdensome, for instance because we do not need to count all of a party’s votes to establish that they cross the threshold; but once we accept these costs then we may have less reason to give up on a complete and accurate counting. Moreover, people may have a fairly good idea about the split of public opinion anyway, especially if debate takes place before the vote, so this need not solve the problems. Therefore, it is not entirely clear that the problem can be avoided by secrecy at this stage. 185 Assuming that votes are counted, it is a further question whether these numbers should be made public. One might think there should be a presumption towards openness in politics, which would favour letting people know how many people voted for each option. The danger is that if we say something like ‘90% voted for A, 10% for B and B won’ then we risk destabilizing the outcome, since those in the majority may feel hard done by and that the outcome is less legitimate. But this may only be a problem if people’s expectations have been shaped by majority-rule, so they assume it natural that the majority get their way and the minority lose. This would, of course, be a difficulty for any attempted transition to lottery-voting, but it presumably would not affect people who were brought up under lottery-voting and accepted its fairness. Under a system of lottery-voting all should be aware they may be in a majority and lose, just as they may be in a winning minority. All should be prepared to accept this, knowing that majorities will win more often than not. Just as two parties may agree to resolve their claims by a lottery, and each accept the outcome, so, if all accept the reasons for lottery-voting, as opposed to simple counting procedures, they can accept its outcome as legitimate, regardless of the split of opinions. Thus, even if those in the majority do think they have a stronger claim, they may still accept defeat after being given a greater chance. Obviously if the majority tried to have their way by sheer weight of numbers or force then they are unjustly usurping the democratic process. Nonetheless, there may be legitimate worries that those who lose a decision – whether a majority or minority – may be less motivated to comply with the outcome. If it is publicly known that most people opposed the outcome, then it may be feared that there will not be sufficiently widespread compliance, so the selected outcome will not be implemented or sustainable. This need not be a problem, however; it is so only if people are not sufficiently motivated by a sense of justice or fair play. We 186 need to distinguish people’s private preferences from what they think is legitimate. If all accept that B was the outcome of a fair procedure, then ideally all should accept and comply with B417. Any democratic resolution of disagreement will require the compliance of losers and, while this will be easier the fewer losers there are, we should recognize that even majority-rule may leave many dissatisfied – particularly when there are more than two options, those counted part of the ‘majority’ may in fact only be getting their second worst option. While, because it relies on chances, lottery-voting cannot quite guarantee that the same people will not be persistently defeated, it can at least guarantee that they will always have a chance of victory next time, and thus reason to continue playing the game. A permanent minority, therefore, may be reconciled to lottery-voting, though the odds are stacked against them, while they have little reason to accept majority-rule, which means they will never get their way unless they happen to agree with the majority. The other worry is that if a defeated group know they were actually a majority, they may have a greater incentive to call for a revote. This is in fact a general problem, that if losers know that they always have another chance to win ‘next time’ then they will be all too keen for next time to come soon, that is dealt with in section 5.10-11 below. That a decision has gone against one is not, however, justification for seeking to change it, or disregarding it in the meantime. I argue, below, that we may have to impose time limits before the issue can come to a vote again; so the minority will get their way, at least for now. Ultimately, whether we want to announce the distribution of votes will depend on the empirical consequences of doing so, which will depend on the characters of the people involved and cannot be determined a priori. In an ideal polity, there is no reason why we should not; but in practice it may 417 We might make exception for conscientious objection or civil disobedience if B is clearly substantively unjust. I assume unjust options are excluded by the constitution, so the difference between A and B is only a matter of preferences. 187 be more expedient not to reveal these facts for fear of unsettling the chosen outcome. For now, it suffices to say lottery-voting is compatible with either transparency or non-disclosure of these results. (5.9) Scrutiny Whether or not the numbers of votes are made public, we can predict that people will be particularly keen to scrutinize the random selection machinery to ensure that all is indeed fair and above board. Those who lose the draw – particularly if they are in a substantial majority or it happens repeatedly – may wonder whether it was somehow stacked against them. Hopefully this worry will be dismissed over a run of decisions, which should see all groups get their way sometimes. However, it is in the nature of randomness that outcomes will not be exactly proportional or predictable (as pointed out in section 5.11, below, the case for lottery-voting does not rely on them being so). While this is to some extent an advantage (see, in particular, 4.5 and 4.10-11, above) because people often make irrational predictions – e.g. ‘if the coin was heads the last three times, it is more likely to be tails now’418 – it may lead to concerns about whether the process is truly random. This is all the more troublesome because, while the number of votes can be recounted and verified, one cannot repeat the random draw and get the same answer. Despite these worries, lotteries have been used for many important decisions (see chapter 3.2). If a lottery is to make voting fairer though, then it needs to be a fair one itself. To some extent, people’s trust in the impartiality of the process can only be built up after a reasonably lengthy period of successful operation. Nonetheless, we 418 People are generally bad at randomness. If you ask people to give you a random number from 1-100 few are likely to give round numbers (37 seems ‘more random’ than 40), and if you ask the same person for a second one they are likely to choose one further away from their first number (high then low, or vice versa, seems more random than two high or low numbers). 188 should give all concerned ample opportunity to inspect whatever randomizing device is used, for example a number of ‘trial draws’ to illustrate lack of bias. Just as the UK National Lottery uses a number of different machines and balls, supervised by independent adjudicators, all of these safeguards can be in place. Further, we can follow the example of making the draws public and televised or, at least, recorded in case of any later dispute about results (since this resolves worries about being unable to repeat the process). Provided that the randomizing part of the process is sufficiently open, so all can see that it is fair, this should help foster public confidence in the process as a whole. (5.10) No Repeats One problem that we will have to face, however, is how ‘final’ decisions have to be. When one side win an election, they are likely to want the matter settled ‘once and for all’, but the losers have an incentive to bring the issue to the vote again and again until they win. This is true under majority-rule too; some issues are likely to keep recurring until one group is satisfied, and even then the other side may begin trying to reverse the decision. Majority-rule, however, offers relative stability because if one option was chosen, in virtue of winning say 60% of the vote, then there is no point re-introducing the issue unless a significant number of voters can first be swayed to change their minds (or a significant number of former non-voters can be persuaded to vote for one side). If the same motion comes in again tomorrow, it will merely be rejected again. Because lottery-voting makes the outcome depend partly on chance, however, a defeated group – whether minority or majority – always has an incentive to try again as soon as they can for, even on the next day, they may win, even if no one changes 189 their vote, simply by the fortune of the lottery. In some sense, this is a tragedy of the commons. It is individually rational for the defeated party to keep proposing a motion until successful, yet it would be better for everyone if the issue was settled rather than constantly being re-introduced – because, indeed, as soon as the defeated party get their way, the former victors are only likely to begin a constant campaign to overturn the decision and restore the former status quo (if possible). It seems that we need some check on motions being proposed again and again. One option is to have a constitutional limit that simply says something of the form ‘no motion, once decided, can be re-introduced within the space of a year’. This will prove somewhat inflexible, however – there might well be radical changes in circumstance that would induce a majority to change their minds and perhaps make it reasonable to re-vote. Moreover, any such limit will require some means of deciding what counts as the same motion. Clearly it need not be exactly the same, as effectively the same proposal could be reintroduced with a minor change to achieve the same effect – for example, the book buying motion from section 4.6 could be reintroduced in a modified form as ‘This GCR proposes to spend £150 on books and the president must sing the national anthem’. Clearly we want to exclude motions intended to achieve the same substantive effects. Another option might be to leave the duration of each decision within the scope of the decision itself. Since all voting on a particular outcome may want it to last with some stability, it may be open to decide on a case-by-case basis that ‘we will adopt policy X, and this will be so for 5 years without change’. Again, this might be too inflexible, and is particularly problematic if some want to make the policy permanent. If all agree to a protected duration for the decision before it is taken, however, then all have to make a trade off between how much protection they want if successful and 190 how long they must endure if unsuccessful, thus maybe a reasonable compromise can be found (for general remarks on prior agreement and compromise, see 5.4, above). If unanimous agreement is not reached, then the duration of the decision can be settled by a separate lottery-vote. Obviously making the duration part of the same decision will not work, as everyone is likely to vote for permanency, knowing that if they get their way they will want it protected. We can, however, have people cast separate votes for the duration and have one of these determine how long the decision will remain in effect for, while a different vote determines the substantive policy. In this way, everyone has to suggest a duration independently of knowing whether or not they will win and therefore has to balance the protection they will want if they do win against the prospect of a re-vote they will want if they do not. If the decision is one that calls for a certain duration, then we can expect that all voters will agree to accept such, independently of whether the substantive decision goes their way. Matters would be more complicated if certain substantive options called for longer durations than others, and then it may be that duration would have to be left part of the decision itself, but such cases do not seem likely to be common. (5.11) Dividing Decisions The duration of a particular decision could be seen as one way of making it more momentous. Deciding that a given policy is to be in effect for a week is obviously not equivalent to deciding the policy for the coming year. The preceding discussion of duration is therefore simply a particular instance of the more general issue concerning the ‘size’ or importance of any particular decision. It might be thought fairest if we could make all decisions at least roughly equivalent. After all, if we were to take two decisions, one of which was of great moment and the other 191 relatively trivial, and suppose two people had opposite views on each, it might be objected that it would be unfair for one to get their way on the very weighty issue and the other only on some insignificant matter. If the decisions have to be made in this way, however, then perhaps the fairest outcome is probably for the more important one to be decided randomly, and for whoever does not get their way there to have the other as ‘consolation’. This is not perfect equality of satisfaction, but it is the closest possible in the circumstances. It is no part of the argument for lottery-voting that it will produce proportionate outcomes. Indeed, this it should be remembered that lottery-voting does not even guarantee that the two decisions will go in opposite ways. If we toss a coin on each, then it is quite possible (one-half chance, in fact) that one of the people involved will get their way on both. Lottery-voting realizes fairness and proportionality at the stage of chances, so the equality of voters is a procedural matter, not necessarily reflected in outcomesatisfaction. Therefore it does not matter if some of the decisions to be made are more important than others, since all ex hypothesi are made fairly, and lottery-voting would still be fair even if only one decision was ever made, so there could be no proportionality of outcomes419. This latter possibility shows that lottery-voting could be used to select a dictator, who decides over all issues420. If all agreed to this then it would be fair, but I assume that most people would not be willing to risk very likely total exclusion from all decisions for the sake of a tiny chance of getting everything their way421 – thus they would not agree to put so much power into a single vote, but would prefer to break decisions into discrete packages. 419 Assuming a proportional compromise over that one outcome is impossible, of course – see 3.13, above. 420 See remarks on Arrovian dictators in chapter 6.11. 421 Compare the ‘strains of commitment’ described by Rawls (1999 [1971]) pp.153-4. 192 There is no need to worry about the fact that some of these decisions will be more important than others, but one might still wonder how the issue terrain is to be broken up into these packages. It might be that one can still assert undue influence on decision-making by manipulating the framing of the agenda. For example, return to the earlier example of buying reference books (sec 4.6). There it was suggested that lottery-voting offered a solution that respected each person’s preferences equally. But over what questions should we hold such a vote? We might perhaps hold a lotteryvote to decide whether to buy any books, in which say there would be an 80% chance of ‘yes’, in which case we could hold another lottery-vote to determine how much to spend (where all were committed to spending something, and each has an equal chance of being the one to set their preferred amount). We saw, however, that it was indeterminate whether we should proceed this way or start by deciding an amount and then voting whether or not to go ahead and spend it. If these two decisions are treated as separate issues, to be resolved serially, then lottery-voting does not help order our concerns. If we reconsider what is intuitively appealing about lottery-voting, the equal chance for everyone to get their way, then it seems that the natural approach is to hold a single vote, where the winner gets to decide both how much is spent and how. This gives everyone an equal chance of their preference being decisive. As we just saw, the ‘size’ of the issue is unimportant. What matters is the procedural equality of chances, rather than outcome-satisfaction, which need not be equal, so it does not matter if we combine potentially separable issues; indeed, it may be the best way to produce a coherent result. If we separate the votes, then we may get one person being able to decide the amount they want spent, for example, but then dissatisfied at how it is spent. If we were to start with voting whether or not to buy books, and decided to 193 do so, how could we constrain those who had voted against this from putting forward a prohibitively small amount (e.g. £0.01) in a second vote, to effectively have their way? Two separate votes could therefore give these people two chances to secure their preferences. Thus it seems sensible to decide the whole issue in a single vote, whereby one individual gets to decide how much is spent on what books. This is all very well, but one might object that ‘a single issue’ is still too vague. After all, why should we take the purchase of a set of reference books as one issue, rather than vote on each separately? In this case, I would suggest that we turn to the underlying preferences of the people. Suppose everyone agrees that they are better off with either a complete reference library or none, then it is reasonable to decide the matter by a single vote. If, however, there are some who want, say, a dictionary but nothing else, then we have two options – we can either have a separate vote over the dictionary, or we can table the purchase of reference books as a single motion but let those people express their preference to, for example, spend £25 on a dictionary. In this case, I would favour the latter, as having separate votes again runs into the problem of giving some people ‘second chances’. If it is already decided that we will have a dictionary, it may affect what people think on the latter question about the other books. Now suppose it is suggested that we also buy some bookshelves to house the possible reference library. If all agree that we want the bookshelves if and only if we buy the books, then it is reasonable to amalgamate their purchase into the same issue, which will guarantee the consistency of the decision (see chapter 7.4-5, below). But, conceivably, some might want the shelves for another purpose or think that the books are best kept somewhere else without cluttering the place with more furniture. In this case, it would not matter if there was a separate vote on the shelves. While it would 194 still be fair to let one person determine both issues, it would be reasonable to separate these decisions in order to spread decision-making, so that more people get their way on something. There might well be a trade off between the possible increase in consistency if one person was to decide both issues and the desirable dispersion of power and consequent increase in proportionality that would result from two separate votes, in which the same person was unlikely to decide both. Since no one is guaranteed victory on any particular vote, it seems no one has any particular ex ante reason to favour either approach, so which is adopted should itself be a matter of democratic debate (see section 5.4, above for the possibility of prior agreement and compromise). (5.12) Conclusion While the previous chapter described, in the abstract, how lotteries could be used as part of a democratic decision-mechanism, the task of the present chapter has been largely to flesh this out with a more detailed discussion of practical problems and issues. My concern has only been to show how lottery-voting could work in practice, and thus I have not felt obliged to spell out every possible detail – in many cases, I have simply laid open possible options, believing the final decisions best left to the people concerned in light of their circumstances. Nonetheless, I hope this discussion has gone a considerable way to showing lottery-voting practical, and dispelling many doubts about how (or whether) it would work in the real world. Now that we have a clearer idea of how lottery-voting could operate, we can better assess its desirability. The task of the next chapter is to assess it against the conditions many have felt a decision-mechanism should possess, as spelled out by the 195 axioms adopted in social choice literature, while the seventh and final chapter considers the objection that it is irrational to trust important decisions to chance. 196 6 Minimal Conditions of Social Choice “[R]ules must take into account neither the different content of alternative policies nor the characteristics of those who favor them. Two types of decisional rules satisfy these conditions: random rules and numerical rules”422 “[T]he only voting procedure which is Pareto-optimal, nondictatorial and strategy-proof is ‘random voting’”423 (6.1 Introduction) I have argued that democracy must be based on procedural fairness, and the last two chapters have set out a novel alternative to majority-rule, called lottery-voting. The discussion of lotteries in chapter 3 was supposed to show that this system of proportional chances is, indeed, one way of treating people equally. Chapters 4 and 5 further described how lottery-voting could work in practice. However, it may still be objected that lottery-voting fails to satisfy some of the basic conditions we expect from a satisfactory decision mechanism. The present chapter takes up this objection, drawing a number of axiomatic conditions from the social choice literature – specifically, from May and Arrow – and assessing lottery-voting by each in turn. Each requirement is assessed for its independent plausibility first before I turn to how well lottery-voting satisfies it. The conclusion is that lottery-voting clearly satisfies the more normatively important conditions, while those it fails to meet it violates in relatively unproblematic ways or because of its different aims. The upshot of this chapter, then, is that lottery-voting passes the minimal standards that we should demand of a decision mechanism. The related, but wider, question of its rationality is postponed to the next and final chapter. The present discussion ends, however, with two more results of interest to the social choice approach: that lottery-voting is non- 422 423 Berg (1965) p.130 [not emphasized in original]. Elster (1989) p.87 [not emphasized in original]. 197 manipulable (a voter cannot do better by strategically misrepresenting their preferences) and that it facilitates, should we desire it, weighted voting. (6.2) May’s Conditions May gives four necessary and sufficient conditions for simple majority-rule: that the procedure be decisive, anonymous, neutral and positively responsive424. While he does not explicitly claim their desirability, and is concerned only with definitional questions, they have – not unreasonably – been taken as providing reasons to favour majority-rule425: we generally want the procedure to be anonymous and neutral to satisfy political equality, while decisiveness and positive responsiveness are practically desirable features of any decision mechanism. Each shall be assessed in turn but, before examining May’s conditions, it should be noted that he limits himself to a two-option framework. This is not itself a criticism of his proof, but limits its practical relevance, since real-life decisions almost always involve more than two options, and Riker goes so far as to say there is no fair way of reducing them to a binary choice426. While one can use various techniques to artificially narrow or reduce alternatives to two or to deal with a wider range of options, such as STV or AV, part of Riker’s point is that these may yield different results, so none can be identified as the will of the people427. As argued in chapter 2, democracy cannot in general aspire to produce a single ‘best’ or ‘true’ outcome. Many questions are at least partly distributive or indeterminate, so what matters is that we have a fair procedure to settle who gets what from cases of conflicting interests. Democracy is, on this view, a form of pure 424 425 May (1952) pp.681-2. E.g. McGann (2006) pp.80, 89, 134, 168, and 174. 426 Riker (1988 [1982]) pp.59-60. 427 See Riker (1988 [1982]) pp.21-40, and Shepsle and Bonchek (1997) pp.167-70. 198 procedural justice, and so the path dependence of the outcome is not a problem, since it is what legitimates the outcome. Finally, it should also be noted, because it will become important in what follows, that May is only concerned with determinate procedures. Because lottery-voting relies on chance, it is not always easy to see how well it fits May’s technical definitions; sometimes, therefore, the argument has to rely on the intuitive normative force of his conditions, rather than formal tests. (6.3) Decisiveness 1.“[T]he method must be decisive and universally applicable, or more briefly always decisive, since it must specify a unique decision (even if this decision is to be indifferent) for any individual preferences.”428 May’s first condition appears uncontroversial – we want a procedure that always yields outcomes for any two options X and Y. Sometimes, in life, we are not sure that all values are comparable, for example, ‘who is the greater artist, Shakespeare or Mozart?’ or ‘which of your children do you love more?’ It does not always seem possible to say either that X is better than Y, Y is better than X, or that they are exactly equal. This would be a problem if we expected our democratic procedure to yield some independent truth about this comparison; however, decision procedures are practical rather than theoretical matters. The purpose of the procedure is to determine what we should do, not believe, so a more appropriate question therefore is ‘should we go to see a Shakespeare play or hear a Mozart concerto?’. While, in theoretical matters, we can always suspend judgement, in the practical domain we always have to do something (even if it is nothing – which counts as an action in this context), so we cannot permit such ‘gappiness’. We need an answer; we must do either without thereby committing ourselves to saying either is better. 428 X or Y, May (1952) p.681. 199 It is this very necessity of an outcome, however, that highlights a shortcoming in May’s requirement. When he says that the method must be decisive between two options he explicitly includes indifference as a unique decision – that is, in formal language, the procedure must yield either XPY, YPX, or XIY (so not XRY or the conclusion that the two are incomparable). Indifference is, however, of no use in a practical context, where we must choose either X or Y, so in fact we want a stronger decisiveness condition than May suggests429, that includes a tie-breaking rule. One obvious way to break ties is to toss a coin430 – and, indeed, this is the only option compatible with both anonymity and neutrality (obvious alternatives include giving the chair casting vote, which is non-anonymous, or preserving the status quo, which is non-neutral)431. Lottery-voting obviously satisfies decisiveness, even in this stronger form, since all votes must register either XPY or YPX so whichever is randomly- selected will yield either X or Y as the preferred outcome. (6.4) Anonymity 2.“The group decision function is a symmetric function of its arguments… D is determined only by the values of the Di that appear, regardless of how they are assigned to individuals as indicated by subscripts (names). A more usual name is equality.”432 429 This is how Risse (2004) pp.44-5, esp. fn.6 (mis)interprets the condition, leading him to say it is only reasonable with an odd number of votes. 430 This solution is suggested, among others, by Barry (1965) p.88, Mackie (2003) pp.5, 50, 84 and McGann (2006) pp.17-8. Goodwin (2005 [1992]) points out lotteries are accepted in English law pp.198 and 255, and is in fact used to break electoral ties p.55, citing the Representation of the People Act 1983 Schedule 1 (parliamentary election rules). Her example of its use comes from July 1984, but in May 2006 a tied vote for a St Albans council seat was also resolved randomly, by drawing straws (actually pencils), as reported http://news.bbc.co.uk/1/hi/england/beds/bucks/herts/4974304.stm (last accessed 26/01/07) and in June 2007 the Ulster Union party used the flip of a coin to decide which of its two candidates would become Lord Mayor of Belfast, http://news.bbc.co.uk/1/hi/northern_ireland/6720647.stm (last accessed 20/08/07). Note any such random tie-breaking will also fail Sen’s neutrality test (see 6.5 below). 431 McGann (2006) pp.17-8, 62-3, and 80-1. 432 May (1952) p.681. 200 Anonymity requires that the social decision depend only on the value of individual votes, and not on who cast them, so a permutation of votes does not affect the outcome. That is, if when I voted XPY and you voted YPX YPX the outcome was XPY X, then, had the situation been reversed – i.e. I voted and you – the outcome would still have been the same (X). As May observes, this condition is basically equality between voters, as is easily seen if we consider its violation – suppose in the previous example my vote counts for twice yours, then when our votes are reversed the outcome may change to Y, which shows that our votes are not equal. Some theorists of democracy deny that voters must always be counted equally (see the examples in chapter 2.5b) – that is, they think that there can be reason to give one person more weight if, for example, their interests are more affected. There are, however, two ways of giving one person more weight – one can either give them a single vote that counts for more or one can give them more votes. While these will come to the same thing in the final analysis, it is better to think of these people as having more votes; that way, we can still accept that we want all votes to be equal, regardless of who casts them. This means that, if all my votes were swapped for all your votes, the outcome may change (since one of us may justly have more votes), but if one of my votes is swapped for one of yours then the outcome should be unaffected. Although weighted voting is not proposed here, section 6.14 below, shows how lottery-voting actually facilitates giving some more voting power than others, if it is thought desirable. In the meantime, however, we can continue to assume that equality, and so anonymity, is desirable. The question is whether lottery-voting satisfies anonymity. One may argue that it does not, because there comes a point – when let us suppose my vote has been randomly drawn – when a permutation would affect the outcome. That is, if I voted 201 XPY and you voted YPX, and my vote was then selected, the outcome would be X, but if our votes had been reversed, and mine had still been selected, then the outcome would be Y. This seems to violate anonymity, because my vote matters more than yours – indeed, mine is decisive, whatever yours (and everyone else’s) is. It is too hasty, however, to conclude that lottery-voting does not treat everyone equally. The equality in lottery-voting is that everyone has an equal chance of being selected as decisive (see chapter 3.13 and 4.5). Complete retrospective equality is possible only if there is unanimity or a perfectly proportional outcome, so equality usually has to be given up at some point, and this is true also of majority-rule – even if it is the case, beforehand, that everyone is equally likely to be in the winning majority, once the decision is made the losers are not treated equally except in that they had a chance of victory. If we consider the ex ante case in lottery-voting, before the draw is made, then swapping my vote and yours makes no difference to the likely outcomes, each remains equally probable, so it is clear that lottery-voting does treat each vote equally and captures what is normatively important about anonymity. Each person’s vote has an equal chance of being decisive, regardless of who cast it. (6.5) Neutrality 3.“[T]he method of group decision does not favor either alternative. A precise way of stating this is that if the names of x and y are reversed, the result is not changed.”433 While anonymity means that no voter is favoured, neutrality means that no option is favoured. Neutrality rules out supermajority requirements, such as the need for a twothirds majority. This can be seen if we consider Sen’s formalization of the requirement: 433 May (1952) p.681. 202 “Neutrality demands that if two alternatives x and y, respectively, have exactly the same relation to each other in each individual’s preference in case 1 as z and w have in case 2, then the social preference between x and y in case 1 must be exactly the same as the social preference between z and w in case 2”434 This means that if x and z each have 60% support and y and w each have 40%, then either x and z or y and w must win. If x wins the former and w the latter the decisionrule appears to be non-neutral, as would be the case for example if the z decision was a proposed constitutional amendment requiring a two-thirds majority to pass. Neutrality is violated because the 60% who favour x get their way, while the 60% who favour z do not, thus it can be seen as another aspect of equality, because if some options are favoured it makes it easier for some people to get what they want than others. It is not enough for all to have equal votes if, for some, the goalposts are smaller or further away. There are two points to note about neutrality, before we turn to assessing lottery-voting. Firstly, it requires that if all voters reversed their preferences, any nonindifferent result would be reversed, while indifference would remain. Thus, technically, a rule that says D=0 (i.e. the social choice is always indifference, regardless of people’s preferences) is neutral, albeit undesirable because of its total lack of responsiveness435. Secondly, this is a case where it is particularly clear that May’s conditions are descriptive rather than normative. It is of course obvious that anything with a supermajority requirement cannot be simple majority-rule; however, it is less obvious that supermajority requirements are always undesirable because they are non-neutral. Often we think certain matters, such as constitutional protections, should be more difficult to amend. While this may mean compromising democracy for the sake of other values, it is not, all-things-considered, wrong. Thus we should 434 Sen (1970) p.72. He acknowledges this also includes IIA, (1970) p.68, as does May’s formulation; but since IIA is assessed in 6.10, below, it seems misleading to consider it a requirement for neutrality. 435 Kelly (1988) pp.9-10. 203 not attach over-riding importance to neutrality in general or to technical definitions of such. Lottery-voting certainly fails Sen’s test. It is possible for x to get 60% of the vote and win while z gets 60% and loses. This does not, however, mean that lotteryvoting is actually non-neutral in an intuitive sense. Sen’s test may be appropriate for preferences, but not for decision procedures. We can see this if we consider random tie-breaking, as introduced in section 6.3, above. Note that votes could be split 50/50, in which case it could be said that there is no social preference. If the winner is determined by tossing a coin, however, this decision procedure fails Sen’s neutrality test because x gets 50% of the vote and wins while z gets 50% and loses (indeed, it could be that x=w and z=y). Since tossing a coin seems an obviously fair way to resolve a tie, despite failing Sen’s test for neutrality436, it seems that either the formal test fails to capture the intuitively compelling idea of neutrality or that it is simply inappropriate for assessing decision procedures. What matters normatively is that no option is favoured a priori, as is the case when the status quo is privileged. While lottery-voting could be combined with constitutional protection (see 5.5 above), which would not be neutral, the desirability of such is a separate matter. The basic procedure is neutral, because it does not privilege any particular outcome. Each voter’s preference is equally likely to be decisive, regardless of what option it is cast for. (6.6) Positive Responsiveness 4.“[I]f the group decision is indifference or favourable to x, and if the individual preferences remain the same except that a single individual 436 E.g. Barry (1965) p.88, Bordes (1985) p.185, Duxbury (1999) p.20, Kelly (1978) p.75, Mackie (2003) pp.5, 50, 84, Berg (1965) pp.130, 132. 204 changes in a way favourable to x, then the group decision becomes favourable to x.”437 May’s final condition, positive responsiveness, is a strong one, requiring that a single vote be able to break any tie. Note, however, that just as neutrality could be trivially satisfied by an unresponsive rule, positive responsiveness can be trivially satisfied by a non-neutral rule: a rule where the social decision is always in favour of x satisfies this requirement, because whatever changes take place in individual votes it will remain favourable to x438. The problem is, of course, the lack of neutrality, but this interdependence between May’s conditions highlights the caution needed in assessing each separately. We can distinguish two features of May’s condition. Firstly, non-negative responsiveness: if the decision is favourable to x and an individual changes in a way favourable to x then the group decision must remain favourable to x. It would indeed be perverse if an individual’s change in favour of one option could actually harm that option’s chances of winning439. Non-negative responsiveness therefore has obvious intuitive attraction – indeed, it has been argued that it captures what is essential and plausible about the requirement440. Lottery-voting satisfies this part of the condition. Suppose individual i’s vote has been selected. If this vote is for x then the social decision is x; if it is not, then the social decision is not x. If any individual other than i switches their vote, in favour or against x, it makes no difference. If i’s vote was initially other than x, but i then switches their vote to x, the social decision becomes x. Putting it from the perspective prior to the random selection of the decisive vote, any individual’s movement in 437 438 May (1952) p.682. Kelly (1988) p.11. 439 This is one major problem of STV. 440 McGann (2006) p.62 cites Nurmi (1987) Comparing Voting Systems (Dordrecht: D. Reidel) p.67 as suggesting such; unfortunately this book is not in any Oxford library, so I have been unable to examine the argument first-hand. 205 favour of x increases x’s chances of winning. Since this is slightly stronger than nonnegative responsiveness (which would require merely not harming x’s chances), it may suggest that positive responsiveness is satisfied, but that is not quite so. While what Arrow calls the ‘positive association’ of individual and collective orderings actually only requires monotonicity, or the non-negative responsiveness discussed so far, May’s positive responsiveness is more stringent441. We can see this if we isolate the second part of his condition: if the group decision is indifference and a single individual changes in a way favourable to x, then the group decision becomes favourable to x. Let us call this the tie-breaking element. It can immediately be seen why this is stronger than mere non-negative responsiveness, because it requires that a single vote is always able to break ties. Incidentally, one may think this rules out the non-neutral rule ‘always favour x’, introduced above, but in fact this is not so – such a rule does not permit indifference to be the initial situation, so the antecedent of the conditional (‘if the group decision is indifference…’) is never met. It seems that lottery-voting satisfies this requirement only on a similar technicality, since it never allows the initial group decision to be indifference either. Since, whichever vote is picked must be for either x or y – and hence the decision is strongly decisive in the way defined earlier (see 6.3, above) – there can never be a tie to break. One may think that lottery-voting therefore passes on a technicality, but this should not trouble us. Firstly, it is not clear that this tie-breaking element is normatively desirable. Many have thought if there are large numbers on either side, it is not obvious that a single vote should completely tip the balance442. Secondly, we have already seen that lottery-voting does satisfy the more obviously important nonnegative responsiveness element of May’s condition. While no one individual can 441 442 See Arrow (1963 [1951]) pp.25-6 and May (1952) p.682, fn.8. In another context, I seem to be joined by Kamm (1993) p.103 – who describes an extra person in this case as an ‘irrelevant’ utility’ – Broome (1998) pp.956-7, and Hirose (2004) pp.73-8. 206 break a tie, ultimately some individual is decisive, and every extra vote for x increases x ’s chances of victory (see 4.11 above). Thus the switch of one individual’s vote in favour of x can never harm x, and always increases its chances of success, which seems enough responsiveness to demand. Since these conditions, as May defines them, are supposed to be not only necessary but sufficient for simple majority-rule, and lottery-voting obviously is not simple majority-rule, it may be supposed that lottery-voting must fail to meet at least one of the conditions. In fact, however, May was only considering determinate social rules and this makes it hard to tell whether lottery-voting technically meets some of the conditions. Lottery-voting does, however, satisfy what is intuitively normatively appealing in each case: it produces a decisive outcome, treats all options and voters equally, and is responsive to changes in preferences in that, though they do not automatically decide the outcome, a switch in favour of any option is always better for that option (increasing its chances of victory). I now turn from May’s conditions, which are ostensibly definitional, to Arrow’s explicitly normative axioms. (6.7) Arrow’s Conditions It should first of all be noted that Arrow was not concerned with all decision mechanisms, but a special sub-class which he calls social welfare functions: “By a social welfare function will be meant a process of rule which, for each set of individual orderings R1, …, Rn for alternative social states (one ordering for each individual), states a corresponding social ordering of alternative social states, R”443. 443 Arrow (1963 [1951]) p.23 [emphasis original]. Note that Arrow’s social welfare function is a choice rule specifying orderings, which must by definition be transitive and complete. A complete relation is one with no gaps, i.e. xRy or yRx. A transitive relation means xRy and yRz implies xRz. These are Axioms I and II, respectively, Arrow (1963 [1951]) p.13. Arrow’s social welfare function is related to, but different from, a Bergson-Samuelson social welfare function, which is simply an ordering; see Arrow (1963 [1951]) p.23, Sen (1970) pp.33-6, and Arrow (1984) pp.14, 50, 68-9. In later work, 207 Arrow focuses on such rules because he believes, by analogy to individual choice, than such a preference ordering is necessary for rational decision-making. I have already argued that democracy is about fair distribution, rather than realizing a socially-best outcome (see chapters 1-3), and I will dispute this conception of rationality in the following chapter. What concerns me here are the conditions Arrow claims that any reasonable social welfare function for turning individual preference orderings into a social preference ordering should satisfy, namely: Universal domain, weak Pareto, Independence of irrelevant alternatives (IIA) and non-Dictatorship444. Arrow’s conclusion is that no social welfare function can simultaneously satisfy all of these requirements, which I take to be one good reason to question any imperfect procedural conception of democracy aiming to discover an independentlyexisting social preference or ‘general will’445. Since the proof of this theorem is technical and complicated it is omitted here, but various proofs can be found in the literature, for example following Sen446. My concern is not with this broader conclusion; rather I focus on the desirability of the conditions that Arrow imposes on a social welfare function. While lottery-voting is not a social welfare function, since it aims merely to pick a winner rather than produce a social preference ordering, it seems prima facie that many of the same conditions will be desirable. In fact, it is not clear how or whether some of them could be applied to a procedure like lotteryvoting. I shall, however, proceed to comment on each in turn, in order to illustrate what is desirable about each of them and how far lottery-voting goes towards meeting these supposedly minimal conditions. Remember, since Arrow’s result is that it is Arrow acknowledged the scope for confusion and employed the terms constitution, e.g. Arrow (1963 [1951]) pp.103-4, or social choice function, e.g. Arrow (1984) p.50. 444 I take these labels from Sen (1970) pp.37-8. 445 Others, most notably Riker (1982), have drawn sceptical conclusions about the prospects of democracy generally. As McGann (2006) observes, however, it is only a particular conception of democracy that is discredited. C.f. Mueller (2003) p.595. 446 McLean (1987) p.176. 208 impossible to satisfy all of these conditions simultaneously, no procedure should be expected to do so – the question facing us is which requirements we should weaken or abandon, which gives us reason to examine their normative plausibility447. I shall argue that lottery-voting captures what is normatively important in each case, but rejects Arrow’s basic understanding of democracy, and so both his ‘collective rationality’ requirement and strong form of IIA. In each case, except IIA, discussion comes in two sections, first exploring the normative attractiveness or plausibility of the requirement, and then assessing how well lottery-voting performs against it. (When it comes to IIA, a third section is needed to clarify its meaning, before turning to the other two tasks). (6.8) Universal Domain 6.8.a Plausibility The condition often referred to as Universal (or Unrestricted) Domain in the secondary literature448 follows Arrow’s Condition 1 “Among all the alternatives there is a set S of three alternatives such that, for any set of individual orderings T1,…,Tn of the alternatives in S, there is an admissible set of individual orderings R1,…,Rn of all the alternatives such that, for each individual i, xRiy if and only if xTiy for x and y in S”449; which is more plainly stated as, “The social welfare function is defined for 447 C.f. Samuelson (1977) p.938: “what Kenneth Arrow proved once and for all is that there cannot possibly be found… an ideal voting scheme: The search of the great minds of recorded history for the perfect democracy, it turns out, is the search for a chimera, for a logical self-contradiction”; also quoted in Mackie (2003) p.10. 448 E.g. McLean (1987) pp.173-4, Sen (1970) p.37. 449 Arrow (1963 [1951]) p.24. 209 every admissible pair of individual orderings, R1, R2”450 or in the modified Condition 1' “All logically possible orderings of the alternative social states are admissible”451. It is often objected that certain orderings ought to be excluded, for instance preferences that violate the rights of others or are simply ‘external’452. Arrow can, in fact, allow for these restrictions – he notes that we can confine ourselves to individual orderings deemed admissible by some such standard, provided that there remain three alternatives such that any ordering is possible453. Thus, suppose out of all alternatives a to z, certain preferences are inadmissible – for instance, person i cannot express a preference that j sleeps on his side. So long as there are three alternatives, x, y and z that can be ranked in any order – i.e. (x,y,z), (x,z,y), (y,x,z), (y,z,x), (z,x,y) or (z,y,x) – then Arrow’s proof will hold. This means that individual preferences are not restricted to, for example, those that are single-peaked. Preferences are single-peaked if options can be ordered along a single dimension, such as a left-right economic spectrum or amount of money spent, in such a way that all voters have a single ideal or ‘bliss point’ (which may be central or to either extreme) and less-prefer alternatives further away from it454. What this excludes is someone with V-shaped preferences, i.e. who prefers either extreme to the middle-ground, however options are ordered. It has long been known that if preferences were single-peaked then Condorcet’s majority-cycling could be avoided455. However, while it sometimes seems irrational to prefer either extreme to the middle ground, we cannot assume a priori that it is always so. Consider, for example, the book-buying decision from chapter 4.6. There, it could be reasonable for 450 451 Arrow (1984) p.15. Arrow (1963 [1951]) p.96. 452 Dworkin (1977) pp.234-8, Sen (1970) pp.87-8. 453 Arrow (1963 [1951]) p.24. 454 Black (1998 [1958]) pp.8-13. 455 Black (1998 [1958]) p.23 observes that, if voters’ “curves are singled-peaked, Omed. will be able to get a simple majority over any of the other motions a1,…, am put forward”. 210 someone to want high-spending or none at all – regarding low-spending as a waste of money, without achieving any benefit, and so literally worse than nothing – but this is not single-peaked along the dimension of money spent. If the spending possibilities high, low or none correspond to x, y and z, then it is possible for voters to express any of those orderings, and so cycling is possible. While I have rejected the populist idea that we are seeking the ‘will of the people’, any conception of democracy requires power to be in the hands of the people, and thus that the people should be able to say what to do with it. Coherence can be ensured if we restrict what preferences people are able to express, but to exclude certain orderings seems to violate something like the neutrality discussed above (section 6.5). As Riker says, “Any rule or command that prohibits a person from choosing some preference order is morally unacceptable (or at least unfair) from the point of view of democracy”456. If we want society to be responsive to individual preferences, then it seems that we cannot legitimately place restrictions on those preferences. Thus, the universal domain requirement seems reasonable. 6.8.b Lottery-Voting Assessed In assessing lottery-voting, it is important to remember that Arrow was seeking a social welfare function that defines a social preference ordering on the basis of individual preference orderings. Lottery-voting does not seek a social preference ordering, it merely seeks to establish one alternative as the winner. It seems, however, that it does allow individuals to express whatever preferences they wish – in that they can vote for x, y or z. One might, however, object that only considering voters’ first preferences does restrict what voters can express, since either (x,y,z) or (x,z,y) will simply come out as 456 Riker (1982) p.117. 211 a vote for x and we will not be able to tell them apart. In fact, however, there is no reason why we cannot allow voters to list as many preferences as they like – though ordinarily only the first we count, we could count the second preference if the first is deemed unconstitutional or otherwise eliminated from consideration457. Further, that only part of the voters’ preference ordering is used does not mean that our decision mechanism fails to operate for any preference ordering458. It is still the case that lottery-voting can deliver an outcome, whichever preference ordering individuals have. Moreover, lottery-voting is not unique in only considering first-preferences. The familiar British electoral rule, plurality-voting, for instance, only enables voters to pick a first-choice, and is vulnerable to cycles, manipulation (i.e. strategic misrepresentation of preferences) and violates IIA, since the entry of a third candidate C can change the results between A and B by taking more votes from the one who would have won otherwise, but it is held to satisfy the universal domain condition. Moreover, while comparing all options at once means that those whose first preference is for x cannot express a preference between y and z, there is something to be said for this. All voters are restricted to expressing a single preference, but all can say what they truly prefer, whereas a binary voting procedure may allow such a voter to choose between y and z but not register that she actually detests both. Once we exclude the possibility of cardinal comparison, the case for requiring a complete preference ordering is weak, since the way procedures like Borda treat lower-ranked preferences is effectively arbitrary and not obviously appropriate. In any case, voters may not even have clear preferences over low- or middle-ranked alternatives, which offers a further reason not to require such (though some voting rules merely allow them to specify such). Thus, while lottery-voting does not allow voters to express all 457 458 C.f. Arrow (1963 [1951]) p.26 on the death of a candidate. One may make a separate objection that it is unfair or sub-optimal to exclude second-preference information. 212 of their preferences (i.e. a complete ordering) – a property that it has in common with other non-positional decision rules – it does allow them to express any of their preferences and can produce a collective decision regardless of the individual orderings, so it satisfies the universal domain condition. (6.9) Pareto 6.9.a Plausibility Arrow’s Pareto condition was not in his original proof, but follows from his conditions 2' (positive association), 3 (IIA) and 4 (citizen sovereignty)459. Scepticism about the possibility of interpersonal comparison led Pareto to formulate a more modest principle of ordinal comparison – state x is better than state y if (or ‘if and only if’) it is better for at least one person and worse for no one460. This principle of comparison had been enshrined by economists seeking efficiency, though it is in fact very modest. State x may be one of great inequality, but may still be Pareto optimal provided that the poor cannot be made worse off without any loss to the rich461. This shows that Pareto optimality is not in fact a very strong condition, and is compatible with social states that seem intuitively undesirable. Moreover, the idea that a Pareto improvement always makes the social state better could be questioned. Suppose that we have the souls of Hitler and Mother Teresa in the afterlife, where Hitler is being punished for his deeds, and currently has a welfare of -10, while Mother Teresa is rewarded with +5. If it was possible to bestow an extra twenty units of welfare on Hitler, without making Mother Teresa 459 460 Arrow (1963 [1951]) pp.96-7. Arrow (1984) pp.122-3. 461 Arrow (1984) p.122. 213 worse off, resulting in a distribution (Hitler 10, Mother Teresa 5), that would be a Pareto-improvement, but few would believe the resultant social state preferable462. Arrow actually requires more than an improvement for one person to make the state socially-preferred. His Pareto condition is one of unanimity, “If xPiy for all i, then xPy”463, or “If alternative x is preferred to alternative y by every single individual according to his ordering, then the social ordering also ranks x above y”464. Unanimity is required rather than a single preference because Arrow’s positive association in fact only requires non-negative responsiveness, it is possible for x to improve in i’s ranking without becoming more-preferred socially465. Arrow therefore calls this the weak Pareto requirement, because it requires something stronger (i.e. the unanimous strong preference xPy; rather than merely unanimous xRy with one xPiy) for one state to count as better; so it is easier for a decision rule to respect it466. Note however that this would still count (Hitler 10, Mother Teresa 6) as a socially-preferred state, because better for both, so it is not entirely clear that it is always normatively compelling. It seems, however, that if the social welfare function is only responsive to individual preference orderings it cannot track impersonal good. Indeed, many have denied that a world can be better though better for no one467, and so would presumably think it no better that Hitler suffers. I cannot here attempt to resolve this controversy, so merely note the point of possible tension. Further, Arrow’s Pareto condition is weak in another way. It is not actually concerned with what is better but what people prefer. Being an economist, Arrow 462 This example modifies, but is inspired by, Temkin’s ‘saints and sinners’ example; see Temkin (1992) pp.260 and 273-7. 463 Arrow (1963 [1951]) p.96. 464 Arrow (1984) p.70. 465 Arrow (1963 [1951]) pp.25-6; c.f. May (1952) p.682, and McGann (2006) pp.18, and 62-3. 466 As we shall see, lottery-voting satisfies this weak Pareto requirement, though it does not satisfy the stronger one that social preference tracks any Pareto improvement. 467 Broome (1991) pp.165-73; c.f. Temkin (1992) pp.248-55, 256-8, and 277-81. 214 tends to conflate these, but it may be that Mother Teresa would not prefer the distribution (10,6) in the previous section, in which case Arrow is not committed to saying it is socially better, though it is actually better for each individual. I have argued that we should not think about decision-making in terms of betterness, and if we are trying to make a choice according to individual preferences, then it does seem compelling that we should choose what all people prefer. After all, voters’ preferences may fail to track what is good, but if everyone prefers x to y, and society’s preferences are to be a democratic function of individual preferences, then it would be odd for society not to prefer x to y. Suppose, for example, we considered restaurants in terms of quality of food, range of food, price, décor and location. If we prefer restaurant x to restaurant y for each of these five criteria, it would be very odd not to prefer x to y overall. Similarly, if I prefer x to y, you prefer x to y and everyone else prefers x to y, it would be odd not to say that we collectively prefer x to y468. In interpreting the Pareto condition, it is not assumed that a social preference or decision implies anything about what is actually better. A decision-procedure, such as flipping a coin between two options, x and y, is only supposed to tell us what to do, not what is best. If voting is conceived of likewise, then even if there is unanimity, xPiy and xPjy, we cannot infer that x is actually better than y. A fortiori, where there is disagreement, e.g. xPiy and yPjx, we cannot say anything about whether x or y is ‘socially better’, but – at best – that x is better for individual i and y is better for individual j469. The reason we can eliminate a Pareto-dominated option (y), however is because we have agreement, so this can be done without conflict or anyone’s 468 Of course, it is not always true that groups possess the properties of their parts. For instance, if I am large and you are large, the two of us do not make a large group. However, this example does not seem to commit the ‘fallacy of composition’. 469 ‘At best’ because a preference for x does not imply that x actually is better for one. C.f. MacKay (1980) p.52. 215 objection470. If all prefer x to y, then no one opposes the move from x to y471. Thus, if we are concerned only with making responsive decisions, we should require our decision procedure to respect Pareto in the weak sense only. 6.9.b Lottery-Voting Assessed It is easy to show that lottery-voting respects the weak Pareto condition. If all votes are for x, then necessarily any randomly selected vote will be for x. If we know everyone is in agreement, we need only ask a single random person, to give us a representative sample of the population472. Indeed, Rawls suggests such in his A Theory of Justice; assuming rational convergence (and hence unanimity) behind the veil of ignorance, he says, “we can view the agreement in the original position from the standpoint of one person selected at random”473. This is, to my knowledge, the closest he ever comes to endorsing a system like lottery-voting, and he does so only on the assumption of unanimity (Pareto). I think the Original Position is flawed because it assumes a single rationally correct answer that all would reach, if only any bias was removed. I think, if we recognize value pluralism, then we should realize that even reasonable and well-motivated deliberators could reach different conclusions (which is, I think, something Rawls moves towards in Political Liberalism). Nonetheless, if people hold different reasonable opinions about what 470 471 Bordes (1979) p.196ff. Note, however, that this reasoning would support a stronger Pareto requirement on the decision procedure: One that says so long as there exists an i such that xPiy, and for all others xRjy, then xPy. That is, one would only need a single individual to prefer x, so long as all others were at least indifferent – no one prefers y. Such a stronger requirement may be controversial, however: Would we really want to say a single person’s preference should decisively dictate a social preference for x if there were millions of other, all indifferent? I need not resolve this issue here; any reasoning that supports a stronger Pareto requirement will a fortiori support a weaker one. 472 Indeed, in such cases, one may think we do not need proper randomness: if everyone prefers x to y, then there is no problem with a dictator – she will prefer, and therefore implement, x, and so everyone will get what they want; c.f. Arrow (1963 [1951]) p.90. 473 Rawls (1999 [1971]) p.120; c.f. McGann (2006) p.162. 216 should be done, all equally permissible by justice, it would still be fair to select and implement one such conception of justice at random474. While lottery-voting may respect this weak Pareto requirement, however, it fails to respect a stronger version of the requirement. Suppose that xPiy while all other individuals are indifferent, xIjy. In this case, x is Pareto superior to y in Pareto’s original weak sense, because one person is better-off while no one else is worse off. If, however, some of those who are indifferent between x and y vote for y – perhaps to ensure that some third alternative, z, does not win475 – then the social choice may well be y. This shows that lottery-voting may be inefficient, and illustrates a good reason for deliberation (see sections 5.6-7, above) – because if i can inform people of her preferences before the vote, then those who are indifferent will presumably vote x. Nonetheless, if it happens that y is chosen because people voted y, this is not obviously unreasonable because we can only ever respond to people’s revealed, not actual, preferences or interests. Section 6.13, below, points out that lottery-voting has one advantage, in that it encourages the sincere expression of preferences. Moreover, we have already seen the sense in which it is positively response; if i prefers x to y then she can assign her chance to x, and her chance is all she can demand – there is no reason why society must prefer x, simply because it makes her better-off. In any case, the fairness of such cases is not the immediate concern. For now, all we need to note is that if there is universal agreement, lottery-voting will respect it, because the randomly-chosen individual must be part of a truly universal consensus. While Arrow’s unanimous Pareto condition is so weak that we may well want something stronger, it is at least uncontroversial because its only restriction on the 474 Though note I do not necessarily accept that matters of justice should be decided democratically or, therefore, by lottery-voting. 475 These people have the preference ordering xIjyPjz. To ensure the case really is a Pareto one, however, we may suppose that nobody actually prefers z to either x or y, though this wasn’t known to the voters at the time. 217 decision-procedure is that it must respect what all agree to. Any stronger version of Pareto could, I think, be contested and my aim here is only to show that lottery-voting meets the minimal conditions we should demand from a decision-procedure. Therefore it seems enough to show that lottery-voting satisfies Arrow’s Pareto requirement. (6.10) Independence of Irrelevant Alternatives Probably the most complicated and controversial of Arrow’s conditions is the Independence of Irrelevant Alternatives (IIA) condition – therefore this section proceeds in three parts, first clarifying the condition, before turning to its plausibility and an assessment of lottery-voting. 6.10.a Clarification Arrow states his Condition 3 as follows: “Let R1,…,Rn and R1',…,Rn' be two sets of individual orderings and let C(S) and C'(S) be the corresponding social choice functions. If, for all individuals i and all x and y in a given environment S, xRiy if and only if xRi'y, then C(S) and C'(S) are the same (independence of irrelevant alternatives)”476. As he puts the point less formally, “the social choice among a set of candidates should depend on the individual preferences for those candidates and those candidates only. In other words, the collective choice made from a given set of options should be invariant with respect to changes in the individual preferences concerning options outside the set”477. The intuitive idea has a long history, going 476 477 Arrow (1963 [1951]) p.27; c.f. Arrow (1984) p.16. Arrow (1984) p.51. Note that, though I am concerned with decision-making, Arrow seems to switch focus to electing representatives here, so for simplicity I follow him in some of this section. I do not consider this a major change, because the election of a representative is a direct decision taken by a constituency; however, in the final analysis, I recognize we may want to use different voting rules for electing representatives and making decisions, see e.g. McGann (2006) pp.18-24. 218 back at least to Condorcet’s complaints against the Borda system478; however, it is unclear which of two understandings of independence Condorcet was appealing to, and the confusion persists through to Arrow and discussion of his theorem479. While the underlying idea, that choice between x and y should not depend on z, is simple enough, specifying the condition, its rationale and its implications is far less straightforward. Radner and Marchak use Arrow’s term, independence of irrelevant alternatives, to describe Sen’s property α or contraction consistency480. This means that if x is chosen from set S then x should be chosen from any subset of S of which it is still a member. Arrow gives examples of this in his discussion, pointing to what would happen if one candidate were to die or be disqualified and claiming that said candidate should be blotted from the lists but (unless they are the winner), the result should not be changed481. Thus, if x is chosen from w, x, y and z, then this should not be changed by eliminating y; but ranking systems, such as the Borda count, may violate this, as Arrow illustrates482: Fig. 6.1: Initial Preference Rankings Rank (points) Voter 1 Voter 2 Voter 3 1st (4) x x z 2nd (3) Y Y W 3rd (2) Z Z X 4th (1) w w y 478 McLean (1995) pp.111-3, and 118. Condorcet protested that “The points [Borda] method confuses votes comparing Peter and Paul with those comparing either Peter or Paul to James. As long as it relies on irrelevant factors to form its judgments, it is bound to lead to error”, in McLean and Urken (1987) p.34; also quoted McGann (2006) p.21. 479 Arrow (1963 [1951]) pp.26-8, 110; the same confusion exists in McLean (1987) pp.167-8, though it is pointed out in McLean (1995) p.108 c.f. pp.109-10, and 118. 480 McLean (1995) p.108; c.f. Sen (1970) p.17. Consider a choice between cheese, chicken or tuna salad in a restaurant. If someone chose cheese, but then changed their mind to tuna on being told chicken was no longer available, we’d certainly think it odd… 481 Arrow (1963 [1951]) p.26. This claim can, of course, be disputed – for instance, Dummett (1984) p.57 argues that rankings can serve as an approximation of preference strength and that, even if one candidate were to die, keeping their name on the list gives us more relevant information. 482 Arrow (1963 [1951]) p.27. 219 In this case, w=5 points, x=10 points, y=7 points, and z=8 points, so x wins. Yet if y were eliminated, then the results would be different: Fig. 6.2: Candidate Death: y Eliminated Rank (points) Voter 1 Voter 2 Voter 3 1st (3) X X Z 2nd (2) z z w 3rd (1) w w x Now w=4 points, x=7 points, and z=7 points, so x and z tie! This is indeed one form of independence – which, following McLean, we might term IIA(RM)483. In fact, however, there is a sense stronger than contraction and expansion consistency484. Arrow wants the choice to be totally independent of what happens to other options. As he says, “Condition 3 tells us that the choice between x and y is determined solely by the preferences of the members of the community as between x and y. That is, if we know which members of the community prefer x to y, which are indifferent, and which prefer y to x, then we know what choice the community makes”485. This means that choice between x and z should be unaffected not only be the presence of y but also the relative rankings of y486. We can therefore give another example where, rather than y dying, y remained in the ballot, but scandalous accusations emerged about his personal life, so he dropped to the bottom of everyone’s preference rankings: Fig. 6.3: Scandalous Accusations: y Dis-preferred Rank (points) 483 484 1st (4) 2nd (3) 3rd (2) 4th (1) McLean (1995) p.108. For these, see Sen (1970) p.17. If someone is now told there’s also ham salad, this should not make them prefer tuna to cheese either (though they may, of course, prefer ham to either). 485 Arrow (1963 [1951]) p.28. Somewhat confusingly, this follows his example where the elimination of y is affecting the choice between x and z. 486 Continuing the salad example, if someone is told that tuna, chicken and cheese are indeed available, but that the chicken isn’t very nice, that shouldn’t affect their preferences between tuna and cheese either. 220 Voter 1 Voter 2 Voter 3 X X z Z Z W W W X Y Y Y Now w=7 points, x=10 points, y=3 points and z=10 points. Again, this results in x and z tying. While contraction consistency requires that, if xPz from any S, then xPz from any subset of S, of which x and z are members, IIA(A)487 requires that if xPz then this remains so whatever happens to y – whether y moves up or down people’s preferences, or is eliminated altogether. IIA(A) is therefore stronger than even the conjunction of contraction and expansion consistency, and has the consequence that we can choose between x and z knowing only people’s preferences between x and z: we do not need to know anything about people’s preferences over other options; in other words, the social ranking is decomposable into pairwise choices488. This is because Arrow assumes a complete ordering over all logically possible alternatives, which is then applied to those feasible options to hand. For instance, if I rank all logically possible alternatives u-z of which the feasible options are x, y and z then all that matters in my ordering are the positions of x, y and z, and any change in my preference over non-feasible options u or v should make no difference. If we have only ordinal preferences, then there is no difference between x,u,v,y,z and x,y,u,v,z. It is to the plausibility of these assumptions that we now turn, although comments on Arrow’s notion of rational choice are reserved for chapter 7.7-9. 6.10.b Plausibility Arrow finds IIA attractive because he assumes that y should not affect the outcome between x and z since it is irrelevant to this comparison. One might, however, object that it simply begs the question to suppose that third alternatives are 487 488 Again, following McLean (1995) p.108. This is, of course, how a Condorcet winner is identified, which is appropriate given Condorcet’s support for IIA, cited above. 221 irrelevant. Some proponents of ranking systems, for example, defend them on grounds that – though imperfect – they offer some indication of preference intensity, which is relevant information489. Certainly, the fact that everyone gets their second preference does not mean that they are equally satisfied – for example, suppose two conservatives and one socialist were running, conservative voters would regard their second preference as little worse than their first, while socialists would regard their second preference as little better than their worst. Nonetheless, some argue, it is better to use preference rankings as an imperfect proxy for intensity than be without such information altogether. There are occasions where we think that comparison to third alternatives is relevant. Perhaps, instead of thinking of finishing orders in a race, we should think of another example, say a football league table. The fact that x finishes above z in the final table depends not only on matches between x and z, but how each performs against all the other teams; thus it is quite possible for x to be above z in the table despite losing the matches (pairwise comparisons) between the two. This makes sense to us because results between x and z are not the only information we have with which to judge the two teams; here it does seem reasonable to consider the fact that x beat y comfortably, while z could only draw490. The same reasoning carries over to voting: as McLean and Urken remark, in response to Condorcet’s criticism of comparing Peter and Paul to James rather than each other (quoted above), “when v [vérité] is not much greater than 0.5, the Borda rule is more likely to select the correct winner than is a search for the Condorcet winner, because Paul’s total score counts 489 490 Dummett (1984) pp.38-8 and 50-7, and Saari (2001) passim, especially pp.81, 179, 190-3. Thus elimination of y could move us from x>z to x=z, as in the tables from the previous sub-section. 222 for more than Peter’s precarious and unreliable majority over him”491. While this reasoning assumes (with Arrow) that there is, in fact, some unique ‘best’ option that we are trying to identify – something denied here – it does suggest that comparisons to y may be relevant in comparing x and z. Further, it should be noted that any actual voting system that compares three or more alternatives while only allowing voters to express a single preference appears to violate both senses of IIA. Suppose I rank m,n,o,p and that, in the social choice, m beats o. Had n moved up my preference ranking, however, so my ordering became n,m,o,p and correspondingly I voted for n, then the system would not register that I prefer m to o and this may affect the overall outcomes. This appears to violate IIA(A). If other voters rank p,o,n,m and then p withdraws, their vote will switch to o, which may lead to o beating m and violates IIA(RM). Much, however, depends on what is taken to be relevant. Arrow supports IIA with examples, for instance pointing out that a community’s choice between building a stadium and a museum should not be affected by preferences over a university if that is not feasible492. If the university is feasible, however, under a first-preference plurality voting rule, it may detract enough support from the museum that the stadium wins when it otherwise wouldn’t have done so, violating expansion consistency. One might, at this point, offer a more pragmatic justification for IIA – if outcomes can be affected by the entry or removal of other candidates, then there is a danger that they can be manipulated. In the example from tables 6.9.1-2, for instance, y could choose to drop out, so that z tied with x. The fact that the Borda count could 491 McLean and Urken (1995) p.34. In Condorcet’s formula, vérité is the probability of each individual’s judgement being right. Where e (erreur) = v-1 and h and k respectively the majority and minority, then the probability that the h voters are right is: h-k vh-k . v +eh-k See, e.g., McLean and Urken (1995) p.28, and Levine (2002) p.87. 492 Arrow (1984) pp.51-2, and 70-1. 223 be manipulated not only by misrepresentation of preference rankings but the entry of spurious ‘stalking horse’ candidates has long been recognized493. One reason to make the choice between x and z depend only on their respective rankings is that it eliminates such worries. As McGann argues, however, that this is primarily a worry when it comes to decision-making contexts, where there are a potentially infinite number of alternatives494. Ranking systems may be appropriate when the number of alternatives is relatively fixed, as in football leagues, comparing job applicants or electing representatives. Further, he points out that such binary independence seems inappropriate if we are concerned with distributive matters, such as allocating seats495. The entry of a New Left party into an election should not leave the distribution of seats between the original Left and Right parties unaffected – rather they should take support from those close to them. While lottery-voting is a decisionmechanism, it is also concerned with a distributive issue – who gets what their way – so similar considerations apply. It is far from obvious that we want our decisionmechanism to be unaffected by the range of options available, so it is unclear that IIA, as Arrow defines it, really is a minimal condition of choice, at least applied to such decision-mechanisms496. 6.10.c Lottery-Voting Assessed It is difficult to assess lottery-voting by the principle of IIA, not only because of the ambiguity of this requirement, but because it is designed for deterministic mechanisms. As Arrow says, “Condition 3 tells us that the choice between x and y is 493 494 E.g. MacKay (1980) p.36. McGann (2006) p.23. 495 McGann (2006) p.24. 496 MacKay (1980) is more cautious, speaking of “a defeasible requirement of desirability” (p.31) and “a weak presumption in its favor” (p.35), but admitting that it does not, at least prima facie, seem a necessary condition of social choice, and introducing a weaker ‘ordinality’ (pp.99-100). Indeed, Arrow himself admits that it is the condition he is most inclined to weaken or drop, see Arrow (1963 [1951]) p.110 and, more explicitly, Arrow (1984) p.76. 224 determined solely by the preferences of the members of the community as between x and y. That is, if we know which members of the community prefer x to y, which are indifferent, and which prefer y to x, then we know what choice the community makes”497. This is the simplest case, where R1,…,Rn = R1',…,Rn', and is obviously violated by lottery-voting, since the need for a randomizing element means that we can never predict the choice simply from knowing individual preferences, and may indeed get different choices on two occasions from the same profile of preferences. Thus IIA technically excludes all random mechanisms. Riker explicitly notes that randomness is excluded, observing that IIA requires that: “[A] method of amalgamation, F, picks the same alternative as the social choice every time F is applied to the same profile, D… From the democratic point of view, one wants to base the outcome on the voters’ judgments, but doing so is clearly impossible if the method of amalgamation gives different results from identical profiles… Then it is the device, not voters’ judgements in D, that determines outcomes… [which is] simply a way to by-pass voters’ preferences”498 While he is right about the consistency requirements of IIA, it is unclear why he thinks that the lottery simply by-passes voters’ preferences; it is, rather, a fair way of choosing between them. It is true that the chosen outcome will not depend only on these preferences, but lottery-voting need be more embarrassed by the fact the same procedure could produce a different result than by the fact that a different procedure could produce a different result (though there are practical problems taken up in sections 5.8-10 and 7.4-5)499. Riker claims that “[i]f, for any choice that is supposedly fair because it comes out of a fair procedure, there is another choice from another procedure that is fair in a different and conflicting way, then it is difficult to justify 497 498 Arrow (1984) p.17. Riker (1982) p.118. 499 Riker (1982) pp.21-40 and 58-113, Shepsle and Bonchek (1997) pp.167-71. 225 the fairness of any choice”500. However, while this may be an effective criticism of any procedure that claims to uncover ‘the will of the people’, it seems to miss the point of pure procedures. Tossing a coin between two equal claimants is plainly fair, precisely because it can produce different outcomes. If we are effectively dealing with a case of pure procedural justice501, as argued above, then we merely need to designate a fair procedure and accept whatever outcomes it produces. While lotteryvoting may be condemned as ‘arbitrary’, what Arrow and Riker have in fact shown is the implicit arbitrariness in any choice procedure, so this can be no objection to lottery-voting. The question remains, however, whether the normative considerations justifying IIA should lead us to reject any random mechanism. Setting aside the problem of lottery-voting’s indeterminacy, let us consider whether it is indeed affected by ‘irrelevant alternatives’. What seems the logical extension of IIA to probabilistic electoral methods is to say that x’s chances of victory should not be affected by irrelevant third alternatives, but the question remains as to what alternatives are relevant. If we are considering a binary election between x and z, then lottery-voting indeed holds that only preferences between x and z are relevant (although, as said, they do not alone determine the social decision – the lottery will also play a part). Preferences regarding any third option, y, which is not in fact under consideration are irrelevant to the decision. Lottery-voting, however, holds that y becomes relevant if and when y enters the contest; if voters are now faced with the three-way choice, then y can affect the relative chances of x and z winning. It does not, however, seem normatively problematic that one alternative being available can detract support from another. To use McGann’s example, suppose that 500 501 Riker (1982) pp.58-9. Rawls (1999 [1971]) pp.74-5. 226 initially 60% of the population supported the Left candidate x, and 40% the Right candidate z, so x is more likely to win502. Now, if a New Left candidate y enters the race, assume that the left-wing vote is evenly split – so x and y each have 30% of the total vote. In this case, x is now less likely to win than z, which would seem to violate IIA because their chances of success have been affected by the third party. However, while this is a violation of IIA, it seems to be the fairest outcome. We presumably would not want to give Left and New Left separate chances, as that would make the procedure manipulable and give parties incentive to fragment (see 3.10 and 4.5, above). Rather, we should want left-wing support to be split between the relevant parties, which is what happens. Lottery-voting means that the left-wing parties, taken together, still have a 60% chance of winning, but this chance is divided between those who support New Left and those that still back the original Left candidate. Moreover, if we were to ask voters to rank all three alternatives, and subsequently found New Left ineligible and redistributed those votes to Left, then we would presumably find that the contest between Left and Right was unaffected. As McGann concludes, “Binary independence [IIA] is not a quality that we should require of seat allocation rules. Indeed, it is a quality that any reasonable seat allocation rule should violate”503. McGann’s argument concerns distributing seats, but if democracy is generally conceived of in terms of distributing, rather than maximizing, power, then parallel conclusions apply to cases of direct decision-making504. 502 McGann (2006) pp.22-4 argues that we require binary independence in decision-making because otherwise we are vulnerable to manipulation since there are infinite possibilities that may be introduced. Lottery-voting, however, limits alternatives to those people will vote for and is, as I show in section 6.13, below, non-manipulable, so these problems do not arise. 503 McGann (2006) p.24. 504 Using the book-buying example from section 4.6, it seems right that the proposals £100 and £101 should split the votes of those whose ideal point is in that region, since giving them each an independent chance would effectively give someone whose ideal was £100.50 twice the chance of getting an outcome that was practically her ideal (see again chapter 3.10 and 4.5, above). 227 One problem with lottery-voting is that it does not allow us to decompose a three-way choice into pairwise comparisons – knowing that x beats y and y beats z isn’t enough to know that x beats z. Arrow demands that this should be so, because he is seeking a transitive ordering – if x beats y in a race and y beats z in the same race then x must also finish before z – thus, “The overall ranking of any given set of competitors depends only on the order of finish of those competitors”505. However, the idea of a race or transitive betterness ordering does not seem to be the most appropriate way to think about democratic decision-making; what we should be concerned with is a fair distribution of decision-making power. If our aim is to distribute this power fairly, then it does not seem that the choice between x and z should be independent of the presence of y – for, if some voters choose to bestow their chance on y instead of either x or z, then obviously things should not be left as they were between x and z. Thus, while lottery-voting does violate Arrow’s formal condition, it does so because of different assumptions about the nature and purpose of democracy, and in a way that seems not only normatively unproblematic but most reasonable. (6.11) Non-Dictatorship 6.11.a Plausibility Arrow’s Condition 5 is that the social welfare function not be dictatorial. Since we are dealing with democracy, this seems uncontroversial – indeed, it seems definitional that democracy (rule of the people) involves many, whereas a dictatorship 505 MacKay (1980) p.31. 228 is literally monarchical (rule of the one)506. Once one person always determines the ‘social’ ordering we seem to have left behind the area that we are interested in507. There might be many reasons why a dictatorship may be advantageous, at least in certain circumstances (e.g. emergency powers), but we are no longer concerned with collective choice as understood here – we would be dealing only with making choices for a collective, rather than decisions by a collective. Mere verbal agreement that the rule should be non-dictatorial is not enough, however, we need to know what this means in practice. Arrow elaborates his condition by adding: “A social welfare function is said to be dictatorial if there exists and individual i such that, for all x and y, xPiy implies xPy regardless of the orderings R1,…,Rn of all individuals other than i, where P is the social preference relation corresponding to R1,…,Rn”508 The key point here is that an Arrovian dictator decides for all x and y, her preferences determine all issues. One may perhaps want to say that if i can solely determine the choice over a single issue (say, kPil implies kPl even if all other individuals rank lPjk) then i is a dictator over that issue, but this is not how Arrow uses the term. MacKay describes such rules, that allow one person’s preferences to prevail over all others, as ‘vicious’509 but – despite this morally-loaded term – it is not clear ‘vicious’ rules are necessarily bad; those who accept some form of liberal private sphere or selfregarding rights510 generally accept that each individual should be decisive over some set of decisions. In any case, for one individual to be able to decide a given issue is not dictatorial in Arrow’s technical sense. Again, therefore this is a weak condition, 506 Aristotle (1988) p.61 [III.7 1279a26-40]; c.f. Hobbes (1985 [1651]) pp.239-40 [original pp.94-5, ch.19]. 507 Pratchett (1987) p.176, fn: “Ankh-Morpork had dallied with many forms of government and had ended up with that form of democracy known as One Man, One Vote. The Patrician was the Man; he had the Vote”. 508 Arrow (1963 [1951]) p.30 509 MacKay (1980) p.105. 510 C.f. Sen (1970) pp.79-81, and 87-8. 229 leaving open the possibility that i decides for all x and y, except for k and l. Two further details remain: Firstly, what if everyone wants a dictator, and secondly how is the conditional implication to be understood? On the first, Arrow says, “the desires of those individuals include a liking for having social decisions made by a dictator or at least a liking for the particular social decisions which they expect the dictator to make”511. In Chapter VII, he goes on to briefly consider the possibility of higher-order preferences for a decision-mechanism conceived of as a value itself – for example, whatever our first-order preferences between x and y, we might prefer x-arrived-at-democratically to y-arrived-at-in-otherways. In such circumstances, we cannot rule out that “the desire for a dictatorship or for a particular dictator may be overwhelming”, in which case “our social welfare problem may be regarded as solved since the unanimous agreement on the decision process may resolve the conflicts as to the decisions themselves”512. It is, of course, quite possible that all would prefer to defer to an authority they accept, e.g. the Archbishop of Canterbury513, and this can be understood either as them having a higher-order preference for what he decrees, or as them revising their lower-order preferences between x and y in light of what he says (‘I must have been wrong, x is better for me after all’). I think Arrow is right to regard such as an acceptable solution to the social choice problem, however. While they may formally appear to be dictatorships – because the Archbishop’s preference over x and y determines the social ordering – this is so only because others (unanimously) accept the authority514. 511 512 Arrow (1963 [1951]) p.30. Both Arrow (1963 [1951]) p.90. 513 This example comes from Barry (1991) pp.30, and 60. ‘Authority’ can be understood in terms of Raz (1985) pp.9-18 and Raz (1986) p.21ff; but note deference occurs to those that other people think are authorities, it does not necessarily follow that the chosen dictator is better at decision-making, merely accepted as so. 514 Ross (1952) pp.81-2 points out that all accepting an absolute monarch doesn’t make such democratic, but it can still be a legitimate solution to the social choice problem – Arrow (1963 [1951]) 230 As such, it seems a consequence of Pareto that the social choice must conform to these preferences515. This brings us to the related point, how one should understand Arrow’s “if there exists and individual i such that, for all x and y, x Pi y implies x P y”516. One natural understanding is as a material conditional: it is true iff in all circumstances where xPiy then xPy. Sen, however, points out that Arrow’s formulation is ambiguous, and would be satisfied by an individual indifferent on all issues, since the antecedent is never met517. Moreover, this misses the ordinary idea of causation. We think i is a dictator if it is because xPiy that xPy; we do not necessarily think there is anything untoward simply because whenever xPiy it is also the case that xPy. Consider the example of an individual with higher-order democratic preferences; it may be that she can transform her preferences, perfectly and instantaneously, so that whenever xPy (as the democratic outcome of social choice), her preferences are xPiy518. In this case, it will also be true that whenever xPiy so also xPy, but this is not intuitively a dictatorship519. The implication that Arrow is talking about has to be causal; that is, xPy because xPiy, rather than vice versa, so if i changed to yPix then the social decision would also change (violating May’s anonymity). Understood in this sense, it is clear that – subject to the considerations of the previous paragraph – a social choice procedure must be non-dictatorial, or it is not a social (collective) choice at all. 6.11.b Lottery-Voting Assessed p.90 – and, provided this consent is revocable, it does ensure decisions are mediated through everyone’s reason. 515 Arrow (1963 [1951]) p.74 recognizes that “the condition of nondictatorship loses its intrinsic desirability” in certain conditions, e.g. complete unanimity. Thus his objection to dictatorship is not absolute. And, as I will argue, the ‘random dictator’ of lottery-voting is not an Arrovian dictator, for he holds if an individual is decisive on a pair of issues they will be on all (pp.99-100), but this isn’t so if the decisive individual is randomly determined for each vote. 516 Arrow (1963 [1951]) p.30. 517 Sen (1970) p.42. 518 This may have been Rousseau’s ideal, see Rousseau (1997 [1762]) p.124 [Social Contract IV.2]. 519 Risse (2001) p.710, fn.6 claims that genuine dictatorship cannot be expressed in Arrow’s formal framework. 231 In lottery-voting, a single person is picked to decide a given issue. Since this effectively disregards all other preferences, the rule is what MacKay would call vicious. Despite the fact that this is often informally called a ‘random dictator’ method520 though it is not dictatorial in Arrow’s sense: there is no individual whose preferences automatically determine all social choices. While i might be drawn to decide this issue, next time it might be j. Arrow’s proof involves the lemma that any individual or group that is decisive over one decision must be so over all521, and therefore dictatorial in his strong sense; yet the random lottery means, even though there is a sense in which the one drawn is a ‘quasi-dictator’, no one will be so over all decisions and each has an equal chance to be so on each occasion. Random selection is not, however, taking turns. While over a large number of persons and issues, it is likely power will change hands regularly – even if not everyone gets a turn – we might worry that one individual might receive more than her fair share of power as a result of mere fluke. For instance, suppose we have ten individuals (a-j) and ten issues to resolve. It might be convenient if each individual got to decide one of the issues522. If we leave who decides to random chance, however, then it is logically possible that an individual – say, d – may get to decide all ten issues523. Does this mean that, because on each issue xPdy determines the social choice, d is a dictator? No, this is merely a case where d’s preferences happen to coincide with the social choice, but that does not make d a dictator – for d’s preferences would also coincide with all the social choices if there was unanimity over each issue. For d to be a dictator, her preferences have to prevail not only 520 521 E.g. Estlund (1997) pp.191-4. Arrow (1963 [1951]) pp.54-5, and 98-100. 522 Though, of course, different issues might differ in importance to different people; and some may be more concerned simply that they get their way (e.g. x) and not that they get it because it is their way (or their vote is decisive). 523 The probability is (0.1)10. 232 whatever others’ preferences were but whatever the lottery-outcome had been524. Even if it happens that d’s preferences do determine each and every social choice, it is not the case that they always do so simply because they are d’s, since anonymity is respected (see 6.4, above)525. It could have been that someone else’s vote had been drawn on any of the issues. Thus, even in this vanishingly rare possibility, lotteryvoting does not violate non-dictatorship; for while only one person’s vote ultimately counts, it gives each an equal chance of being the decisive one over any given issue. While the fact that lottery-voting may be called a ‘random dictatorship’ may lead to prima facie suspicion that it violates Arrow’s axiom, this is not in fact so because no one need decide all issues. The normative force of this axiom derives from the idea of equality, and that is satisfied by the equal random chances. Since we have already seen that lottery-voting satisfies anonymity, no individual can be a dictator. (6.12) Arrovian Conditions Concluded Assessing lottery-voting by Arrow’s conditions has been complicated by the fact that lottery-voting does not aim to do what Arrow requires a social welfare function to do: lottery-voting is merely a practical decision-mechanism and, because it is concerned with distributing decision-making power, it does not aspire to a complete and transitive betterness ordering. Nonetheless, if Arrow’s axioms are really normatively attractive minimal conditions to impose on a social welfare function, then one might at least expect them to be desirable properties of other decision- This comes close to saying d’s preferences must prevail in all possible worlds, but that is not quite accurate as d is not a dictator in all possible worlds. 525 Further, note that since ballots will not have voters’ names on, no one need ever know it was in fact d’s preferences that decided any issue. This would be knowable only if it could be found that, on each decision, the other nine all voted the other way. 524 233 mechanisms too. Indeed, this generally seems so; certainly in so far as we are democrats, we want our decision-mechanism to be non-dictatorial (by definition), to allow voters to express any preferences they like526 and to respect unanimous preferences – all conditions fairly obviously satisfied by lottery-voting. While lotteryvoting does allow a single individual to decide any given issue, the randomizing element ensures equality because all have an equal chance of being decisive for any issue and no one is guaranteed to be so (or, in practice, will be for all issues). Further, the procedure allows each individual to express any preference they like, and to have their 1/n chance of getting their way, yet if all are unanimous then the decision will be the same, no matter whose vote is picked. The problematic condition for lottery-voting is IIA. However, this failure is simply a consequence of the different aims of a decision procedure and true also of many other not obviously implausible procedures, such as the Borda count. While, if we are after a betterness ordering, it makes sense that Left and New Left will be ranked similarly (if ranked similarly by voters); but, when it comes to allocating chances of victory, we do want the presence of New Left to diminish those of Left (assuming it takes votes from Left). There is thus nothing normatively problematic about lottery-voting in this regard; the desirability of Arrow’s axiom simply does not extend across to decision-procedures in this case. Taken together, my assessment of lottery-voting by May’s and Arrow’s conditions implies that it does satisfy the minimal properties we should require of a decision-mechanism. There is one major further consideration to discuss, however – whether it is a rational decision-procedure. This demands longer discussion, which is postponed to the next chapter. Before concluding this one, however, I want to point to 526 Of course, we may hope that voters do not express certain preferences that we consider irrational, undesirable, etc. 234 two further advantages of lottery-voting that are of interest to those concerned with social choice mechanisms: firstly, that it is non-manipulable and, secondly, it facilitates weighted-voting, should we want to give some more power than others. (6.13) Non-Manipulability We saw above (in section 6.10) that one motivation for IIA may be to avoid strategic manipulation of the decision-procedure, as when one voter deliberately misrepresents their preferences in order to obtain a more-preferred outcome. Further, it was argued that lottery-voting’s failure to satisfy IIA was normatively unproblematic, one may worry that it will consequently be manipulable, and that this is an undesirable property for a decision-procedure. This section will argue that one advantage of lottery-voting is actually its non-manipulability. As we saw above, the introduction of a new alternative must, if it is to have any chance, take that chance away from existing ideologically-similar alternatives, and there is no reason for any voter to vote for other than their most-preferred option, since their vote bestows the same chance on whatever it is they vote for, so to do so would merely mean that if their vote is chosen they get a less good outcome than they could have had if they voted sincerely. Gibbard and Satterthwaite independently proved the susceptibility of almost all voting procedures to manipulation. As McLean summarizes: “A voting procedure would be strategy-proof if it satisfied the following conditions. For all individual preference profiles it would ensure that whenever an option became more popular its chances of success would at least get no less; and it would ensure that the result could not be manipulated by adding or withdrawing options. But this turns out to be the same as saying that such a procedure must satisfy conditions U, P, and I in Arrow’s Theorem. Therefore if there are 235 more than two options any strategy-proof voting procedure might throw up a dictator”527. Whatever one thinks about the widespread existence or otherwise of cycling and manipulation of alternatives528, it is well-known that strategic-voting is common. In British General Elections, for example, if a voter ranks the parties (Labour, Lib Dem, Conservative), but thinks that the Lib Dems and Conservatives are the two most likely winners, he can vote Lib Dem – effectively reporting (Lib Dem, Labour, Conservative). This is by no means unique to FPTP plurality elections. Indeed, the Borda count is even more susceptible, as here voters give a complete ordering, so there is more scope for shifting an option up our down one’s ranking. One might report (Lib Dem, Labour, Green, UKIP, BNP, Conservative) – even though one actually preferred the Conservatives to UKIP or the BNP – if one thought they were a bigger threat, to minimize their points score. Though these examples focus on electing representatives, where ‘tactical voting’ is well-known, they do not rely on anything distinctive about that case, and similar examples could be invented of other direct decisions, e.g. choosing a restaurant for a group meal. If the choice was between Indian (I), Chinese (C), pizza (P) or McDonalds (M), for example, then someone who ranked them I,P,C,M, but was fairly certain M would lose (because of being low in others’ preferences) but worried C might win, could decrease the chances of C winning by reporting the ranking I,P,M,C. Strategic-voting may be less familiar in such contexts only because there is often no formal decision mechanism or voting. Nonetheless, even if the group reach their decision simply by discussion and agreeing on a reasonable compromise, 527 McLean (1987) p.168. Note that it is in fact expansion/contraction consistency appealed to here, not IIA(A). C.f. McLean (1995) p.108, and Sen (1970) p.17, fn. 528 For varying views, see Riker (1988 [1982]) pp.17-8, 158-9, 186-8, Mackie (2003) pp.197-377, Regenwetter et al (2006) pp.23-51, and McGann (2006) pp.70-6. 236 one could still influence the decision by misrepresenting one’s preferences – e.g. saying ‘not Chinese, I’d even prefer McDonald’s to that’. The overall consequences of strategic-voting are unclear. Sometimes coordinated strategic voting (e.g. log-rolling) can produce better outcomes for all involved in it, and maybe for everyone. Alternatively, winners can gain at the expense of losers and, if what they gain is less than what others lose, then the overall consequences can be bad529. Aside from that, the overall consequence, the purpose of strategic-voting is for one person to get what they want, when they otherwise would not have done so, which raises the worry that it gives manipulators greater influence, undermining political equality and unfairly distorting winners and losers. Theoretically, perfectly-informed strategic voters on either side can cancel out, but in practice those who are better-informed will be better able to manipulate the system to get their way. Not everyone is so concerned by strategic-voting – provided everyone has a vote, that they make differential use of it does not obviously undermine equality530 – some people fail to make any use of their vote by abstaining, while others fail to maximize its use by ‘unsophisticated voting’. If one wanted to draw an analogy, one could suppose all were given equal sums of money, and some spent it frivolously, while others invested it well or badly. Nonetheless, many have thought that it would be desirable to prevent strategic manipulation, and even those who do not regard it as particularly problematic would not necessarily complain if we designed institutions to make it impossible (this would be like forcing all investments to pay the same interest, to avoid some people investing better than others). 529 Mackie (2003) p.65 adds “the possibility that strategic voters might outfox one another and unintentionally end up selecting an outcome that almost no one wants”. 530 C.f. Rawls (1999 [1971]) p.179, where he distinguishes between equal liberties for all and the unequal worth of such liberty. 237 Lottery-voting easily solves any problem here, as Mackie points out, after quoting Hinich and Munger’s summary of the Gibbard-Satterthwaite theorem, ““No voting rule that can predictably choose one outcome from many alternatives is strategy-proof unless it is dictatorial” (A voting lottery such that each voter marks a ballot and then one ballot is chosen randomly is strategy-proof, here there is no reason for Deborah to mark her ballot insincerely, thus the Gibbard-Satterthwaite theorem is limited to predictable voting rules.)”531. Riker too notes the possibility of probabilistic solutions532, and rejects them because they violate the Independence axiom, but we have seen it is not clear why this should be desired of a decisionprocedure, especially if strategic manipulation can be avoided without it. Since every vote for A increases A’s chances, there is no possible way a voter can be better off by voting for B if he sincerely prefers A. Since every vote, for any option, increases the chances of that option winning, there is nothing to be gained from voting for one’s ‘second choice’533, or from introducing a new alternative that one does not prefer to those on the table. As we saw in discussing IIA above (section 6.10) a new alternative is likely to take votes from those that were already similar, so if the Left introduce a New Left alternative it will not increase the overall chances of their winning, while a New Right option, though it will split the right-wing vote, will not diminish the overall chances of a right-wing solution. Note, however, that there is still the possibility of collective manipulation, e.g. log-rolling. That is, one group who want policy a and do not care much either way about policy b may agree (with those who want b and do not care so much about a) to vote for b if those others vote for a. Arguably, compromise is best promoted by such 531 Mackie (2003) p.161. The quotation is from Hinich and Munger (1997) Analytical Politics (Cambridge: Cambridge UP) p.165. 532 Riker (1982) p.143. 533 The one exception is if we impose a minimum threshold, as suggested in section 5.3b, and one’s first-preference is unlikely to meet it. 238 discussion rather than voting rules (see 3.13) and this bargaining is simply part of trying to build up the widest support possible for a given proposal, nor is there anything obviously wrong with such arrangements. Therefore, lottery-voting can accommodate such a possibility, as described in sections 2.5c-d and 5.8, above, because such collaboration may be defensible on the grounds that it allows some influence for intensities and produces better overall decisions534. Moreover, it may even promote such trades, as it is no longer enough to have just 50%+1 votes, so the incentive is there to build the widest support possible. Some do worry, however, that not all groups will have ready collaborators and that log-rolling permits greater influence to organized interest groups. If this is so, however, then it is easy to forestall attempts at such collaboration by making voting and the results secret (see 5.9), which means neither group can ever know for sure whether the others complied with the bargain, and so trust is undermined because each has an incentive to defect when the vote comes. Thus, one can design lottery-voting institutions to either permit or discourage log-rolling, depending on how one assesses its effects. Any other manipulation, based on misrepresenting preferences or introducing other alternatives, though, is seemingly doomed to fail, which seems to be a significant advantage of lottery-voting in decision-making. (6.14) Weighted-Voting Finally, before concluding the present chapter, there is one final potential advantage of lottery-voting that is worth pointing out – viz. that it facilitates weighted-voting. While proposals for weighted-voting have a long and distinguished 534 Buchanan and Tullock (1962) pp.131-45. 239 history, e.g. Aristotle535 and J. S. Mill536, it generally has undemocratic associations. That is not, however, necessary – there can be occasions where some form of ‘proportionate equality’537 seems appropriate, e.g. if decision-makers are representatives, we may want to weight their vote according to the numbers they represent538, or we may think that those with greater interests in a particular issue deserve greater say over it539, as when share-holders have differential control over companies proportional to their share holdings. Thus, weighted voting has been recommended by some democratic theories, as discussed in chapter 2.5b, above. Whatever we think of its normative justification, weighted-voting has been used in practice, for instance in 1958, the EEC employed weighted voting to reflect differences in size and economic power between its member countries – France (4), Germany (4), Italy (4), Belgium (2), Netherlands (2) and Luxembourg (1). Political scientists studying such institutions, however, have found that weighting votes intuitively, e.g. according to population or income, can produce surprising results. It is easy to see that if votes are split 3, 3, 3, and 1 (with 6 needed for a motion to pass) the fourth person has no voting power – they are never vital to the passing of any motion, as it always requires at least two of the others, who are sufficient themselves. The above EEC case was effectively like this, twelve votes were needed to pass a motion, which required, as a minimum, either the three large countries (France, Germany and Italy), or two of those plus Belgium and the Netherlands. Luxembourg was never pivotal to a winning coalition; because all the others had an even numbers of votes, they could never combine to give eleven (i.e. one short of the threshold), so 535 536 Aristotle (1988) pp.145-6 [Politics VI.iii (1318a11-18b5)]. Mill (1998 [1861]b) pp.334-9 [Considerations on Representative Government ch.8]. 537 Aristotle (1999 [1985]) pp.71-2 [Nicomachean Ethics V.3 (1131a21-b20)]. 538 Either between national constituencies or different countries in internal bodies. 539 See the proposals by Brighouse and Fleurbaey (2006) and Heyd and Segal (2006), discussed in 2.5b, above. 240 any winning coalition including Luxembourg would still be winning without. We can also invent even more fanciful examples of weighted-voting gone wrong, for instance a division of votes 50/49/1, with 51 (a majority) needed to pass bills. Here we notice that any winning coalition must include the first voter (since 49+1=50, not enough). However, either the second or third voter is enough in addition to the first (i.e. it does not matter whether we have 50+49=99 or 50+1=51, either passes). Thus we see that, despite the very unequal weights, voters two and three actually have just as much voting power as each other! Further, the effects of changing voting weights are hard to predict intuitively. Suppose that we double the third voter’s weight to two. If we increase the winning threshold to 52, then we have made no difference to voting power, as victory still requires the first voter and either of the other two. On the other hand, if we leave it at 51, then now voters two and three are sufficient, so any two of the three can pass a bill and voting powers are equal! One way of instituting an adequate weighted-voting system would be to find some measure of voting power, and weight the votes so as to achieve the desired proportionality in voting power. The problem with this approach, however, is that there is no agreed measure of voting power540. Lottery-voting, however, offers a solution, because each vote has an equal chance of victory, so two votes always have twice the chance of one vote. To use the EEC example again, there will be a one-inseventeen chance that Luxembourg’s vote is decisive and a four-in-seventeen chance that France’s is. It is therefore easy to proportion each party’s chance as we wish. Indeed, it has long been recognized that random selection offers an easy escape from this problem, but again it tends to be quickly dismissed – for instance, Felsenthal and Machover remark, in a footnote: 540 See A. D. Taylor (1995) pp.63-90, and 205-37. 241 “[T]he problem of voting power can also be trivialized by using a probabilistic decision rule. Suppose each voter is assigned a weight, as in weighted voting; and each bill is decided by a single voter chosen at random by weighted lottery, in which the probabilities of being chosen are proportional to the members’ weights. Clearly, these probabilities can be regarded as the members’ respective voting powers.”541 They seem to believe a solution that trivializes the problem is too easy. Of course, sometimes an easy way of defusing a problem is not what we want, e.g. if we reduce crime by legalizing things that are currently crimes, such as theft or murder. Other times, however, easy solutions simply reveal the problem was somewhat spurious and never as intractable as supposed. One could say the ‘problem’ of getting from Oxford to London, for example, is ‘trivialized’ by taking the train, but this simply shows it was never that serious a problem – it is easily solved. As G. A. Cohen remarks, while offering a ‘trivial’ solution to a different problem: “[I]f the solution is indeed trivial, why doesn’t that show that there was something wrong with the problem, to wit, that it was always easy to solve, and could have been thought otherwise only because of elementary errors… Why must an acceptable solution to the trilemma imply that the problem that the trilemma formulates is hard to solve?”542 So it is in this case; a probabilistic decision rule simply shows that weighted-voting need not be problematic. If we want to solve the ‘problem’, then this is a reason to favour lottery-voting, and – unless one can advance other arguments against such a probabilistic procedure – there is no reason to simply reject it. (6.15) Conclusion Chapters 1 through 3 defended an account of equality that is satisfied by proportional chances. The previous two chapters described how this could be 541 542 Felsenthal and Machover (1998) p.5 fn.8. Cohen (forthcoming) [Rescuing Justice… unpublished draft, ch.5]. 242 operationalized, in a democratic decision-making method known as lottery-voting. It remained to be shown, however, that such a mechanism would be desirable, beyond the fact that it is one way of satisfying equality. The present chapter has offered some supposedly minimal normative conditions for a decision-mechanism, drawing from May’s analysis of majority-rule and adapting the axioms that Arrow requires of any acceptable social welfare function. It has been argued that, insofar as these conditions really are minimal and desirable, lottery-voting appears to satisfy them. Lotteryvoting obviously respects equality of voters (anonymity and neutrality) and, despite sometimes being called ‘random dictatorship’, does not violate Arrow’s nondictatorship requirement. It is compatible with a universal domain, non-negatively responsive and respects the weak Pareto (unanimity) requirement. The one condition that lottery-voting seems to clearly fail is IIA, but it is unclear whether we should want this of a decision-mechanism, which has different objectives than a social welfare function. Moreover, if what is normatively desirable about IIA is that it limits manipulation, then lottery-voting should be attractive because it is non-manipulable (except, possibly, by allowing collaboration if thought desirable). Thus, it seems that lottery-voting is a promising possibility. There remains, however, one further worry, viz. that it is simply irrational to leave decisions to chance – this is the objection taken up in my final chapter. 243 7 Rationality “How can random machinery be rational?”543 “[T]he rationality of casting lots is related to their fairness”544 (7.1) Introduction The previous chapter found that lottery-voting satisfies those conditions of social choice that relate to equality, which supports the earlier argument that it is fair (chapter 3.8-11). It was also shown, however, that lottery-voting violates IIA and does not produce a complete and transitive ordering of possibilities. This may simply be because lottery-voting does not share Arrow’s conception of democracy, as a procedure for bringing about most socially preferred outcomes, but there remains the objection that deciding in this way is simply irrational. Since lottery-voting is rarely discussed in print (see introduction, 0.4, for examples), this claim is not made in print, but it seems a widespread reaction to lottery-voting, perhaps reflecting a general hostility to lotteries545. This chapter begins by discussing the nature of rationality, seeking to clarify the objection, focusing particularly on the key issues of maximization and consistency, before moving on to assess the rationality of lotteryvoting and arguing that, while no decision-mechanism is inherently rational, it is one that – in appropriate circumstances – all individuals can rationally and reasonably agree to. 543 544 Dick (1955) p.58 [not emphasized in original]. Heyd (2000) p.59 [not emphasized in original]. 545 Wolfle (1970). See also the largely negative reaction to the decision to use lotteries to allocate school places in Brighton and Hove, e.g. Andalo (2007), and Laville and Smithers (2007). 244 (7.2) The Nature of Rationality It is worth pointing out that rationality is itself a contested concept546. Some radical critiques suggest either that our notion of rationality is seriously inadequate or that there can never be an adequate standard because rationality itself is not neutral (e.g. feminists who argue that it privileges male impartial reason over empathy547). Even amongst those who subscribe to the value of rationality, there are considerable disagreements about its nature, and some of these can be almost equally threatening. Not everything that has been called rationality in the literature can be covered here. Mouffe, for example, claims that deliberative democrats understand rationality as defence of liberal rights548. It is hard to see how this could be what is meant by calling a system rational – though one might reasonably hold that any rationally acceptable system would be one that did guarantee certain basic rights. It will help to clarify the nature of rationality if we begin by considering four contrasts, between theoretical and practical rationality, ends and means, objective and subjective senses of rationality, and maximizing and satisficing conceptions of rationality. 7.2.a Theoretical versus Practical It is important to distinguish between theoretical and practical domains, since different norms may apply to each. Theoretical rationality is concerned with the formation and justification of beliefs, whereas practical rationality is directly actionguiding. Sometimes the two can concern the same matters – for example, if I am undertaking a journey from A to B, my deliberations about which route to take are practical, while if you are merely thinking about which route is shorter or predicting 546 547 C.f. Gallie (1956). E.g. Lloyd (1993 [1984]) pp.16-7, 26-7, 36-7, 75-8 and 105-10, Gilligan (1993 [1982]) pp.24-32, and Jones (2004) pp.301-19. 548 Mouffe (2000) p.83. 245 which I will take, then you are engaged in theoretical reflection549. The two can, however, diverge. Deciding which route to take, for example, is not the same as deciding which is shorter, or more scenic, or takes me past a newsagent’s, but necessarily an all-things-considered judgement. Moreover, while one can suspend theoretical judgement – e.g. hold no opinion on which route is shorter – one cannot refrain from acting, since doing nothing counts as a relevant action here. The necessity of action means that it can be rational to act on reasons that would not be sufficient to justify a belief. Suppose, for example, that it was difficult to judge which of two routes was shorter, which was all that mattered to me – in that case, it would probably be quicker, and therefore reasonable, for me to toss a coin to decide which way to go. It would not be reasonable to toss a coin to decide which route to believe was actually shorter, because the coin does not give epistemic reasons. Since democracy is about making practical decisions, however, this suggests a random element in the decision need not be problematic. The aim is not to decide whether option X or Y is better, but merely to decide which course of action to take; and losers need not therefore revise their judgement on this issue simply because they are defeated in the vote. 7.2.b Ends versus Means The second contrast concerns whether rationality has anything to say about ends or is, as Hume insisted, merely the “slave of the passions”. On the Humean view, rationality is almost silent about desires550, and merely concerns the instrumental means one takes to satisfy them. As he famously put it: “‘Tis not contrary to reason to prefer the destruction of the whole world to the scratching of my finger. ‘Tis not contrary to reason for 549 550 Harman (2004) p.45. Or, at least, most desires. It may still be irrational to desire the logically impossible, for instance, and Hume (1978 [1739-40]) p.416 [Treatise 2.3.3] admits a desire can be called unreasonable if based on a false supposition. 246 me to chuse my total ruin to prevent the least uneasiness of an Indian or person wholly unknown to me. ‘Tis as little contrary to reason to prefer even my own acknowledg’d lesser good to my greater”551 Others, however, have held that rationality can concern ends. Certain desires, for instance a non-instrumental desire for a pin or saucer of mud, seem unintelligible552, while Parfit claims it would be irrational to be indifferent to pain on future Tuesdays553. This long-running dispute cannot be settled here but, when it comes to a democratic decision-mechanism, it seems reasonable to focus only on instrumental rationality. Since the ends that we are to pursue are provided by the people’s expressed preferences or votes, and no decision-mechanism can enforce rational ends. This still leaves the question of what means we should be concerned with. One approach would be to identify social or collective ends, and then hold that the whole society should adopt rational means towards furthering these ends. This, however, is contrary to the theory of democracy developed here. In cases of disagreement, there is no comprehensive ‘general will’ or ‘common good’. Such cases of disagreement do not mean that there is no shared end or overlapping consensus at a very general level, e.g. on Pareto efficiency, but that such as there is does not determine answers to all political questions. There remain areas of conflict in which people have fundamentally opposed (though not necessarily zero-sum) interests. If we can only assess individuals for rationality, then we will have to focus on the means that each individual adopts towards the fulfilment of their own ends. One matter that will importantly affect the achievement of any individual goal, of course, is collective cooperation and decision-making. As argued in chapter 1.2, every individual can benefit from centralized coordination, even if the decisions sometimes 551 552 Hume (1978 [1739-40]) p.416 [Treatise 2.3.3]. Anscombe (1957) p.70, Crisp (2000) p.459; c.f. Hooker and Streumer (2004) p.67. 553 Parfit (1984) pp.123-6. 247 go against their ideal preferences. We can therefore assess whether individuals are rational in adopting any particular social decision-mechanism554. 7.2.c Objective versus Subjective We can also distinguish what might be called ‘objective’ and ‘subjective’ senses of rationality. In the theoretical domain, what it is objectively rational to believe is the truth, while it is subjectively rational to believe what one has best evidence for. In the practical domain, the objectively rational course is the best thing to do (what one actually has most reason to do) – or, more precisely, since reasons may be tied, a nondominated course of action – while the subjectively rational action is that bestsupported by the reasons that one knows that one has. Thus, in either the theoretical or practical domain, one can be subjectively rational without complying with the actual balance of reasons one has, provided that one responds rationally to the information one possesses. Again, no decision-mechanism – democratic or otherwise – can ensure the objective rationality of decisions. Moreover, if we were concerned only with such, then it is not obvious that we would adopt a democratic procedure – that would be the case only if it was shown to be epistemically better than an enlightened guardianship555. A democratic decision-mechanism could be conceived of as subjectively rational, if votes are taken as evidence of reasons. However, this still assumes that there is a single, objectively correct ‘will of the people’ or common good. In a world of plural values and conflicting interests, voters need not be tracking the same thing. Some will vote for X and some for Y, and all that this tells us is that X is presumably or subjectively better for the former and Y is subjectively better for the latter. This reveals that there are in fact at least two obstacles to the search for the 554 The rationality of any given choice will depend, of course, on whether they are concerned only to further their self-interest or also motivated by fairness. 555 Dahl (1989) pp.52-64 and 262, and Estlund (1997) pp.181-3. 248 objectively best outcomes. Firstly, even if everyone voted the same way, what they vote for need not actually be objectively best. Secondly, if their objective interests are opposed then we cannot easily say that either option is necessarily better ‘for the group’ because of the indeterminacy about Pareto non-comparable cases discussed in chapter 2.4 and 2.8. If, however, one voter thinks that X is best and another that Y is best, then we have some subjective reason to value either. 7.2.d Maximizing versus Satisficing The final distinction considered here is that between maximizing and satisficing conceptions of rationality. Traditionally, many have assumed that rationality requires maximizing something. This has been challenged, however, by others who argue one only needs to adopt satisfactory means556. The standard objection the latter is that it would seem irrational to forego a significant benefit that could be achieved at little cost557. One could, however, respond that what is ‘satisfactory’ must be relative to the options available, not an absolute level (or one might be faced with no satisfactory options, so be unable to comply with rationality), and therefore that – whatever one’s absolute level – it is not satisfactory if one could achieve a significant gain for little cost. Nonetheless, many have assumed that satisficing is really best seen as a decision-mechanism, and (objectively) rational only when it is actually optimal, e.g. because it reduces decision costs558. The next section will assess lottery-voting against the charge that it fails to maximize. Even if we accept maximization as a desideratum of rationality though, it is not clear that a group must maximize anything. If we are methodological individualists, then we see individual persons as the only agents and group decisions as supervenient upon the actions of individuals. As such, talk of group actions or 556 557 E.g. Slote (1989) pp.7-31 and 47-81; c.f. Mele (2004) p.261. Sorensen (2004) p.261. 558 Mele (2004) pp.261-2, Harman (2004) p.49. 249 decisions is, if not simply confused or metaphorical, elliptical and a group consisting of individual rational maximizers need not itself maximize anything559. Thus, again, we must be cautious in trying to apply the standards of rationality to a collective decision-mechanism. (7.3) Maximization Even assuming that rationality really does require maximizing something, it is not obvious that this gives us reason to favour majority-rule over lottery-voting. Admittedly, prima facie, majority-rule may appear more rational, because it obviously maximizes something, viz. the number of people satisfied by that decision560. As we saw in chapter 2.4-5, however, this is not the same as maximizing total satisfaction, since it neglects intensities of preference and those in the minority. Moreover, even if rationality does require maximizing something, it does not follow that maximizing something implies rationality. One could respond, rhetorically, that there are ways in which lottery-voting is maximizing. Firstly, because it will bring about someone’s first preference, it always produces an outcome that someone regards as maximal. Secondly, because it gives everyone a fair chance (as argued in chapter 3.8-10), it can be seen as maximizing fairness. Returning to the critique of majority-rule, the number of satisfied voters is simply not an appropriate maximand but, at best, an imperfect proxy or indicator of what should be maximized – which is presumably something like aggregate net 559 560 Gauthier (1990) pp.181, 190, 199-201, and 205. It is sometimes claimed that majority-rule maximizes the number of people self-governing, e.g. Risse (2004) p.44. This is not necessarily so. While it is true that majority-rule does maximize the number of people self-governing in a single case, if we consider a series of decisions in a society with a permanent majority/minority split, then majority-rule deprives the minority of any share of selfgovernment. Lottery-voting, in contrast, would tend to give each group its way sometimes, meaning that all people were self-governing over at least some issues. 250 satisfaction. If votes do indicate what is objectively rational or good, however, then it may be subjectively rational to go with the majority of votes, if this gives one the best chance of complying with objective reasons. At this point, one may try to defend majority-rule by appeal to Condorcet’s Jury Theorem, which supposes that voters cast judgements on the appropriate maximizing strategy for society to follow and, if more heads are epistemically better than one, we should assume that what the majority have identified is most likely to be objectively rational. Again, however, we have already seen reasons to reject this argument (see chapter 2.7-8). We cannot assume either that there is a single objectively rational or best end for the whole of society or that, if there is, the majority are most likely to identify it. The idea of maximizing is threatened at two levels by pluralism: firstly, the plurality of values or ends and secondly the plurality of separate agents involved in collective choice. If we believe all ends can be reduced to one super-value, such as utility, then we can use this to decide all choices. For instance, if we must choose between spending an evening reading or in the pub, we simply have to ask which gives us most utility. It is not clear, however, that all experiences are commensurable in this way. Finnis suggests seven basic human goods, including knowledge and sociability561. If our choice is between reading that promotes knowledge or sociability in the pub, then there is no master-value by which to compare562. Thus we cannot simply aim to maximize any one value; the choice in a particular instance is whether to promote knowledge or sociability, and overall we may prefer a balance of each to maximizing either. One might wonder whether there really are plural, incommensurable values – after all, we can and do choose between different possible activities, such as reading 561 562 Finnis (1980) pp.59-75, and 87-8. Finnis (1980) pp.92-5. 251 or going to the pub, so it might be held that we do so in virtue of some over-arching value – or that utility is simply a construct based on our choices. However, when it comes to collective action, the plurality of ends is guaranteed by the plurality of agents. Even if I can choose between any two activities by reference to my happiness or utility, my happiness is a different end from your happiness, and – because of the separateness of persons – there is no over-arching value that allows society to choose between my happiness and yours. Sometimes, the difference between what I have at stake and what you have at stake may be so great that pairwise comparison will tell us that one claim trumps or outweighs the other. Often, however, all we will be able to say is that I gain and you lose from one policy, while the situations are reversed for another – nothing particularly informative can be said about what’s ‘better overall’563. Chapter 1.2-3 argued that, while central coordination can make all better off, democracy should be seen as primarily a distributive matter: a way of fairly deciding who gets their way over which contested issues. As such, it should not surprise us that it does not obviously maximize anything – the aim of distribution is more usually equality, as when we divide a cake between different persons564. It is rational to adopt a fair distributive procedure, regardless of whether it maximizes anything. (7.4) Consistency between Decisions Perhaps it is unreasonable to expect a collective decision procedure to maximize any single value, at least unless it is a goal that all members share (for proof that lottery-voting respects unanimity, or weak Pareto, see chapter 6.9). We may, instead, 563 564 C.f. Taurek (1977) pp.299-300. Gauthier (1990) p.205: “Morality is not concerned with maximizing some quantity analogous to individual good; rather, morality is concerned with the way in which the benefits society makes possible are distributed among individuals, each pursuing his own good… And so the rationality of moral choice is assured, not by modeling it on the rationality of individual, personal choice under risk, but rather by modeling it on the rationality of a bargain”. 252 hold that our procedure should at least give consistent decisions. There are different forms of consistency that we may want, however. Firstly, we may want consistency between different decisions, which is the subject of this section, and secondly – seemingly more minimally – we may want consistency over a single decision, which is addressed in the next section. Note that consistency between different decisions is not a problem only for lottery-voting. Such consistency is likely only if there is either a single dictator or a permanent majority, so that all decisions are made by the same people. When there are different majorities, then majority-rule may also lead to inconsistency, as illustrated by separate decisions for lower taxes and higher spending. Suppose, for example, we face three decisions: H: raise health spending. E: raise education spending. T: raise taxes. All know that, with current tax revenues, they can raise either health or education spending, but to do both they must raise taxes. Now suppose there are three individuals whose preferences are as follows565: Tab. 7.1: Group Inconsistency 565 The structure of this problem mirrors what Christian List calls the ‘discursive dilemma’, although I am not necessarily making the same point. “Suppose that a three-member court has to make a judgment on whether a defendant is liable for a breach of contract. According to legal doctrine, the defendant is liable (proposition R) if and only if the defendant did some action X (proposition P) and the defendant had a contractual obligation not to do action X (proposition Q). Thus legal doctrine requires R<->(P&Q). Suppose that the individual judgments of the three judges are as in table 1. P Q R<->(P&Q) R Judge 1 Yes Yes Yes Yes Judge 2 Yes No Yes No Judge 3 No Yes Yes No Majority Yes Yes Yes No Table 1. The "Doctrinal Paradox" or "Discursive Dilemma" (Conjunctive Version)”. This table and the preceding paragraph are taken from Christian http://personal.lse.ac.uk/LIST/doctrinalparadox.htm (14/10/06) List’s website 253 H Individual 1 Individual 2 Individual 3 Majority Yes No Yes Yes E No Yes Yes Yes T No No Yes No Majoritarian voting on the three decisions separately yields the conclusions that we should raise both health and education spending without raising taxes, though all acknowledge that this is impossible. It seems that this reflects the kind of conflict between the fair decentralization of political power and consistency that motivates Arrow’s impossibility theorem. The inconsistency occurs because there is no fixed majority making all the decisions, so it arises from the very ‘rotation’ of power that makes the procedure fair. If all accept the logical entailment (if H and E then T), then making two of these decisions will ordinarily logically determine the third, e.g. if we vote to raise education spending but not taxes then we no longer need to vote on health spending as it cannot be raised. Note, however, that this may, like the cycling case, simply deepen controversy over the order in which votes are taken, since that may effectively determine the overall settlement. Moreover, if this solution is deemed acceptable, then it is equally available to lottery-voting – we can, having held lottery-votes on two of these issues, use their conclusions to determine the third logically, without having to put it to a separate vote. It has already been argued that, where all agree that two decisions are complementary, they can simply be combined into a single vote (see chapter 5.11). Suppose, for example, that we are already committed to raising health spending. Then, given acceptance of the logical relation between spending and taxes, we effectively only have one decision left, viz. whether to raise education spending and 254 taxes or not. It may be thought that problems will arise if there is not agreement on the complementary nature of two issues, for instance (to use the example from chapter 5.11) if someone holds that we should buy bookshelves despite not buying books. This would only be inconsistent, however, if there was no use for the bookshelves without the books. If the decision on the bookshelves is taken in light of the decision not to buy any books though, then it can only come up in favour of buying the shelves if someone still sees a use for them without the books566. Generalizing, if even one person’s preferences are such that they do want X without Y, then there is no reason for us to suppose that the combination of X without Y is somehow inconsistent. On the other hand, if it is true that no one wants X without Y then, having decided against Y, no randomly-selected vote will be for X (since lottery-voting respects unanimity). (7.5) Consistency over Single Decisions One might accept the possibility that separate decisions over connected issues may lead to risk of inconsistency, and that lottery-voting need be no worse here than other decision methods, but still hold that there should at least be some consistency over individual decisions – that is, if X beats Y then X should always beat Y and the outcome should not be reversed by an immediate recount or new lottery draw. Lottery-voting violates this condition, because the decision is not solely determined by the votes but also relies on a lottery, so a different outcome could have been produced from the very same profile of votes (see the related discussion of IIA, in section 6.10, above). Again, the inconsistency problem is closely-tied what makes 566 One problem would be is if someone wanted to buy the shelves to increase the chances of buying books in a later vote. 255 lottery-voting fair, so we need to investigate how reasonable it is to impose such a consistency requirement. This inconsistency will always be, in a sense, counter-factual because, while it is known that any given outcome could in fact have been different, once the lottery has taken place the issue is settled and there is no re-vote/draw on the issue. While there may be a practical problem, because the very real possibility of success ‘next time’ gives defeated groups (whether majorities or minorities) an immediate incentive to call for another decision, this is something dealt with in chapter 5.10. The present objection supposes that, even if this practical incentive problem could be resolved, it is simply undesirable that the procedure could give different outcomes from the same inputs, but it is not clear why this is. One possible explanation is a persisting belief that we should be striving to realize an ideal, best outcome, because a procedure that gives different outcomes obviously cannot always realize such. Chapter 2.8, however, argued that any social ideal is likely to be, at best, indeterminate so democracy should be conceived of as a fair procedure for adjudicating conflicts of interest. This does not mean that it necessarily produces fair or otherwise optimal outcomes, but that voting can be compared to tossing a coin between two equal claimants to an indivisible good. In a given conflict, either A is satisfied or B is, so what we are dealing with is effectively a zero-sum game. In such cases, we cannot say that either outcome is ‘objectively better’ for society as a whole. We must recognize that each party’s conflicting interests are not to simply be traded off and tossing a coin is fair precisely because it could produce either outcome, rather than consistently favouring either one. The focus on fair distribution, and alternation, of power also explains why lottery-voting fails to produce a transitive ordering – it simply is not concerned with 256 ranking the ‘objective betterness’ of outcomes. The lottery is not intended to judge it better that A gets satisfied – of course, that would be a surrender of judgement or superstitious trust in divination. The lottery merely bestows legitimacy on A’s claim to satisfaction rather than B’s. It is quite consistent for us to say that A should therefore get her way, though we think it would have been better if B had got his way (indeed, this is what B will presumably say)567. It is possible that, if we make three pairwise comparisons, we will get an apparent cycle (A chosen out of A and B, B chosen out of B and C, and C chosen out of A and C), but this is not really the puzzling intransitivity A>B>C>A, because none of those choices actually judge one option better than the other – it is merely a generalization of the fact lottery-voting may one time choose A out of A and B and another time choose B. Should we be bothered by this? It seems that Arrow wants transitivity to avoid path dependence568. However, while one should be concerned that a bias in favour of the status quo will arbitrarily favour some instead of others, it does not seem that one should preclude a pure procedure such as a lottery. Again, the difference is that Arrow seeks a welfare or betterness function, and it would be odd for that to be path dependent, yet it would be odd for democratic legitimacy not to depend on the path by reached it is reached. Any outcome from a pure procedure is legitimate, so what matters is that the procedure is neutral (see chapter 6.5) and that it is applied properly, not that A could have been chosen had the lottery come out differently. C.f. Wollheim (1962) pp.76-87. Arrow (1963 [1951]) p.120; c.f. Chapman (1998) p.297: “Arrow did not believe, if social choice was collectively irrational, that actual cycling would be observed. Rather, he feared that collectively irrational social choice would show itself as a kind of arbitrary dependence of the final social choice on the choice path, something which is now commonly referred to as path dependence”. I shall have more to say on this in section 7.9, below. 568 567 257 Moreover, in accusing all path dependent methods of ‘arbitrariness’, Arrow seemingly conflates the negative connotations of being dependent on an arbitrary or capricious will with the more neutral idea of being independent of any will569. Lotteries seem fair because they are arbitrary in the latter sense – for example, in cases of equal claims to indivisible goods, Rawls recommends that we “select one claim as meriting satisfaction by an impartially arbitrary method, e.g., by seeing who draws the highest card… [This] is impartial because prior to the drawing of the cards each person has an equal chance to acquire in his person the characteristic arbitrarily taken to be relevant”570. Where we face a tie or cycle, we must either (i) risk indecision/paralysis, (ii) settle for the status quo, (iii) allow one person to determine the outcome, e.g. through agenda control, or (iv) resolve the issue randomly. The first is inadequate since something must be done, even if it is nothing (which may, of course, be one of the options on the table). The second and third options violate neutrality and anonymity respectively, and so also equality571. It seems that the only fair way to resolve deadlock is randomly572. As Mueller notes, “Although obviously arbitrary, the general popularity of random decision procedures to resolve conflictual issues suggests that “fairness” may be an ethical norm that is more basic than the norm captured by the transitivity axiom for decisions of this sort”573. One way to do this would be to randomly-determine the order in which votes are taken, since this 569 OED online “The quality of being arbitrary or uncontrolled in the exercise of will; a. capriciousness; b. despotism” http://dictionary.oed.com/cgi/entry/50011307?query_type=word&queryword=arbitrariness&first=1& max_to_show=10&single=1&sort_type=alpha (last accessed 08/07/07). Under ‘arbitrary’ and ‘arbitrarily’ the OED also includes “dependent upon will or pleasure”, “[d]erived from mere opinion or preference” and “at will”. 570 Rawls (1951) p.193 [emphasis added]. 571 McGann (2006) pp.17-8. 572 McGann (2006) pp.17-8. Bordes (1979) p.185: “the most ‘democratic’ way to break the ties seems to be the lottery”. 573 Mueller (2003) p.588. 258 means no one has agenda-control and effectively settles a cycle574. Lottery-voting, however, directly randomizes who gets their way. Note that neither approach makes the decision wholly random – the only randomness is who gets their way, but we can still insist that that person’s preferences should be rational and informed, according to whatever standards we find appropriate. Thus, the use of random devices is not a complete abdication of rationality, and any inconsistency should not bother us, as it is simply the consequence of fairness. (7.6) Individual Rationality The general line of response to all of the above objections has been to insist on a difference between individual and collective decision-making, and thus to claim that it is inappropriate to hold a group to standards of rationality defined for individuals. One may, however, wonder whether this is true or whether we should still want our group to choose and act as if it were a rational individual. This section attempts to illustrate the disanalogy between individual and collective choice in order to support the claim that what is rational for an individual need not be appropriate for a group and vice versa. It seems fruitful to begin by considering individual rationality. It may be appropriate for an individual to simply write a list of reasons for and against any particular course of action, and then go with the preponderance of reasons, which is essentially the approach of Benjamin Franklin’s famous ‘prudential calculus’575. It may seem that this offers prima facie support to the rationality of majority-rule, but note that it is not exactly analogous, because an individual deciding rationally takes 574 575 Arrow (1997) p.5. This is basically the method of ‘Prudential Algebra’ Benjamin Franklin recommended to Joseph Priestley in a letter of 19th Sept 1772 (Franklin (1975) vol.19 pp.299-300). He also offered the same advice to Jonathan Williams Jr. in a letter of 8th April 1779 (Franklin (1975) vol.29 pp.283-4). 259 account of the weight of different reasons; whereas voting ordinarily reduces all reasons to +/-1 (we saw problems with attempts at weighting votes according to interests in chapter 2.4-5 and 2.8). In any case, the rational individual aims to maximize his or her utility, or personal balance of pleasure over pain, thereby behaving like a utilitarian in a society of one. This picture of individual maximizing rationality can be contrasted to an individual who decides what to do by a random process somewhat like lotteryvoting576. Such an individual is characterized in Luke Rhinehart’s novel, The Dice Man577, which concerns an individual who decides between his desires using dice. Whereas Franklin’s method involved balancing competing reasons and so acting on the preponderance of reasons, Rhinehart’s method is to list all of the various actions that he desires to do, assign each possibility to one or more numbers, and roll dice to decide between them578. This seems to be an individual analogue of lottery-voting, with each option given a chance proportional to the strength of desire and the dice being a means of choosing fairly between them. Rhinehart explicitly justifies this randomizing as giving each desire a fair chance of expression. He refuses to do those things that he most wants to do because small parts of him do not want to do it them, saying “I gave a minority self a chance to be heard”579 and telling a patient that she should “Treat all of your desires as if they had equal value” rather than letting one inhibit the rest580. Giving an account of his 576 Note, I am not saying that it is never rational for an individual to decide randomly – it may, for example, have advantages of unpredictability, lower decision costs or exciting variety. The following example is supposed to illustrate the irrationality of adopting a randomizing device as part of a general decision mechanism. 577 Rhinehart (1999 [1971]). 578 It should be noted that this allows for numerous possibilities, whereas Franklin’s seems confined to choosing between two courses of action. Also Rhinehart did sometimes use two or more dice and assign some possibilities greater chances than others. 579 Rhinehart (1999 [1971]) p.224. 580 Rhinehart (1999 [1971]) p.232. Note that this conjunction seems contradicted by the fact that he sometimes gives some desires greater chances. 260 method to colleagues, he says, “we all have minority impulses which are stifled by the normal personality and rarely break free into action… We refuse to recognize that a minority impulse is a potential full man, and that until he is granted the same opportunity for development as the major conventional selves, the personality in which he lives will be divided”581, describing Western psychology as encouraging a totalitarian personality that suppresses other selves582. The aim of his dice-therapy is to make the person “free to be fully all of his selves – as the dice dictate”583 and “to express his host of minority selves clawing for life”584. Unsurprisingly, many think Rhinehart is insane. In part, this is the general hostility to chance that concerns me here, as when a Rabbi criticizes Rhinehart’s method, saying that it is “a resignation from the status of man: it is a worship of chance”585. Rhinehart, in contrast, defends his approach trying to explain to a colleague, who thinks he’s schizophrenic, “my eccentricities, inconsistencies, absurdities, and breakdowns of the last year have all been the logical consequences of a highly original but highly rational approach to life”586. It is not my purpose here to comment on the psychology, although certainly there is reason to doubt Rhinehart’s sanity, from the way in which he begins to worship the Dice587 and the options he gives it588. The aim of this section is to show why we regard this way of deciding as irrational for the individual. The next section shows why what we intuitively reject for the individual may nevertheless be justifiable on a societal level – that is, why a 581 582 Rhinehart (1999 [1971]) pp.314-5. Rhinehart (1999 [1971]) p.316. 583 Rhinehart (1999 [1971]) p.406. 584 Rhinehart (1999 [1971]) p.436. 585 Rhinehart (1999 [1971]) p.513. 586 Rhinehart (1999 [1971]) p.301. 587 Rhinehart (1999 [1971]) e.g. pp.237, and 525. 588 In fact, the first choice the Dice make is to rape his neighbour Arlene (p.69). Later, he admits he would even obey the dice if they said to kill his son (p.456), though he had earlier said he would not give them such an option (p.285) – which proves his point that the dice increase possibilities (p.317). Ironically, he recognizes “the dice can show almost as poor judgment as a human” (p.75). 261 collective may rationally adopt a method (lottery-voting) that appears analogous to such dice-therapy. While many have been tempted to treat the mind or ‘self’ as if itself composed of miniature individuals589, it is not clear that each part is to be treated as an equal in the way Rhinehart supposes – rather, psychological harmony may well be a matter of proper ordering, as has been held at least since Plato590. It seems Rhinehart thinks that each part of him must be given chance of satisfaction, but it is not clear why. He refuses to identify with any single dominant desire, insisting his weaker or minority selves are just as much part of him and so deserving expression. However, if he is right that all his various desires are equally, or at least truly, his own, then it does not seem to matter intrinsically which is expressed. Whichever desire he acts on – be it a strong or weak one – he is expressing some aspect of himself; indeed he even suggests such to a patient, reasoning since “each of your desires is as arbitrary, meaningless and trivial as the next… In some sense it makes absolutely no difference what you do… Then why not let the flip of dice – chance – decide what you do?”591. But if we grant that it does not matter, cosmically, which of our desires is satisfied then it seems to make sense to satisfy one’s stronger desires. In other words, utilitarian reasoning seems unobjectionable applied to a single person592. The idea of fairness seems irrelevant when it comes to strictly personal or selfregarding decisions. Of course, I may decide to act fairly or not towards others, but it 589 Psychologists, so far as I can make out, are hostile to conceiving of the individual as a collection of selves (or homunculi), see Gregory with Zangwill (1987) p.313. Nonetheless, this image seems to go back at least to Plato (1992) e.g. pp.110-8 and 260-3 [Rep 435b-42d and 588b-92a], and has also been influential more recently; e.g. Hardin (1982) p.10, and Kaplan (1964) p.58: “we are individuals only by courtesy; in truth the individual is a congress of selves, each pursuing values to which the other selves may be indifferent or hostile”. This, of course, suggests social choice problems may apply to individuals, see May (1954), and Arrow and Raynauld (1986). 590 Plato (1992) pp.119-20 and 260-2 [Rep 443c-4a and 588b-91a]. 591 Rhinehart (1999 [1971]) p.232. Note that this is only an argument that there is no reason not to choose randomly; rather than a positive argument that there is reason to do so, as suggested by his idea of fairness to each desire. 592 Rawls (1999 [1971]) p.21. 262 is not clear that I can act unfairly towards myself593. If I desire to eat both an apple and a banana, then I may toss a coin simply to decide which to have, to avoid being caught in indecision like Buridan’s famous ass. There is, however, no need for me to be fair to each of these desires. If I have an apple, then I am satisfied, and if I have a banana then it is also me who is satisfied, so it makes sense for me to choose whichever most satisfies me. Since it is me satisfied, either way, the choice effectively reduces to how much satisfaction I gain, e.g. 5 units or 4 units. To repeat what was said above, the rational individual maximizes his or her own utility, acting as if a utilitarian in a society of one. What we cannot necessarily do, however, is extend this principles of rational choice for an individual to society as a whole594. The next section turns to the case of a collective made up of separate individuals. (7.7) Collective Rationality Recall the reason that Rawls rejects utilitarianism; because it seeks to maximize social satisfaction, balancing gains and losses for separate individuals, as if all were part of some social super-entity595. In other words, it applies the maximizing principle of individual choice to society as a whole, with no regard to distribution. Intuitively, however, this is to ignore the claims of fairness. While it seems that there is no reason for a single individual to be fair to her different desires, matters are importantly different when distinct people are involved. While a single individual may rationally prefer three units of good to two, it is not obvious that society should prefer (3,0) to 593 This is not to deny that I can be, as Kant held, under duties to myself. Nor is it necessarily to deny that issues of justice may arise between temporal parts of myself – this raises questions of diachronic personal identity, whereas I am focusing on the self at a given time. 594 Rawls (1999 [1971]) p.24. 595 Rawls (1999 [1971]) pp.19-24. 263 (1,1)596. When we are dealing with separate persons, we must pay attention not only to aggregate good but how it is distributed. Persons in society are separate and distinct in ways different from the desires of an individual. Whereas it makes no difference whether I satisfy my desire for an apple or a banana, and it is therefore reasonable for me to satisfy the stronger desire, this is not so if the decision is a collective one. Suppose that you and I must agree on a fruit, where for some reason we must have the same. Now, if I want an apple and you want a banana, it is not – as Arrow supposes – a matter of indifference which is chosen. If it is an apple then I am satisfied and you are not, while you are satisfied and I am not by a banana. My satisfaction with an apple in no way compensates or benefits you, assuming that you are not altruistic, so there is no reason why you should waive your interest in having bananas even if I have a stronger desire for apples. Even if we can make meaningful interpersonal comparisons, and thereby know that I get slightly more pleasure from an apple than you do from a banana, if the choice could be represented (9,5) or (5,8), there is no reason why your claim to three units of good should be outweighed simply because I could get four. We are both separate persons with our own good to consider, rather than parts of some social whole, so it seems that the fairest way to treat both of us equally is to toss a coin to decide between our competing claims, giving me the same chance of getting an apple as you of getting a banana. A further way to illustrate the difference between an individual and society is to point out that a society made up of individual rational maximizers need not itself be trying to maximize anything597. Even if all agree on some standard of good, such as happiness or money, individuals need not be impartially concerned with this good. 596 Part of my case for lotteries, however, is that both may prefer tossing a coin between (3,0) and (0,3) to the equal distribution (1,1) or even (1.5,1.5). 597 Gauthier (1990) pp.181, 190, 199-201, and 205. 264 Utilitarians have often argued that what I hold good for me I must acknowledge good for others, from the point of view of the universe, and thus that it is rational to maximize happiness generally, rather than merely my own598. However, there is no incoherence in the egoist wanting only to maximize her own pleasure, indifferent to that of others, preferring (5,1,1) to (4,6,6), where her pleasure is the first number. If my concern is to maximize my happiness and your concern is to maximize your happiness, then there is no social total that either of us – or the two of us together – is trying to maximize. I would prefer (7,4) and you (4,7) even if (6,6) is also available. If we must coordinate on one of these outcomes, then it may be rational for both of us to agree to (6,6), but the point is that this ‘social maximum’ is not maximizing for either of us individually599. On this account, the idea of maximizing is appropriate to individual rationality, but not appropriate to a group, which should consider fairness between different members, and thus has reason to favour something like an equal distribution or a lottery to allocate goods. Buchanan and Tullock also focus only on individual rationality, remarking that: “Under the individualistic postulates, group decisions represent outcomes of certain agreed-upon rules for choice after the separate individual choices are fed into the process. There seems to be no reason why we should expect these final outcomes to exhibit any sense of order which might, under certain definitions of rationality, be said to reflect rational social action… In this book we shall not discuss social rationality or rational social action as such. We start from the presumption that only the individual chooses, and that rational behavior, if introduced at all, can only be discussed meaningfully in terms of individual action”600 598 Mill (1998 [1861]a) pp.64, 81, and 105-6 [Utilitarianism II.18, IV.3, and V.36], and Sidgwick (1913 [1874]) pp.382, and 418-21 [Methods of Ethics III.13.§3, and IV.2]. 599 If the best equal alternative had been (5,5), we might each prefer inequality at the higher level, but be unable to agree on who should get more – this is the coordination problem I began with in chapter 1.2. 600 Buchanan and Tullock (1962) p.32. 265 Similarly, Gauthier writes, “I presuppose that it is primarily to the individual that we ascribe rationality. To speak of a rational action, or rational activity, or a rational morality, or a rational society, is to speak of rationality in a way which must be derived from our conception of a rational individual or person”601. This is broadly the line taken here. It is primarily individuals that act, and the combination of their individual actions determines the collective602. While the consequences of all individuals acting rationally may be undesirable, as in the Prisoners’ Dilemma, we should not think of this as a lack of ‘collective rationality’, but if anything a failure of individuals to coordinate to better advance everyone’s ends. The focus on individual rational action – and, more importantly, consent – leads on in the direction of the contractualism advocated in chapter 1.5. As Gauthier puts it, “The idea of society as a cooperative venture for mutual advantage links rational social choice with justice. And this aim of justice is to be achieved not by decision making that embodies a single, social, maximizing procedure, but rather by decision making through agreement among the individual participants in the cooperative venture”603. If the appropriate question to ask is simply whether individuals are acting rationally, then, when it comes to social decisions, we need to ask whether they are made by a process that all individuals could rationally agree to using for collective decisions. (7.8) Rational Use of Decision-Mechanisms Another way of arriving at contractualism is to argue that no procedure is inherently rational or irrational; rather all decision-mechanisms are simply tools that 601 602 Gauthier (1990) p.210. This assumption contradicts Arrow, who takes social action as – in a sense – primary, because no individual can act without others. See Arrow (1984) pp.63-5. 603 Gauthier (1990) p.175. 266 are themselves arational. We should adopt the mechanism that best suits our purposes, which will depend on the circumstances, and it is this choice of decisionprocedure – rather than the procedure chosen itself – that may properly be judged as rational or irrational, depending on how it serves our purposes. For a charge of irrationality to prevail against lottery-voting then, it would have to be shown either that it is never rational to use a lottery or that we could always do better by purposeful choice, but both of these claims seem dubitable. Those who attack lotteries as an abdication of responsibility or intelligent judgement604 are themselves often criticized for their excessive faith in rational judgement, or ‘hyper-rationality’605. Of course, it is often better to choose between options rather than simply toss a coin; that is not in dispute. The debate is whether it can ever be rational to use a lottery. Those who oppose such presumably subscribe to something like the principle of sufficient reason606 – that is, they think that it is wrong to choose one option over another without reason – yet this principle is only admissible in the theoretical domain, where it is possible to withhold any judgement. In the practical domain, something has to be done, even if it is nothing (which may be one of the options). Moreover, to refuse to take either course, on the basis there is no reason for either, may lead to a state of paralysis that is clearly worse than either, as in the case of Buridan’s fabled ass, which starved because unable to choose between two equally-appetizing bales of hay. The claim here is that sometimes reasons for choice run out. The ass had no reason to choose either bale of the other, but plenty of reason to choose either rather than starve due to indecision, so it would be perfectly rational to choose arbitrarily (whether by explicit randomization or not). 604 605 E.g. Wolfle (1970). Elster (1989) pp.17-38, Neurath (1983) pp.7-10, and Heyd (2000) pp.66-7. 606 This is generally attributed to Leibniz, e.g. Sorensen (2004) p.270, and Murphy (1986) pp.143-4. C.f. Leibniz (1998) p.272 [Monadology §32]. 267 Even if we want our choice to be dictated by reasons, we should note that the choice of a decision-procedure is itself a choice, and there may be reasons to favour a totally or partially random one. One reason to use a lottery, as in the Buridan’s ass case, is simply to break a tie or deadlock607. It may also be rational to use a lottery in cases where options are not strictly tied, but where any gain from discrimination is likely to be outweighed by the decision-costs. For example, suppose we did not know that the two bales were equally-sized, but the only way to establish which was larger was to count the straws or weigh them. In this case, given that each bale is big enough anyway, counting would be more trouble than it was worth, so it would be rational to decide randomly. There are occasions where any quick decision is better than time spent deliberating, even if more deliberation could lead to an outcome that would, decision-costs aside, be better, and in these cases a lottery is attractive because it is a low-cost decision-mechanism. There may also be cases where unpredictability is a virtue, for example in strategic interaction where one wants to avoid one’s opponent or in mixed-strategy equilibria, so a lottery may be useful in deciding what to do in certain games. If democratic outcomes should be unpredictable (see chapter 4.10-11, above), for instance because such uncertainty gives everyone incentives to participate, both in deliberation and voting, then a lottery may be desirable. Perhaps the most powerful reason to employ a lottery, however, is to be fair between competing claims. This is not simply a case of tie-breaking, because no-one could complain if Buridan’s ass had simply chosen one bale without randomizing. However, if we are distributing an indivisible good between two rival claimants, then either may protest if the other is arbitrarily favoured. When it comes to adjudicating between separate people, then it seems that often all we can say is that X is better for 607 C.f. Goodwin (20005 [1992]) pp. 55, Barry (1965) p.88, Mackie (2003) pp.5, 50 and 84, Duxbury (1999) p.20, and Berg (1965) p.130 and 132. 268 one and Y is better for the other, and there is no rational way to decide between their competing claims. This is why the case of collective choice is importantly different from individual choice, because these considerations of fairness come into play. Where our interests conflict, it may be rational for both of us to accept a lottery that gives each of us an equal chance of satisfaction. Thus, for example, if we could achieve either (6,3) or (3,6) – let us suppose that the best equal option is (3,3) but that we are both agreed not to accept this Pareto sub-optimal outcome – it would be perfectly rational for each of us to agree to tossing a coin to decide which of these efficient outcomes would obtain. In other words, it is rational to use lotteries to decide between different patterns of coordination, where all are agreed that they want coordination but have conflicting interests over which pattern of such obtains (see chapter 1.2, above). Showing that random decision methods can be rational is, of course, not yet fully to vindicate choosing lottery-voting. To do that, we would have to show that, in a three person case, it would be rational to accept a proportional lottery that gave a two-thirds chance to the outcome (6,6,3) and a one-third chance to an outcome (3,3,6)608. Nonetheless, I assume it can be rational to offer fair terms of cooperation to others, and thus if my preceding argument for the fairness of lottery-voting (see primarily chapter 3) is plausible, then it can be rational to adopt lottery-voting. In Rawlsian terms, one might put this ‘it can be rational to be reasonable’609. If this is disputed, then I would simply say that reasonableness is more important than rationality. If one assumes that rationality is simply about promoting one’s selfinterested ends, then one may argue that it is not rational for members of a permanent 608 These figures are purely illustrative. If we showed the rationality of lottery-voting in this quite specific case, that would show it is sometimes suitable. I believe that it is also appropriate in other cases, for example to decide between (4,4,3) and (3,3,6), or even perhaps (6,6,3) and (3,3,4), although we really need to know what these numbers represent to be sure. 609 C.f. Rawls (2007) pp.54-7. 269 majority to accept lottery-voting when majority-rule would better serve their interests. In this case, there is little merit to rationality when it simply allows such selfishness. What we really want is reasonable agreement, which can be achieved either by direct appeal to a more Scanlonian form of contractualism610 or by constraining rationality through some device like Rawls’ veil of ignorance611. Further, it should be emphasized that lottery-voting does not completely surrender human judgement. To leave decision totally to a lottery may indeed be rationally criticisable612 – and, indeed, unworkable, if it proved impossible to identify all possible options. Remember that the lottery comes in only at the second stage. First, everyone makes what is hopefully an intelligent choice about what should be done when they cast their vote613. The lottery only decides between different people’s votes, but whichever outcome is picked is one that has already been chosen by at least a single voter, and can therefore be deemed presumptively rational, on the grounds that at least that person considers it optimal, assuming them not clearly mistaken. The lottery over votes does not deny any role to reason – indeed, in a sense it forces voters to think more carefully about how they cast their vote, knowing that could be decisive, even if in a minority of one, so they had better vote responsibly (see chapter 5.3d). (7.9) Path Dependence So far, it has been argued that it is inappropriate to condemn lottery-voting by standards of rationality appropriate to single individuals, because the context of a 610 611 Scanlon (1998) esp. pp.189-247. Rawls (1999 [1971]) pp.118-24, Rawls (2001) pp.14-8, and Rawls (2007) pp.17-9 and 152. 612 This is the kind of lottery Vernon (2001) p.43 attacks. 613 Similarly, we do not feel Luke Rhinehart escapes responsibility for his actions because ‘the dice chose them’, when we know he defined the possibilities and indeed their odds. 270 group of separate persons requires different norms, which include fair consideration of each person’s interests. If Arrow was simply guilty of an illegitimate extension of individual rationality to a collective group – as some utilitarians supposedly were – then this might be the end of the matter. However, in response to criticism like that of Buchanan and Tullock614, above, he explicitly denies that this is what he is doing. Rather, he argues that transitivity is ensures an escape from ‘democratic paralysis’ without privileging the status quo: “Those familiar with the integrability controversy in the field of consumer’s demand theory will observe that the basic problem is the same: the independence of the final choice from the path to it. Transitivity will insure this independence; from any environment, there will be a chosen alternative, and, in the absence of a deadlock, no place for the historically given alternative to be chosen by default… Collective rationality in the social choice mechanism is not then merely an illegitimate transfer from the individual to society, but an important attribute of a genuinely democratic system capable of full adaptation to varying environments”615 But if this is the only reason to value transitivity, then transitivity is not required – for this argument is merely that transitivity is sufficient to avoid such problems, not that it is necessary. As we saw in chapter 6.3 and 6.5, lottery-voting is both decisive and neutral. Therefore, lottery-voting is never paralyzed, it is always able to produce a decision, and the status quo is not arbitrarily favoured. This is not, however, to say that the decision is independent of the path to it. If democracy is conceived of as a pure procedure, then this path independence is an impossible ideal. Arrow’s concern is that the winner of a cycle will depend on the order in which votes are taken. Lottery-voting proposes putting all the alternatives together, and giving each a fair chance according to the number of votes it attracts. It will still matter how motions are packaged, but the aim is to ensure fairness to everyone however this is done. This 614 615 Buchanan and Tullock (1962) p.32, quoted above (section 7.7). Arrow (1963 [1951]) p.120; this key passage is also quoted in Kelly (1978) p.22 and Chapman (1998) p.297 fn.8. 271 seems to resolve the indeterminacy of the cycle in a way that is fair to each person, but the solution depends on following the fair procedure rather than independently identifying some option as socially preferred. (7.10) Conclusion It has been argued that lottery-voting treats all voters fairly, by giving each an equal chance of being decisive on any given issue (see chapter 3). This gives us a powerful presumptive reason in its favour, but it could still be outweighed if it was to have disastrous effects. On the contrary, however, lottery-voting seems to satisfy minimal conditions for a social choice procedure (chapter 6) and cannot lead to outcomes that are bad for everybody – at least one person must consider them best – and it may therefore encourage moderation and compromise (see chapter 5.4). As for the process itself, it gives all involved reasons to participate in deliberation (trying to persuade as many as possible to vote their way), to vote (to maximize chances of their option, regardless of whether they’re in a majority or minority) and to do so responsibly (thus realizing the developmental benefits sometimes associated with participation616). These considerations show that we may have many reasons to choose lottery-voting, and I earlier sketched some examples of contexts where it may seem an appropriate decision-mechanism all-things considered (see chapter 4.6-7). Of course, many of the claims made are at least partly empirical. It could be that people see no reason to compromise, and prefer to make extreme demands knowing that they will have a chance of victory. One virtue of majority-rule is arguably that it encourages convergence on ‘middle ground’ – yet, taken too far, this means that 616 E.g. Mill (1998 [1861]b) esp. pp.226-7 and 243-56 [Considerations on Representative Government chs.2 and 3], and Pateman (1970) pp.103-11. 272 many never get their way and can even complain that – in effect – their votes do not count for anything since they know that they will always lose. Lottery-voting gives everyone a chance of satisfaction, but may risk polarization and so dramatic changes of policy. If these negative possibilities are borne out, then of course it may not be rational to employ lottery-voting. The claim here is not that lottery-voting is always appropriate – indeed, it was admitted in chapter 3.12 that, where it is fair, majorityrule may be a better procedure. However, it is not necessarily appropriate to assess the practical effects lottery-voting according to its likely results given our present political culture – a fair comparison would require us to consider people who had grown up under such a system and have had their expectations and demands shaped by it (to avoid this problem, I discuss a utopian example in my conclusion, 8.4). All that has been said is that, if at least some of my preceding arguments stand, then it may be rational to adopt lottery-voting. As with any other decision-mechanism its allthings-considered appropriateness depends on the specifics of the context. There is, however, no reason to simply dismiss lottery-voting out of hand as irrational, any more than there is to reject majority-rule because there is nothing especially rational about counting. If lottery-voting seems an appropriate way to realize political equality in our circumstances then it may be both democratic and rational. 273 Conclusion “[T]here are myriad ways of making group choices… each has some merit in some circumstances, but the rationale for none seems decisive or compelling in all situations”617 “A voting procedure and an electoral system are not… mere pieces of mechanism: they enshrine general ideals of what democracy is”618 (8.1) Summary of the Argument To recap, it was first argued that democracy is a matter of political equality and that there is nothing naturally privileged about majority-rule. It follows that, in designing decision rules that treat each person equally, it may be helpful to think in contractualist terms. Chapter 2 rejected a range of ‘utilitarian’ arguments purporting to justify majoritarianism. We cannot be sure that majority-rule will produce better outcomes or the ‘will of the people’. Although it may be that all generally benefit from democracy, because central coordination is better than the ‘state of nature’, there is often disagreement over what the collective arrangements should be. Political equality is therefore justified on the grounds of fair distribution. Chapter 3 argued that we treat individuals equally when we give each group proportionate chances of success. While majority-rule may be fair if no-one can predict the composition of the majority or minority in advance, its fairness is undermined if there are persistent losers who know if advance of any vote that they will never get their way. This completes the first part of the thesis, which offers a general normative analysis of democracy without any immediate institutional implications. Chapter 4 begins the second part of the thesis, describing how lotteries can be incorporated within a chance-based democratic procedure called lottery-voting. This 617 618 Shepsle and Bonchek (1997) p.167 [emphasis original] Dummett’s (1998) p.xvii [not emphasized in original]. 274 prescription follows from the normative considerations developed in chapters 1-3, so treats everyone equally. Chapter 5 outlines some further practicalities and takes up some immediate objections, illustrating how lottery-voting could be institutionalized. Having described this alternative to majority-rule, the last two chapters offer a defence. Chapter 6 considers whether lottery-voting meets even minimal conditions of social choice, and argues that it does generally satisfy the normatively appealing conditions, though – following the argument of the first part – lottery-voting should be seen simply as a pure decision-making procedure, rather than a way of implementing a social welfare function. Finally, chapter 7 takes up the objection that relying on a lottery is somehow irrational. It was argued that no decision-procedure is inherently (ir)rational; all we can assess is the rationality of our adopting a given procedure to resolve a given disagreement and my earlier arguments, if successful, give us rational reasons to adopt lottery-voting in certain contexts. (8.2) The Importance of Thought Experiments Note that the preceding argument has been entirely theoretical or philosophical. While there has been some speculation about the possible effects of lottery-voting, for instance the incentives it creates for deliberation, I have not offered any empirical evidence as to its likely effects. It would, of course, be impossible to say how lotteryvoting would work in practice without widespread trials619. Further, it would be somewhat unfair to test lottery-voting in a political culture shaped by majoritarian institutions (where people expect to win simply because they are in the majority). Any proposal to adopt lottery-voting would therefore have to be very cautious. If 619 I am aware of few instances of lottery-voting being used in practice, though it was used to decide whether to go to the King’s Arms or University Club following my presentation to the Graduate Political Theory Workshop 10/05/07. The decision was the King’s Arms. 275 widespread attitudinal change is required first, it is not clear how the revolution could come about, and transitional difficulties are something else I have not addressed at any length here. The theoretical argument – that lottery-voting can, in suitable conditions, be an appropriate (and probably best) decision-mechanism – should not be mistaken for practical advocacy of lottery-voting in any actual context, though I do think there may be actual cases where it could work. Given the lack of practical implications, one may be excused for wondering what is important or interesting about this thesis. Political philosophy is, however, a branch of philosophy and, as such, values knowledge intrinsically rather than for merely instrumental benefits. Conceptual analysis and clarification is interesting in its own right, even if devoid of immediate practical implications. Moreover, we have seen that democracy does not require majority-rule, and this may open the possibility for other non-majoritarian forms of democracy. Perhaps more importantly, however, it also means that majority-rule needs theoretical defence, that should be distinguished from a more general defence of democracy, and should not beg the question against rival forms of democracy such as lottery-voting. Lottery-voting is rarely considered at all, and when it is it is usually done so very briefly620. If we remain aware of all democratic possibilities, however, then this complacency seems to limit many works of democratic theory. Amar is explicit that his advocacy of lottery-voting, the only prior such defence known to me, is only a thought experiment, not a practical policy proposal621. Thus, while the claim that there may be a few, perhaps relatively trivial, cases (such as those in chapter 4.6-7) where lottery-voting seems a particularly appealing decisionmechanism will no doubt strike many as rather modest, the importance of lottery620 621 See introduction, section 0.4. Amar (1984) p.1308, and Amar (1995); c.f. Duxbury (1999) p.78. 276 voting extends beyond its practical role – the argument for it also challenges the way we think about democracy in general. That two different ideals, say utilitarianism and Rossian deontology, or egalitarianism and prioritarianism, will converge in a given case – or even all practically likely cases – does not end philosophical discussion. While the practical impossibility of an ideal certainly invalidates it, if we accept the principle that ought implies can, the mere fact that an ideal would be difficult to implement does not diminish its validity as an ideal, only mean that it cannot be put into practice as is622. But one need not find the idea appealing in order to find it philosophically important; we cannot reject an unpalatable idea simply because of prejudice – we need reasons. As Duxbury records in his Acknowledgements: “I was not especially interested in… randomized social decisionmaking. In the course of teaching, however, I found myself increasingly asking students why a randomized solution to one or another legal issue would not have been appropriate. Little by little, the notion of decision-making by lot was reining me in”623 To take an analogy, Michael Tooley argues for a moral equivalence between abortion and infanticide624, since neither foetus nor young child have a self-conception of themselves as continuing subjects of experience that, he claims, grounds the right to life. One need not be in favour of infanticide to recognise the importance of this argument as a challenge to one who wants to justify abortion but not infanticide. So it is with many other philosophical arguments, such as claims that utilitarianism could sanction slavery625, or in its ‘negative’ (suffering-minimising) form would sanction the destruction of the world626. One need not endorse any of these conclusions, for the 622 Cohen, in speech, makes the same response to many of Anderson (1999)’s objections to luck egalitarianism. See also Estlund (2007) pp.12-5 and 258-75. 623 Duxbury (1999). There are no page numbers given prior to the introduction, but this would be p.vii. 624 Tooley (1972). 625 Rawls (1999 [1971]) pp.137, Kymlicka (2002 [1990]) pp.140-1. 626 Smart (1958). 277 arguments can be employed as reductios of opposing positions, or at least challenges to opponents to identify a relevant difference and thus explain why the undesirable consequences do not follow from their position. So it is with lottery-voting. One does not need to be attracted to the idea, all things considered, to see it as stimulating argument. Of course, one can draw conclusions either by modus ponens (if P then Q, P, therefore Q) or modus tollens (if P then Q, not-Q, therefore not-P) – i.e. one can either accept the ‘equal chances’ criterion and so consequently lottery-voting or one may find lottery-voting unappealing and therefore reject ‘equal chances’ as a goal. However, those who argue that democracy requires political equality, and accept that this should be understood in terms of each vote having an equal chance of determining the election, must either accept lottery-voting or find some further grounds on which they can reject it. Thus the rejection of lottery-voting requires further argument, which has not generally been given because the possibility has not been envisaged or – if it has – taken seriously. The main findings of this research are that democracy should not be confused with majority-rule. There is certainly no analytic or conceptual identity between the two and, further, the possibility of lottery-voting as a way of treating all voters equally, even if not all-things-considered desirable, shows that political equality does not require majority-rule as a synthetic or practical matter either. (8.3) Practical Possibilities I have offered lottery-voting as a theoretical argument, yet chapter 5 described how it might work in practice. One may justifiably wish to hear more about potential practical implications, but the effects of implementing lottery-voting would, of course, depend on the people involved and their preferences, as should presumably be 278 expected of any democratic theory that puts power in the hands of the people. Institutions do not operate in a vacuum, but interact with the political culture and a system that suits some people may be radically unsuitable for others (indeed, this might even apply to democracy as a whole). It should be remembered that lotteryvoting is primarily a theoretical, rather than practical, proposal, designed to shed light on arguments in democratic theory. What I claim here is that lottery-voting could be a workable and attractive democratic method, not that it is something we – here and now – should consider adopting. As a practical policy proposal, lottery-voting suffers several serious defects. Firstly, any attempt at constitutional voting reform would face significant transition costs. Even though there may not be much need to replace or update much capital infrastructure – as there was in the face of, say, the Y2K bug or currency changes – there is still need for widespread public consultation and later awareness and information campaigns. Secondly, and more significantly, however, people are very much embedded in the present system and resistant to new ideas. While I have emphasized that randomization is not in fact anything novel when we look at the longer history of democracy (see 4.2, above), lottery-voting would still seem a strange, alien system to many. Those who have been brought up to believe that democracy requires – or even simply means – majority-rule would find it hard to accept it if a member of a minority was selected. From the fact that possible people, in other ideal circumstances, could accept this and operate the system quite successfully, it does not follow that lottery-voting would work for the people as we have them – at least, not until after a lengthy period of adaptation in which they had seen it work. One problem, for instance, is that sensible voting behaviour may be 279 endogenous to the system (see chapter 5.3d), so if people vote as they are accustomed to under lottery-voting, it may have disastrous consequences. It should be remembered, therefore, that lottery-voting is not advocated as a practical proposal. I am not proposing that we should do away with our present electoral systems in favour of such a radical change. It may be, however, that we could experiment with lottery-voting in some of the small group decision making contexts that I have described. For instance, if five friends are trying to decide on a restaurant to go to for dinner, then a ‘random dictator’ method should be seen not only as a simple decision mechanism, but a fair and democratic one. Further, it may be worth giving lottery-voting serious consideration in certain decision-making contexts, particularly those that seem to raise problems for our present procedures, as illustrated by the examples in chapter 4.6-7, above. If we familiarize people with the system through such small, local examples, then maybe in time it could be extended further and further, and perhaps ultimately our national politics could operate according to such a system – though further questions concerning representation would have to be answered that I do not address here. Any such transition to widespread use of lottery-voting would have to be carefully and gradually implemented, however. Whatever one thinks about these practical speculations, and I stand by the claim that lottery-voting can be practical in certain circumstances, even if the transition costs would usually rule out any calls for implementing it where unfamiliar, I put lottery-voting forward as a theoretical possibility to illuminate – and refute – certain arguments about democracy and majority-rule, for instance that only majority-rule treats each vote equally, or perhaps as a reductio ad absurdum of the claim that each 280 vote should have an equal chance of determining the outcome of an election627. The implications the reader wishes to draw will depend on how attractive they find lottery-voting in particular and democracy in general. Nonetheless, while lotteryvoting may not be immediately practical for us, it is important to show that it could be used in practice, because democracy is inherently a way of making decisions. I wish to show that it is potentially practical with a utopian sketch of how lottery-voting could operate in some other ideal society, where all accept it as not only a possibility, but the (most) fair, democratic decision method. (8.4) A Utopian Example The society of Aleatoria628 had employed lottery-voting as long as Felicity629 could remember. She was aware from her history books that, once upon a time, representatives had simply been appointed directly by lot. Indeed, there were stories that said the introduction of the vote had caused some controversy, when it became clear that majorities could lose – but the disputes had soon been resolved when everyone realized there was no reason why the majority had to win all the time anyway. Now no one believes in the old superstition that there is any inherent superiority in mere numbers and all accept the ancient lottery as the fairest way to adjudicate between the votes of equals. Even though Aleatoria was marked by clear divisions on religious lines, the lottery had allowed both Sun- and Moon-worshippers the chance to have their way on any given issue, with the long run result not only that each sometimes got what they 627 For references to all having some form of equal chance, see Guinier (1994) p.72, Christiano (1996) p.55, Beitz (1983) p.72, Nelson (1980) p.19; see my introduction, sections 0.1 and 02. 628 This is the name used for a futuristic lot-based society in Goodwin (2005 [1992]) p.3ff, but it is derived from the word ‘aleatory’ (dependent on chance), which comes from the Latin for dice (alea). 629 This existing girl’s name is associated with happiness or good fortune, i.e. luck. 281 wanted but also that each tasted defeat, and this encouraged a degree of moderation and compromise. Indeed, when it came to the controversies over religious education in schools that had taken place around 30 years ago, neither side had wanted to risk being excluded as a result of the lottery, and the result was that all had supported a proposal to expose children to both religious views, but not allow them to be confirmed into either until they reached adulthood. Felicity had thus been taught the central tenets of both – including how the Moon-worshippers believed the lottery to be controlled by Luna’s divine hand, thus directly guiding public policy, while the Sun-worshippers (and, of course, atheists and others) did not believe in any need for Godly intervention, but saw the lottery as simply a sensible invention of human reason to deal with earthly, human problems. The rationale that justified the lottery to each group was not important, however, given that all accepted it. Felicity had recently been asked for her vote on a number of issues in her local constituency, one concerning a new road, one about public transport and a third about whether to build a children’s play area on a patch of existing parkland. Being a nondriver with young children, she had voted for improved bus services and the play area, but against the road. As it happened, most people wanted the new road, and – with the blessing of the lottery – they got their way. Though only a minority were in support of public transport links, they too got their way, so new bus services were to be implemented. Finally, the proposed play area had so divided opinion that in the end, after much discussion, a new suggestion had been advanced – that the park was preserved but a previous ‘brownfield’ site be developed into a youth and activity centre. This had won majority support, but a substantial minority objected to the cost, and in the end it was one of their votes that prevailed. Still, Felicity consoled herself 282 that the money saved was that being spent on the improved bus links, and she would still be able to enjoy the existing park with her son. Of course, lottery-voting had not always run smoothly in Aleatoria, but then, what political system does? Felicity had been taught in school about a time long ago when, after a run of significant decisions all going the way of the same majority, there had been accusations that the machinery must be somehow rigged – ‘how could it be democratic that the others never got their say?’ people asked. Much scrutiny of the random number generators had taken place, and it had been alleged in some quarters that not all of the minority votes had been placed in the ballot. Nonetheless, nothing had ever been proven, and most likely these complaints were merely the grievances of those who did not get their way. Extensive trials had suggested that the machines were indeed fair, and full disclosure of voting details produced no evidence of any tampering. In any case, in time the issue died down, as several later decisions went the way of the formerly aggrieved group. While some conspiracy theorists continued to maintain that this was, in fact, a deliberate cover-up and that the numbers had always been controlled by secretive forces who would accept a few defeats only in order to retain their undiscovered control, few people believed these wild rumours and, at the end of the day, provided that all got their way sometimes, perhaps it hardly mattered. As she went to cast her latest vote, praying that one from her side was selected, Felicity reflected on the benefits that the lottery brought. No group in Aleatoria could complain of being permanently excluded by the system, and conversely none could complacently assume that any given decision was to go their way unless their case was so clearly reasonable as to win over everyone. If anyone dared suggest that the Sun-worshippers should have exclusive control over education – or any other policy – 283 simply because they were more numerous, then they would have been laughed away. Of course, it was accepted that de facto more decisions might go their way, but this was only because the lottery was duly responsive to the proportionate weight of numbers – as had proved sensible many times in the past, when Moon-worshippers had been able to persuade enough of their rivals to change sides to win a majority, and thus had more usually won the day. The idea that any other system would treat all on both sides equally would seem absurd. (8.5) The Place of Democracy To return to theoretical considerations, however, it should be stressed that I am concerned with investigating one particular notion, viz. democracy. My claim is that it would be perfectly democratic for a given group to make decisions about any given policy by lottery-voting (without necessarily excluding the claim that there may be other – and possibly better – democratic ways of making this decision). This is not, however, intended as an all-things-considered endorsement of the practice for, despite what some may have you believe, democracy is not all good things. It would, for example, be democratic to decide whether to invade a hostile Arab state by lotteryvoting630, but this need not be even a prima facie reason for holding such a vote. I have said nothing about what decisions should be taken democratically, and it remains open that this research will actually, by introducing a more radical conception of democracy, make us less keen on such. 630 Technically for it to be democratic requires a wide franchise, but I have not here commented on who should have a vote. Democracy involves both a wide franchise and the equality of voters, and I have been concerned with only the second of these. My claim would be that, for any given group (even an elite oligarchy), decision-making is more democratic when that group decide as equals. Thus, there would be something democratic about the decision being made by a small group of MPs or generals, deciding by lottery-voting. 284 I should, therefore, perhaps say something about why democracy might be thought of as important or valuable. I have rejected the idea that democracy is to be thought of as simply instrumental to justice, as some imperfect proceduralists would have it. The relation between justice and democracy is a complex one, that cannot be set out here, but hopefully some brief remarks on how I see the matter will clarify my position. Justice makes certain demands of us, which might for instance include equal opportunity for members of racial minorities, whatever the population think of them. This may run counter to democratic decision-making (even where that is not simply majoritarian), but this is an occasion where it is permissible – indeed, generally good – to limit democracy, e.g. through the establishment of a bill of rights and court that upholds non-discrimination. The requirements of justice, as I see them, are however relatively minimal or at least too indeterminate to settle many issues on which we want co-ordinated solutions (see chapter 1.2). This means there is considerable scope for democratic decision-making within the (wide) limits set by justice and, within this sphere of discretion, democratic decision-making can be legitimately treated as a ‘pure’ (or, because of the limit, ‘quasi-pure’) procedure631. While any outcome in this range can be regarded as ‘permitted by justice’ in the abstract, this does not make the choice one of indifference, since different individuals may have different preferences or interests in the matter (see chapter 1.2, 2.8 and 3.2). Democracy is therefore a requirement of (procedural) justice, because it is important that each person’s interests are considered equally. To give a more concrete example: suppose that the two of us have to distribute five beans between us. Other things (such as desert) being equal, we could rule out the distributions (4,1) or (0,5), either way round, as unjust, and these would be the kind of things prohibited by 631 Rawls (1999 [1971]) pp.176 and 318, McGann (2006) pp.80-3. 285 constitutional rights, taking such outcomes off the democratic agenda. Nonetheless, the demands of justice are not fully determinate, because the outcomes (3,2) and (2,3) might be regarded as equally correct – and even levelling down to (2,2) might also be regarded as reasonable. Since it matters to me whether I get two or three, it’s important that there is a fair process to decide between these equally just possibilities. In this case, since there are only two of us, it seems fair to toss a coin between (3,2) and (2,3). Some may regard this as only a matter of tie-breaking, and hold that an extra person’s claim would break such a tie – so, for instance, we should favour (3,3,2) to (2,2,3) – but I think it is only a particular instance of lottery-voting and that, in this three person case, it would be fair to give the third individual a one-third chance (see chapter 3.6-8). On this conception, while justice sets limits within which democracy can operate, it is not wholly independent and, in fact, democracy is a constitutive (rather than merely instrumental) part of a fully just society, one that respects each individual person equally. This may seem to subordinate democracy to justice, but it is important to note that justice cannot be achieved without democracy, for instance by Platonic Guardians. Moreover, while I have argued that democracy may be a requirement of justice, this is not to say that its value is exhausted by its contribution to justice. It is consistent with my argument here – and, indeed, something I believe – that democracy has educative effects on citizens. I believe (though, again, these views cannot be argued here) that agency is an important part of human well-being, broadly understood. Since we are inter-dependent, this agency must be not only individual but collective; all of us can benefit from exercising agency in questions of how we can live together. While lottery-voting is supposed to alleviate the excessive demands of complete impartiality (see chapter 2.7), it does require us to exercise a certain 286 responsibility in how we use our vote (see chapter 5.3d), knowing that whatever outcome we vote for could eventuate, and allow us to trace a connection between collective outcomes and our inputs632. Mill claims that every “voter is under an absolute moral obligation to… give his vote to the best of his judgement, exactly as he would be bound to do if he were the sole voter, and the election depended on him alone”633 and lottery-voting is one way of requiring them to take this seriously, since any one individual’s vote could in fact determine the outcome. 632 Though I will never know if it was actually my vote that led to a decision, if it goes the way I voted then I will be able to assess the outcome knowing it could have been mine, and therefore be able to judge whether or not my vote was wise. 633 Mill (1998 [1861]b) p.355 [Considerations ch.10]; c.f. Rawls (1999 [1971]) p.120: “we can view the agreement in the original position from the standpoint of one person selected at random”. 287 Bibliography (B.1) Books and Articles This is not an exhaustive list of works on an extensive topic. Rather, this records works I have consulted, either in the course of this project itself or beforehand and deemed relevant. However, while some have been read cover to cover, in others only a few pages have been consulted. Nor have they all been cited within the main text. R. P. Abelson (1995) ‘The Secret Existence of Expressive Behavior’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 25-36 B. Ackerman (1980) Social Justice in the Liberal State (New Haven: Yale UP) M. Allingham (2002) Choice Theory: A very short introduction (Oxford: Oxford UP) G. E. M. Anscombe (1976) ‘On the Frustration of the Majority by the Fulfilment of the Majority’s Will’ Analysis 36:4 161-168 A. R. Amar (1984) ‘Choosing Representatives by Lottery Voting’ The Yale Law Journal 93 1283-1308 A. R. Amar (1995) ‘Lottery Voting: A Thought Experiment’ University of Chicago Legal Forum 193-204 D. Andalo (2007) ‘Tories attack council's plan to allocate school places by lottery’ The Guardian (Wed 28/02/07). Available at: http://education.guardian.co.uk/admissions/story/0,,2023209,00.html (last accessed 04/03/08) E. Anderson (1999) ‘What is the Point of Equality?’ Ethics 109:2 287-337 J. Anderson with S. Done (2004 [2002]) The Official Liverpool FC Illustrated History 2nd edition (London: Carlton) Anonymous (2006) ‘The Human Touch: Selecting junior doctors online makes patient care a lottery’ [leading article] The Times (Sat 04/03/06) 23. Available at: http://www.timesonline.co.uk/tol/comment/leading_article/article737222.ece (last accessed 15/04/07) G. E. M. Anscombe (1957) Intention (Oxford: Basil Blackwell) Aristophanes (1919) The Frogs of Aristophanes translated B. B. Rogers (London: G. Bell & Sons) Aristotle (1959) Aristotle’s Politics; and the Athenian Constitution translated J. Warrington (London: Dent) Aristotle (1988) The Politics translated B. Jowett, revised J. Barnes, edited S. Everson (Cambridge: Cambridge UP) Aristotle (1991) On Rhetoric: A theory of civic discourse translated G. A. Kennedy (Oxford: Oxford UP) Aristotle (1992) The Politics translated T. A. Sinclair, revised T. J. Saunders (London: Penguin) Aristotle (1996) Nicomachean Ethics translated H. Rackham (Ware, Herts: Wordsworth) Aristotle (1999 [1985]) Nicomachean Ethics 2nd edition. Translated T. Irwin (Indianapolis and Cambridge: Hackett) Aristotle (1999) Politics Books V and VI translated and commentary D. Keyt (Oxford: Clarendon) R. J. Arneson (1990) ‘Liberalism, Distributive Subjectivism, and Equal Opportunity for Welfare’ Philosophy and Public Affairs 19:2 158-94 288 R. J. Arneson (1997) ‘The Priority of the Right Over the Good Rides Again’ Ethics 108:1 169-196 G. Arrhenius (2007) ‘The Boundary Problem in Democratic Theory’ [Draft 070312] Available at: http://www.scasss.uu.se/pdf/the_boundary_problem_in_democratic_theory.pdf (last accessed 13/08/07). This updates an earlier paper, of the same title, published in F. Tersman (ed.) (2005) Democracy Unbound: Basic Explorations I (Stockholm: Filosofiska institutionen, Stockholms Universitet) K. J. Arrow (1963 [1951]) Social Choice and Individual Values 2nd edition (New Haven: Yale UP) K. J. Arrow (1973) ‘Some Ordinalist-Utilitarian Notes on Rawls’ Theory of Justice’ The Journal of Philosophy 70:9 245-263 K. J. Arrow (1984) Collected Papers of Kenneth J. Arrow. Volume 1 Social Choice and Justice (Oxford: Basil Blackwell) K. J. Arrow (1997) ‘The Functions of Social Choice Theory’ in K. J. Arrow, A. Sen and K. Suzumura (eds.) (1997) Social Choice Re-Examined. Volume 1 (Basingstoke: Macmillan Press) K. J. Arrow and H. Raynauld (1986) Social Choice and Multicriterion DecisionMaking (Cambridge, Mass.: MIT Press) K. J. Arrow, A. Sen and K. Suzumura (eds.) (1997) Social Choice Re-Examined. Volume 1 (Basingstoke: Macmillan Press) A. J. Ayer (1954) Philosophical Essays (London: Macmillan) M. L. Balinski and H. P. Young (1982) Fair Representation: Meeting the Ideal of One Man, One Vote (New Haven: Yale UP) J. F. Banzhaf III (1965) ‘Weighted Voting Doesn’t Work: A Mathematical Analysis’ Rutgers Law Review 19 317-343 E. Barker (1918) Greek Political Theory: Plato and his predecessors (London: Methuen & Co.) E. Barker (1942) Reflections on Government (Oxford: Oxford UP) A. Barnett and P. Carty (1998) The Athenian Option: Radical reform for the House of Lords (London: Demos) B. Barry (1965) Political Argument (London: Routledge and Kegan Paul) B. Barry (1991) Democracy, Power and Justice: Essays in political theory volume I (Oxford: Clarendon) L. Becker and C. B. Becker (eds.) (2001) Encyclopedia of Ethics 2nd edition (New York: Routledge) D. Beetham and K. Boyle (1995) Introducing Democracy: 80 questions and answers (Cambridge: Polity Press) C. R. Beitz (1983) ‘Procedural Equality in Democratic Theory: A Preliminary Examination’ in J. R. Pennock and J. W. Chapman (eds) (1983) NOMOS XXV: Liberal Democracy (New York: New York UP) J. Bentham (1831) Parliamentary Candidate’s proposed Declaration of Principles: or say, a test proposed for parliamentary candidates (London: Robert Heward) J. Bentham (1843 [1827]) ‘Rationale of Judicial Evidence, specially applied to English practice’ in The Works of Jeremy Bentham volume VII, edited J. Bowring (Edinburgh: William Tait) J. Bentham (1973) Bentham’s Political Thought edited B. Parekh (London: Croom Helm) J. Bentham (1983 [1822-32]) Constitutional Code volume I, edited F. Rosen and J. Burns (Oxford: Clarendon Press) 289 B. R. Berelson, P. F. Lazarsfeld and W. N. McPhee (1954) Voting: A study of opinion formation in a presidential campaign (Chicago: U of Chicago P) E. Berg (1965) Democracy and the Majority Principle: A study in twelve contemporary political theories (Göteborg, Sweden: Akademiförlaget) F. Berger (1984) Happiness, Justice and Freedom: The moral and political philosophy of John Stuart Mill (Berkeley: U of California P) I. Berlin (1964) ‘Rationality of Value Judgments’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) C. Bicchieri (2004) ‘Rationality and Game Theory’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) A. Birch (2001) The Concepts and Theories of Modern Democracy 2nd edition (London: Routledge) D. Black (1998 [1958]) The Theory of Committees 2nd edition, edited I. McLean, A. McMillan and B. L. Monroe (Boston: Kluwer Academic Publishers) A. Blais (2000) To Vote or Not to Vote?: The merits and limits of rational choice theory (Pittsburgh: U of Pittsburgh P) J. Bohman and W. Rehg (eds.) (1997) Deliberative Democracy: Essays on reason and politics (Cambridge, Mass.: MIT) G. Bordes (1979) ‘Some more Results on Consistency, Rationality and Collective Choice’ in J.-J. Laffont (ed.) (1979) Aggregation and Revelation of Preferences (Amsterdam: North-Holland Publishing Company) J. L. Borges (1970 [1941]) ‘The Lottery in Babylon’ trans. J. M. Fein, in D. A. Yates and J. E. Irby (eds.) (1970) Labyrinths: Selected stories and other writings (Harmondsworth: Penguin) D. Bostock (2000) Aristotle’s Ethics (New York: Oxford UP) R. B. Brandt (1979) A Theory of the Good and the Right (Oxford: Oxford UP) G. Brennan and L. Lomasky (1993) Democracy and Decision: The pure theory of electoral preference (Cambridge: Cambridge UP) H. Brighouse (2000) ‘How lotteries could make school admissions a whole lot fairer’ The Independent (Fri 08/06/00). Available at: http://education.independent.co.uk/news/article258775.ece (last accessed 15/04/07) H. Brighouse & M. Fleurbaey (2006) ‘On the Fair Allocation of Power’ [Draft of Feb 2006] Available at: http://mora.rente.nhh.no/projects/EqualityExchange/Portals/0/articles/brighouse fleurbaeymarch2006.pdf (last accessed 13/08/07). I also benefited from discussion with Marc Fleurbaey when he presented an earlier version to the Nuffield Political Theory Workshop (April 2005) D. W. Brock (1998) ‘Aggregating Costs and Benefits’ Philosophy and Phenomenological Research 58:4 963-967 J. Broome (1984a) ‘Uncertainty and Fairness’ The Economic Journal 94:375 624-632 J. Broome (1984b) ‘Selecting People Randomly’ Ethics 95:1 38-55 J. Broome (1991) Weighing Goods: Equality, uncertainty and time (Oxford: Blackwell) J. Broome (1998) ‘Kamm on Fairness’ Philosophy and Phenomenological Research 58:4 955-961 J. Broome (1999) Ethics Out of Economics (Cambridge: Cambridge UP) J. Buchanan and G. Tullock (1962) The Calculus of Consent (Ann Arbor: U of Michigan P) J. Burley (ed.) (2004) Dworkin and his Critics (Oxford: Blackwell) 290 J. Burnheim (1985) Is Democracy Possible?: The alternative to electoral politics (Cambridge: Polity Press) J. Butler (1906 [1736]) The Analogy of Religion, Natural and Revealed (J. M. Dent: London) E. Callenbach and M. Phillips (1985) A Citizen Legislature: a modest proposal for the random selection of legislators (Berkeley, CA: Banyan Tree Books) J. Carens (1990) ‘Difference and Domination: Reflections on the Relation between Pluralism and Equality’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) E. F. Carritt (1947) Ethical and Political Thinking (Oxford: Clarendon) L. Carson and B. Martin (1999) Random Selection in Politics (Westport, Conn.; Praeger) S. Carter (1994) ‘Foreword’ to L. Guinier (1994) The Tyranny of the Majority: Fundamental fairness in representative democracy (New York: Free Press) R. Chang (ed) (1997) Incommensurability, Incomparability and Practical Reason (Cambridge, Mass.; Harvard UP) B. Chapman (1998) ‘More Easily Done than Said: Rules, Reason and Rational Social Choice’ Oxford Journal of Legal Studies 18 293-330 B. Chapman (2000) ‘Chance, Reason, and the Rule of Law’ University of Toronto Law Journal 50:4 4469-492 J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) D. Chong (1995) ‘Rational Choice Theory's Mysterious Rivals’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 37-58 T. Christiano (1990) ‘Political Equality’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) T. Christiano (1996) The Rule of the Many: Fundamental issues in democratic theory (Boulder, Colo.: Westview Press) T. Christiano (ed.) (2003) Philosophy & Democracy: An anthology (Oxford: Oxford UP) G. Christos (1995) ‘Democracy by Random Selection’. Available at: http://www.maths.curtin.edu.au/~christos/randomdemocracypaper.shtml (accessed 26/06/04) W. Churchill (1974) Winston S. Churchill: His Complete Speeches, 1897-1963 volume VII (1943-9), edited R. R. James (New York: Chelsea House Publishers) M. Clayton and A. Williams (eds.) (2000) The Ideal of Equality (Basingstoke: Palgrave Macmillan) J. Cobb (1994) ‘The Real Threat to U.S. Sovereignty’ Heritage Lecture #497 01/08/94. Available at: http://www.heritage.org/Research/InternationalOrganizations/HL497.cfm (accessed 30/01/05) G. A. Cohen (1989) ‘On the Currency of Egalitarian Justice’ Ethics 99:4 906-944 G. A. Cohen (1995) Self-Ownership, Freedom, and Equality (Cambridge: Cambridge UP) G. A. Cohen (1997) ‘Where the Action is: On the Site of Distributive Justice’ Philosophy and Public Affairs 26:1 3-30 G. A. Cohen (2004) ‘Expensive Tastes Ride Again’ in J. Burley (ed.) (2004) Dworkin and his Critics (Oxford: Blackwell) G. A. Cohen (forthcoming) Rescuing Justice and Equality 291 J. Cohen (1997) ‘Deliberation and Democratic Legitimacy’ in J. Bohman and W. Rehg (eds.) (1997) Deliberative Democracy: Essays on reason and politics (Cambridge, Mass.: MIT) L. Colley (1992) Britons: Forging the nation 1707-1837 (New Haven: Yale UP) J. Colomer and I. McLean (1998) ‘Electing Popes: Approval Balloting and QualifiedMajority Rule’ Jounral of Interdisciplinary History 29:1 1-22 W. E. Connolly (1993 [1974]) The Terms of Political Discourse 3rd edition (Oxford: Blackwell) D. Copp, J. Hampton and J. Roemer (eds.) (1993) The Idea of Democracy (Cambridge: Cambridge UP) P. Crawford (2001) ‘‘The poorest she’: women and citizenship in early modern England’ in M. Mendle (ed.) (2001) The Putney Debates of 1647: The army, the Levellers and the English state (Cambridge: Cambridge UP) R. Crisp (2000) ‘Value… And What Follows’ [book review] Philosophy 75 458-62 R. A. Dahl (1985) A Preface to Economic Democracy (Berkeley: University of California Press) R. A. Dahl (1989) Democracy and its Critics (New Haven: Yale UP) R. A. Dahl (1998) On Democracy (New Haven: Yale UP) R. A. Dahl (2006) On Political Equality (New Haven: Yale UP) R. A. Dahl and C. Lindblom (1976 [1953]) Politics, Economics and Welfare: Planning and politico-economic systems resolved into basic social processes (Berkeley: University of California Press) N. Daniels (ed.) (1989 [1975]) Reading Rawls: Critical studies on Rawls’ A Theory of Justice (Stanford, CA: Stanford UP) N. Daniels (1998) ‘Kamm’s Moral Methods’ Philosophy and Phenomenological Research 58:4 947-954 S. L. Darwall (1983) ‘Equal Representation’ in J. R. Pennock and J. W. Chapman (eds.) (1983) NOMOS XXV: Liberal Democracy (New York: New York UP) S. L. Darwall (1998) Philosophical Ethics (Boulder: Westview) H. de Quetteville (2006) ‘The Regretful Fatah Voters who took Poll ‘Game’ too far’ The Daily Telegraph Fri 27/01/06 p.5. Available online at: http://www.telegraph.co.uk/news/main.jhtml;jsessionid=F3IU5T1CE5DKTQFI QMFCFGGAVCBQYIV0?xml=/news/2006/01/27/wmid427.xml (last accessed 19/08/07) A. De Tocqueville (1994 [1835]) Democracy in America (London: Everyman’s Library) A. Dechert (2003) talk given at the UC Santa Cruz forum on electronic voting held 26/10/03. Available at: http://openvotingconsortium.org/ad/oct26forum.html (accessed 30/01/05) P. K. Dick (2003 [1955]) The Solar Lottery (London: Gollancz) D. Diermeier (1995) ‘Rational Choice and the Role of Theory in Political Science’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 59-70 G. Dietze (1964) ‘The Limited Rationality of Law’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) K. Dowding, R. E. Goodin and C. Pateman (eds.) (2004) Justice and Democracy: Essays for Brian Barry (Cambridge, Cambridge UP) O. Dowlen (2004) ‘Sortition in Democratic Theory and Practice’ (unpublished D.Phil material) 292 O. Dowlen (2005) ‘Sortition in England and America in the Seventeenth and Eighteenth centuries’ (unpublished D.Phil material) A. Downs (1985 [1957]) An Economic Theory of Democracy (Boston: AddisonWesley Publishing) J. Drier (2004) ‘Decision Theory and Morality’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) J. Dryzek (1990) Discursive Democracy: Politics, policy, and political science (Cambridge: Cambridge UP) J. Dryzek (2000) Deliberative Democracy and Beyond: Liberals, critics, contestations (Oxford: Oxford UP) M. Dummett (1984) Voting Procedures (Oxford: Clarendon) M. Dummett (1997) Principles of Electoral Reform (Oxford: Oxford UP) M. Dummett (1998) ‘Foreword’ to P. J. Emerson (1998) Beyond the Tyranny of the Majority (Belfast: Da Borda Institute) N. Duxbury (1999) Random Justice: On Lotteries and Legal Decision-Making (Oxford: Clarendon Press) R. Dworkin (1977) Taking Rights Seriously (London: Duckworth) R. Dworkin (1986) Law’s Empire (Cambridge, Mass.: Belknap Press of Harvard UP) R. Dworkin (2000) Sovereign Virtue: The theory and practice of equality (Cambridge, Mass.: Harvard UP) R. Dworkin (2004) ‘Hart’s Postscript and the Character of Political Philosophy’ Oxford Journal of Legal Studies 24:1 1-37 R. Dworkin (2005) reply to Guest (2005). Available http:// www.ucl.ac.uk/spp/download/seminars/ 0405/Integrity-equality-and-justice.rtf (accessed 15/03/05) J. Elster (1979) Ulysses and the Sirens: Studies in rationality and irrationality (Cambridge: Cambridge UP) J. Elster (1989) Solomonic Judgements: Studies in the limits of rationality (Cambridge: Cambridge UP) J. Elster (1993) ‘Majority Rule and Individual Rights’ in S. Shute and S. Hurley (eds.) (1993) On Human Rights. The Oxford Amnesty Lectures 1993 (New York: Basic Books) J. Elster and J. Roemer (eds.) (1991) Interpersonal Comparisons of Well-Being (Cambridge: Cambridge UP) P. J. Emerson (1998) Beyond the Tyranny of the Majority (Belfast: Da Borda Institute) D. Estlund (1997) ‘Beyond Fairness and Deliberation: The Epistemic Dimension of Democratic Authority’ in J. Bohman and W. Rehg (eds.) (1997) Deliberative Democracy: Essays on reason and politics (Cambridge, Mass.: MIT) D. Estlund (ed.) (2002) Democracy: Blackwell readings in philosophy (Oxford: Blackwell) D. Estlund (2007) Democratic Authority: A philosophical framework (Princeton, NJ and Oxford: Princeton UP) J. P. Euben, J. R. Wallach and J. Ober (eds.) (1994) Athenian Political Thought and the Reconstruction of American Democracy (Ithaca: Cornell UP) H. Eulau (1964) ‘Logics of Rationality in Unanimous Decision-Making’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) M. Farrand (ed.) (1966 [1911]) The Records of the Federal Convention of 1787 volume 1, revised edition (New Haven: Yale UP) 293 J. Feinberg (1973) Social Philosophy (Englewood Cliffs: Prentice Hall) D. S. Felsenthal and M. Machover (1998) The Measurement of Voting Power: Theory and practice, problems and paradoxes (Cheltenham: Edward Elgar) J. Ferejohn and D. Satz (1995) ‘Unification, Universalism, and Rational Choice Theory’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 71-84 C. Finkelstein (2004) ‘Legal Theory and the Rational Actor’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) J. Finnis (1980) Natural Law and Natural Rights (Oxford: Clarendon) M. P. Fiorina (1995) ‘Rational Choice, Empirical Contributions, and the Scientific Enterprise’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 85-94 J. Fishkin (1975) ‘Justice and Rationality: Some Objections to the Central Argument in Rawls’s Theory’ The American Political Science Review 69:2 615-629 R. E. Flathman (ed.) (1973) Concepts in Social & Political Philosophy (New York: Macmillan) B. Franklin (1975) The Papers of Benjamin Franklin edited W. B. Willcox (Yale UP) M. Freeden (1991) Rights (Milton Keynes: Open UP) S. Freeman (1994) ‘Utilitarianism, Deontology, and the Priority of Right’ Philosophy and Public Affairs 23:4 313-349 S. Freeman (2001) ‘Deontology’ in L. Becker and C. B. Becker (eds.) (2001) Encyclopedia of Ethics 2nd edition (New York: Routledge) P. A. Freund (1964) ‘Rationality in Judicial Decisions’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) C. Fried and D. Parfit (1979) ‘Correspondence’ Philosophy and Public Affairs 8:4 393-397 J. Friedman (1995) ‘Introduction: Economic Approaches to Politics’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 1-24 C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) W. Gaertner (1997) ‘Discussion of Arrow’s Paper [Arrow (1997)]’ in K. J. Arrow, A. Sen and K. Suzumura (eds.) (1997) Social Choice Re-Examined. Volume 1 (Basingstoke: Macmillan Press) W. B. Gallie (1956) ‘Essentially Contested Concepts’ Proceedings of the Aristotelian Society 56 167-98 W. A. Galston (1983a) ‘Review: Nelson, William N. On Justifying Democracy’ Ethics 93:3 600-1 W. A. Galston (1983b) ‘Review: Levine, Andrew Liberal Democracy: A Critique of Its Theory’ Ethics 93:3 601-3 R. Garner (2007) ‘Thousands of schools set to follow Brighton with admission lotteries’ The Independent (Thur 01/03/07). Available at: http://education.independent.co.uk/news/article2314218.ece (last accessed 15/04/07) G. F. Gaus (2000) Political Concepts and Political Theories (Boulder, Colo.: Westview) D. Gauthier (1962) Practical Reasoning: The structure and foundations of prudential and moral arguments and their exemplification in discourse (Oxford: Clarendon) D. Gauthier (1986) Morals by Agreement (Oxford: Clarendon) 294 D. Gauthier (1990) Moral Dealing: Contract, ethics & reason (Ithaca: Cornell UP) D. Gauthier (1993) ‘Constituting Democracy’ in D. Copp, J. Hampton and J. Roemer (eds.) (1993) The Idea of Democracy (Cambridge: Cambridge UP) L. Gevers (1997) ‘Discussion of Sen’s Paper [A. Sen (1997)]’ in K. J. Arrow, A. Sen and K. Suzumura (eds.) (1997) Social Choice Re-Examined. Volume 1 (Basingstoke: Macmillan Press) G. Gigerenzer, Z. Swijtink, T. Porter, L. Daston, J. Beatty and L. Krüger634 (1989) The Empire of Chance: How probability changed science and everyday life (Cambridge: Cambridge UP) C. Gilligan (1993 [1982]) In a Different Voice: Psychological theory and women’s development (Cambridge, Mass. and London: Harvard UP) J. Glover (1975) ‘It Makes No Difference Whether Or Not I Do It’ in P. Singer (ed) (1986) Applied Ethics (Oxford: Oxford UP) J. Glover (1977) Causing Deaths and Saving Lives (Hammondsworth: Penguin) R. E. Goodin (2007) ‘Enfranchising All Affected Interests, and its Alternatives’ Philosophy and Public Affairs 35:1 40-68 B. Goodwin (1997 [1982]) Using Political Ideas 4th edition (Chichester: Wiley) B. Goodwin (2005 [1992]) Justice by Lottery 2nd edition (London: Harvester Wheatsheaf) H. Greely (1977) ‘The Equality of Allocation by Lot’ Harvard Civil Rights-Civil Liberties Law Review 12 113-141 D. P. Green and I. Shapiro (1995) ‘Pathologies Revisited: Reflections on our Critics’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 235-76 R. L. Gregory with O. Zangwill (eds.) (1987) The Oxford Companion to the Mind (Oxford: Oxford UP) J. Griffin (1986) Well-Being: Its meaning, measurement and moral importance (Oxford: Clarendon) S. Guest (2005) ‘Integrity, Equality and Justice’ (Forthcoming in Internationale Revue de Philosophie. Available http:// www.ucl.ac.uk/spp/download/seminars/ 0405/Integrity-equality-and-justice.rtf [accessed 15/03/05]) L. Guinier (1994) The Tyranny of the Majority: Fundamental fairness in representative democracy (New York: Free Press) A. Gutmann (1999 [1987]) Democratic Education revised edition (Princeton: Princeton UP) A. Gutmann and D. F. Thompson (1996) Democracy and Disagreement (Cambridge, Mass. and London: Belknap Press of Harvard UP) J. Haigh (2000) Taking Chances (Oxford: Oxford UP) P. J. Hammond (1986) ‘Consequentialist Social Norms for Public Decisions’ in W. P. Heller, R. M. Starr and D. A. Starrett (eds.) (1986) Social Choice and Public Decision Making: Essays in honor of Kenneth J. Arrow volume 1 (Cambridge: Cambridge UP) R. Hardin (1982) Collective Action (Baltimore: John Hopkins UP) R. Hardin (1990) ‘Public Choice versus Democracy’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) 634 N.B. their introduction, p.xvi: “Dutiful subjects of the empire of chance, we used a lottery to order our names on the title page”. 295 G. Harman (2004) ‘Practical Aspects of Theoretical Reasoning’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) J. Harris (1975) ‘The Survival Lottery’ in P. Singer (ed.) (1986) Applied Ethics (Oxford: Oxford UP) J. C. Harsanyi (1953) ‘Cardinal Utility in Welfare Economics and in the Theory of Risk-Taking’ The Journal of Political Economy 61:5 434-5 J. C. Harsanyi (1955) ‘Cardinal Welfare, Individualistic Ethics, and Interpersonal Comparisons of Utility’ The Journal of Political Economy 63:4 309-321 J. C. Harsanyi (1975) ‘Can the Maximin Principle Serve as a Basis for Morality? A Critique of John Rawls’s Theory’ American Political Science Review 69:2 594606 H. L. A. Hart (1975) ‘Rawls on Liberty and its Priority’ in N. Daniels (ed.) (1989 [1975]) Reading Rawls: Critical studies on Rawls’ A Theory of Justice (Stanford, CA: Stanford UP) E. Hauptmann (1996) Putting Choice Before Democracy (Albany: SUNY Press) F. A. Hayek (1976) Law, Legislation and Liberty. Volume 2: The mirage of social justice (London: Routledge and Kegan Paul) J. W. Headlam (1933 [1891]) Election by Lot at Athens 2nd edition (Cambridge: Cambridge UP) J. Heim (1990) ‘The Demise of the Confessional State and the Rise of the Idea of a Legitimate Minority’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) W. P. Heller, R. M. Starr and D. A. Starrett (eds.) (1986) Social Choice and Public Decision Making: Essays in honor of Kenneth J. Arrow volume 1 (Cambridge: Cambridge UP) P. Hesketh (2005) Life’s A Game So Fix The Odds (Chichester: Capstone Publishing) D. Heyd (2000) ‘When Practical Reasons Plays Dice’ in E. Ullmann-Margalit (ed.) (2000) Reasoning Practically (Oxford: Oxford UP) D. Heyd and U. Segal (2006) ‘Democratically Elected Aristocracies’ Social Choice Welfare 27 103-27 I. Hirose (2001) ‘Saving the Greater Number without Aggregation’ Analysis 61:4 341-342 I. Hirose (2004) ‘Aggregation and Numbers’ Utilitas 16:1 62-79 I. Hirose (2005) ‘Against Weighted Lottery’ (unpublished; presented in the CEPL workshop on aggregation and numbers 05/02/05. Available from http://users.ox.ac.uk/~corp1012/item/weighted%20lottery.pdf [accessed 01/02/05]) I. Hirose (2007) ‘Weighted Lotteries in Life and Death Cases’ Ratio 20:1 45-56 T. Hobbes (1985 [1651]) Leviathan (Harmondsworth: Penguin) B. Holden (1988) Understanding Liberal Democracy (Oxford: Philip Allan) T. Holland (2005) Persian Fire: The first world empire and the battle for the west (London: Little, Brown) B. Hooker and B. Streumer (2004) ‘Procedural and Substantive Practical Rationality’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) D. L. Horowitz (2000) ‘Constitutional Design: An Oxymoron?’ in I. Shapiro and S. Macedo (eds.) (2000) NOMOS XLII: Designing Democratic Institutions (New York: New York UP) 296 J. Hospers (1961) Human Conduct: An introduction to the problems of ethics (New York: Harcourt, Brace & World) J. Houghton (1988) Does God Play Dice? A look at the story of the universe (Leicester: Inter-Varsity) D. Hume (1978 [1739-40]) Treatise of Human Nature 2nd edition, edited L. A. SelbyBigge and P. H. Nidditch (Oxford: Calrendon) M. Humphrey (2004) ‘Democratic Legitimacy, Public Justification and Environmental Direct Action’ (unpublished) S. Hurley (2003) Justice, Luck, and Knowledge (Cambridge, Mass.: Harvard UP) J. Hyland (1995) Democratic Theory: The philosophical foundations (Manchester: Manchester UP) M. James (1981) ‘Public Interest and Majority Rule in Bentham’s Democratic Theory’ Political Theory 9:1 49-64 K. Jones (2004) ‘Gender and Rationality’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) P. Jones (1983) ‘Political Equality and Majority Rule’ in D. Miller and L. Siedentop (eds.) (1983) The Nature of Political Theory (Oxford: Clarendon) J. M. Joyce (2004) ‘Bayesianism’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) F. M. Kamm (1985) ‘Equal Treatment and Equal Chances’ Philosophy & Public Affairs 14:2 177-194 F. M. Kamm (1993) Morality, Mortality, Vol. I: Death and whom to save from it (Oxford: Oxford UP) F. M. Kamm (1998a) ‘Precis of Morality, Mortality, Vol. I: Death and Whom to Save From It’ Philosophy and Phenomenological Research 58:4 939-945 F. M. Kamm (1998b) ‘Replies’ [to N. Daniels (1998), J. Broome (1998) and D. W. Brock (1998)] Philosophy and Phenomenological Research 58:4 969-974 F. M. Kamm (2002) ‘Owing, Justifying and Rejecting’ Mind 111 323-354 F. M. Kamm (2005) ‘Aggregation and Two Moral Methods’ Utilitas 17:1 1-23 I. Kant (1948 [1785]) The Moral Law [a.k.a. Groundwork of the metaphysic of morals] translated H. J. Paton (London: Hutchinson) A. Kaplan (1964) ‘Some Limitations on Rationality’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) G. Kavka (1979) ‘The Numbers Should Count’ Philosophical Studies 36 285-294 S. Kelley Jr. (1995) ‘The Promise and Limitations of Rational Choice Theory’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 95-106 J. S. Kelly (1978) Arrow Impossibility Theorems (London: Academic Press) J. S. Kelly (1988) Social Choice Theory: An Introduction (Berlin: Springer-Verlag) L. Kohlberg (1979) ‘Justice as Reversibility’ in P. Laslett and J. Fishkin (eds.) (1979) Philosophy, Politics and Society: fifth series (Oxford: Blackwell) S.-C. Kolm (1996) Modern Theories of Justice (Cambridge, Mass.: MIT Press) L. Kornhauser and L. Sager (1988) ‘Just Lotteries’ Social Science Information 27 483-516 A. Kuflik (1977) ‘Majority Rule Procedure’ in J. R. Pennock and J. W. Chapman (eds.) (1977) NOMOS XVIII: Due Process (New York: New York UP) R. Kumar (2001) ‘Contractualism on Saving the Many’ Analysis 61:2 165-170 W. Kymlicka (1988) ‘Rawls on Teleology and Deontology’ Philosophy and Public Affairs 17:3 173-190 W. Kymlicka (1989) Liberalism, Community and Culture (Oxford: Clarendon) 297 W. Kymlicka (2002 [1990]) Contemporary Political Philosophy: An introduction 2nd edition (Oxford: Clarendon) J.-J. Laffont (ed.) (1979) Aggregation and Revelation of Preferences (Amsterdam: North-Holland Publishing Company) R. E. Lane (1995) ‘What Rational Choice Explains’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 107-26 G. Lang (2004) ‘On Hedging One’s Bets’ (unpublished) H. J. Laski (1933) Democracy in Crisis (London: George Allen & Unwin) P. Laslett (ed.) (1956) Philosophy, Politics and Society (Oxford: Blackwell) P. Laslett and W. G. Runciman (eds.) (1962) Philosophy, Politics and Society: second series (New York: Barnes & Noble) P. Laslett and J. Fishkin (eds.) (1979) Philosophy, Politics and Society: fifth series (Oxford: Blackwell) S. Laville and R. Smithers (2007) ‘Council runs lottery for school places’ and ‘War over school boundaries divides Brighton’ The Guardian (Thur 01/03/07), pages 1 and 4 respectively. Available at: http://education.guardian.co.uk/schools/story/0,,2023791,00.html (last accessed 15/04/07) J. Leca (1994) ‘Democratization in the Arab World: uncertainty, vulnerability and legitimacy. A tentative conceptualisation and some hypotheses’ in G. Salamé (ed.) (1994) Democracy without Democrats? The renewal of politics in the Muslin world (London: I. B. Taurus) G. W. F. Leibniz (1998) Philosophical Works edited and translated R. S. Woolhouse and R. Francks (Oxford: Oxford UP) A. Levine (1990) ‘Electoral Power, Group Power and Democracy’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) A. Levine (2002) Engaging Political Philosophy: From Hobbes to Rawls (Malden, Mass.: Blackwell) A. Lijphart (1977) Democracy in Plural Societies: A comparative exploration (New Haven: Yale UP) A. Lijphart (1999) Patterns of Democracy: Government forms and performance in thirty-six countries (New Haven and London: Yale UP) A. Lijphart (2004) ‘Constitutional design for divided societies’ Journal of Democracy 15:2 96-109 A. Lincoln (1991) Abraham Lincoln: Great Speeches edited S. Appelbaum (New York: Dover) C. E. Lindblom (1964) ‘Some Limitations on Rationality: A Comment’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) C. List (2001) Mission Impossible?: The problem of democratic aggregation in the face of Arrow’s theorem (unpublished Oxford D.Phil, deposited in the Bodleian library) C. List and R. E. Goodin (2001) ‘Epistemic Democracy: Generalizing the Condorcet Jury Theorem’ The Journal of Political Philosophy 9:3 277-306 J. Lively (1975) Democracy (Oxford: Basil Blackwell) D. Locke (1981) ‘The Principle of Equal Interests’ The Philosophical Review 90:4 531-559 J. Locke (1980 [1689]) Second Treatise of Government (Indianapolis: Hackett) 298 S. Lohmann (1995) ‘The Poverty of Green and Shapiro’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 127-54 C. G. Lord, L. Ross and M. R. Lepper (1979) ‘Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on Subsequently Considered Evidence’ Journal of Personality and Social Psychology 37:11 2098-109 D. Luban (1988) Lawyers and Justice: An ethical study (Princeton: Princeton UP) P. Lucas (2001) The Dice House (London: Faber and Faber) G. Lloyd (1993 [1984]) The Man of Reason: ‘Male’ and ‘female’ in Western philosophy 2nd edition (London: Routledge) D. Lyons (1972) ‘Rawls Versus Utilitarianism’ Journal of Philosophy 69:18 535-545 A. F. MacKay (1980) Arrow’s Theorem: The Paradox of Social Choice. A Case Study in the Philosophy of Economics. (New Haven: Yale UP) G. Mackie (2003) Democracy Defended (Cambridge: Cambridge UP) C. B. Macpherson (1966) The Real World of Democracy (Oxford: Oxford UP) C. B. Macpherson (1973) Democratic Theory: Essays in retrieval (Oxford: Clarendon) B. Malkin (2006) ‘Graduate’s case illustrates asylum ‘lottery’, say campaigners’ Sunday Telegraph (23/04/06) 10. Available at: http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2006/04/23/nasylum2 3.xml&sSheet=/news/2006/04/23/ixhome.html (last accessed 15/04/07) B. Manin (1997) The Principles of Representative Government (Cambridge: Cambridge UP) B. Martin (1992) ‘Demarchy: A democratic alternative to electoral politics’ Kick It Over 30 (Fall) 11-13. Available http://www.uow.edu.au/arts/sts/bmartin/pubs/92kio.html (accessed 13/03/05) B. Martin (1993) ‘Are Elections the Ultimate in Democracy?’ Green Left 14/07/93 p.12. Available at http://www.uow.edu.au/arts/sts/bmartin/pubs/93gl1.html (accessed 13/03/05) B. Martin (1995) ‘Democracy without Elections’ Social Anarchism 21 18-51. Available http://www.uow.edu.au/arts/sts/bmartin/pubs/95sa.html (accessed 13/03/05) B. Martin (2001 [1989]) ‘Demarchy: A democratic alternative to electoral politics’ pamphlet. Available http://www.uow.edu.au/arts/sts/bmartin/pubs/89demarchy.html (accessed 13/03/05) M. Martin and L. MacIntyre (1994) (eds.) Readings in the Philosophy of Social Science (MIT Press) A. P. Martinich (1996) Philosophical Writing 2nd edition (Oxford: Blackwell) K. O. May (1952) ‘A Set of Independent Necessary and Sufficient Conditions for Simple Majority Decision’ Econometrica 20:4 680-684 K. O. May (1954) ‘Intransitivity, Utility, and the Aggregation of Preference Patterns’ Econometrica 22:1 1-13 N. Mayer and P. Perrineau (1992) ‘Why Do They Vote for Le Pen?’ European Journal of Political Research 22: 123-41 B. Mayo (1956) ‘Is there a Case for the General Will?’ in P. Laslett (ed.) (1956) Philosophy, Politics and Society (Oxford: Blackwell) H. B. Mayo (1960) An Introduction to Democratic Theory (New York: Oxford UP) G. McColl (1998 [1995]) Celtic: The Official Illustrated History 1888-1998 (London: Hamlyn) 299 A. McGann (2006) The Logic of Democracy: Reconciling equality, deliberation, and minority protection (Ann Arbor: U of Michigan P) I. McLean (1987) Public Choice: An Introduction (Oxford: Blackwell) I. McLean (1995) ‘Independence of Irrelevant Alternatives before Arrow’ Mathematical Social Sciences 30:2 107-26 I. McLean (2001) Rational Choice and British Politics: Rhetoric and manipulation from Peel to Blair (Oxford: Oxford UP) I. McLean (2002) ‘Review Article: William H. Riker and the invention of Heresthetic(s)’ British Journal of Political Science 32 535-558 I. McLean and A. Urken (1995) Classics of Social Choice (Ann Arbor: U of Michigan P) E. F. McClennen (2004) ‘The Rationality of Being Guided by Rules’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) A. R. Mele (2004) ‘Motivated Irrationality’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) M. Mendle (ed.) (2001) The Putney Debates of 1647: The army, the Levellers and the English state (Cambridge: Cambridge UP) D. Meyers (1990) ‘Democratic Theory and the Democratic Agent’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) J. S. Mill (1998 [1861]a) ‘Utilitarianism’ in J. S. Mill (1998 [1859-69]) On Liberty and other essays edited J. Gray (Oxford: Oxford UP) J. S. Mill (1998 [1861]b) ‘Considerations on Representative Government’ in J. S. Mill (1998 [1859-69]) On Liberty and other essays edited J. Gray (Oxford: Oxford UP) J. S. Mill (1998 [1859-69]) On Liberty and other essays edited J. Gray (Oxford: Oxford UP) D. Miller (1992) ‘Deliberative Democracy and Social Choice’ Political Studies 40 54–67 D. Miller and L. Siedentop (eds.) (1983) The Nature of Political Theory (Oxford: Clarendon) F. D. Miller Jr. and D. Keyt (eds.) (1991) Companion to Aristotle’s Politics (Oxford: Blackwell) F. D. Miller Jr. (1997) ‘Aristotle and American Classical Republicanism’ in L. G. Rubin (ed.) (1997) Justice V: Law in Greek political thought (Lanham, Md: Rowman & Littlefield Publishers) Montesquieu [Baron de, Charles Louis de Secondat] (1989 [1748]) The Spirit of the Laws translated and edited A. M. Cohler, B. C. Miller and H. S. Stone (Cambridge: Cambridge UP) C. W. Morris (1993) ‘On Contractarian Constitutional Democracy’ in D. Copp, J. Hampton and J. Roemer (eds.) (1993) The Idea of Democracy (Cambridge: Cambridge UP) C. Mouffe (2000) The Democratic Paradox (London: Verso) D. C. Mueller (2003) Public Choice III (Cambridge: Cambridge UP) J. B. Murphy (1995) ‘Rational Choice Theory as Social Physics’ Critical Review 9:12 [special double issue on Rational Choice Theory and Politics] 155-74 300 J. G. Murphy (1986) ‘Rationality and Constraints on Democratic Rule’ in J. R. Pennock and J. W. Chapman (eds.) (1986) NOMOS XXVIII: Justification (New York: New York UP) T. Nagel (1987) ‘Moral Conflict and Political Legitimacy’ Philosophy and Public Affairs 16:3 215-240 T. Nagel (1991) Equality and Partiality (New York and Oxford: Oxford UP) T. Nagel (2000) ‘Equality’ in M. Clayton and A. Williams (eds.) (2000) The Ideal of Equality (Basingstoke: Palgrave Macmillan) W. N. Nelson (1980) On Justifying Democracy (London: Routledge & Kegan Paul) O. Neurath (1983) ‘The Lost wanderers of Descartes and the Auxiliary Motive (On the Psychology of Decision)’ [originally 1913] in his Philosophical Papers 1913-46 edited and translated R. S. Cohen and M. Neurath (Leicester: Reidel) R. Nozick (1974) Anarchy, State, and Utopia (Oxford: Blackwell) J. Ober and C. Hedrick (eds.) (1996) Demokratia: A conversation on democracies, ancient and modern (Princeton: Princeton UP) K. Oberman (2007) ‘Pupils do win in a lottery’ [letter] The Observer (11/03/07). Available at: http://observer.guardian.co.uk/letters/story/0,,2031118,00.html (last accessed 15/04/07) O. O’Neill (2004) ‘Kant: Rationality as Practical Reasoning’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) S. O’Neill, F. Gibb and H. Brooke (2005) ‘Justice by postcode: the lottery revealed’ The Times (Wed 23/11/05) 1-2. Available at: http://www.timesonline.co.uk/tol/news/uk/article593173.ece (last accessed 15/04/07) F. E. Oppenheim (1964) ‘Rational Decisions and Intrinsic Valuations’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) P. C. Ordeshook (1995) ‘Engineering or Science: What Is the Study of Politics?’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 175-88 M. Otsuka (2000) ‘Scanlon and the claims of the many versus the one’ Analysis 60:3 288-293 M. Otsuka (2006) ‘Saving Lives, Moral Theory, and the Claims of Individuals’ Philosophy & Public Affairs 34:2 109–135 D. Parfit (1978) ‘Innumerate Ethics’ Philosophy and Public Affairs 7:4 285-301 D. Parfit (2000 [1991]) ‘Equality or priority?’ The Lindley Lecture. Reprinted in M. Clayton and A. Williams (eds.) (2000) The Ideal of Equality (Basingstoke: Palgrave Macmillan) D. Parfit (2003) ‘Justifiability to Each Person’ Ratio 16:4 368-390 S. Parsons (2005) Rational Choice and Politics: A critical introduction (London: Continuum) C. Pateman (1970) Participation and Democratic Theory (Cambridge: Cambridge UP) G. Paton (2007) ‘Schools can use lotteries for ‘fairer’ admissions, says minister’ The Daily Telegraph (Thur 01/03/07) 15. Available at: http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2007/03/01/nschools4 01.xml (last accessed 15/04/07) J. R. Pennock (1964) ‘Reason in Legislative Decisions’ in C. J. Friedrich (ed.) (1964) NOMOS VII: Rational Decision (New York: Atherton Press) 301 J. R. Pennock (1979) Democratic Political Theory (Princeton: Princeton UP) J. R. Pennock (1990) ‘Democracy and its Critics’ [Review of R. Dahl (1989)] Political Theory 18:3 512-518 J. R. Pennock and J. W. Chapman (eds.) (1977) NOMOS XVIII: Due Process (New York: New York UP) J. R. Pennock and J. W. Chapman (eds.) (1983) NOMOS XXV: Liberal Democracy (New York: New York UP) J. R. Pennock and J. W. Chapman (eds.) (1986) NOMOS XXVIII: Justification (New York: New York UP) P. Pettit (2000) ‘Democracy, Electoral and Contestatory’ in I. Shapiro and S. Macedo (eds.) (2000) NOMOS XLII: Designing Democratic Institutions (New York: New York UP) J. Piven (2003) As Luck Would Have It: Incredible stories from lottery wins to lightning strikes (Chatham: Hodder & Stoughton) J. Plamenatz (1938) Consent, Freedom and Political Obligation (London: Oxford UP) J. Plamenatz (1992 [1963]) Man and Society: political and social theories from Machiavelli to Marx. Volume 2: From Montesquieu to the Early Socialists. 2nd edition, revised M. E. Plamenatz and R. Wokler (London: Longman) Plato (1992) Republic translated G. M. A. Grube, revised C. D. C. Reeve (Indianapolis: Hackett) Plato (1997) Complete Works various translations, edited by J. M. Cooper and D. S. Hutchinson (Indianapolis: Hackett) T. Pratchett (1987) Mort (London: Corgi) A. Przeworski (1991) Democracy and the Market: Political and economic reforms in Eastern Europe and Latin America (Cambridge: Cambridge UP) A. Przeworski (1999) ‘Minimalist Conception of Democracy: A Defense’ in I. Shapiro and C. Hacker-Cordan (eds.) (1999) Democracy’s Value (Cambridge: Cambridge UP) A. Przeworski, S. Stokes and B. Manin (eds.) (1999) Democracy, Accountability and Representation (Cambridge: Cambridge UP) A. Quinton (ed.) (1967) Political Philosophy: Oxford readings in philosophy (Oxford: Oxford UP) E. Rakowski (1991) Equal Justice (Oxford: Clarendon) A. Ranney and W. Kendall (1956) Democracy and the American Party System (New York: Harcourt, Brace and Company) D. D. Raphael (2001) Concepts of Justice (Oxford: Clarendon Press) J. Rawls (1951) ‘Outline of a Decision Procedure for Ethics’ The Philosophical Review 60:2 177-197 J. Rawls (1975) ‘Fairness to Goodness’ The Philosophical Review 84:4 536-554 J. Rawls (1985) ‘Justice as Fairness: Political not Metaphysical’ Philosophy and Public Affairs 14:3 223-251 J. Rawls (1988) ‘The Priority of Right and Ideas of the Good’ Philosophy and Public Affairs 17:4 251-276 J. Rawls (1996 [1993]) Political Liberalism paperback edition (New York: Columbia UP) J. Rawls (1999 [1971]) A Theory of Justice 2nd edition (Cambridge, Mass.: Harvard UP) J. Rawls (2001) Justice as Fairness: A restatement edited E. Kelly (Cambridge, Mass. and London: Harvard UP) 302 J. Rawls (2007) Lectures on the History of Political Philosophy edited S. Freeman (Cambridge, Mass. and London: Belknap Press of Harvard UP) J. Raz (1985) ‘Authority and Justification’ Philosophy and Public Affairs 14:1 3-29 J. Raz (1986) The Morality of Freedom (Oxford: Clarendon) J. Raz (2003) ‘Numbers, with and without Contractualism’ Ratio 16:4 346-367 J. Rees (1971) Equality (London: Pall Mall) M. Regenwetter, B. Grofman, A.A.J. Marley & I. M. Tsetlin (2006) Behavioral Social Choice: Probabilistic models, statistical inference, and applications (Cambridge: Cambridge UP) A. Rehfeld (2005) The Concept of Constituency: Political representation, democratic legitimacy, and institutional design (New York: Cambridge UP) L. Rhinehart (1999 [1971]) The Dice Man (London: Harper Collins) P. J. Rhodes (2003) Ancient Democracy and Modern Ideology (London: Duckworth) W. H. Riker (1982) Liberalism Against Populism: A confrontation between the theory of democracy and the theory of social choice (San Francisco: W. H. Freeman) W. H. Riker (1988 [1982]) Liberalism Against Populism: A confrontation between the theory of democracy and the theory of social choice (Prospect Heights, Illinois: Waveland Press) M. Risse (2001) ‘Arrow’s Theorem, Indeterminacy, and Multiplicity Reconsidered’ Ethics 111 706-34 M. Risse (2004) ‘Arguing for Majority Rule’ Journal of Political Philosophy 12:1 4164 A. Rorty (ed.) (1980) Essays on Aristotle’s Ethics (Berkeley: University of California Press) F. Rosen (1983) Jeremy Bentham and Representative Democracy: A study of the Constitutional Code (Oxford: Clarendon) F. Rosen (1990) ‘Majorities and Minorities: a Classical Utilitarian View’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) A. Ross (1952) Why Democracy? translated D. Gatley-Philip (Cambridge, Mass: Harvard UP) J.-J. Rousseau (1994 [1762]) The Social Contract translated C. Betts (Oxford: Oxford UP) C. Rovane (2004) ‘Rationality and Persons’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) L. G. Rubin (ed.) (1997) Justice V: Law in Greek political thought (Lanham, Md: Rowman & Littlefield Publishers) M. Russell (2000) Reforming the House of Lords: Lessons from overseas (Oxford: Oxford UP) A. Ryan (2000) ‘Oxford Blues’ The Guardian (Wed 24/05/00). Available at: http://education.guardian.co.uk/higher/story/0,,255380,00.html (last accessed 15/04/07) A. Ryan (2007) ‘Alan Ryan’ [regular column] The Times Higher Education Supplement (Fri 19/01/07). Available at: http://www.thes.co.uk/search/story.aspx?story_id=2034924 (subscription required) (last accessed 15/04/07) D. G. Saari (2001) Decisions and Elections: Explaining the unexpected (Cambridge: Cambridge UP) G. Salamé (ed.) (1994) Democracy without Democrats? The renewal of politics in the Muslim world (London: I. B. Taurus) 303 P. A. Samuelson (1977) The Collected Scientific Papers of Paul A. Samuelson volume IV, edited J. E. Stiglitz (Cambridge, Mass.: MIT Press) J. Sanders (1988) ‘Why the Numbers Should Sometimes Count’ Philosophy and Public Affairs 17:1 3-14 G. Sartori (1965) Democratic Theory (New York: Praeger) G. Sartori (1987) The Theory of Democracy Revisited (Chatham, N.J.: Chatham House) T. M. Scanlon (1975) ‘Preference and Urgency’ Journal of Philosophy 72:19 655-69 T. M. Scanlon (1998) What we Owe to Each Other (Cambridge, Mass.: Harvard UP) T. M. Scanlon (2001) ‘Symposium on Amartya Sen’s Philosophy: 3 Sen and Consequentialism’ Economics and Philosophy 17 39-50 S. Scheffler (1994 [1982]) The Rejection of Consequentialism revised edition (Oxford UP) N. Schofield (1995) ‘Rational Choice and Political Economy’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 189-211 J. Schumpeter (1967) ‘Two Concepts of Democracy’ in A. Quinton (ed.) (1967) Political Philosophy: Oxford readings in philosophy (Oxford: Oxford UP) B. Schwartz (2007) ‘Make college admissions a crapshoot: Top schools are already too selective, so why not draw names from a hat?’ LA Times (Sun 18/03/07). Available at: http://www.latimes.com/news/opinion/la-opschwartz18mar18,0,7478180.story?coll=la-opinion-rightrailHe (last accessed 15/04/07) A. Sen (1970) Collective Choice and Social Welfare (Edinburgh: Oliver and Boyd) A. Sen (1987) On Ethics & Economics (Oxford: Basil Blackwell) A. Sen (1992) Inequality Re-examined (Oxford: Clarendon Press) A. Sen (1995) ‘Rationality and Social Choice’ American Economic Review 85:1 1-24 A. Sen (1997 [1982]) Choice, Welfare and Measurement (Cambridge, Mass.: Harvard UP) A. Sen (1997) ‘Individual Preferences as the Basis of Social Choice’ in K. J. Arrow, A. Sen and K. Suzumura (eds.) (1997) Social Choice Re-Examined. Volume 1 (Basingstoke: Macmillan Press) A. Sen (2001) ‘Symposium on Amartya Sen’s Philosophy: 4 Reply’ [to Scanlon (2001)] Economics and Philosophy 17 51-66 R. Sewell, D. MacKay and I. McLean (2004) ‘Probablistic Electoral Methods, Representative Probability, and Maximum Entropy’ (unpublished, available www.nuff.ox.ac.uk/Politics/ papers/2004/Sewell%20McLean.pdf [accessed 18/05/04]) I. Shapiro (1990) ‘Three Fallacies Concerning Majorities, Minorities and Democratic Politics’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) I. Shapiro (2003a) Moral Foundations of Politics (New Haven: Yale UP) I. Shapiro (2003b) The State of Democratic Theory (Princeton: Princeton UP) I. Shapiro and C. Hacker-Cordan (eds.) (1999) Democracy’s Value (Cambridge: Cambridge UP) I. Shapiro and S. Macedo (eds.) (2000) NOMOS XLII: Designing Democratic Institutions (New York: New York UP) K. A. Shepsle (1995) ‘Statistical Political Philosophy and Positive Political Theory’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 213-222 304 K. Shepsle and M. Bonchek (1997) Analyzing Politics: Rationality, behavior, and institutions (New York: W. W. Norton & Company) G. Sher (1980) ‘What Makes a Lottery Fair?’ Nous 14:2 203-216 S. Shute and S. Hurley (eds.) (1993) On Human Rights (The Oxford Amnesty Lectures 1993) (New York: Basic Books) H. Sidgwick (1908 [1891]) Elements of Politics 3rd edition (London: MacMillan & Co.) H. Sidgwick (1913 [1874]) The Methods of Ethics 7th edition (London: MacMillan & Co.) A. J. Simmons (1993) On the Edge of Anarchy: Locke, consent and the limits of society (Princeton: Princeton UP) R. Simon (1990) ‘Pluralism and Equality: the Status of Minority Values in a Democracy’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) P. Singer (ed.) (1986) Applied Ethics (Oxford: Oxford UP) M. Slote (1989) Beyond Optimizing: A study of rational choice (Cambridge, Mass. and London: Harvard UP) J. J. C. Smart and B. Williams (1973) Utilitarianism: For and Against (Cambridge: Cambridge UP) R.N. Smart (1958) ‘Negative utilitarianism’ Mind 67 542-3 H. R. Smith (1960) Democracy and the Public Interest (Athens: University of Georgia Press) M. Smith (2004) ‘Humean Rationality’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) R. Sorensen (2004) ‘Paradoxes of Rationality’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) F. Spagnoli (2003) Homo Democraticus: On the universal desirability and not so universal possibility of democracy and human rights (London: Cambridge Scholars) I. Stewart (1997 [1989]) Does God Play Dice? The new mathematics of chaos 2nd edition (London: Penguin) J. Still (1981) ‘Political Equality and Election Systems’ Ethics 91:3 [Special issue: Symposium on the Theory and Practice of Representation] 375-394 P. Stone (2007) ‘Why Lotteries Are Just’ The Journal of Political Philosophy 15:3 276-95 C. Sunstein (2001) Designing Democracy (Oxford: Oxford UP) A. Swift (2006 [2001]) Political Philosophy: A beginners’ guide for students and politicians 2nd edition (Cambridge: Polity Press) J. Taurek (1977) ‘Should the Numbers Count?’ Philosophy and Public Affairs 6:4 293-316 A. D. Taylor (1995) Mathematics and Politics: Strategy, voting, power, and proof (New York: Springer-Verlag) M. Taylor (1995) ‘Battering RAMs’ Critical Review 9:1-2 [special double issue on Rational Choice Theory and Politics] 223-34 L. Temkin (1993) Inequality (Oxford: Oxford UP) G. Testa (2003) ‘Contractualism and the Significance of Numbers’ (unpublished paper presented to the 7th Annual Oxford Philosophy Graduate Conference 16/11/03) J. J. Thomson (1997) ‘The Right and the Good’ The Journal of Philosophy 94:6 273298 305 H. D. Thoreau (1993 [1849]) ‘Civil Disobedience’ in his (1993) Civil Disobedience and Other Essays edited P. Smith (New York: Dover) J. Timmerman (2004) ‘The Individualist Lottery: How People Count, but not their Numbers’ Analysis 64:2 106-112 C. Titus (1931a) ‘A Nomenclature in Political Science’ The American Political Science Review 25:1 45-60 C. Titus (1931b) ‘A Nomenclature in Political Science: Part II’ The American Political Science Review 25:3 615-627 M. Tooley (1972) ‘Abortion and Infanticide’ Philosophy and Public Affairs 2 37-65 G. Tullock (1967) Towards a Mathematics of Politics (Ann Arbor: University of Michigan Press) E. Ullmann-Margalit (ed.) (2000) Reasoning Practically (Oxford: Oxford UP) N. Urbinati (2002) Mill on Democracy: from the Athenian polis to representative government (Chicago: University of Chicago Press) H. Van den Doel and B. Van Velthoven (1993 [1979]) Democracy and Welfare Economics 2nd edition (Cambridge: Cambridge UP) D. Van Mill (2006) Deliberation, Social Choice and Absolutist Democracy (Abingdon: Routledge) P. Van Parijs (1995) Real Freedom for All (Oxford: Clarendon) J. Van Steenis (2002) ‘Alternating Minorities versus a Democracy based on Majorities’ Available at: http://members.chello.nl/jsteenis/catastrophes10.htm (accessed 27/02/05) R. Vernon (2001) Political Morality: A theory of liberal democracy (London: Continuum) W. K. Viscusi (1983) Risk by Choice: Regulating health and safety in the workplace (Cambridge, Mass.: Harvard UP) C. Von Clausewitz (1997 [1832]) On War translated J. J. Graham, revised F. N. Maude (Ware: Wordsworth) J. Waldron (1990) ‘Rights and Majorities: Rousseau Revisited’ in J. Chapman and A. Wertheimer (eds.) (1990) NOMOS XXXII: Majorities and minorities (New York: New York UP) J. Waldron (1999) Dignity of Legislation (Cambridge: Cambridge UP) M. Walzer (1983) Spheres of Justice: A defence of pluralism and equality (Oxford: Blackwell) D. Wasserman and A. Strudler (2003) ‘Can a Nonconsequentialist Count Lives?’ Philosophy and Public Affairs 31:1 71-94 M. Weber (1994 [1904]) ‘Objectivity in Social Science and Social Policy’ in M. Martin and L. MacIntyre (1994) (eds.) Readings in the Philosophy of Social Science (MIT Press) P. Weirich (2004) ‘Economic Rationality’ in A. R. Mele and P. Rawling (eds.) (2004) The Oxford Handbook of Rationality (Oxford: Oxford UP) F. G. Whelan (1983) ‘Prologue: Democratic Theory and the Boundary Problem’ in J. R. Pennock and J. W. Chapman (eds.) (1983) NOMOS XXV: Liberal Democracy (New York: New York UP) W. Whitman (1982) Poetry and Prose (New York: Literary Classics of America/Viking Press) B. Williams (1980) ‘Justice as a Virtue’ in A. Rorty (ed.) (1980) Essays on Aristotle’s Ethics (Berkeley: University of California Press) B. Williams (1973) ‘A Critique of Utilitarianism’ in J. J. C. Smart and B. Williams (1973) Utilitarianism: For and Against (Cambridge: Cambridge UP) 306 P. Wilson (2007) ‘Mawhinney's big idea has as much appeal as American cheese’ The Observer (18/03/07). Available at: http://football.guardian.co.uk/Columnists/Column/0,,2036542,00.html (last accessed 17/04/07) L. Wittgenstein (1958 [1953]) Philosophical Investigations 2nd edition, translated G. E. M. Anscombe (Oxford: Blackwell) R. P. Wolff (1976) In Defense of Anarchism (New York: Harper & Row) D. Wolfle (1970) ‘Chance, or Human Judgement?’ Science (new series) 167:3922 (Feb 27) 1201 R. Wollheim (1962) ‘A Paradox in the Theory of Democracy’ in P. Laslett and W. G. Runciman (eds.) (1962) Philosophy, Politics and Society: second series (New York: Barnes & Noble) A. S. P. Woodhouse (ed.) (1938) Puritanism and Liberty: Being the Army Debates (1647-9) from the Clarke Manuscripts with supplementary documents (London: J. S. Dent & Sons Ltd) P. Woodruff (2005) First Democracy: The challenge of an ancient idea (Oxford: Oxford UP) Xenophon (1923) Memorabilia and Oeconomicus translated E. C. Marchant (London: Heinemann) E. Yuksel (1998) ‘Lottery Elections: Disinfecting Democracy from Lobbies’ Available at: http://www.yuksel.org/e/law/lottery.htm (accessed 27/02/05) H. Yunis (1996) Taming Democracy: Models of political rhetoric in classical Athens (New York: Cornell UP) L. Ziff (1981) Literary Democracy: the declaration of cultural independence in America (Harmondsworth: Penguin Books) (B.2) Other Acknowledgements This thesis began out of an MPhil (2003-05), original acknowledgements to which included: The AHRB, Olly Dowlen, Marc Fleurbaey, Christopher Hanges, Iwao Hirose, Robert Jubb, Gerald Lang, Dan McDermott, David Miller, Sarah Fine, Toby Ord, Derek Parfit, Philip Schofield and Adam Swift. Since then, I have incurred many further debts of gratitude, including to the audiences and convenors of the following presentations. Introduction ‘Lottery-Voting: A Neglected Solution’ to Manchester Brave New World Graduate Conference 28/06/06 Ch.1 to Simon Caney. ‘Democracy: Procedure and Contract’ to Oxford University Political Theory Graduate Workshop (convened by Dr Daniel Butt) 26/01/06 (Hilary term) ‘Random Democracy: An Introductory Argument’ to the Eighth Annual Graduate Conference in Political Theory in the Department of Politics and International Studies at the University of Warwick 11/02/06 Ch.2 to Pavlos Eleftheriadis, Marc Fleurbaey, and Magnus Jedenheim. ‘Against Majority Rule’ to Oxford University Political Theory Graduate Workshop (convened by Dr Daniel Butt) 11/05/06 (Trinity term) 307 ‘Maximizing Arguments for Majority Rule’ to David Miller’s Nuffield Political Theory Workshop (15/10/07) (Michaelmas term) Ch.3 to Johann Frick, Iwao Hirose, Toby Ord, and Stuart White. ‘Numbers, Fairness and Lotteries’ to Oxford University Political Theory Graduate Workshop (convened by Dr Adam Swift) 24/02/05 (Hilary term) Student presentation for Dr Iwao Hirose’s seminar on Saving the Greater Number (Oxford University) 16/11/05 (Michaelmas term) ‘Using Lotteries to Adjudicate Between People’ to David Miller’s Nuffield Political Theory Workshop (22/01/07) (Hilary term) ‘Using Lotteries to Adjudicate Between People’ to the Ninth Annual Graduate Conference in Political Theory in the Department of Politics and International Studies at the University of Warwick 17/02/07 (Hilary term) Ch.4 to Olly Dowlen, Alfonso Moreno, Robert C. T. Parker, and Stuart White. Ch.5 to Clare Chambers. ‘Putting Lottery-Voting into Practice’ to Oxford University Political Theory Graduate Workshop (convened by Daniel Butt) 24/10/07 (Michaelmas term) Ch.6 to Clare Chambers, G. A. Cohen, Keith Dowding, Iwao Hirose, and Iain McLean. ‘Social Choice and Lottery-Voting’ to David Miller’s Nuffield Political Theory Workshop 05/06/06 (Trinity term) Ch.7 to John Broome, and Iwao Hirose. ‘The Rationality of Random Decision-Making’ to Oxford University Political Theory Graduate Workshop (convened by Prof G. A. Cohen) 18/10/06 (Michaelmas term) Overall, I am again particularly thankful to the AHRC for continued funding and David Miller for continued supervision (and Iain McLean for covering David’s one term sabbatical). For further financial support, I thank my parents, the ViceChancellors’ and Andrew Smith Memorial Funds, Jesus College, the Bahram Dehqani-Tafti Memorial Fund and those who have offered me teaching work over the last three years, particularly Edward Kanterian. I would also like to thank Andrew Melling, Carole Thomas and Jane Sherwood, along with the rest of the administrative staff in Jesus College (including porters), the DPIR and Wellington Square. I am also grateful, for various reasons, to Gustaf Arrhenius, Conall Boyle, Dan Butt, Krister Bykvist, Olly Dowlen, Sarah Fine, Peter Hawkins, Robert Jubb, Nick Lees, Pavel Ovseiko, Fabienne Peter, Glyn Prysor, Keith Sutherland, and Stuart White. And, to anyone whose name I have forgotten, I am both thankful and sorry.