Which Preferences Can Democracy Serve?*

Alon Harel**** and Moses Shayo******

ABSTRACT

Democracy is often perceived or justified as a means to realizing voters' preferences. Two major

difficulties have received much attention: how to aggregate votes and how to align the interests of

representatives with the preferences of voters. This paper identifies a third difficulty which stems from

the fact that some voters care not only about the electoral outcome (Outcome preferences) but also about

the act of voting itself (Voting preferences). If Voting and Outcome preferences are incongruent, then

democracy may not be able to satisfy voters’ preferences even if all voters have identical preferences and

if all representatives scrupulously implement the policies voters choose. Indeed, as no voter is likely to be

pivotal, rational voters would tend to vote primarily on the basis of their Voting preferences. In this case,

the person elected (or the policy executed) may not be the one voters wish to see elected (or executed).

Taking Voting preferences seriously sheds new light on classical controversies in political theory, e.g.,

the delegate/trustee controversy, the democratic deficit, and judicial review. We show that, at times,

seemingly undemocratic practices could be understood as a means to facilitate the simultaneous

satisfaction of both Outcome and Voting preferences.

* We are grateful to Yuval Eylon, David Enoch, David Heyd, Erez Nitzan, Re’em Segev, and Hagai Wygoda for excellent comments on previous drafts of this paper. **** Mizock Professor of Law, Hebrew University Law School, Professor of Law, Boston UniversityLaw School.****** Department of Economics, The Hebrew University.

1

1. Introduction

Political theorists often perceive democracy as a means of realizing voters' preferences or giving

effect to their judgments concerning the common good.1 Two major difficulties in realizing these

goals have received extensive attention. The first has been how to aggregate votes2 and the

closely related difficulty in devising a system where voters have an incentive to vote according

to their true preferences.3 The second difficulty has been the agency problem, namely, how to

align the interests of representatives with the preferences or judgments of voters.4 This paper

identifies and examines a third difficulty in realizing voters’ preferences. This difficulty stems

from the fact that some voters may care not only about the electoral outcome of the elections but

also about the act of voting itself. Furthermore, as no voter is likely to be pivotal,5 rational voters

would vote in a way that reflects primarily (or exclusively) their preferences concerning the act

of voting. In such a case even if all voters have identical preferences concerning the electoral

outcome (hence no aggregation problems) and even if all representatives scrupulously implement

the policies voters choose (no agency problem), the person elected or the policy executed may

not be the one voters wish to see elected or executed.

Consider a voter facing a choice between a hawkish and a dovish candidate. Voting potentially

confers two distinct types of benefits. The first is derived from the act of voting itself: the voter

gains utility from the mere expression of a patriotic or nationalist sentiment. The second is

derived from the outcome of the elections, namely, from the hawkish or dovish policies. Notice

that it is not inconceivable that the voter prefers the dovish policy to be executed, despite the

1 Indeed, the correlation between democracy and voters’ preferences or judgments is a standard (if not exclusive) argument justifying democratic procedures. For an equality-based argument that does not depend on the correlation between voters’ preferences or judgments, see Thomas Christiano, “The Authority of Democracy,” Journal of Political Philosophy 12 (2004), 266, 284-287.2 Kenneth J. Arrow, Social Choice and Individual Values (New York: Wiley, 1951).3 Allan Gibbard, “Manipulation of Voting Schemes: A General Result“ Econometrica 41 (4) (1973), 587–601; Mark A. Satterthwaite, “Strategy-proofness and Arrow’s Conditions: Existence and Correspondence Theorems for Voting Procedures and Social Welfare Functions,” Journal of Economic Theory 10 (1975), 187–217.4 Madison, The Federalist No. 10 (John Madison); John Stuart Mill, Considerations on Representative Government (New York: Prometheus Books, 1991), chap. 6; Bernard Manin, The Principles of Representative Government (Cambridge: Cambridge University Press, 1997), 163-167; Timothy Besley, Principled Agents?The Political Economy of Good Government (Oxford: Oxford University Press, 2006).5 It has been estimated that the highest probability (across states) that a single vote would determine the national election outcome in the 2008 U.S. presidential elections was approximately 1 in 10 million. The estimated probability for the average voter was 1 in 60 million. Andrew Gelman, Nate Silver, and Aaron Edlin, “What Is the Probability Your Vote Will Make A Difference?,” NBER Working Paper w15220 (2009).

2

warm patriotic feelings resulting from voting for the hawkish candidate. As a second example,

consider a voter whose family has, for generations, been voting Labor and has been active in

various struggles for workers’ rights. The voter identifies himself as a Labor voter. Failing to

vote for Labor is like converting to a different religion. Yet our voter is also convinced that

Labor is corrupt and he hopes that it loses the election (perhaps for the sake of inducing reform

or he simply believes that their policies are misconceived). One natural way for the voter to

conceptualize his predicament is to feel that while he genuinely wants Labor to lose, he does not

want to contribute to its defeat by failing to vote for it. We call the preferences concerning the

outcome of the elections (dovish candidate/policy, defeat of Labor) “Outcome preferences,” and

preferences over the act of voting itself (voting for a hawkish candidate, voting for Labor)

“Voting preferences.”6 As our examples demonstrate, these two types of preferences may not

always coincide.

It could be argued that some (or all) of the Voting preferences do not merit attention as they

reflect hypocrisy on the part of the voter. Should not we dismiss the Voting preferences of a

voter, as they do not reflect her “real” preferences – preferences concerning policies/candidates

resulting from the elections?7 Or perhaps we ought to ignore the Outcome preferences, at least in

those situations where the Voting preferences reflect one’s moral convictions? We believe that at

least sometimes both types of preferences merit consideration. One justification is purely

utilitarian: both types of preferences affect the welfare of the voter. Second, one cannot simply

maintain that any incongruity between Outcome and Voting preferences necessarily reflects

hypocrisy on the part of the voter, as some of these Voting preferences are grounded in concerns

about self-image, moral convictions, or expressive motivations. The Labor party voter in our

example is not hypocritical; his Voting preference (to vote for Labor) expresses genuine

solidarity with his family and community. Hence our analysis will not dismiss one type of

preferences and will assume that both Outcome and Voting preferences should, at least to some

6 Empirically distinguishing Voting from Outcome preferences in actual voting behavior is extremely hard, but experimental evidence suggests that a significant percentage of voters are indeed affected by Voting concerns. For a review of the literature and experimental results see Moses Shayo and Alon Harel, “Non-consequentialist Voting,” Discussion Paper 545, Center for the Study of Rationality, Hebrew University, Jerusalem 2010.

7 As Tullock has pointed out, rich people may vote for high redistributive taxation merely as a way of maintaining a charitable self-image at a cheap price (knowing well that their vote or the expression of support is immaterial as it does not affect their material interests). Gordon Tullock,”The Charity of the Uncharitable,” Economic Inquiry 9 (1971), 379-392.

3

extent and in some contexts, be honored.8 Yet, as we show below, even if only Outcome

preferences should ideally count, the fact that individuals have Voting preferences has important

implications for democratic theory. Our analysis is therefore also relevant for those who would

wish to see elections as reflecting only Outcome preferences.

The possible incongruence between Outcome and Voting preferences can lead to a bias toward

Voting preferences in large elections. Rational voters who are unlikely to be pivotal would vote

on the basis of their Voting preferences even when their Outcome preferences are much stronger

than their Voting preferences.9 But our account points to a deeper challenge to democratic

theory, namely, whether it is at all possible to design a scheme that honors both types of

preferences and, if so, how. Further, taking Voting preferences seriously sheds new light on

classical controversies in political theory, e.g., the delegate/trustee controversy, the democratic

deficit, and the “countermajoritarian difficulty” triggered by judicial review. In particular, it may

provide a novel justification for vesting power in the hands of politicians, public officials, and

judges. What is labeled “the countermajoritarian difficulty” (namely, vesting power in the hands

of the judiciary) or “the democratic deficit” (namely, vesting power in the hands of non-elected

officials) need not necessarily be explained as compromising democracy for the sake of

efficiency. Instead, it could be understood as a means to facilitate the simultaneous satisfaction

of both Outcome and Voting preferences. Voters may vote in accordance with their Voting

preferences and yet, legislators, judges, or executive bodies may, ideally, select policies in

accordance with voters’ Outcome preferences. Ironically, we also show that when Voting

preferences differ from Outcome preferences, unelected officials (judges or executive bodies)

may be more likely to promote policies favored by the electorate than are elected politicians

(e.g., legislators).

We proceed as follows. Section 2 defines the terms Outcome and Voting preferences. Section 3

examines the difficulties resulting from the co-existence of Outcome and Voting preferences. We

8 In this vein, we also do not consider the possibility or desirability of actively changing voters’ preferences. Thus claims of the type “If we do not honor Voting preferences individuals will be induced to vote on the basis of their Outcome preferences rather than on the basis of their Voting preferences” are not explored here.9 Geoffrey Brenan and Loren Lomasky, Democracy and Decision: The Pure Theory of Electoral Preference (New York: Cambridge University Press, 1993); Timothy Feddersen, Sean Gailmard, and Alvaro Sandroni, “Moral Bias in Large Elections: Theory and Experimental Evidence,” American Political Science Review 103 (2009), 175-192; Shayo and Harel 2010 supra note 6; John Morgan and Felix Várdy, “Passion over Reason? Mixed Motives and the Optimal Size of Voting Bodies,” Mimeo, University of California, Berkeley (2010).

4

also explore ways in which one can satisfy both Outcome and Voting preferences and point out

the limitations of institutional techniques designed to do so. In Section 4 we examine some

policy implications of our analysis with particular attention to the delegates/trustees debate in

democratic theory, the democratic deficit, and the debate concerning judicial review. Section 5

concludes.

2. Basic Concepts

Consider a voter who chooses from a set of possible options (policies, candidates) corresponding

to the possible electoral outcomes.10 We say that the voter has only Outcome preferences if her

preferences over her possible actions – namely which option to vote for – are induced

exclusively by her preferences over electoral outcomes. In other words, she prefers voting for

option A to voting for option B if and only if she prefers the electoral outcome A to the electoral

outcome B and voting for A increases the likelihood that outcome A will be elected.

We say that a voter has Voting preferences if her preferences over which option to vote for are

not induced exclusively by her preferences over electoral outcomes. That is, she may prefer

voting for option A to voting for option B even if such a vote does not affect the likelihood of A

being elected, or even if she prefers electoral outcome B to electoral outcome A. To give an

obvious example, a voter in a presidential election who is certain that candidate A is going to win

regardless of her vote, might still not be indifferent between voting for candidate A and voting

for candidate B. More generally, Voting preferences mean having preferences over the act of

voting itself, possibly unrelated or (as we discuss below) even opposed to one’s preferences over

the electoral outcomes.

Voters can have both Outcome and Voting motivations. One may thus distinguish between

congruent and incongruent preferences according to whether or not one’s Outcome and Voting

preferences coincide. A voter who benefits from expressing his patriotic inclinations by voting

for a hawkish candidate and also prefers a hawkish policy has congruent preferences. In contrast,

10 We stress that when we say “electoral outcome,” we refer to the policy or candidate elected, and not to other potential outcomes of the act of voting (such as the margin of victory or the feelings resulting from the act of voting in the voter or in others). For a recent attempt to solve the so-called paradox of voting namely to explain why non-pivotal voters vote on the basis of the conjecture that voters care not only about the outcome of the elections but also about the margins of victory, see Alexander A. Guerrero, “The Paradox of Voting and the Ethics of Political Representation,” Philosophy and Public Affairs 38 (2010), 272-306.

5

if the voter prefers to vote for a hawkish candidate (as a means of expressing her patriotism, say)

and, yet, prefers the dovish candidate to the hawkish candidate, the voter has incongruent

preferences.

The case of congruent preferences may be of interest in explaining the so-called paradox of

voting. A non-pivotal voter may still decide to vote not because she prefers a particular candidate

to win (as she recognizes that her vote does not matter) but because she has a Voting preference

to vote for that candidate. Yet, our primary interest is in the case of voters with incongruent

preferences. It is the incongruent preferences that give rise to challenges to democratic theory

and to the possible frustration of voters’ preferences.

Before establishing the difficulties resulting from incongruent preferences, a few examples of the

types of Voting motivations discussed in the literature may be useful.11 The first is that of

expressive voting.12 Voting under this view is like cheering for sports teams: “neither the act of

voting nor the direction of a vote cast can be explained as a means to achieving a particular

political outcome, any more than spectators attend a game as a means of securing the victory of

their team.”13 The second example is the case of moral reasons and duties. People may care

about the morality of their actions, which is sometimes evaluated independently of their

consequences. Rule utilitarianism and deontological duties are prominent examples. Thus for

instance voters may vote for higher taxes (as they believe that it is morally required of them to

vote for a just society) and yet prefer that no such taxes be ultimately imposed (as they would

suffer monetary costs as a result or they believe high taxes have undesirable macro-economic

consequences). Finally, people may want their actions to be consistent with their self-image. For

example, the Labor party voter regards supporting Labor as part of his identity even if he

believes that a conservative administration is preferable. In such cases he may want Labor to lose

but he is reluctant to be the one that participates or contributes to its defeat.

The next section illustrates the difficulties that arise when Voting preferences are incongruent

with Outcome preferences.

11 Note that the term “Voting preferences” is one that is defined by us and it applies to several different types of preferences identified in the literature. 12 Brennan & Lomasky, supra note 9.13 Geoffrey Brennan and James Buchanan, “Voter Choice: Evaluating Political Alternatives,” American Behavioral Scientist 28 (1984), 185-201, 187.

6

3. The Incongruent Voter’s Predicament

To better understand the implications of incongruent preferences, it is useful to temporarily turn

our attention away from standard political examples. Meet Larry and Sophie. Larry is a

restaurant owner and Sophie is a customer. Sophie decides whether to order wine or a beer. In

our discussion we shall distinguish between two types of Larry: the responsive type, and the

benevolent type. The responsive Larry hands his customers the drink they ask for. The

benevolent Larry hands his customers the drink he believes maximizes his customers’ welfare.

3.1 The Responsive Larry

Sophie & Larry 1: Sophie orders wine from Larry. Larry dutifully serves her wine. Sophie is

unhappy. She prefers to get a beer. She uttered “wine” simply because she likes to express a

desire for wine (or dislikes to express a desire for beer). Perhaps liking beer seems vulgar to her

while wine seems like the right thing to order in a fancy restaurant. In other words, it is not the

(likely) consequence of her utterance – the drink she receives – that moved Sophie to utter

“wine.” Thus, expressing a desire for wine entails both a gain and a loss. The gain is satisfying

her preference to utter “wine”; the loss results from the fact that uttering “wine” results in

receiving wine rather than receiving the drink she craves for, namely, beer.

Perhaps Sophie could be made better off if she were a bit more sophisticated. In Sophie & Larry

2, Sophie understands that Larry is a responsive server who will give her the drink she asks for.

She therefore compares the utility she gains from uttering “wine” (and getting one), to the utility

she gains from uttering “beer” (and getting one). If the former utility is greater than the latter,

Sophie asks for wine. If it’s the other way around, she asks for a beer.

Notice that from a purely utilitarian perspective the result in Sophie & Larry 2 – while possibly

improving on Sophie & Larry 1 – is suboptimal: Sophie loses either the payoff from getting a

beer or the payoff from uttering “wine.” She cannot benefit from both. This problem might be

insignificant if Sophie’s benefits resulting from uttering “wine” (her Voting preferences) are

minor relative to her benefits resulting from drinking beer (her Outcome preferences), i.e., if

Sophie – perhaps like most customers – is chiefly interested in the drink she gets. However, even

in the case of minor Voting preferences, the problem becomes severe once we shift from a

7

setting with a single decision-maker to a setting where there are many decision-makers voting on

a common outcome. 14

To see this, consider Sophie & Larry 3. Larry only serves one type of drink in a given night.

Larry has 1001 customers and he makes a decision whether to serve wine or beer on the basis of

a vote among his customers. Assume that, like Sophie, all customers have a preference to express

a desire for wine and also a preference to drink beer (aggregation of heterogeneous preferences is

not the issue). Larry takes a poll among his customers and makes a decision on the basis of the

outcome of this poll. In this case, Sophie would place a much higher weight on her Voting

preference than in Sophie & Larry 2. The single Sophie is a pivotal voter. Consequently, she

compares her benefits from expressing a desire for wine and getting one with her benefits from

expressing a desire for beer and getting one. In contrast, in Sophie and Larry 3 Sophie knows

that she is not likely to be pivotal: unless the vote is exactly split (500 other customers vote beer

and 500 vote wine), Larry will serve her the same drink regardless of what she votes. By voting

beer she forsakes the Voting utility that she would have gained by voting wine for little benefit,

as her vote is unlikely to affect the outcome.15 Notice that the same reasoning applies to all of

Larry’s customers. Thus, even if all of Larry’s customers care a lot about the drink they actually

get and only very little about expressing a desire for one drink over the other, they may still end

up with the drink they dislike.

Responsive Larry and his customers find themselves in a difficult situation. Even though Larry is

very well-behaved, and delivers exactly what his customers vote for (perhaps unanimously), all

customers end up with the inferior outcome. Would the situation be improved if Larry were less

responsive to Sophie’s utterances, and more attentive to her well-being?

3.2 The Benevolent Larry

Benevolent Larry cares only about his customers’ satisfaction: his preferences perfectly coincide

with theirs (agency is not the issue). Consequently, unlike responsive Larry, he does not

14Perhaps Sophie could solve her problem if she could somehow hint to Larry that when she says ”I want wine,” she does not really want him to give her wine but would rather have a beer (e.g., she could wink). We assume here that any such attempt on her part is equivalent to uttering ”beer.”15 Another way to see the effect of Voting preferences is to notice that, unlike a purely Outcome voter, Sophie does not have a weakly dominant strategy to vote for the option she wants implemented.

8

automatically serve them the drink they ask for, but rather the drink that he believes will make

them happiest. Can a benevolent Larry achieve a better outcome than a responsive Larry?

Notice first that in our story so far, Larry only affects Sophie’s Outcome payoff. Whether or not

she receives her Voting payoff is completely determined by Sophie’s request; Larry’s decision as

to what drink to hand Sophie has no impact on Sophie’s Voting payoffs. This observation

provides an opportunity for a benevolent Larry to improve his performance relative to a

responsive Larry. Consider Sophie & Larry 4, in which benevolent Larry ignores Sophie’s

request altogether, and hands her the drink he believes she prefers. In this case, Sophie’s best

course of action is to make the request that gives her the most pleasure (satisfying her Voting

preferences), regardless of what drink she actually wants to consume. Notice that given this

behavior by Sophie – and absent any knowledge about the relationship between Sophie’s

Outcome and Voting preferences – benevolent Larry would indeed be justified in ignoring

Sophie’s request, as it conveys no information about her Outcome preferences. Thus, if Larry’s

beliefs are correct, Sophie can both express a desire for wine and enjoy a good beer.

There are two main difficulties however resulting from this course of action. The first problem is

that Larry may be wrong with respect to Sophie’s preferences. In this case Sophie receives only

her Voting payoff. This may well be a worse outcome than the one obtained in Sophie & Larry 2

(where Sophie weighed the Voting payoff against the Outcome payoff and chose the better

option knowing she cannot have both). In Sophie & Larry 2 Sophie at least chose what

preferences she wished to satisfy. In contrast, here Sophie cannot make such a choice as Larry’s

actions are unaffected by her request.

This problem might be solved if there existed another customer, Jack, whose Outcome

preferences are known to be identical to Sophie’s but whose Voting preferences are negligible.

In this case, Larry can proceed as follows. Having heard Sophie’s request, Larry delegates the

authority to choose Sophie’s drinks to Jack, on the condition that his choice will determine the

drink that both Sophie and he (Jack) will actually get. Since Jack cares a lot about the drink he

gets and only a little about the mere expression of a desire for one drink or the other, he will

choose the drink both he and Sophie prefer to consume. The outcome may be similar to that in

Sophie & Larry 2. But unlike Sophie & Larry 2, in this case Sophie gets to both express a desire

9

for wine and drink beer.16 As we shall see later, this solution might shed light on the concept of

representative bureaucracy, namely, “a bureaucracy in which the bureaucrats resemble the

general public in terms of social origins and general values.”17

But, as this proposed solution underscores, there is a second problem. So far, we have described

Sophie as having a Voting preference for uttering “wine.” which is independent of Larry’s

response to that utterance. But would Sophie gain anything from expressing a desire for wine

knowing full well that Larry would simply ignore the expression of her desires (either by

following his own convictions or by delegating all authority to someone else, e.g., Jack)? Can

Sophie take pride in her expressed refined taste for wine if her expression is an empty ritual,

having no impact on what is being served (and she knows it)? In other words, it is possible that

Sophie is keen to express a desire for wine only if she believes she is being listened to and her

expressed desire is, at least to some extent, being acted upon. If Voting payoffs are conditional

on being listened to, even an omniscient and benevolent Larry cannot attain the optimal

outcome.18

3.3 On Respecting Incongruent Preferences

Before we turn to examine what implications this analysis has on electoral politics let us examine

one primary objection to this analysis, in particular, an objection to the presumption that ideally

we wish to satisfy both Voting and Outcome preferences. As mentioned in the Introduction, from

a utilitarian perspective this is not an issue, as both types of preferences affect Sophie’s welfare.

16 Notice that, again, both Sophie and Larry are acting optimally given what others are doing. Knowing that what she says does not affect the outcome, it is optimal for Sophie to demand “wine.” And given Sophie’s course of action, benevolent Larry indeed finds it optimal to delegate authority to Jack (as Sophie’s utterance conveys no information while Jack’s choice satisfies Sophie’s Outcome preferences).17 See Kenneth John Meier, “Representative Bureaucracy: An Empirical Analysis,” American Political Science Review 69 (2) (1975), 526. 18 A seemingly similar point could perhaps be made with respect to Sophie’s Outcome preferences. One could ask whether Sophie would derive the same pleasure from drinking beer independently of whether that beer has been served to her in compliance or in contradiction to her expressed will. In the extreme case, Sophie enjoys receiving a particular drink if and only if she has asked for it. Notice, however, that in this case Sophie’s preferences are not incongruent (and can be well served by the responsive Larry). This is not the case in the example given in the text and we do not discuss this possibility here although it may deserve attention. Sophie may conceivably derive pleasure from asking for wine only when she believes she is being listened to, and yet enjoy drinking beer more than drinking wine.

10

Nonetheless, a non-utilitarian might question the merit of such preferences and, consequently,

may be reluctant to give such preferences any weight. 19

Arguably, our Larry & Sophie tale may support the conjecture that Voting preferences are not

particularly important and valuable and therefore do not merit attention. Sophie’s joy resulting

from uttering “wine” may be grounded in frivolous eccentricities (perhaps she likes the sound of

the word “wine”). More plausibly she may simply be snobbish; her preference to utter “wine”

may be simply grounded in her perception that preferring wine to beer indicates a superior taste.

In contrast, politics is a serious matter since voters are expected not only to satisfy preferences

but to be committed to political justice, take their vote seriously, and make an effort not to be

manipulative. Hence it may be argued that such preferences ought simply to be ignored in the

realm of politics.

A diametrically opposite claim can also be made. Arguably what should count are only the

Voting preferences rather than the Outcome preferences, as Sophie ought to be “a woman of her

word.” Hence the political system ought to treat Sophie as if she makes no differentiation

between word and act. Perhaps complying with one’s utterance is the only way to show respect

towards the person as an autonomous agent. Ignoring it is humiliating, as it does not respect

Sophie’s expressed preferences.

Either of these views may be plausible in the Larry & Sophie case and we make no normative

claims concerning this question. Yet, we believe that such reactions cannot be generalized to all

cases of incongruent preferences. The question whether Outcome preferences, Voting

preferences, or both merit attention depends on the context. It is easy to describe circumstances

under which both Outcome and Voting preferences merit respect. The Labor party supporter who

is keen to vote for Labor (as such a vote conveys his solidarity with the working class) and, yet,

also hopes that Labor will be defeated (“to teach them a lesson”) cannot be accused of hypocrisy;

neither is such a vote irrational. Voting serves for him as a badge of belongingness to a

community. It seems that such cases indicate that, at least sometimes, both Voting and Outcome

preferences merit respect.

19 This is particularly true for those who believe that voting represents a judgment of voters as to the public or common good. See, e.g., David Estlund, Democratic Authority (Princeton: Princeton University Press, 2008).

11

To bolster this argument, think of a football fan whose team is competing in some tournament. It

is sometimes in the interest of such a team to lose a game; if it loses it would play against a

weaker team in the next round and, as a result, would be more likely to win in a game that has

greater significance. Its fans however are in a dilemma. They feel the urge to cheer for their team

even if they may, at the same time, hope it will lose. Such fans can be accused of neither

hypocrisy nor irrationality. Cheering for their team is an expressive gesture that is part of what

identifies them as fans of a team. At the same time, their “Outcome” preference that their team

loses is also perfectly natural for fans that care about their team’s achievements.

Furthermore, with respect to actual situations of voting, the widespread use of secret ballots

helps weed out the kind of Voting preferences that might seem least worthy of respect.

Specifically, the institution of secret ballots helps preclude voters from voting simply as a way of

“showing off” (as Sophie might be doing when expressing a preference for wine). The Voting

preferences likely to manifest themselves in actual (secret) elections may thus have more to do

with genuine concerns (e.g., moral convictions or identity concerns) than with merely putting on

a show of the type possibly motivating Sophie in our example.20

Last, even if one is persuaded that only Outcome preferences should count and that Voting

preferences should be ignored, one ought to take into account the fact that individuals do have

Voting preferences21 and that given that voters are almost never pivotal, Voting preferences

dictate the electoral outcome. Hence, to the extent that we care about Outcome preferences, it is

crucial to design political mechanisms that give them a voice. This is the task of Section 4.

4. Back to Elections: Incongruent Preferences and Political Institutions

It seems that neither Sophie nor voters in a regular election could dream up a better Larry than

the ones that have been described in the examples given above. Neither a fully responsive Larry

facing a unanimous electorate nor an omniscient and benevolent Larry can satisfy both Outcome

and Voting preferences. Can political institutions accomplish what Larry cannot? Even if the

20 As David Estlund once put it, “It is hard to see how any further expressive gains, beyond those available outside the voting booth, could depend on what is done in that shrouded cubicle. One could postulate expressive benefits in displaying one’s character to oneself, but what reason is there to give that expressive explanation rather than a directly moral one?” See David Estlund, “Review of ‘Democracy and Decision’ by G. Brennan & L. Lomasky,” Economics & Philosophy 12 (1996), 113, 116. 21 See Feddersen et al., supra note 9, and Shayo and Harel, supra note 6, for empirical evidence.

12

answer is negative the problems resulting from incongruent preferences can justify on

democratic (or perhaps quasi-democratic) grounds entrenched institutional structures that, so far,

have been regarded as compromising democratic values.

If the political system is responsive, the predicament of the voters is similar to the one described

in Sophie & Larry 3. Since no voter is pivotal they all vote in accordance with their Voting

preferences. Thus in the hawk/dove example noted in the Introduction, hawkish policies will be

executed even if all voters prefer the dovish policy yet derive some patriotic satisfaction from

expressing hawkish sentiments by voting for the hawkish candidate. This requires one to

consider seriously the possibility of benevolent Larry as a solution to the problem of incongruent

preferences.

We saw that benevolent and omniscient Larry could – subject to the second qualification

mentioned above – ameliorate Sophie’s predicament. Benevolent Larry acted on the basis of his

judgment concerning Sophie’s Outcome preferences. In doing so he ignored Sophie’s “vote.” In

the world of politics this is equivalent to letting decision-makers ignore the voters’ expressed

will (as reflected in the outcome of the elections) and turning instead to judge what the voters’

Outcome preferences are. This can take at least two forms. The first is to allow politicians to

renege on their electoral promises and the second is to transfer power to nonelected judges or

civil servants. Such institutional arrangements obviously pose serious risks, but these risks are

rather different from the ones emphasized in traditional discussions of democratic theory.

Many theorists condemn these (nonresponsive) institutional arrangements as undemocratic and

some of them even suggest that the system ought to be reformed to give greater voice to the

people at the expense of politicians or officials. Politicians, who renege on electoral promises, so

it is argued, violate their political obligations and undermine the integrity of the democratic

process.22 A similar critique is leveled against officials (including judges) who are empowered to

act against the majority’s judgments as expressed in the elections.23

Even theorists who do not condemn these institutions and maintain that representatives ought to

be able to renege on their promises or those who believe that officials and judges ought 22 Jane Mansbridge, “Rethinking Representation,” American Political Science Review 97 (2003), 515-528, 516.23 For a critique of the power of officials, see Joseph Weiler et al., “European Democracy and Its Critique” in J. Hayward ed. The Crisis of Representation in Europe (Frank Cass, 1995). For a critique of judicial power, see Jeremy Waldron, “The Core of the Case Against Judicial Review,” Yale L.J. 115 (2006), 1346-1406.

13

sometimes to act against the collective will of the polity believe so for reasons that have to do

with efficiency, efficacy, or the alleged expertise of judges or officials. In their view these

mechanisms compromise democracy for the sake of promoting conflicting values. In contrast, the

existence of Voting preferences can, at least in theory, provide a democratic justification for

these institutional mechanisms. Indeed, such mechanisms can potentially be more rather than less

democratic in the sense that they give voice to the voters’ Outcome preferences which might

otherwise be ignored. The primary challenge then is not how to make political institutions more

responsive to the expressed will of the electorate as reflected in the election, but rather to

guarantee that they are (a) benevolent, and, in particular, compliant with the Outcome

preferences of voters and (b) reasonably well-informed about the Outcome preferences of the

electorate.

Notice that we do not argue that a mechanism that aligns politicians/officials/judges’ decisions

with the Outcome preferences of the electorate presently exists or can easily be designed.

Obviously, politicians who ignore electoral promises may be manipulative, and non-elected

officials may frustrate voters’ Outcome preferences rather than promote them. Nonetheless, we

point out that non-responsive mechanisms are potentially valuable as tools to promote the voters’

Outcome preferences and that this function should hence be taken into consideration when

designing or evaluating political institutions.

In Section 4.1 we explore the potential value resulting from allowing benevolent politicians to

renege on their promises and in Section 4.2 we examine the case of non-elected officials such as

judges and members of the executive branch.

4.1 On the Delegate-Trustee Debate

Under current democratic systems, representatives have some independence from their

constituents’ preferences – independence that arguably allows the frustration of the constituents’

preferences.24 The power of representatives to act independently of the expressed will of their

constituencies is considered to be a challenge to democratic theory. We agree that this is a

challenge but believe that the nature of the challenge differs from the traditional description

given to it in the literature. Traditionally, the very existence of a gap between representatives’

24 Manin, supra note 4, 163.

14

electoral promises and the decisions they make in office is perceived to be a problem for

democracy. This is because it is often assumed that voting conveys information concerning

Outcome preferences. Voters who vote for a candidate on the basis of the electoral promises are

interested in the implementation of these promises. But notice that our argument establishes that

the gap necessarily implies a challenge to democracy only in the special case of congruent

preferences. In the case of incongruent preferences, the gap can be beneficial to the extent that it

facilitates the simultaneous satisfaction of voters’ Outcome and Voting preferences. The

challenge is not the gap per se but, instead, it is how to align representatives’ decisions with the

Outcome preferences of voters.

To put it bluntly, when preferences are incongruent, it may sometimes be desirable that

representatives issue fraudulent electoral promises, i.e., promises they have little intention to

honor. Voters who vote for such representatives on the basis of such promises may under certain

conditions satisfy their Voting preferences (as the promises facilitate such satisfaction) but, at the

same time, such voters also satisfy their Outcome preferences (as the representative may

eventually violate these promises and act in accordance with the Outcome preferences of the

electorate).

This discussion sheds new light on a classical question in political theory concerning the nature

of political representation. Political theorists distinguish between two views of representatives:

trustees (free agents following what they believe is right, just, or simply correct factually), and

delegates (who follow the instructions of their constituents).25 Let us briefly examine the

underlying rationales for each position.26

Some political theorists believe that representatives are delegates and ought to be bound by the

expressed preferences. Most famously this was Rousseau’s view in Considerations on the

25 See Suzanne Dovi, “Political Representation” in Stanford Encyclopedia of Philosophy (http://plato.stanford.edu/entries/political-representation/); Heinz Eulau, John C. Wahlke, William Buchanan, and Leroy C. Ferguson, “The Role of the Representative: Some Empirical Observations on the Theory of Edmund Burke,” American Political Science Review 53 (1959), 742, 750-751.26 The terms trustees and delegates have different meanings in different contexts. For a recent investigation of the ambiguities of the terms trustees and delegates and an accusation that lying beneath this classical distinction there are three different and independent distinctions, see Andrew Rehfeld, “Representation Rethought: Delegates, and Gyroscopes in the Study of Political Representation and Democracy,” American Political Science Review 103 (2009), 214-230. In this paper we focus our attention on the second distinction of Rehefeld, namely, the distinction based on the “source of judgment.” The question is “whether the representative lawmaker relies on his or her own judgment or of a third party (to determine … the good at which he or she aims).”

15

Government of Poland. In Rousseau’s view: “the nation does not send deputies to the Diet to

give voice to their own sentiments but to declare the nation’s own will.”27 To avoid the problem

of “corruptible representatives” Rousseau suggested drawing detailed binding instructions to the

deputies and also maintained that a representative who did not follow these imperative mandates

be excluded from any subsequent mandate. Other proposals designed to narrow the gap between

representatives’ decisions and the electorate included creating a scheme of discretionary

revocability of representatives if they failed to follow the instructions of the electorate. Karl

Marx identified such a scheme in his description of the Paris commune: “The rural communities

of every district were to administer their common affairs by an assembly of delegates …each

delegate to be at any time revocable and bound by the mandat imperatif (formal instructions) of

his constituents.”28

Yet none of these proposals has prevailed; representatives have, by and large, retained the

freedom to decide as they wish.29 Representatives are trustees rather than delegates and one of

the defining features of trustees is the potential (and often actual) existence of a gap between the

representatives’ decisions and the expressed will of the electorate. Typically, the electorate

cannot force its representatives to make decisions in accordance with its will; its will is always

filtered by and, perhaps, at times, frustrated by the representatives’ decisions.

But why should representatives serve as trustees rather than as delegates? How can this

established institutional principle of contemporary democracies be justified? Two main answers

have traditionally been proposed to this question. Under the first, an ideal government would

have no such gap; representatives would operate as delegates. Yet it is difficult and costly to

design institutional mechanisms to bridge the gap effectively. Under the second answer,

representatives are often wiser or better informed than the general public and better understand

the long-term consequences of their actions. While popular input is important, it ought

sometimes to be mediated by the deputies’ decisions based on their judgments concerning the

27 See Rousseau, The Government of Poland (Indianapolis, Indiana: Hackett Publishing Company, 1985, translated by Willmoore Kendall), p. 37. 28 Karl Marx, “The Civil War in France,” in Karl Marx and Friedrich Engels, The Communist Manifesto: A Road Map to History’s Most Important Document (ed. Phil Gasper, Chicago: Haymarket Books).29 Manin, supra note 4, 167; Hanna Pitkin, The Concept of Representation (Berkeley & Los Angeles: University of California Press, 1967), chap. 6.

16

public good. Furthermore, circumstances often change in ways that necessitate choices that

conflict with the choices and preferences made by the electorate at the time of the elections. 30

However, as we have argued, there is a third possibility. The partial independence of

representatives is a desirable feature of representative governments as it facilitates the

satisfaction of both Voting and Outcome preferences. We also argue that properly speaking, a

representative who operates in the way we propose has both features of a delegate and of a

trustee. She is a trustee in the sense that she is institutionally free from the expressed dictates of

the electorate; the electorate cannot force the representative to comply with its will. Yet she is a

delegate in the sense that her choices ideally ought to be dictated by the Outcome preferences of

the electorate. Ironically, if the directives of the electorate bind her (as is the case with Rousseau-

type delegates), she could not in fact serve as a successful delegate as she would not be able to

comply with the Outcome preferences of the electorate (in cases of incongruent preferences).

Complying with the Outcome preferences may require defying the expressed dictates of the

electorate and violating electoral promises. It is precisely the representative’s independence that,

at least theoretically, enables her to be a faithful delegate, i.e., to comply with the Outcome

preferences of the electorate while, at the same time, allowing the electorate to satisfy their

Voting preferences. To illustrate: our patriotic voter is better served by being provided with the

opportunity to express his nationalist sentiments (by voting for an outspoken hawkish candidate)

while, at the same time, benefiting from the dovish policies that he actually prefers to see 30 While it is not our purpose here to dispute these explanations, the traditional explanations were in fact exposed to compelling objections. Under the first rationale it is too costly to design institutions that replicate voters’ preferences. Representatives ought to be independent because monitoring their promises is often too complicated. Yet, as Bernard Manin has shown persuasively, this rationale fails. Political theorists designed sophisticated institutional mechanisms for bridging the gap between the will of the electorate and the votes of representatives. The “partial independence of the representatives” is not an inevitable institutional feature of contemporary democracies. To the extent that it is justified it is because it is a normatively desirable feature of representative governments. See Manin, supra note 4, 163-167.

Under the second view representatives ought to be able to deviate from the will of the electorate because representatives can deliberate better or are better placed to make decisions. Edmund Burke famously argued that representatives ought to promote the public good rather than the sectarian interests of their constituencies. The will of the constituencies cannot provide appropriate guidelines to the representatives. Madison developed a slightly different version of this view. In his view representatives are necessary to "refine and enlarge" the public view. Furthermore “it may well happen that the public voice, pronounced by the representatives of the people, will be more consonant to the public good than if pronounced by the people themselves convened for this purpose.” While this view is a dominant one, it suffers from a serious flaw. Most importantly, it suffers from elitism. While this elitism may be justified at times, it conflicts with some of the premises underlying democracy. Voters are called upon to vote on the grounds that they are trustworthy and yet, representatives act as trustees precisely because voters’ judgments cannot be trusted.

17

implemented. In the language of the Sophie & Larry tale, the benevolent politician can provide

beer to the voters without depriving them of the joy of asking, demanding, or requesting wine.

By pointing out this theoretical possibility we do not argue that this is the way the political

system operates and we do not wish to underestimate the difficulties in maintaining a system in

which politicians have the relevant information concerning voters’ Outcome preferences and are

inclined to comply with it. We only maintain that the problem is not necessarily how to limit or

minimize the gap between electoral promises and the political decisions; the problem is how to

align the political decisions with the voters’ Outcome preferences.

4.2. On Empowering Benevolent Institutions: The Role of Unelected Officials/Judges

A persistent political controversy concerns the power to be given to judges or other unelected

officials. The underlying issue is similar to the one discussed in Section 4.1: judges or officials

(like representatives who act as trustees) act against the will of the electorate as expressed in

elections and, therefore, arguably, act undemocratically. Perhaps they serve the public interest

better by acting in accordance with their judgment but they do it against the expressed will of the

majority. It seems evident to most theorists that to the extent that it is justified to empower

judges and officials, it is justified not for the purpose of complying better with the preferences of

the electorate but for the reason that judges and officials are sometimes better decision-makers:

they are less prone to public hysteria, have greater expertise than the electorate in certain fields

or are particularly motivated to promote the public interest (rather than their own personal

interests), etc.

In the context of judges, this debate is framed in terms of “judicial activism” or “the counter-

majoritarian difficulty.”31 In the context of other officials this concern is sometimes labeled “the

democratic deficit.”32 In both cases the challenge is framed as balancing the need to guarantee

high-quality decisions (which arguably can be made only by judges or officials) with the need to

take into account legitimacy-based concerns (which, arguably, can only be made by the public as

a whole or by its representatives). In contrast we maintain that, at least sometimes, empowering 31 For a discussion, see Waldron, supra note 22; Alon Harel and Tsvi Kahnana, “The Easy Core Case for Judicial Review,” Journal of Legal Analysis 2 (2010), 227-256.32 See Weiler, supra note 22.

18

judges and other officials can be justified as a means of satisfying voters’ incongruent

preferences. Let us first examine the way the debate is traditionally framed in each one of these

contexts.

One major debate in democratic theory concerns the desirability of judicial review. The standard

argument for judicial review is a two-stage argument. At the first stage, the theorist addresses the

question of what the point of the Constitution is and, consequently, how it should be interpreted.

Once the “point” of the Constitution is settled, the theorist turns to identify the institutions best

capable of realizing the point of the Constitution.33 The institution in charge of interpreting the

Constitution is simply the institution most likely to interpret the Constitution “rightly” or

“correctly,” or whose decisions are the most conducive to the constitutional goals or values as

defined at the first stage of analysis. Under this view what justifies courts’ powers to strike down

legislation is (most typically) their epistemic or motivational superiority: either judges are more

informed or they are more likely to be motivated to decide correctly. The institutions that ought

to have the power to make constitutional decisions are those institutions that are most capable

and are most likely of interpreting it rightly or correctly in a way that protects human rights,

promotes democracy, or realizes other values.34

The standard argument against judicial review raises an evident and intuitive legitimacy-based

concern. Under this concern even if judges are better decision-makers they have no justification

to interfere in the decisions of the polity as such interference infringes upon political freedom

and, in particular, infringes the “right to participate on equal terms in social decisions on issues

of high principle.”35 The right to “equal participation” is under this view a precondition for the

legitimacy of the decision.

A similar debate is conducted in the context of empowering officials. The concern often raised

by theorists is sometimes labeled “the democratic deficit.” A democratic deficit occurs when

ostensibly democratic organizations or institutions fall short of fulfilling what are believed to be

the principles of democracy.36 The term “democratic deficit” has been used in many different

33 For a detailed analysis, see Harel & Kahana supra note 30, 231. 34 See, e.g., Adrian Vermeule, Judging Under Uncertainty: An Institutional Theory of Legal Interpretation (Cambridge MA: Harvard University Press, 2006), 5. 35 Jeremy Waldron, Law and Disagreement (Oxford: Clarendon Press, 1999), 213. 36 Sanford Levinson, “How the United States Constitution Contributes to the Democratic Deficit in America,” Drake L. Rev. 55 (2007), 859, 860.

19

contexts for different purposes. It indicates lack of legitimacy resulting from detachment and

alienation of executive and legislative institutions from the wishes of the citizens. One of the

standard instances of “democratic deficit” results from an increase in executive power at the

expense of parliamentary power.37 The accusation that contemporary politics suffers from a

“democratic deficit” is raised in the context of European institutions as well as the U.S. and

international organizations.38 Sometimes scholars use the term “democratic deficit” to denote “a

sense of alienation and a lack of confidence in the ability of the economic and political system to

solve our problems.”39 They also purport to establish the claim that such alienation is present by

pointing out the failure of the political system to replicate the voters’ expressed preferences.

Such theorists identify therefore the alienation with the existence of institutions that distort or

misrepresent the will of the people as was expressed in the elections.40

For the purposes of this paper we need not examine in detail the debates concerning the counter-

majoritarian difficulty or the democratic deficit. We only note that the existence of incongruent

preferences offers a new angle on these debates. Politicians could be elected on the basis of

Voting preferences. Yet the commitments politicians make have little impact as the relevant

decisions can be made only by non-elected bodies. Politicians have the power to promise and

therefore satisfy Voting preferences. Non-elected bodies have the power to act and (at least

ideally) to satisfy the Outcome preferences of the electorate. Ideally such non-elected institutions

operate like benevolent Larry; they identify the voters’ Outcome preferences and act

accordingly. Meanwhile the electoral process facilitates the satisfaction of the Voting

preferences.

This possibility implies that it may be misguided on the part of those who discuss democratic

deficit to automatically link the failure on the part of the executive branch to replicate the

expressed preferences of voters with alienation and lack of legitimacy. If voters have incongruent

preferences, institutions may fail to replicate the policies that voters expressed without

necessarily triggering alienation, as the expressed will of voters (as reflected in the elections)

37 Andreas Follesdal and Simon Hix, “Why There is a Democratic Deficit in the EU: A Response to Majone and Moravcsik,” JCSM 44 (2006), 532-62, 534.38 Andrew Moravcsik, “Is there a ‘Democratic Deficit” in World Politics? A Framework for Analaysis,” Government and Opposition 39 (2004), 336-363.39 This phrase appeared in the 1977 manifesto on European politics. See JEF Manifesto (1977) http://www.federalunion.org.uk/the-first-use-of-the-term-democratic-deficit/40 See Weiler, supra note 22.

20

conveys only their Voting preferences and those preferences may deviate from the Outcome

preferences.

We repeat here the qualifications made in Section 4.1. We do not maintain that the incongruity

between Outcome and Voting preferences necessarily provides a democratic justification for the

powers of officials and judges. Instead, our analysis shifts the emphasis towards the question of

how to design mechanisms that align decisions of non-elected officials, not with the preferences

of the electorate as reflected in the elections, but with the voters’ Outcome preferences.

4.3 Elected Trustees and Non-Elected Officials

We have established that both legislators serving as trustees and non-elected officials have the

potential to ameliorate the problems resulting from voters’ incongruent preferences. It is

worthwhile at this stage to consider the differences between the two types of institutions, in light

of the considerations raised in this paper.

At first blush, it seems that the prospects of reelection provide better incentives for

representatives to attend to voters’ Outcome preferences. After all, if elected politicians fail to

comply with voters’ Outcome preferences they may be punished by the voters. In contrast, non-

elected officials are not subject to such incentives and, consequently, care less about complying

with voters’ Outcome preferences. This argument is not as self-evident as it seems. Recall that if

voters are unlikely to be pivotal, their rational voting behavior is governed by their Voting

preferences and not by their Outcome preferences. Hence, in the presence of incongruent

preferences, voting is unlikely to reflect judgment about the performance of the incumbent

politician vis-à-vis the voters’ Outcome preferences. Rather, it will reflect the Voting preferences

(and perhaps the voter’s judgment on how well the incumbent satisfied these preferences). If

elected trustees understand that Outcome preferences do not significantly affect voting behavior,

then reelection considerations are not likely to provide them with strong incentives to comply

with voters’ Outcome preferences.

This discussion suggests a criterion for comparing the suitability of decision-makers – whether

elected or appointed – to make decisions. In both cases, the more the decision-maker shares the

21

Outcome preferences of voters the more suitable she is to make the decision. As we have just

argued, it may well be the case that in the presence of incongruent preferences non-elected

officials are more likely to act in conformity with the Outcome preferences of the electorate than

elected officials. As rational voters in large elections tend to vote on the basis of Voting

preferences (rather than Outcome preferences), elected decision-makers are likely to be

especially attentive to Voting preferences. Non-elected officials, on the other hand, are far less

bound by this constraint. To the extent that their preferences coincide with the population’s

Outcome preferences (not a trivial matter) they are in a better position to execute policies that

satisfy the population’s Outcome preferences.

In practical terms, this argument offers support to the view of political theorists who advocate

“representative bureaucracy,” namely, a bureaucracy that is “reflective of the interests, opinions,

needs desires, and values of the general public.”41 The literature distinguishes between two ways

to realize this ideal: passive representation and active representation.42 Passive representation

requires the recruitment of a “bureaucracy that mirrors the demographic composition of the

entire population”43 or that “reflects the diversity of the general population.”44 To the extent that

the bureaucracy satisfies this requirement, it shares its Outcome preferences with the citizens

and, consequently, promotes the Outcome preferences of the population. Even if bureaucrats

have Voting preferences similar to those of the population at large (that are incompatible with

the Outcome preferences), they are unlikely to act on the basis of such preferences as in many

cases bureaucrats are pivotal; their decision determines the policies executed. Alternatively the

bureaucracy can actively and consciously promote the preferences of the citizens not because it

has the same preferences as citizens but because citizens can actively influence bureaucracy or

because they regard promoting such preferences as part of their mission.45 Such representation

requires “bureaucrats’ advocacy of their constituents’ interests, making policy decisions that

41 Lael R. Keiser, , Vicky M. Wilkins, Kenneth J. Meier, and Catherine A. Holland, “Lipstickand Logarithms: Gender, Institutional Context, and Representative Bureaucracy," American Political Science Review 96 (3) (2002), 553-564. 42 See Fredrick C. Mosher, Democracy and the Public Service (New York: Oxford University Press, 1968), p. 12. 43 See Rhys Andrew et al., “Representative Bureaucracy, Organizational Strategy, and Public Service Performance: An Empirical Analysis of English Local Government,” Journal of Public Administration Research and Theory 15 (2005), 489-504, 491. 44 Sally Coleman Selden, The Promise of Representative Bureaucracy: Diversity and Responsiveness in a Government Agency (London: M.E. Sharpe, Inc. 1998).45 For a discussion of active representation see, e.g., Fredrick C. Mosher, Democracy and the Public Service (New York: Oxford University Press, 1968), p. 12.

22

benefit a given group.”46 In terms of this paper, a representative bureaucracy is a bureaucracy

that promotes the Outcome preferences of the electorate, like Jack in the Sophie & Larry fable.

4.4 On Being Listened to

The institutional proposals designed to satisfy both Outcome and Voting preferences raise a

concern similar to the one raised in the context of benevolent Larry. Benevolent Larry who

consistently ignores Sophie’s request cannot satisfy Sophie’s Voting preferences as (at least

sometimes) those preferences hinge on Sophie’s conviction that her request will be honored. If

Sophie is consistently ignored, she may lose the sense that uttering “wine” is, as a matter of fact,

nobler than uttering “beer” as the utterance “wine” counts only if this utterance affects what she

is served. Similarly it could be the case that voters’ Voting preferences hinge on voters’ power to

make a difference or on their feeling that their vote really counts. If voting does not matter,

voters may well perceive elections as a charade and, consequently, may not derive Voting

satisfaction from their vote.

The issue raises an empirical question. Do voters have Voting preferences that are sensitive to

the degree to which voting makes a difference to the outcomes of the elections? To the extent

that they do, there seem to exist two possible mechanisms that may ameliorate the problem: one

psychological, the other institutional.

First, voters may engage in self-deception and believe in the credibility of electoral promises,

even after a long experience of politicians reneging on their promises: “This leader is different;

she really is loyal to her voters.” However widespread this phenomenon is in practice, it may

well be perceived as inherently undesirable on grounds of political morality, even if such self-

deception serves voters’ welfare. The second mechanism seems less objectionable. Politicians

ought not to violate promises too often even when such violations may serve Outcome

preferences. Benevolent politicians who are aware of this dilemma ought to violate their electoral

promises only rarely. In making a decision to comply with the Outcome preferences and violate

their electoral promises, they ought to consider the possibility that they may undermine the

credibility of electoral promises and thus undermine the satisfaction of Voting preferences.

5. Conclusion

46 See Keiser et al., supra note 40, 553.

23

Elections are traditionally perceived as a means to satisfy voters’ Outcome preferences. Yet it

has been recognized again and again that people often vote for reasons other than promoting

their Outcome preferences (moral, expressive, etc.). This paper raises a conjecture that naturally

follows from this observation, namely, that it is possible that voters have incongruent Outcome

and Voting preferences. Voters may wish to vote for one candidate/policy and yet also wish that

another candidate or another policy be implemented. As we have argued, the existence of such

preferences poses a grave challenge to democracy, one that is quite distinct from the agency and

aggregation challenges prominent in the literature. Last, we demonstrated that these challenges

offer a new perspective for evaluating democratic institutions, and, in particular, it sheds new

light on the delegate-trustee debate and the powers granted to unelected officials.

The paper gives rise to two sets of questions. The first consists of empirical questions that have

heretofore received little attention from social scientists. Most importantly: how often do

incongruent preferences occur in reality and are Voting preferences sensitive to the vote

potentially having some effect on the resulting outcomes?

But the paper also raises some conceptual questions. Suppose there exist no differences in

preferences across different people and that a benevolent and omniscient entity came down to

earth, equipped with perfect knowledge of people’s Outcome preferences. Would such an entity

make elections redundant? If the purpose of elections is merely to satisfy voters’ Outcome

preferences (or to equally advance their interests) then the answer is trivial. Yet perhaps there are

other reasons we want to hold elections. One such purpose, which has been much discussed, is

that elections engender discourse and deliberations that can alter (presumably for the better)

people’s preferences. Our analysis, however, suggests another reason. Perhaps elections have

intrinsic value as they facilitate the realization of Voting preferences. Revisiting our previous

example, elections provide an avenue for the working class voter to convey his commitment to

the labor movement, give voice to his moral convictions, express his solidarity with family and

community, or develop class identity – regardless of his Outcome preferences. In this sense,

elections may have value beyond being an instrument of decision-making designed to satisfy

voters’ Outcome preferences.

24

25