the good, the bad and the populist: a model of political agency with emotional voters

European Journal of Political Economy 27 (2011) 611–624

Contents lists available at ScienceDirect

European Journal of Political Economy

j ourna l homepage: www.e lsev ie r.com/ locate /e jpe

The good, the bad and the populist: A model of political agency withemotional voters

Colin Jennings⁎Department of Economics, University of Strathclyde, Glasgow, G4 0GE, UK

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +44 141 548 2910.E-mail address: [email protected].

1 Early examples are Barro (1973) and Ferejohn (12 For an overview of Besley's framework, see Liu (2

0176-2680/$ – see front matter © 2011 Elsevier B.V.doi:10.1016/j.ejpoleco.2011.03.005

a b s t r a c t

Article history:Received 8 January 2010Received in revised form 9 March 2011Accepted 23 March 2011Available online 8 April 2011

This paper extends the political agency approach to an environment in which voting iscategorised into informed and instrumental, informed and ‘expressive’, and uninformed due to‘rational irrationality’. Politicians may be “good”, “bad”, or “populist”. Initially the existence ofonly good and populist politicians is assumed: the incentives for good politicians to pool withor separate from populists are investigated and the implications for voter welfare are explored.Then the consequences of the inclusion of bad politicians are considered. The paper makesthree main contributions. First, I provide a rational choice analysis of populism as populism iscommonly understood. Second, I locate a potential role for government as a persuasiveprovider of information regarding the quality of policy. Third, when bad politicians are added tothe analysis, it is shown that a little potential corruption can improve voter welfare.

© 2011 Elsevier B.V. All rights reserved.

JEL classification:D72

Keywords:Political agencyExpressive votingRational irrationalityDemocratic inefficiencyPopulism

1. Introduction

This paper incorporates a broader depiction of voter rationality and politician types than that which is normally used in apolitical agency framework and studies how this depiction feeds into the analysis of economic inefficiency as a product of thedemocratic political process.

Political agency models emerged in response to models of electoral competition stemming from Downs (1957).1 Electoralcompetitionmodels were open to the criticism that campaign promises are cheap talk. Political agencymodels developed from theperspective that voters decide to vote on the basis of incumbent performance rather than electoral promises. Besley (2006) hasprovided a comprehensive overview of this area of political economics. In these models voters base their decision on theperformance of the incumbent, so voting is retrospective and voters update their beliefs regarding the quality of the incumbentaccording to Bayes Rule. Voters in these models are therefore highly rational.

In thebaselinepolitical agencymodel, lowquality incumbents (either in termsof competenceor character) faceadecisionwhetherto behave well or not. If they behave well and provide good governance, the reward is that they are more likely to be re-elected andenjoy the opportunity to engage in rent extraction in the future. If they behave badly, the reward is that they enjoy rents today but thecost is that they are less likely to be re-elected and enjoy rents in the future. Besley labels these politicians as ‘dissonant’, as theirpreferences are not aligned with the interests of the general public. High quality incumbents are ‘congruent’ because they arerewarded for doing the right thing,which is their desire in any case as their interests are alignedwith thoseof thegeneral public. Besleysurveys many extensions to this basic model.2 In this paper I focus on one extension, to models of pandering or populism.

986).007).

All rights reserved.

http://dx.doi.org/10.1016/j.ejpoleco.2011.03.005

mailto:[email protected]

http://dx.doi.org/10.1016/j.ejpoleco.2011.03.005

http://www.sciencedirect.com/science/journal/01762680

612 C. Jennings / European Journal of Political Economy 27 (2011) 611–624

Voters do not always seem to vote their interests, and in this situation it is the good politicianwho faces a dilemma. Does he or shegive the voters what theywant (butwhich is not in their interests) in order to do the ‘right thing’ in the second period or do the ‘rightthing’ today and risk not being re-elected and allowing the possibility of a bad politician taking charge in the second period? This styleof problemhas been explored in Canes-Wrone et al. (2001),Morris (2001) andMaskin and Tirole (2004), who assume that voters arefully rational and Bayesian, and as a result emerge with what are arguably quite complicated stories as to why we might observepopulism. For example, in Maskin and Tirole the idea is that voters decide that a certain policy is one that would be enacted by a badpolitician (based on experience) and vote against it. A good incumbent discovers that this policy is, in fact, one that is in the interests ofvoters but in order to be re-elected must implement the inefficient, but popular, policy instead. This view of pandering/populism israther inconsistent with the general use of those terms in popular publications.3 When used in journalistic articles by economiccommentators, ‘populism’ tends to identify policies that all economists in general would recognise as bad.4 If such populism iscommon, the theoretical problem is that this suggests that a degree of irrationality amongst voters exists that cannot easily bedismissed. Some policies are supported that voters should be expected to oppose, given the effect of the policies on their utility.

However, is voting against one's self-interest actually irrational? In this paper I provide an analysis of populism in which somevoters are uninformed due to rational irrationality (Caplan, 2002, 2007) and others are informed and expressive (Brennan andLomasky, 1993), in addition to the conventionally informed and instrumental voters. The concepts of rational irrationality andexpressiveness will be discussed in more detail in the next section, but it is sufficient for now to state that voters may rationallyexpress an electoral preference for policies that are not in their economic interests. Recognition that voters may be rationallyirrational and/or expressive suggests that, in addition to the dissonant (bad) and a congruent (good) politician as defined byBesley, a third type is likely to exist, the populist. The populist is not corrupt in terms of seeking to maximise personal rents, butwill do whatever the majority of voters want, even if he or she knows it is not in the voters' interests. I thus expand the set ofpolitician types to include good, bad, and populists. Unlike good and bad politicians who face dilemmas regarding the merits ofholding on to political power, the populist faces no such dilemma and is thus the familiar Downsian office-seeking politician.

Given that I allow for the existence of uninformed voters, paternalistic information provision regarding good policy wouldpotentially be worthwhile for the good politician. I also analyse the strategic choices facing different types of politician whenmaking policy decisions, the conditions that will determine their choice, and the efficiency implications of these choices.

The next section discusses related literature. In Section 3 the basic model is outlined: the setting of the game is introduced; thestrategies and payoffs for the politicians and voters are outlined, as is the timing of the game. The motivations of the three types ofpolitician and the different types of voter are discussed in detail. At first there are no ‘bad’ politicians and the focus is on threepossible options for incumbent good politicians — to choose good policy, populist policy, or in between through education of thepublic. The welfare effects (in terms of the quality of policy measured across two periods) of these strategies are analysed and theconditions under which they may emerge are explored. Section 4 introduces bad politicians and shows that some potential“badness” is potentially welfare-improving ex ante because the potential for a good politician to be replaced by a bad politician at alater stage makes the good politician more attractive in the eyes of the voters and frees him or her to set optimal rather thanpopulist policy and win the election. Also, a possible strategy for good and populist politicians is described that would allow themto separate from bad politicians. Section 5 offers concluding comments.

2. Related literature

This paper is concerned with the extent to which full rationality can be assumed on the part of voters and the implications ofthis for the efficiency of the democratic process. Downs (1957) introduced the problem of ‘rational ignorance’. The idea is that,given the very low probability of determining the outcome of an election, there is little incentive for voters to become well-informed about the links between policy and outcomes. The concept of rational ignorance became an important component of the‘Virginia School’ of public choice. In, for example, Brennan and Buchanan (1980), the idea of voters having great difficulty holdingpoliticians to account became a key contribution to their emphasis on democratic inefficiency and the importance ofconstitutionally set fiscal restraints. Voters may suffer from ‘fiscal illusion’ or they may not be cognisant of costly ‘sneaky transfers’to rent-seeking interest groups (Tullock, 1983).

This led to a response by the ‘Chicago School’ (notably Becker ,1983 and Wittman,1989, 1995). Wittman attempted to dismissconcerns regarding voter irrationality and argued that democratic institutions are, in fact, efficient. He recognised the fact thatindividual voters are insignificant and that this could lead to individual biases away fromwhat the voter would understand to be acorrect policy if fully informed, but overall we should expect individual biases to cancel out. In addition, if voters are being fooled,why would we expect this to persist over time? To a large extent, Wittman was confronting an issue that many politicaleconomists had already assumed away. The vast majority of papers in political economics do not worry about the implications ofvoter insignificance and assume full voter rationality. Nonetheless, whilstWittman's contention that a strong version of rationalityshould be assumedwould generally meet with vigorous approval within the economics profession, the emphasis on the efficiencyof democracy would not find a common consent.

3 See for example ‘The return of populism’, The Economist, April 12, 2006.4 There is, of course, a different use of the term ‘populism’ in the sense that it relates to movements of the people standing up to powerful elites (see Canovan

1999 and the Economist cited above). This perspective treats populism as a term related to distributional conflict and zero-sum outcomes. The derogatory use othe term populism, as used in this paper, implies inefficiency and negative-sum outcomes. Mejía and Posada (2007) provide a treatment of populism as adistributional conflict, but which also contains the potential for inefficient outcomes.

,f

613C. Jennings / European Journal of Political Economy 27 (2011) 611–624

Surveys such as Rodrik (1996), Robinson (1998) and Besley (2006, chapter 2) present various explanations for the existence ofdemocratic inefficiency. However, where inefficiency does seem clearly to exist, the source of the problem should not be identifiedas related to voter rationality but rather as the equilibrium outcome of the political game played between strategic agents (orgroups of agents).5 This point is made very forcefully by Rodrik (1996) who after surveying rational models of badmacroeconomicpolicy writes ‘they confirm that we can do better than resort to myopia or irrationality when explaining social phenomena’ (p.25).Examples of bad macroeconomic policy could be excessively large budget deficits and inflationary finance. In microeconomicpolicy, bad policy could refer to excessive protectionism and other market interventions that create deadweight losses that arenot justifiable on grounds of social justice. Outside economics, inefficiency can arise through aggression where communities orcountries are in conflict. The more efficient, but non-popular policy approach would often be to work towards peace.6

I contend that irrationality may play a role in the democratic process but that it should be understood as rational irrationality,as developed by Caplan (2002, 2007). A crucial finding in Caplan's empirical work is that citizens untrained in economics havesystematically biased beliefs with respect to citizens trained in economics. He shows that these biases can be seen in four mainareas — anti-market bias, anti-foreign bias, make-work bias and pessimistic bias. The biases exist because people want to havebeliefs that may run contrary to evidence. Consider, for example, the belief in creationism. Beliefs are a normal good and, whentheir price is low, demand for themwill be high. The price will be low in situations where, if the individual were to alter his or herbelief, it would make no (or little) difference to his or her material existence. In democratic elections, the probability of beingdecisive is extremely low, so changing beliefs is unlikely to make any difference to the outcome of the election, and as a result tothe life of the individual voter. Furthermore, there is, in addition, no incentive to even become acquainted with evidence thatwould suggest that there is an alternative and more accurate view of the world than the one a person holds. Rationally irrationalvoters do not possess information that would conflict with their belief system, as they have no incentive to acquire it.

Note that rational irrationality differs from rational ignorance in that an ignorant voter cannot be presumed to have a particularopinion. There was no role within rational ignorance for a demand for beliefs. A rationally irrational voter, on the other hand, hasan opinion because there is a role for a demand for beliefs. When the majority of voters hold biases such as those identified byCaplan, a fairly direct rule of thumb is provided for the policies incumbents should implement should they wish to be re-elected.This allows for the common sense understanding of populism as an identifiable policy package that is defined as always inferiorand one that fully rational instrumentally motivated voters would never select.7

The norm in the political agency literature is to present policy as a simple binary choice of good or bad. I present amore realisticdepiction of policy choice and assume that it is continuous and that there is an option for the incumbent to engage in publiceducation of good policy. This introduces a novel perspective to both the seriousness of the inefficiency (for exampleprotectionism, fiscal deficits and violent conflict are not generally either/or issues but rather issues of howmuch) and to the waysin which government can send signals. One way in which a government can send signals is to engage in costly acts of publiceducation or persuasion regarding the nature of good policy. Romer (2003)makes the point that education is very important whenbad policy is caused by misconceptions. If the source of inefficiency in a model is held to be unrelated to misconceptions, thenclearly education would be irrelevant because the voters are already fully informed. Like Caplan, Romer argues thatmisconceptions are real and widespread, and thus form a major contribution towards bad policy. As an appeal to the reality ofbiases, both Caplan and Romer point to the clear biases that university students arrive with and which teachers in economics andmany other subjects try to correct. Romer emphasises that biases also exist at the higher and perhaps more worrying level ofpublic officialdom and policy-makers.

The model will be further enriched by the presence of expressive voters. These are closely related to Caplan's rational irrationalvoters but there is a crucial difference. Expressive voters are fully informed. Expressive voters hold beliefs but these beliefs arechallenged by the voter possessing an understanding that choosing expressively may result in outcomes that are against theirinstrumental interests. In market settings where individuals are decisive, we expect people to choose instrumentally because thecost of choosing expressively is too high. In mass elections, however, expressive voters may not vote their instrumental interestsbecause they know their votes will not affect the outcome of the election. If expressive voters define themselves as protectionist orreligious fundamentalists, an understanding that protectionism or religious fundamentalism is not in their instrumental interestsmay have no effect upon their deciding to vote for protectionist or fundamentalist positions for the very reason that these votersunderstand that their vote is very unlikely to entail any instrumental consequence.

Some might find that the idea of expressively voting for policy X when the voter knows that Y is instrumentally better is arather strong assumption. Aweaker versionwould be that the voter ex post rationalises that X is better than Y to suit the emotionalattachment to X. As Westen (2007) writes ‘what passes for reasoning in politics is more often rationalization, motivated by efforts toreason to emotionally satisfying conclusions.’ (p. xi). Is the rationalisation of emotional beliefs compatible with expressive voting?How is this different to rationally irrational voting? An expressive voter might say that policy X is best (although Y is, in fact, best),but in reality knows that Y is best, but his or her insignificance in democratic participation does not force him or her to beconfronted with any consequence from denying the truth. In contrast, rationally irrational voters actually believe X is best, as they

5 The identification of inefficiency does not imply that it can be eliminated. A seeming inefficiency may be constrained efficient.6 Examples of papers that attempt to provide fully informed, rational explanations for socially undesirable fiscal deficits are Persson and Svensson (1989) and

Alesina and Tabellini (1990). Examples of papers that attempt to account for inefficient redistribution are Coate and Morris (1995), Acemoglu and Robinson(2001) and Aidt (2003b). Fearon (1995) provides a review of rationalist explanations for war.

7 An important question that remains to be answered is where does the bias come from?


have successfully avoided processing information that would lead them to understand the truth that Y is a superior policy, even ifthis is a truth they would subsequently attempt to rationalise away.

Hillman (2010) links expressive behaviour to confirmation of chosen identity. It seems clear that many acts confirm identitythat people have chosen, and, unless individuals are faced with a significant material cost of conforming to that identity, they haveno incentive to deviate from behaving expressively and reaching an ‘emotionally satisfying conclusion’. Elections and rhetoric areclear examples of acts for which the cost of choosing expressively is low. If votes or rhetoric were decisive, the cost of expressivebehaviour would be high compared to the benefit of confirming identity (say as a benevolent person) and reaching a materiallysatisfying conclusion would become a priority at the expense of expressive utility.8

The extent to which voters are expressive is, in turn, important in determining the extent to which information provision iscapable of converting rationally irrational voters towards supporting better policies. The depiction of expressive voting, up to now,has been a normatively negative one as it reinforces rational irrationality. An alternative, normatively positive view of expressivevoting is provided by Brennan and Hamlin (1999)who focus on the expressive selection of moral characteristics. Both the negativeand positive perspectives regarding expressive voting will be considered in this paper.9

Rationally irrational and expressive voters are both labelled ‘emotional’ and the extent to which these emotional voters arerationally irrational or expressive is central to the analysis. The hope is that this richer depiction of voters' types and motivationsgoes some way to meeting the challenge set by Besley (2006) to incorporate behavioural approaches into political agencymodelling. “Going forward it would be interesting to understand better what the differences are between behavioral models of politicsand the postulates of strict rationality supposed here. It would be useful to understand when simple and sensible behavioral rules lead tolarge policy distortions.” (Besley, 2006: 172).10

As an explanation for inefficient policies, my depiction of emotional voting is a complement rather than a substitute foranalyses that assume full instrumental rationality and focus on strategic interaction as the source of inefficiency. ConsiderAcemoglu and Robinson's (2001) discussion of inefficient redistribution. They focus on the inability to form binding commitmentsas the trigger that leads interest groups to seek transfers through market interventions rather than the relatively more efficientchannel of cash transfers. Political power comes with maintaining large numbers and inefficient transfers keep current and futuremembers within the interest group, whereas with cash transfers membership of the interest group would decline and thus thegroup would weaken, and with it the group's ability to extract transfers. Therefore, inefficient transfers are in the material self-interest of the members of the interest group. They apply the model to agricultural, labour market, and trade policies.

Interest groups benefit from inefficient transfers, but it is also the case that such policies may be supported by voters who arenot members of the beneficiary interest group and who are actually harmed by the policies. These are the sorts of voters whoappear in this paper and it is unlikely that these ‘emotional’ voters would find cash transfers to interest groups emotionallyappealing. Logically, where an inefficient policy that benefits a small group of people appears to receive a very large level ofsupport, it would seem clear that a large proportion of the supporters are not material beneficiaries of the policy. Evidence for bothtypes of support for protectionism is found in Mayda and Rodrik (2003). In addition to finding support for protectionism fromthose groups that benefit from it, they also find that ‘protectionist attitudes go together with a well-defined set of normative attributes.Individuals who favor trade restrictions tend to have high attachments to their neighborhood and community, have a high degree ofnational pride, and believe that national interest should be paramount in making trade-off’ (p.1395). From the perspective of theanalysis presented here, an interesting issue is whether the desire for protectionismwould persist, if these voters were placed in aposition in which their beliefs impinge on their material utility through decisiveness in determining the outcome of an election.Clearly the support of large numbers of emotional voters who are actually made worse-off by inefficient transfers greatly increasesthe power of the recipient group. The same sort of analysis might also be applied to conflict, where emotional group attachmentprovides the large-scale political support essential for the much smaller group of agents who actually do benefit materially fromthe conflict. Again, we might conjecture that such support would be greatly reduced if the supporters found themselves in aposition in which their support is decisive in determining whether their group will engage in conflict. A key implication is thatinefficient policies that generate emotional attachment (such as nationalism or more generally factionalism of various kinds) havea greater chance of success, as they attract large levels of support both from the relatively small group of individuals whomaterially benefit from the policy and a relatively large group of individuals who do not andwho will not be open to persuasion. Ifan interest group can present its case as emotionally appealing, it is placed in a much stronger political position.11

8 See Hamlin and Jennings (2011) for a review of the theoretical and empirical literature on expressive choice.9 Given the existence of informed and uninformed voters, a link can be made between my paper and the swing voter's curse of Feddersen and Pesendorfer

(1996), wherein less informed voters have an incentive to delegate the decision to more informed voters via abstention. I do not consider abstention, althoughthe implication is that abstention of rationally irrational voters would be no bad thing. The assumption of no abstention is justified by recognising thatuninformed (and informed) voters receive expressive benefits that could outweigh costs of voting. Expressiveness does not play a role in the swing voter's curse.See however Sobbrio and Navarra (2010) for an empirical study of the swing voter's curse and expressive choice in elections.10 Brennan and Hamlin (1998) amend the standard model of electoral competition to include voting that is expressive. The paper presented here could beviewed as a similar attempt to amend the standard model of political agency to incorporate expressive voting. It is also worth noting that the Review of AustrianEconomics has published a symposium on Besley's book. This includes contributions by Brennan (2009) and Caplan (2009) in which they are critical of Besley'sdepiction of voters ignoring expressiveness and rational irrationality. Besley in his reply acknowledges this criticism and repeats his interest in attempting toincorporate behavioural approaches into political agency modelling.11 Kliemt (1986) makes a similar argument.


3. The model

3.1. The setting

It is supposed that there is a policy PS that generates an inefficient social outcome which (for the reasons discussed so far andfurther elaborated upon below) would be favoured by the majority of voters. As discussed, this could relate to policies such asheavy protectionism or other deadweight loss creating interventions that would be unjustifiable on equity grounds. Inmacroeconomic policy it could be an excessively large budget deficit or the policy could be one of pursuing a destructive war.Alternatively, (and more in keeping with the Maskin and Tirole (2004) set-up) the policy could be the status quo approach thatwas appropriate in the past but exogenous shocks have occurred that render this policy approach currently inefficient. PS isintended to cover a very broad range of potential inefficient social policies.

The incumbent government is assumed to have access to informationwhich instructs it as to best policy P⁎. Policy is depicted ascontinuous from PS to P⁎ such that voter welfare (W) is maximised at P⁎where a range of policies could be debated as reasonablemaximisations ofW and the further PS is from P⁎ the greater the initial inefficiency.12Welfare is increasing from PS to P* soW′(P)N0. We may expect that slight moves away from PS towards P* may bring relatively big returns that diminish as policy movestowards P⁎ so that W″(P)b0. It is assumed that there is a maximum level of tax revenue T from which public spending orappropriation is made.

3.2. Politicians

There are three types of politician; good, bad and populist. Let π be the probability that a randomly picked politician is good, κthat he or she is bad and 1−π−κ that he or she is a populist. The objectives of these politicians differ. It is helpful to first identifytheir objective if they were only to be in power for one period and did not face re-election incentives. In a two-period model thisalso, of course, identifies the behaviour of the politicians if they were to be in power in the second period.

All politicians are assumed to receive an ego rent E from being in power. The good politician is concernedwithmaximisingW sowould select P⁎ and receive a payoffW⁎+E. The bad politician (and ‘bad’ is taken tomeanmorally bad rather than incompetent) ismotivated to steal tax revenue for personal gain, so would simply steal T and receive payoff T+E. The populist is concerned onlywith being popular with the electorate. His or her incentive for being in power is to generate ego rents and the populist would haveno interest in stealing from the electorate as this would obviously be unpopular. The populist would never steal but will wish to setwhichever policy is most popular even when he or she knows it is not welfare-maximising. His or her payoff to being in power isassumed to be E. As we shall see, the politicians may be forced to trade-off maximising their single period objective in order toachieve re-election. In this section of the paper it is assumed that κ=0 so that there are only good and populist politicians. Thisassumption will be relaxed in Section 4.

3.3. Voters

Voters are heterogeneous with ω well informed and (1−ω) rationally irrational. The well informed voters correctlyunderstand the link between policy and outcomes so that they know that P⁎maximises their welfare. They are not informed aboutthe identity of the incumbent, but they understand that politicians may behave strategically and, based upon the behaviour of theincumbent politician, they update their beliefs as to his or her identity and use this updated information when making theirdecision to re-elect or not. The rationally irrational votersmisunderstand the link between policy and outcomes and believe that PS

maximises their welfare. Their voting strategy is simple. They re-elect if the incumbent politician implements PS and will vote forthe challenger if the incumbent does otherwise. Rationally irrational voters are thus depicted as using a basic rule of thumb withbias. They do not consider the implications as to why a politician may choose an unpopular policy. The justification for assumingthis kind of voting behaviour is that rationally irrational voters are not likely to devote much time to considering the politicalstrategy of politicians and the reason why they do not is that their vote is very unlikely to be decisive. Well-informed voters alsounderstand that their votes are unlikely to be decisive, but they became informed about the merits of policy because they werewell educated or followed political and economic issues out of personal interest. Rationally irrational voters would become betterinformed about the link between policy and welfare if they were to be educated about it.

Expressive voting is also considered. Recall that expressive voting is compatible with being well-informed. Voters mayexperience an inner tension such that the policy that they know would be in their instrumental interest does not hold a strongexpressive appeal. In situations of decisiveness, we may be justified in ignoring expressive preferences, as we would expect votersto choose their material interests given that they are more likely to be concerned about this than any expressive loss that theymayexperience. So, for example, if a voter is decisive and understands the law of comparative advantage, wemight expect the choice of

12 Trained social scientists obviously disagree over the appropriate level of protectionism and other market interventions such as the minimum wage, theappropriate level for budget deficits, whether violent conflict is on some occasions the correct response to a perceived threat and so on. The point made here isthat the policy debate amongst experts is very likely to have a much smaller variance than the debate between non-experts and more importantly that the meanopinion is very often (as in Caplan's findings) sharply different between the two groups. P⁎ can be viewed as covering the mean policy that would berecommended by the well-informed although it is recognised that there is disagreement among this group, whereas PS can be viewed as the mean policy thatwould be recommended by the uninformed.


free trade rather than protectionism, or peace rather than war although emotionally protectionism and war may be moreappealing. However, because in most voting scenarios the probability of being decisive is very low, the price of choosingexpressively is also very low such that well-informed voters may vote for protectionism or war even though these are policies thatthey actually know (if they are being completely honest with themselves) are not in their material interests. It is assumed that aproportion of well-informed voters may receive purely instrumental benefits from voting. It is assumed that voters incur no cost ofvoting and because there is a non-zero probability of a voter being decisive (potentially an infinitesimally small probability),instrumental voters therefore always vote — as do all voters. I do not consider abstentions.

To formalise this, it is assumed that all voters have a level of emotional, expressive attachment r maximised at PS that isdistributed across all voters on [0, R]. Those with rN0 at PS experience an expressive attachment to PS and those with r=0 do not.As Pmoves from PS, those with an expressive attachment experience an expressive loss such that r′(P)b0, whereas those with noexpressive attachment do not, such that r′(P)=0. Since r is decreasing in P for the attached voters, its valuemay at some point turnnegative over the range [PS, P*]. This will be especially so for voters with r″(P)N0 and the analysis will consider zealots who holdsuch a strong attachment to PS that any deviation from it will cause them to receive negative expressive returns. Only well-informed voters may experience internal tension with movements away from PS. Rationally irrational voters experience notension between their expressive and instrumental preferences, as they believe PS is also the best policy. It is assumed thatexpressive preferences are invariant to education, but instrumental preferences would be revised in the light of new information.

3.4. Timing

Nature determines the type of politician who then picks a preferred action. The incumbent is either re-elected, or, if thechallenger wins, nature determines the type of second period leader. The incumbent has the option to choose as part of policy acampaign of costly public education. It is assumed that voters observe the action but the revelation of the payoff is delayed untilafter the election. The second period incumbent chooses his or her preferred action and at the end of the second period the gameends.

3.5. Equilibrium

Perfect Bayesian equilibria of this game are characterised. In period 2 the politicians do not face re-election so the goodpolitician will choose P⁎, which generates voter welfare W⁎. The populist simply wants to be popular and will select PS. Thedilemma facing the good politician is which policy should be selected in period 1. A key idea here is that setting policy at PS isguaranteed to win the election and any policy other than PS will be defeated.

For now, the possibility of public education is ignored. The good politician will be tempted to choose the populist position PS inorder to be re-elected, although this means sacrificing welfare W⁎ for WS. Alternatively, he or she could choose P⁎ and optimisefirst period voter welfare, but thus accept defeat in the election and hope that the challenger who defeated them is a goodpolitician. The risk for the good politician is that he or shewill be replaced by a populist whowill implement PS thus generating thelower welfare WS for the electorate which, unlike a populist, the good politician cares about. In addition, the defeated goodpolitician would lose his or her ego rents in period 2.

Given that there arewell-informed votersω and rationally irrational voters (1−ω), it does not follow automatically that P S isa guaranteed election winning policy. The conditions for this to be the case need to be demonstrated. Note that (1−ω) of thevoters are rationally irrational and believe P S to be best policy. If presented with a policy of PS they will vote for the incumbent.Clearly if 1−ωN1/2 then populist policy is guaranteed towin. The analysis will, however, cover the broader casewhere 1−ωb1/2 and ω−1/2 votes will need to be won from the set ω of well-informed voters to ensure that PS has majority support and thatany other policy would fail to obtain majority support.

Two types of equilibriumwill be considered. The first describes the conditions required for a separating equilibriumwhere thepopulist chooses PS and wins the election and the good politician chooses P* and loses the election. The second describes theconditions required for a pooling equilibrium where both the good and populist politicians choose PS and win the election.

3.5.1. Separating equilibrium

In order for PS to be awinning policy,ω−1 = 2

ωof well-informed voterswould be required to vote for the incumbent offering this

policy rather than a challenger who will be good with probability π and populist with probability (1−π). That is, a well-informedvoter i has to weigh up the payoff of voting for the incumbent given by

versus

ri PS� �

+ h WS� �

ð1Þvoting for the challenger where the payoff is given by

h πW� + 1−πð ÞWS� �

ð2Þ

Pay-off (1) tells us that, for those voters that experience rN0, there is a positive expressive attachment to the populist policy,but that must be weighed against the potential for deciding the outcome of the election with probability h and incurring the


instrumental payoff of WS. Note that the instrumental component h(WS) of (1) is the lowest payoff attached to policy PS. If theincumbent were known to be good, then the instrumental component of the payoff would be h(W*). If nothing had been learntabout the identity of the incumbent, then the instrumental component would be the same as the payoff in (2). So if (1) is greaterthan (2), then this will be true for all possible cases. Payoff (2) tells us that there is no expressive attachment to a challenger whohas not implemented policy, but the instrumental payoff is greater than WS as with positive probability the challenger may turnout to be a good politician. In the standard political agency analysis where there is no expressive voting and voting is assumed to bedecisive, the challenger would win all the votes of the well-informed voters. The assumption that there is no expressive votingseems reasonable when voters are assumed decisive. In this case we could justifiably argue that emotional attachment to policiesthat are in the past are of trivial significance compared to the material interests of voters in the future and thus it makes sense toignore the former. The problem, as should be clear by now, is that voting is only going to be decisive with a probability of h and thisgreatly reduces the importance of future instrumental welfare for well-informed voters when making their voting decision. Therole of the expressive component in voting increases as h approaches 0.13

A well-informed voter i will vote for the populist incumbent rather than the challenger if

this re

13 Notsuch anbe extrethe valu

ri PS� �

N hπ W�−WS� �

ð3Þ

Ifω−1= 2

ωof the well-informed voters satisfy Eq. (3), then populism implemented by a known populist wins. Clearly this

case is more easily satisfied the closer ω is to 1/2, the closer h is to 0 and that there are a sufficient number of well-informedvoters with a sufficiently strong expressive attachment to produce the necessary majority for populist policy.

The conditions required for any policy P≠PS to be an election loser are now addressed. In this case, the (1−ω) votes of the

rationally irrational are lost due to selecting P≠PS. In order to win the election the good politician will need to attract at least1 = 2ω

of the well-informed votes. A well-informed voter i has to weigh up the payoff of voting for the good incumbent given by

ri P≠PS� �

+ h W�� ð4Þvoting for the challenger where the payoff is given by Eq. (2). Note that in Eq. (4) the instrumental component of the payoff
versus
is maximised. So if Eq. (4) is less than Eq. (2) this will be true for all possible cases. For the challenger to win this election and thuslead to the electoral defeat of any deviation from PS, the following must be true for at least 1− 1 = 2

ω of the well-informed group ofvoters

h 1−πð Þ WS−W�� N r P≠PS� �

ð5Þ

These voters must experience sufficiently large expressive losses from any deviation from PS such that the selection of anypolicy other than PSwill result in the loss of the election. Clearly this is more likely the closer h is to 0 and the greater the number ofwell-informed voters that are zealots so that they experience a fall in expressive benefits so sharp as to become negative with anysmall deviation from PS. A crucial contributor towards allowing Eq. (5) to hold is thatω is close to 1/2. This means that only a smallnumber of well-informed voters are required to be zealots.

The populist politician will always choose to set policy at PS as it provides his maximum payoff of E+βE. The good politicianwill choose P* and lose the election if the following condition holds

E + WS + β E + W�� bE + W� + β πW� + 1−πð ÞWS

� �;

duces to

βEb W�−WS� �

1−β + βπð Þ: ð6Þ

3.5.2. Pooling equilibrium

Given that Eq. (3) is assumed to hold forω−1= 2

ωof the well-informed voters, it automatically follows that ri(PS)N0, thus

ensuring that a majority of voters will vote for the populist policy when they have not received an informative signal as to theincumbent's identity. The populist politician's payoff is once againmaximised and the good politicianwill choose to pool if the signin Eq. (6) is reversed. Therefore, if the future is relatively important, that there is a low probability of being replaced by a goodpolitician and the gap between good and populist social welfare is not too large, then populism will be the preferred strategy.

e that h is treated as exogenous. It could be argued that it should be endogenous as it is determined by the voting strategy of all the voters, but conductinganalysis would be a major distraction from the core of the paper. The reality of any reasonably sized election is that the probability of being decisive willmely close to zero and in keeping with the behavioural nature of this paper well-informed voters are depicted as making some common judgement as toe of that probability.


Significantly, if it were also to be assumed that a good politician has no interest in ego rents, then he or she would produce goodpolicy.

3.5.3. Provision of educationRather than choose good or populist policy, a policy better than the populist one can be implemented but the superiority of

changes from populist policy have to be explained to the section of the electorate that is not well-informed, namely the (1−ω)that are rationally irrational. Due to information costs, policy may not reach the welfare maximising level. The policy is labelled PI

generating voter welfare WI.Information costs are depicted as the sum of two components. The first component I(P) is the cost involved in explaining why

changes in policy towards P* are for the benefit of rationally irrational voters. It is assumed that I′(P)N0 and I″(P)N0. This reflectsthe idea that it becomes increasingly costly to educate individuals to understand that a policy is bad, the closer that policy is to theoptimum. This also means that the case for supporting P* can only be made if there is a sufficient investment in explaining itsmerits. Otherwise, the good politician will be tempted to make a low-cost argument for P*. If he or she were to attempt this, it isassumed that the investment would be completely wasted and those rationally irrational voters that receive the message wouldremain convinced that PS is the superior policy.

I(P) is the cost of constructing the message, the second component is the cost of spreading the message to secure majoritysupport. It is assumed that a government information campaign is less costly, the fewer citizens are required to receive themessage. Ideally the government would like to target the message so that it is only received by those who are not already well-informed. The problem is that the government may be unable to identify the well-informed group, so some of its message will bewasted as it is received by well-informed voters. This increases the number of citizens that will need to be contacted. Onceexpressive concerns are included in the analysis, the cost of spreading the message further increases. As we have seen, formovements from PS somewell-informed voters will vote for the challenger because they have experienced an expressive loss thatoutweighs the instrumental benefit of voting for the good incumbent. Given that these voters will vote for the challenger, thenumber of citizens that need to be contacted further increases. Finally, as rationally irrational voters are contacted and areconvinced that P≠PS is superior to PS, they will also discover whether they experience an expressive loss from being pulled awayfrom PS. For some the expressive loss will be sufficiently large that they will vote for the challenger. This, in turn, increases thenumber of citizens required to receive the message and create a majority.

The preceding discussion can be formalised by labelling the cost of spreading the message d such that

14 CapthroughThis bri

d =1= 2− 1−λ Pð Þð Þω1−ωð Þ 1−φ Pð Þð Þ

� �t ð7Þ

Clearly the smaller isω, the greater the number of uninformed voters that will need to be persuaded regarding improvement inpolicy. The term λ(P) captures the well-informed voters with sufficiently strong expressive preferences that they would insteadvote for the challenger. Likewise, φ(P) captures rationally irrational voters who receive the message but find that they have asufficiently strong expressive attachment such that they would vote for the challenger despite accepting the fact thatW(P≠PS)NW(PS). The term t is a general term that captures modes of information provision such as media outlets, the education system, andalso importantly willingness to listen to the message.

The cost function is therefore C=c(I(P), d(ω, P, t)) and the good politician will invest in information until

∂W∂P =

∂c∂I

∂I∂P +

∂c∂d

∂d∂P :

The good politician will have no incentive to invest in information if

∂W∂PS b

∂c∂I

∂I∂PS +

∂c∂d

∂d∂PS :

Investment in informationwill not be worthwhile if themarginal benefit is less than themarginal cost at PS. This may happen ifPS is close to P*, which reduces the marginal benefit and increases education costs I′(P). This condition could also be driven by lowω, relatively high values of λ′(P) and φ′(P) and a high level of t. Each (or a combination) of these factors may cause the cost ofinformation provision to be prohibitive.14

Through education, the good politician may be able to do better than simply choosing the populist policy, if that were thepreferred approach without the option of education. Through provision of information, he or she may be able to move to higher

lan (2007) argues that the rationally irrational will actively avoid hearing messages that conflict with their beliefs. Formally this could be captureda prohibitively high level of t. The analysis presented here allows for a more optimistic perspective regarding the absorption of information provision.

ngs it more in line with the analysis of Romer (2003).


levels of voter welfare until marginal benefit equals themarginal cost of information at PI. An informational strategy is preferred toa populist strategy if

which

which

15 Thewelfare16 Pop

E + WI−C Ið Þ + β E + W�� N E + WS + β E + W��

:

This reduces to

WI−WSN C Ið Þ ð8Þ

3.6. Welfare

Proposition 1. Populism selected in period 1 by both types of political actor is unambiguously inferior from an ex ante voter welfareperspective compared to the populist choosing populist policy and the good politician choosing good policy . The welfare implication ofan informational strategy is ex ante ambiguous.

Here thewelfare effect of good politicians choosing a good policy versus choosing a populist policy in period 1 is compared

πW� + 1−πð ÞWS + β π2W� + π 1−πð ÞWS + 1−πð ÞWS� �

N WS + β πW� + 1−πð ÞWS� �

yields

W�N WS ð9Þ

The improvement in selection of politicians in period 2 is insufficient to compensate for the distorted discipline shown inperiod 1.15

Now the welfare effect of good politicians choosing to invest in information provision versus choosing good policy is compared.This will only be done where the improvement in voter welfare outweighs the cost of persuading the public WI−WS−C(I)N0.Welfare from choosing good policy will be higher if

πW� + 1−πð ÞWS + β π2W� + π 1−πð ÞWS

+ 1−πð ÞWS

!N πWI + 1−πð ÞWS + β πW�

+ 1−πð ÞWS

� �

yields

W�−WI� �

N β W�−WS� �

1−πð Þ ð10Þ

Now the welfare effect is ambiguous. If information costs are very low then WI will be close to W* in which case welfare isimproved through information provision. In this case, the improved selection effect in period 2 does compensate for the distorteddiscipline in period 1.

Proposition 2. A necessary condition for populism to be implemented in period 1 is that Eq.(6)fails to hold. This implies that either thefuture is relatively important and/or PSis close to P* and/or ego rents are relatively high and/or π is relatively low. Populism will bepreferred to an informational strategy if WI−WSbC(I). This implies that either PSis relatively close to P* and/or C(I) is relatively high.

Many issues are raised here. A first point that might be made is that populism as the pooling strategy in period 1 (although asource of inefficiency)may not be amajor cause for concern. This is the case where PS is relatively close to P*. Muchmoreworryingis where P* and PSare far apart and populism is still the strategy selected by a good politician. This can be driven by large ego rents,a lack of good politicians and costly information provision driven by a dearth of well-informed voters, the existence of a largenumber of strongly expressive voters and weak media outlets. We might also comment on how inefficient policies may differ interms of the emotional attachment they generate. For example, we may expect the myopia associated with supporting largebudget deficits to be more amenable to the provision of information than policies such as protectionism or overinvestment indefence as these latter policies are often inextricably linked to nationalist sentiment. The latter set of policies would more likelygenerate high levels of λ(P) and φ(P) which may imply that any amount of information provision may be prohibitively costly.

In a world with uninformed voters it is natural to askwhether the ability to inform rationally irrational voters is a simplemeansby which welfare can be increased. Paradoxically, the choice to provide information by good politicians may also be a source ofinefficiency.16 Whilst information provision (and the superior policy associated with it) is certainly better than populism from the

expressive payoff is removed from the analysis of voter welfare. This captures the idea that expressive payoffs may be trivial when measured in overall, although they may exercise a major influence on the decision making of voters.ulists have no incentive to provide information and as a result will always implement the inferior populist policy.


perspective of voter welfare, it might be better still if good politicians did not try to provide information and instead implementoptimal policy and accept the electoral defeat that will follow. This is true if Eq. (10) holds. However, a good politician will notchoose electoral suicide if he or she prefers information provision to the provision of optimal policy. This is shown by

which

otherw

wherepopulwouldpoolin

17 It sinformacost of18 It iswould r

E + WI−C Ið Þ + β E + W�� N E + W� + β πW� + 1−πð ÞWS

� �simplifies to

βE + β W�−WS� �

1−πð Þ−C Ið Þ N W�−WI ð11Þ

Note that this result is driven by the existence of ego rents because if Eq. (10) holds then Eq. (11) would certainly not hold ifE=0. To that extent the ability to provide information may actually have an adverse effect on voter welfare.17

4. Including ‘bad’ politicians

In this section the assumption that κ=0 is relaxed so that bad politicians may exist. A bad politician needs to weigh up thebenefit to him or her of extracting rents in period 1 versus providing vote-winning policy in period 1 and delaying the extraction ofrents until period 2. He or she will choose to extract rents today and accept electoral defeat at the end of period 1 if

T N β E + Tð Þ ð12Þ

iseheor shewouldprefer toprovideelectionwinningpolicy inperiod1 andextract rents inperiod2. Each is considered in turn.

4.1. The bad politician prefers to extract rents in period 1 and lose the election

That Eq. (12) holds is assumed to be well known by well-informed voters. The expressive component of the well-informedvoter's utility function is now expanded to incorporate the morality (or otherwise) of the incumbent politician. It is done byexpanding r(P) to include a further termm such thatm={B, NB} where B stands for bad and NB stands for not bad. For a given P, r(P, NB) is obviously greater than r(P, B). Where well-informed voters cannot distinguish whether the incumbent is bad or not theexpressive payoff remains as it was in Section 3. Including morality as an expressive benefit brings the analysis of expressivebehaviour in this paper to be more in line with the approach taken by Brennan and Hamlin (1999). To simplify matters the welfareassociated with a bad incumbent who prefers to extract rents is set at 0. With this we can see that the trade-off shown in Eq. (3) isnow more easily fulfilled. Eqs. (1), (2) and (3) are re-written as follows

ri P S;NB

� �+ h WS

� �ð13Þ

h πbW� + 1−πb−κ� �

WS� �

ð14Þ

ri P S;NB

� �N h πb W�−WS

� �−κWS

� �ð15Þ

π is adjusted to become πb to signify that bad politicians are now included such that π≥πb and 1−π≥1−πb−k. So where aist incumbent would win when the good politician separated from them when there were no bad politicians, the populistcertainly win when rent-extracting bad politicians are included in the analysis. Clearly, populists and good politiciansg is also more easily fulfilled, since the expressive value of the incumbent increases and there is now a κ probability that thenger is bad. In this case, unanimous support is guaranteed.
challe
The focus now turns to whether, given the potential existence of bad politicians, it is possible for a good politician to implementP* in period 1 and win the election? First note that the (1−ω) rationally irrational voters will vote for the challenger as the good

policy fails to meet their benchmark.18 To win the election implementing good policy, the good politician will need to win1= 2ω

of

the informed voters. The payoff to voting for the incumbent is given by

ri P�;NB

� �+ h W�� ð16Þ

e payoff to voting for the challenger is given by Eq. (14). Although Eq. (16) may be reduced in value compared to Eq. (4)
and thbecause P* is even further from PS, the effect of NBmay be to increase the value of Eq. (16) compared to Eq. (4). It is ambiguous as
hould be noted that the existence of ego rents for good politicians might also maximise voter welfare. If the sign in Eq. (10) is reversed so that thetion strategy maximises ex ante voter welfare, the existence of ego rents help to ensure that Eq. (11) holds given that the good politician will incur theinformation provision.thus assumed that the rationally irrational voters are unaware that provision of populist and good policy also entails a signal of moral integrity, as thisequire them to be well enough informed to understand that a bad politician would extract rents in period 1.


to whether Eq. (16) is greater in value than Eq.(4) but is unambiguous that Eq. (14) is lower in value than Eq. (2). The sign in Eq.(5) will be reversed if

and P*

voter

which

19 Not−WS) w20 Lupbad pol

r P�;NB

� �N h 1−πb

� �WS−W��

−κWS� �

ð17Þ

would be an election winning strategy if Eq. (17) holds for at least1 = 2ω

of the informed voters. This is more likely the

r is πb, the larger is κ and significantly the higher is the expressive value of the incumbent demonstrating moral integrity.
smalleThe satisfaction of Eq. (17) may paradoxically imply higher voter welfare than when there were no bad politicians as analysed
in Section 3. Assume that the good politician would pursue an informational strategy to win the election when there are no badpoliticians, but now he or she can pursue a good policy andwin the election when there are bad politicians. Voter welfare will nowequal

πbW� + 1−πb−κ� �

WS + βπbW� + 1−πb−κ

� �WS

+ κπbW� + κ 1−πb−κ� �

WS

0@

1A ð18Þ

welfare will now be ex ante higher if

πbW� + 1−πb−κ� �

WS + βπbW� + 1−πb−κ

� �WS

+ κπbW� + κ 1−πb−κ� �

WS

0@

1A N πWI + 1−πð ÞWS + β πW�

+ 1−πð ÞWS

� �

reduces to

W�−WS� �

πb 1 + β 1 + κð Þð Þ−βπ� �

N π WI−WS� �

+ κWS 1 + βκð Þ ð19Þ

Eq. (19) is more likely to hold the larger πb, the greater is (W*−WS) and the closer WI is toWS.19 This creates a variation on thedebate as towhether theremight actually be somebenefit arising from the existenceof corruption, such as the familiar contention that‘a little bit of corruption greases the wheels’ (see Aidt (2003a) for a discussion). In this paper, ‘a little bit of potential corruptionmayimprove voter welfare.’Well-informed voterswho are not expressively drawn to the policy P* are, however, expressively drawn to themoral integrity of the good incumbent. In addition, they know that voting for the challenger brings a hκ risk of bad policy in period 2 sothere is greater instrumental benefit in voting for a good incumbent. This potentially frees the good incumbent from having toimplement policies other than the optimal one to be re-elected and by implementing good policy can appeal directly to thewell-informed voters both expressively and instrumentally.20 The results of this section are gathered in the following proposition.

Proposition 3. In a world with κ bad politicians where Eq.(12)holds such thatT N βE1−β, a good incumbent who would otherwise lose an

election pursuing a good policy in a world with no bad politicians because Eq.(5)holds, would now win the election if Eq.(17)holds.Welfare will be higher in a world with bad politicians where the good politician pursues a good policy compared to one without badpoliticians where the good politician implements information policy if Eq.(19)holds.

4.2. The bad politician prefers to win the election and extract rents in period 2

In the following it is assumed that Eqs. (5) and (17) hold, which means that a good politician who would lose pursuing goodpolicy in a world without bad politicians, would win the election pursuing good policy in a world with bad politicians. The sign inEq. (12) is now reversed which means that a bad politician would prefer to delay rent extraction until period 2. First, it isdemonstrated that there cannot be a pooling equilibrium with all three types choosing PS. Given the assumption that Eq. (17)holds thenwe know that it is the case that the good politician canwin the election by deviating and choosing P* which is obviouslypreferable to winning the election by choosing PS. It can also be demonstrated that there is no pooling equilibriumwhere all threetypes choose P*. In this case the payoff for voting for the incumbent would be

ri P�� + h πbW� + 1−πb−κ

� �WS

� �ð20Þ

. (14) for the challenger, where it is noted that the moral expressive value in voting for the incumbent is absent because a
and Eqbad politician would pool with the other two types. The challenger will win if
0 N r P�� ð21Þ

We know that Eq. (21) holds because Eq. (5) is assumed to hold.

e that if the good politician had selected populism in a world without bad politicians then the right hand side of Eq. (19) would be lower in value as π(WI

ould be removed.ia and Menning (2009) ask ‘When can politicians scare citizens into supporting bad policies?’ In contrast, this paper asks the question ‘When can fear ofiticians scare citizens into supporting good policies?’


So a pooling equilibrium for all three types is not possible given the conditions previously derived in the model. Is anequilibrium possible where the populist and bad politicians pool on PS and the good politician chooses P*? The payoff to voting forthe incumbent (which could be populist or bad) is

althouso wh

and as

21 Cleaguarantnot holinvestm

ri PS� �

+ h WS 1−πb−κ1−πb

! !ð22Þ

e payoff to voting for the challenger is Eq. (14). For PS to be an election winning strategy for the pooling populist and bad
and th
types, the following must hold forω−1= 2

ωof the well-informed voters

ri PS� �

N hπb W�−WS 1−πb−κ1−πb

! !ð23Þ

gh πb≤π, Eq. (23) is potentially more restrictive than Eq. (3) because of the presence of κ on the right-hand side of Eq. (23),ilst it is consistent with the model for the bad and populist politicians to pool and the good politician to separate, the case

Eq. (23) fails to hold for at leastω−1= 2

ωof the well-informed voters will also be considered.
where
If Eq. (23) does not hold, it immediately follows that an equilibrium where the bad politician does not pool with the populist

and chooses PS cannot exist because h WS 1−πb−κ1−πb

� �� would be deducted from Eq. (22) and the incumbent would be exposed

as corrupt. A bad politician will want to pool. Could he or she pool with a good politician andwin? The payoff in this case for votingfor the incumbent would be

r P�� + h

πbW�

πb + κ

!ð24Þ

usual for the challenger the payoff would be Eq. (14). The challenger would win if

r P�� bh 1−πb−κ� �

WS− πbW�

πb + κ

!: ð25Þ

Given that Eq. (5) tells us that there are at least 1− 1 = 2ω of well-informed voters with negative expressive utility for any

deviation from PS and that the deviation from PS in Eq. (25) is larger than in Eq. (5) then Eq. (25) holds automatically.So given Eqs. (17), (21) and (25) holding and Eq. (23) not holding (whichmeans that bad politicians cannot pool with either or

both of the good and populist politicians and win the election), a pure strategy Nash equilibrium will not exist. Potentially therecould be an equilibrium whereby the good and populist politicians would be willing to signal their integrity by incurring visiblecosts up to a level at which the bad politician would not be willing to do so. For this to be the case, the cost must be such that thebad politician will prefer to extract rents today

βE−C≤T 1−βð Þ ð26Þ

So an equilibrium may exist with the good and populist politicians incurring at least βE−T(1−β) of costs and winning theelection.21 The bad politician extracts rents in period 1 and loses the election. How might these costs be visibly incurred? Onepossibility, available from the analysis in this paper, is information provision. But now the motivation to provide information isvery different. Earlier when it was assumed that there were no bad politicians, the motivation to provide information only existedfor good politicians as a way to improve policy in period 1 andwhere it was clear that such a strategy would win the election. Nowthe purpose of providing information is as a signal from both good and populist politicians that they are not bad. For this reason,information need bear no relation to the truth about policy because the signal is aimed at the well-informed voters who alreadyunderstand that P* is best policy. The populist will engage in false information provision stating that PS is best policy because thatwill be popular with the rational irrational voters and still win the votes of the well-informed. In contrast, the good politician willbe free to argue for P* as best policy. The dishonesty displayed by the populist is paradoxically driven by the desire to prove that heor she is not corrupt.

rly this can only happen if the good and populist politicians are willing to incur these costs. Incurring the costs provides them with the benefit ofeed office because Eqs. (15) and (17) hold, but (unlike in Section 3 for the good politician) if the costs are not incurred it would not imply that they wouldd office because the absence of a pure-strategy equilibrium would imply some positive probability of holding office next period. A way of thinking aboutent in costs here is as a metaphor for effort. By being seen to be working hard good and populist politicians may be able to separate from bad politicians.


The two equilibria identified in this subsection can be summed up in the following proposition

Proposition 4. Suppose Eq.(12)fails to hold so that βE1−β≥T. An equilibrium exists where the bad and populist incumbents choose PSand

the good incumbent chooses P* if Eq.(23)holds. If Eq.(23)fails to hold, an equilibrium may exist where the good and populist politiciansincur costs C so that βE

1−β≥T≥ βE−C1−β and the bad incumbent chooses to extract rents.

Welfare in the first equilibrium identified in Proposition 4 is as follows

πbW� + 1−πb� �

WS + β πbW� + 1−πb−κ� �

WS� �

ð27Þ

The second potential equilibrium identified in Proposition 4 would imply the same welfare analysis as in Section 4.1 given byEq. (18). Inspection of Eqs. (18) and (27) tell us that there is no straightforward welfare ranking of equilibria in a world with badpoliticians. It thus follows that the conclusion at the end of Section 4.1 still holds in a world where bad politicians would ideallywish to delay rent extraction. A little bit of potential corruption may improve voter welfare.

5. Concluding comments

The paper makes three main contributions. First, it provides a rational choice analysis of populism as it is commonlyunderstood. Clearly, a richer depiction of the behaviour of voters exists here than in previous political agency models. This set-upallows a focus on populism as driven by uninformed bias (which could potentially be altered with information provision) andinformed bias with hard expressive preferences. This means that populism in this paper is very different (and arguably more trueto its nature) than the depiction of populism in the existing political agency literature. This analysis also echoes the point made byCaplan (2007) that a neglected, but fairly obvious argument, for the existence of democratic inefficiency exists. Explanationsusually focus on the strategic interaction of special interest politics and in ways that can become complicated because voters areassumed to be fully informed and instrumentally rational. Once this assumption is dropped, a simple explanation for the existenceof inefficiency is that voters might choose inefficient policies because they are emotionally drawn towards them and emotionalappeals can dominate in electoral settings for the rational reason that the likelihood of being decisive in determining the outcomeof most elections is very small.

The second and third contributions aremore optimistic for voter welfare. The paper locates a potential role for government as apersuasive provider of information regarding the quality of policy. Finally, by recognising that expressive choice may focus on therevealed integrity of politicians (rather than reinforcing rational irrationality) it follows that a little bit of corruption may be goodfor voter welfare. If expressive voters take the integrity of politicians into account when deciding how to vote this may allow goodpoliticians to implement good policy despite the fact that voters may not find good policy expressively appealing.

Acknowledgements

I would like to thank Frank Bohn, Thomas Braendle, Geoffrey Brennan, Bryan Caplan, Alan Hamlin, Henrik Jordahl, Kai Konrad,Clare Leaver, Jacob Montgomery, David Myatt, Jens Prufer, Hein Roelfsema, Dana Sisak, three referees, the editors of EJPE andparticipants at the SIRE Meeting in Stirling (2008), ISNIE in Toronto (2008), the Public Choice Meeting in Las Vegas (2009), theSilvaplana Workshop in Political Economy (2009) and seminars at the Universities of Nijmegen, Portsmouth, Strathclyde, Surreyand Wissenschaftszentrum für Sozialforschung Berlin for their helpful comments.

References

Acemoglu, D., Robinson, J.A., 2001. Inefficient redistribution. American Political Science Review 95, 649–661.Aidt, T.S., 2003a. Economic analysis of corruption: a survey. The Economic Journal 113, F632–F652.Aidt, T.S., 2003b. Redistribution and deadweight cost: the role of political competition. European Journal of Politcal Economy 19, 205–226.Alesina, A., Tabellini, G., 1990. A positive theory of fiscal deficits and government debt. Review of Economic Studies 57, 403–414.Barro, R., 1973. The control of politicians: an economic model. Public Choice 14, 19–42.Becker, G.S., 1983. A theory of competition among pressure groups. Quarterly Journal of Economics 98, 371–400.Besley, T., 2006. Principled Agents: The Political Economy of Good Government. Oxford University Press, Oxford.Brennan, G., 2009. Politics, selection and the public interest: Besley's benevolent despot. Review of Austrian Economics 22, 131–143.Brennan, G., Buchanan, J.M., 1980. The Power to Tax: Analytical Foundations of the Fiscal Constitution. Cambridge University Press, Cambridge U.K.Brennan, G., Hamlin, A., 1998. Expressive voting and electoral equilibrium. Public Choice 95, 149–175.Brennan, G., Hamlin, A., 1999. On political representation. British Journal of Political Science 29, 109–127.Brennan, G., Lomasky, L., 1993. Democracy and Decision. Cambridge University Press, Cambridge U.K.Canes-Wrone, B., Herron,M.C., Shotts, K.W., 2001. Leadership and pandering: a theory of executive policymaking. American Journal of Political Science 45, 532–550.Canovan, M., 1999. Trust the people! Populism and the two faces of democracy. Political Studies 47, 2–16.Caplan, B., 2002. Systematically biased beliefs about economics: robust evidence of judgemental anomalies from the survey of Americans and economists on the

economy. The Economic Journal 112, 433–458.Caplan, B., 2007. The Myth of the Rational Voter: Why Democracies Choose Bad Policies. Princeton University Press, Princeton NJ.Caplan, B., 2009. Irrational principals. Review of Austrian Economics 22, 159–167.Coate, S., Morris, S., 1995. On the form of transfers to special interests. Journal of Political Economy 103, 1210–1235.Downs, A., 1957. An Economic Theory of Democracy. Harper and Row, New York.Fearon, J.D., 1995. Rationalist explanations for war. International Organization 49, 379–414.Feddersen, T.J., Pesendorfer, W., 1996. The swing voter's curse. The American Economic Review 96, 408–424.


Ferejohn, J., 1986. Incumbent performance and electoral control. Public Choice 50, 5–25.Hamlin, A., Jennings, C., 2011. Expressive political behaviour: foundations, scope and implications. British Journal of Political Science doi:10.1017/S0007123411000020.Hillman, A.L., 2010. Expressive behaviour in economics and politics. European Journal of Political Economy 26, 403–418.Kliemt, H., 1986. The veil of insignificance. European Journal of Political Economy 2 (3), 333–344.Liu, Q., 2007. How to improve government performance? European Journal of Political Economy 23, 1198–1206.Lupia, A., Menning, J.O., 2009. When can politicians scare citizens into supporting bad policies? American Journal of Political Science 53, 90–106.Maskin, E., Tirole, J., 2004. The politician and the judge: accountability in government. The American Economic Review 94, 1034–1054.Mayda, A.M., Rodrik, D., 2003. Why are some people (and countries) more protectionist than others? European Economic Review 49, 1393–1430.Mejía, D., Posada, C.E., 2007. Populist policies in transition to democracy. European Journal of Political Economy 23, 932–953.Morris, S., 2001. Political correctness. Journal of Political Economy 109, 231–267.Persson, T., Svensson, L., 1989. Why a stubborn conservative would run a deficit: policy with time-inconsistent preferences. Quarterly Journal of Economics 104, 325–345.Robinson, J.A., 1998. Theories of ‘bad policy’. Journal of Policy Reform 3, 1–46.Rodrik, D., 1996. Understanding economic policy reform. Journal of Economic Literature 34, 9–41.Romer, D., 2003. Misconceptions and political outcomes. The Economic Journal 113, 1–20.Sobbrio, F., Navarra, P., 2010. Electoral participation and communicative voting in Europe. European Journal of Political Economy 26, 185–207.Tullock, G., 1983. Economics of Income Redistribution. Kluwer-Nijhoff, Boston.Westen, D., 2007. The Political Brain: The Role of Emotion in Deciding the Fate of the Nation. Public Affairs, New York.Wittman, D., 1989. Why democracies produce efficient results. Journal of Political Economy 97, 1395–1424.Wittman, D., 1995. The Myth of Democratic Failure: Why Political Institutions Are Efficient. University of Chicago Press, Chicago IL.

the good, the bad and the populist: a model of political agency with emotional voters

Documents