The Neuroeconomics of Voting: Neural Evidence of Different

The Neuroeconomics of Voting: Neural Evidence of DifferentSources of Utility in Voting

Ivo BischoffUniversity of Kassel

Carolin Neuhaus and Peter TrautnerUniversity of Bonn

Bernd WeberUniversity Hospital Bonn and University of Bonn

Which motives drive the decision of a voter to approve or reject a policy proposal? ThePublic Choice literature distinguishes between instrumental and expressive votingmotives. We investigated the importance of these motives by analyzing the patterns ofneural activity in different voting situations. We conducted a functional magneticresonance imaging experiment that investigates neural activation at the moment ofvoting and used the altruism scale proposed by Tankersley, Stowe, and Huettell (2007)to differentiate between altruists and nonaltruists. Nonaltruists showed neural activationpatterns that were consistent with expressive voting motives. Among nonaltruists, wealso observed activation patterns that point at egoistic instrumental motives. Bothresults are in line with the corresponding Public Choice literature. On the other hand,we found no evidence for expressive voting motives among altruists. Their neuralactivation pattern was generally much less conclusive with respect to the underlyingmotives.

Keywords: voting behavior, expressive voting, instrumental voting, political decision making,charitable donation

Traditionally, the theory of voter behaviorassumes that voting decisions are driven bypreferences over policy proposals. The voteserves as an instrument to change the voter’sliving environment in the future (e.g., Downs,1957). This so-called instrumental voting canbe thought of as an investment under uncer-tainty: The voter bears the certain and imme-diate costs of voting and expects future utilitygains from more favorable policies. He votesfor his favorite policy proposal to increase theprobability that it becomes accepted and pur-sued (e.g., Downs, 1957; Fiorina, 1976). The

essential shortcoming of the theory of instru-mental voting is that it cannot explain thehigh voter turnout in real-life elections wherethe individual vote has a negligible chance ofchanging election outcomes (e.g., Downs,1957; Tyran, 2004). To dissolve this “paradoxof voting,” it is necessary to assume somedirect utility from the act of voting that doesnot relate to the policy outcomes expected inthe future. According to the theory of expres-sive voting, a voter goes to the booth becausethe act of expressing his approval to a certaincandidate or policy proposal yields immediateutility. Thus, voting is an act of consumption,much like cheering at a football game (e.g.,Brennan & Hamlin, 1998; Fiorina, 1976). Aslong as an expressive voter does not expect tobe pivotal, he votes for the proposal or can-didate for which approving yields the highestimmediate expressive utility. In those caseswhere he expects to be pivotal the instrumen-tal utility outweighs the expressive utility.Given that the chance of being pivotal inreal-world elections is negligible, expressive

Ivo Bischoff, Department of Economics, University ofKassel; Carolin Neuhaus and Peter Trautner, Department ofNeuroCognition, University of Bonn; Bernd Weber, Depart-ment of Epileptology, University Hospital Bonn, and Centerfor Economics and Neuroscience, University of Bonn.

Correspondence concerning this article should be ad-dressed to Bernd Weber, Department of Epileptology, Uni-versity Hospital Bonn, and Center for Economics and Neu-roscience, University of Bonn, Nachtigallenweg 86, 53127Bonn. E-mail: [email protected]

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Journal of Neuroscience, Psychology, and Economics © 2013 American Psychological Association2013, Vol. 6, No. 4, 215–235 1937-321X/13/$12.00 DOI: 10.1037/npe0000016

215

motives may—if existent in a large share ofvoters— drive the collective policy choices(Tullock, 1971). In this case, voting does nolonger aggregate policy preferences. Instead,it promotes policies and candidates that makevoters feel good when voting for them andimpedes policies where the concomitant feel-ing is bad. Whenever expressive and instru-mental motives collide, it is uncertain whetherthe policy chosen promotes the welfare even ofthe majority of voters who approved.1 Moreover,it opens up another channel by which electionpolls can influence voting decisions: Policieswhere expressive and instrumental motives col-lide can be promoted or impeded by makingexpected outcomes seem less or more close.

So far, behavioral experiments provideonly limited empirical support for a promi-nent role of expressive motives (e.g., Carter& Guerette, 1992; Feddersen, Gailmard, &Sandroni, 2009; Fischer, 1996; Kamenica &Egan, 2011; Shayo & Harel, 2012; Tyran,2004).2 In this article, we argue that the ex-isting evidence from behavioral experimentsdoes not necessarily mean that expressivemotives are not as strong as predicted. Thereason is that behavioral data—that is, theobserved decision to vote YES or NO—leavesus with a measure that may be too crude todifferentiate different voting motives. The ob-served voting decision supports the conclu-sion that—among two different conflictingvoting motives—a particular motive domi-nates. However, it does not tell us whether theother motive is relevant at all, nor does itinform us about the relative strengths of bothmotives. To gain deeper insight into the roleof different voting motives, we apply neuro-scientific methods. These provide us with adetailed account of brain activities involvedin the voting process. By providing us withmeasures for neural correlates of the utilityinvolved in the voting decisions, they allowfor additional tests for the presence of expres-sive motives in voting. Over the last years, theapplication of neuroscientific methods in eco-nomics enhanced the knowledge about eco-nomic behavior and decision making. Up tonow, the political sciences have paid littleattention to these new methods (for someexceptions see Amodio, Jost, Master, & Yee,2007; Chorvat, 2007; Farmer, 2007; Har-baugh, Mayr, & Burghart, 2007; Tingley,

2006; Westen, Blagov, Harenski, Kilts, & Ha-mann, 2007). To our knowledge, we providethe first study using functional magnetic res-onance imaging (fMRI) to investigate votingbehavior.3 We find neural activation patternsthat are consistent with the theory of expres-sive voting among subjects classified as non-altruists. For those classified as altruists, wedo not find neural correlates of an expressiveutility from voting. Moreover, our results areinconclusive with respect to the ultimate fac-tors that determine their instrumental utilitywhen voting.

The article proceeds as follows. Aftersketching the incentives involved in expres-sive and instrumental voting, the second sec-tion derives the central hypotheses to betested. The third section describes the exper-imental set-up of our study. The results arepresented in the fourth section and discussedin the fifth section; the sixth concludes.

1 Similarly, Feddersen and Pesendorfer (1996) show thatstrategic voting may undermine the majority vote’s abilityto aggregate private information as proposed in Condorcet’sJury Theorem. The main difference between this strand ofliterature and the one followed here is the purpose of col-lective decisions: In the case these authors have in mind,majority voting is a means of aggregating private informa-tion on an uncertain state of nature. In our article, the aim isto aggregate individual policy preferences. In this context,the ethical voter central to the analysis of Feddersen andPesendorfer votes instrumentally because his aim is im-prove policy outcomes (see also Battaglini, Morton, &Palfrey, 2010). Given only two choices, strategic voting isirrelevant.

2 Most of these studies deliberately create treatments withdifferent degree of pivotality changing the decision makingenvironment (e.g., electorate size; for a review, see Shayo &Harel, 2012). Tyran (2004) is among the few who do notchange the voting environment but elicit voters’ expecta-tions concerning the approval rate and infer the degree towhich voters expect to be pivotal. In this article, we Tyran’sapproach.

3 The study by Rule et al. (2010) on the choice of politicalcandidates in Japan and the U.S. may be regarded as anexception. However, it concentrates on judgments aboutpersonality and culture as drivers of choosing candidates forpolitical offices. In addition, choices are hypothetical andthe interaction of voters in the collective decision makingframework is not captured by the procedures. Thus, weclaim to provide the first fMRI-study on voters’ interactionin voting decisions. We also focus on the choice of policiesrather than candidates.

216 BISCHOFF, NEUHAUS, TRAUTNER, AND WEBER

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Expected Voting Decisions and TheirNeural Correlates

Behavioral Predictions and EconomicExperiments

The set-up of most experimental studies onexpressive voting resembles Tullock’s thoughtexperiment (e.g., Tullock, 1971). It is based ona ballot in which a large number of individualsvote on the following proposal: Tax everyoneand donate the revenues to charity. Throughoutthis article, voting refers to this type of redistri-butional policies; the proposal is to take moneyfrom the broad public and use it to help a clearlydefined minority. This minority consists of in-dividuals that—by a broad consensus in soci-ety—are considered to be in need. To identifythe incentives that voters face when deciding onthis proposal, consider the individual voter i.Let yi denote the utility he witnesses if theproposal is rejected and he can keep the moneyfor private use. Let zi represent the utility fromaltruistic motives that subject i witnesses if theproposal is accepted and the minority in needwitnesses an increase in consumption. The in-strumental utility that subject i witnesses if theproposal is rejected is given by xi � yi �zi.Individuals who draw a higher utility from themoney being donated are hereafter called altru-ists. For them zi � yi and thus, xi � 0 while zi �yi and xi � 0 for nonaltruists who witness ahigher utility when the money is available forprivate use.

Let εi denote voter i’s expressive utility fromvoting for the proposal. By voting in favor ofdonating the money, voter i complies with thepredominant ethical standards in society andmay think of himself as a generous person. Bothaspects arguably yield utility (Mueller, 1986;Hillman, 2010).4 Thus, we follow Tullock(1971) and Tyran (2004) in assuming that theimmediate expressive utility from voting in fa-vor of donating the money is higher than fromrejecting this proposal. We also assume that theinstrumental utility dominates the expressiveutility, (i.e., 0 � εI� |xi|). The expected utilityfrom voting YES is then given by:

Uiˆ (YES) � ri · �xi � εi� � pi · εi � si · εi

� εi � ri · xi

Here, pi denotes the probability at whichvoter i expects to be pivotal. The case wherepi � 0 is hereafter called close-decision case. Inmost elections and ballots the approval rateamong other voters is either sufficiently high toaccept the proposal even if he votes against it ortoo low to accept the proposal even if he votesin favor of it. Hereafter, we denote the first casethe sure-donation case and the second one thesure-private-money case. In both cases, pi �0. The probability at which voter i expects thesure-donation (sure-private-money) case is de-noted ri (si).

The expected utility from voting NO is givenby:

Uiˆ (NO) � ri · xi � pi · xi

The difference between these two utilities is

called the utility differential UDi:

UDi � Uiˆ (YES) � Ui

ˆ (NO) � εi � pi · xi

A risk-neutral voter will vote YES if the ex-pected utility from voting YES is higher than theexpected utility from voting NO (i.e., if the utilitydifferential is positive). He will vote NO if theutility differential is negative. Depending on theinstrumental and expressive utility (xi resp. εi), wecan differentiate four types of voters: (1) altruistswho have expressive motives (zi � yi N xi � 0,εI � 0), (2) altruists without expressive motives(zi � yi N xi � 0, εI � 0), (3) nonaltruists whohave expressive motives (zi � yiN xi � 0, εI �0), and (4) nonaltruists without expressive motives(zi � yiN xi � 0, εI � 0).5 Altruistic voters voteYES because the utility differential is positiveregardless of whether he expects to be pivotal ornot. Whether or not they have expressive voting

4 The possibility that subjects have antisocial preferencescannot be ruled out ex ante (e.g., Coffman, 2010). Tominimize this possibility, we choose a set-up in whichsubjects donate to charitable organizations that have beenchosen for their high acceptance among subjects in pretests.The answers of our subjects in the postexperimental ques-tionnaire support the notion that acceptance for the organi-zations is high among our subjects. None of our subjectsgenerally rejects the importance of the organizations’ workor states that they may be corrupt.

5 Nonaltruists are not defined by the absence of other-regarding motives but by the dominance of selfish motivesover altruistic motives in the case of pivotality.

217THE NEUROECONOMICS OF VOTING

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

motives is irrelevant for their decision as well.Similarly, a nonaltruist who is not motivated byexpressive motives votes NO regardless ofwhether he expects to be pivotal or not. The rea-

son is that his utility differential (UDi � pixi) isalways negative.

On the other hand, the expected probability piof being pivotal is essential for a nonaltruistwho is motivated by expressive motives be-cause he faces a trade-off between expressiveand instrumental motives. By voting YES,he has an immediate and certain gain in expres-sive utility εi. However, this utility comes at thecosts of giving away the chance to tip the scalesagainst donation by voting NO. In the close-decision case where pi is large, the instrumentalmotives dominate his decision (i.e., |pixi| � εi)and he will vote NO. In the sure-donation casewhere ri � 1 as well as in the sure-private-money case where si � 1, the individual cannotexpect to tip the scales (i.e., pi � 0). Therefore,the expressive motive becomes dominant, (i.e.,

|pixi| � εi) and makes the nonaltruistic voter

with expressive motives to vote YES—againsthis instrumental utility.

Expressed in nontechnical terms, the theoryof expressive voting argues that nonaltruisticvoters get the good feeling of approving to aproposal to donate money. In most cases, thisgood feeling comes at zero material costs be-cause the proposal is rejected or approved any-way. As the chance of being pivotal is negligi-ble in virtually all voting decisions, nonaltruistsgenerally behave like altruists even though theyreally do not want to donate money. The truemotivation of the nonaltruists only shows upin those rare cases when they expect to bepivotal and thus, approving bears the dangerof material costs. In this case, they prefer themoney over the good feeling and vote againstdonation. Altruistic voters always vote YESregardless of whether they expect to be piv-otal or not and regardless of whether theyhave expressive motives or not. When itcomes to testing for expressive motives, wecannot use behavioral data from altruists. In-stead, all experimental tests for expressivemotives in voting have to rely on observingthe voting behavior of nonaltruistic subjectsin situations with and without pivotality.

Tyran (2004) performed an economic exper-iment to test for expressive motives in voting.Therein, every participant was given a voucherworth approximately $6. The proposal was todonate the endowments of all participants tocharity (instead of cashing the voucher in at theend of the experiment). Participants voted onthis proposal five times with five differentquora. A quorum defines the minimum approvalrate necessary for the proposal to be accepted.At the end of the experiment, one voting roundwas chosen at random and its decision wasexecuted. By letting candidates vote on thesame proposal using different quora, Tyran(2004) kept the instrumental utility xi and theexpressive utility of approving the proposal εiconstant but induced changes in the expectedprobability pi of being pivotal. He asked candi-date for their estimated approval rates amongfellow-voters for all five decisions. Based onthese estimates, it was possible to identifywhether voter i expected to take his vote in thesure-donation, sure-private-money or close-decision case. A substantial share of partici-pants either disapproved or approved for allquora even when the expected approval ratesuggested that pi � 0. Among those 40% whovoted YES for some quora and switched to NOfor others, only 25% (10% of all participants)showed a switching pattern consistent with thetheory of expressive voting.

On the other hand, Tyran (2004) found astrong positive correlation between subjects’ in-clination to approve and the expected approvalrates among fellow-subjects. He argued that thismay point at the existence of bandwagon mo-tives. However, as Bischoff and Egbert (2013)pointed out, the correlation observed by Tyran(2004) may also result from the so-called falseconsensus effect. The false-consensus effect de-scribes the empirical regularity that subjectsoverestimate the degree to which other subjectshave the same preferences and behavioral inten-tions as the subject himself (e.g., Bischoff &Egbert, 2013; Fields & Schuman, 1976).Bischoff and Egbert ran a series of experimentsusing a similar set-up as Tyran (2004) but pro-vided subjects with social information to influ-ence their expectations with respect to thebehavior of their fellow-subjects. They foundevidence for both bandwagon and expressivemotives. At the same time, they could not rule


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

out the possibility that subjects’ fall victim ofthe false-consensus effect.

The existing empirical evidence on expres-sive motives in voting is mixed (e.g., Bischoff& Egbert, 2013; Shayo & Harel, 2012; Tyran,2004). At the same time, it is not clear whatthe result tells us about the importance ofexpressive motives: Do expressive motivesexist only for a small fraction of subjects orare they relevant for most subjects yet domi-nated by instrumental motives? The essentialshortcoming of behavioral data are that wecan only infer the sign of the utility differen-

tial UDi, but we cannot compare its magni-

tude |UDi| for different situations. Neuroeco-nomics allows for such comparisons. Therefore,neuroscientific methods can help us gain furtherinsight into the process by which choices aregenerated.

Neuroscientific Methods and RelevantExperiments

Neuroeconomics seeks to explain both be-havior and its causes (Kable & Glimcher,2009). By providing measures for the inten-sity of the reward expected from a certaindecision, neuroscientific methods enable us to

test for differences in magnitude |UDi| forsituations where its sign is unchanged. More-over, they allow for a comparison of rewardlevels of different voting decisions in thesame situation. Therefore, neuroscientificmethods can help us to gain more insight intothe decision making process that leads theindividual to vote YES or NO.

In this article, we use fMRI in an experi-ment that is similar to the experiment byTyran (2004). fMRI is a neuroimagingmethod that allows for a measurement ofphysiological changes while a subject per-forms an experimental task. These physiolog-ical changes are correlated with neural activ-ity and can be linked to mental processes. Theunderlying idea is to compare the brain’s con-dition during the exercise of a specific taskwith its condition during a control task or asecond task. The resulting images show acti-vations of the brain areas affiliated with thespecific task and therefore provide informa-tion about differences in the underlying neu-

ral processes responsible for the overt act.fMRI permits discrimination of differentcausal processes even if they result in thesame observable behavior. On the otherhand, it reveals that the brain processes seem-ingly different stimuli in an analogous man-ner (e.g., Camerer, Loewenstein, & Prelec,2005; Izuma, Saito, & Sadato, 2010). Themain research focus in neuroeconomics is onthe neural mechanisms underlying decisionmaking—investigating the expected utilitymodel, decision making under risk and uncer-tainty, intertemporal choice and so forth (seefor an overview: Camerer et al., 2005; Loe-wenstein, Rick, & Cohen, 2008; Sanfey, Loe-wenstein, McClure, & Cohen, 2006). Over thepast years, numerous studies in this field in-creased our knowledge about the neural pro-cesses of decision making extensively. Giventhat voting decisions are based upon the sameprocesses, we can use this knowledge to ex-amine the neurobiology of instrumental andexpressive motives in voting. Although thebrain operates as a network, we also learned alot about the contribution of different brainareas to specific brain functions. We knowthat the brain does not code objective butsubjective reward values and seems to con-vert values of different stimuli (e.g., primaryand secondary reinforcers) to one “neural cur-rency” (Izuma et al., 2010; Kable & Glim-cher, 2009; Saxe & Haushofer, 2008). Thestriatum plays a major role in this encoding ofrewards. Consequently, this brain region is ofspecial interest for our present study. Thestriatum is a subcortical region located in theinterior regions of both cerebral hemispheresof the human brain. It can be divided into twosubregions called caudate nucleus and puta-men. Caudate and putamen are often relatedto as dorsal (upper) striatum (whereas thelower ventral striatum refers to the nucleusaccumbens); both are associated with encod-ing the subjective value of goods and actions.The striatum reveals reward responses refer-ring to forecasted or experienced value(Kable & Glimcher, 2009). Former workshowed that monetary gain but also monetaryloss is processed in the striatum (Delgado,2007; Seymour, Singer, & Dolan, 2007).Therefore, this region seems to be highly im-portant for the encoding of value representa-tion that guides the choice among different


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

options. Neuroeconomics commonly usesmoney as a reinforcer to investigate decision-making, but there is a growing body of re-search suggesting that not only monetary butalso social rewards are represented in thestriatum (Fliessbach et al., 2007; Izuma et al.,2010; Moll et al., 2006; Saxe & Haushofer,2008) and that different types of rewards aretranslated into a common neural currency(Kable & Glimcher, 2009). For instance,Izuma and colleagues (2010) had subjects de-cide whether to donate an amount of money tocharity or to keep the money for themselves.This decision was made in presence or ab-sence of observers. The presence of observersincreased donation rates and fMRI reveals thehighest striatal activations (1) when subjectsdonate in public and they expect high socialrewards and (2) when subjects keep themoney for themselves in absence of observersand thus, expect monetary gain without socialcost. Another study shows that striatal activ-ity is also highest for monetary gain and non-costly donations when anonymity is guaran-teed (Moll et al., 2006). The subjects in ourvoting experiment face a similar task—to do-nate money to charity or to keep it for privateuse. The essential difference between our ex-periment and those reported above is that oursubjects make their decision collectivelyrather than individually. Nevertheless, theabove literature suggests that the striatum isthe primary region of interest for our study.

Harbaugh et al. (2007) analyzed the neuralactivity when individuals are forced to donatecollectively. The subjects in one treatmentwere taxed to finance donations to charitywhile the subjects in the other treatment de-cided voluntarily and on an individual basis todonate an equivalent amount of money tocharity or to keep it for private use. In theirstudies, altruists showed reward-related stria-tal activity even when being forced to give themoney to charity. The activation in the ven-tral striatum was higher when charitable giv-ing is voluntary compared with mandatorytransfers. This supports the notion of warmglow in charitable giving (Andreoni, 1989).Warm glow describes the extra utility derivedfrom the act of giving money for charity. Justlike expressive motives in voting, charitablegiving is then seen as an act of consumptionthat is intended to increase the utility or re-

ward of the donor. The question of whetherothers benefit from the donation is not ofprimary relevance. Contrary to that, altruisticmotives see charitable giving as an investmentwhere the utility or reward results from the con-sequences of being taxed, which is underlined bya recent study by Kuss et al. (2013) who foundneural evidence for outcome orientation in dona-tions. Thus, Harbaugh et al. (2007) showed thatinvestment-related motives (i.e., instrumental mo-tives—altruistic or egoistic) and consumption-related motives (warm glow) are important incharitable giving. In our study, we analyze theimportance of these two types of motives whenindividuals vote on charity. With respect to vol-untary and forced donations analyzed by Har-baugh et al. (2007), voting lies in between thesetwo extremes: Each voter makes an individualdecision when approving or rejecting the proposalbut the final decision is largely determined by theother voters.

Hypotheses

Based on the theoretical considerations dis-cussed earlier in the article and the evidencefrom previous experiments, we can derive anumber of testable hypotheses to help us learnmore about the voting decision. Given thatthis is to some extent an explorative study, wetest for expressive motives as well as for thelargely undisputed instrumental motives. Thetests are based on comparisons of neural ac-tivation patterns for different constellations,depending on the subject’s expectation con-cerning their fellow voter’s behavior and onwhether they vote YES or NO. Most tests areperformed separately for altruists and nonal-truists. There are 12 different constellationsthat can be compared in tests. Each constel-lation is denoted by three-letter abbreviation(see Figure 1). The first letter states whetherthe subject is an altruist (A) or a nonaltruist(N). The second letter captures the subject’sexpectation with respect to the behavior of hisfellow-subjects. It differentiates betweensure-donation (D), sure-private-money (P),and close-decision case (C). The third letter inthe abbreviation states Y (N) when the subjectvoted YES (NO). For example, APY capturesa constellation in which an Altruist sees him-self in a situation in which he expects to keepthe money for Private use and votes YES.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Whenever we do not differentiate betweenaltruists and nonaltruists, the first symbol inthe abbreviation is a wildcard (�).

Hypotheses concerning instrumentalmotives. We first derive a number of hy-potheses to test for instrumental voting mo-tives. The expected instrumental utility fromvoting YES, respectively, NO differs betweensure-donation case (xi � εi, respectively, xi) andsure-private-money-case (0 � εI, respectively,0). For nonaltruists (xi � 0), keeping themoney is more rewarding than donating themoney and the utility from voting YES isalways higher under the sure-private-moneycase than under the sure-donation case (xi �εI � εi). Thus, we arrive at our first hypoth-esis H16: Higher reward-related neural activ-ity in the striatum for NPY than for NDY. Theopposite applies to altruists who want the

proposal to be pursued (i.e., xi � 0). There-fore, H2 states: Higher reward-related striatalactivity for ADY than for APY. In the sure-private-money case, the altruist expects the

6 The same course of reasoning supports the mirror-inverted pendant to H1: Higher reward-related neural activ-ity for NPN than for NDN. However, the small number ofobservations where subjects voted NO in the sure-donationcase made it impossible for us to test this hypothesis. For thesame reason, we cannot provide tests for a mirror-invertedpendant to H2 and H3. In addition, decisions of the categoryC (close decision) were not observed frequently. For thisreason, we cannot test for differences in neural activationsthat result from increased uncertainty when pi � 0 and theindividual voter is not sure about the outcome. For the samereason, we cannot test for possible activations because ofwarm glow as suggested by Harbaugh et al. (2007). Wewould expect activations in the associated regions to behigher for �CY than for �DY because only in �CY does theindividual voter have the chance to give voluntarily.

Figure 1. Constellations and corresponding abbreviations.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

majority to reject the proposal he approves.Evidence from earlier neuroscientific studies(e.g., Abler, Walter, & Erk, 2005; Siegrist etal., 2005) suggest that he may exhibit neuralactivation in areas that are related to “frustra-tion,” especially the insula. Thus, HypothesisH3 states: Neural activation related to “frus-tration” is present in APY but not in ADY. Itis important to note that the three hypotheseson instrumental voting motives apply regard-less of whether or not subjects are motivatedby expressive motives.

Hypotheses concerning expressive motives.When it comes to expressive motives, wehave to compare neural activity patterns insituation where an individual votes YES withpatterns occurring when the individual votesNO. We recall that the expected utility fromvoting YES, respectively, NO is given by xi �εI, respectively, xi for the sure-donation caseand 0 � εI, respectively, 0 for the sure-private-money case. The utility differentialbetween voting YES and voting NO is given

by UDi � �i � 0 for both sure-donation andsure-private-money case. It is important tonote that this result applies to altruists andnonaltruists alike. Thus, we arrive at our finalyet most important hypothesis H4: Higherreward-related activation in the striatum for�PY than for �PN.

Methods

Subjects

The experiment involved 30 healthy malesubjects without any history of neurologicalor psychiatric disease. Because of insufficienttask comprehension or technical problemsduring the fMRI-session, six subjects had tobe excluded so that 24 subjects are finallyanalyzed (age: 26 � 3, range 19 –37 years).7

All subjects were right-handed according tothe Edinburgh Handedness Scale. All subjectsgave written informed consent and the studywas approved by the ethics committee of theUniversity of Bonn.

Task

Before entering the fMRI scanner, the sub-jects were instructed about the experiment.Each participant was informed that he partic-

ipates in a group experiment on voter behav-ior where the total number of voters is 30. Theset-up of our experiment resembles Tullock’sthought experiment as presented the previouschapter. As Tyran (2004) is recognized tooffer a highly suitable yet simple and intuitiveset-up to test for expressive motives that isfrequently used in behavioral experiments, wechoose a set-up that is similar to his.8 In everyround of the experiment, the subjects wereconfronted with the following decision: Eachparticipant is granted X € and is then asked tovote on the proposal to donate the money to acertain charitable organization. In every trialthe proposal is presented naming the organi-zation, the quorum (Q) and the stake (X). Ifmore than Q% of the participants vote YES,the proposal is accepted and thus, the X € ofall participants are donated. Furthermore, allsubjects keep the X €. Given this information,each participant had to approve or reject theproposal by pressing respective buttons on theMRI-response-grips. The next trial began af-ter the response entry (no time-limit) and ablack screen (4 – 6 s) (see Figure 2). Aftersubjects had read the written instructions andall questions had been resolved, we checkedtask comprehension via a questionnaire. Inthe MRI, each participant started with prac-ticing-trials without image acquisition.

The experiment ran over 315 rounds in onesession with three different values of Q (10,50, and 80%) and three different values for X(5, 10, and 20 €). Each combination of Q andX was randomly presented with seven differ-ent charitable organizations (German RedCross, SOS Children’s Villages, Médecinssans Frontières, Welthungerhilfe, UNO

7 Two of the six subjects revealed an insufficient perfor-mance in the prescanning task comprehension questionnaire.Despite further verbal task explanation before entering thescanner, we could not be sure of a sufficient comprehension ofthe different quora, though this was essential for the study.Four subjects were excluded because of incomplete scanningdata—the scanning sessions of two subjects were cut shortbecause of technical problems, the data corresponding to twofurther subjects was logged incompletely, again because oftechnical malfunction.

8 There are, however, a number of differences betweenour set-up and Tyran’s. We use only the first of two treat-ments of his study and use only three (10, 50, and 80) ratherthan five quora (1, 25, 50, 75, and 99). For technicalreasons, we also elicited voters’ expectations only after thedecisions were made.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Flüchtlingshilfe, Deutscher Kinderschutz-bund, Terre des hommes), with five repeti-tions each. The seven organizations were usedto provide some variety in the stimuli to pre-vent boredom. We conducted pretests withsimilar groups of participants and found theseseven organizations to be equally well-knownand similar with respect to reputation andother characteristics. Not differentiating be-tween the seven organizations, the experimentcontained 105 voting decisions per X and Qand 315 voting decisions in total. This num-ber of repeated trials permits analysis for eachvalue of X and Q. Without repetition thesignal-to-noise ratio would be too low to de-tect significant task related differences in neu-

ral activity. The participants did not know thevalues of Q and X and the names of therelevant organizations beforehand. Through-out the experiment, the participants were notinformed about the aggregate voting behaviorof preceding rounds.

To ensure validity, the subjects were in-formed that their decisions will have real con-sequences because after the completion of thegroup experiment one round was chosen atrandom (the same round for all participants)and the decision made in this round was ex-ecuted. In a postexperimental questionnaire,we elicited the participants’ perception of thedifferent charitable organizations using a5-point-scale (“I know this organization,” “I

Figure 2. Task: The experiment runs over 315 rounds. In each round, the proposal ispresented naming the relevant organization, quorum and stake of this ballot. Participants areallowed as much time to respond as desired. Responses are made by pressing one of twobuttons on the response grips corresponding to the location of the options (Yes/No) on thescreen. The location of the options interchange randomly from trial to trial to preventcorrelation between key press and answer. The button press is followed by a black screen for4–6 s (jittered) before the next trial begins.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

attach importance to this organization’swork,” “With this organization, I’m sure thedeserving poor receive the donated money,”“This organization has a good reputation,”and “This organization has to be supported”).We elicited the strength of the subjects’ al-truistic motives using the Personal AltruisticLevel (PAL) proposed by Tankersley, Stowe,and Huettel (2007).9 We then asked the par-ticipants for all values Q, for X � 5 € andX � 20 € and for all organizations (sepa-rately), whether they expected a clear major-ity for or against the proposal and whetherthey expected the decision to be a close one(“For organization XY . . . Do you expect aclear majority to approve the proposal Q/X?”and “There is no clear majority for or againstthe proposal?”).10 We used this informationto elicit whether subjects expected to vote ina sure-donation, sure-private-money, orclose-election case. Thus, we did not use falsefeedback or other forms of deception to arti-ficially create these cases but we kept in linewith the standards of experimental econom-ics—as we do in the other aspects of theexperiment. Finally, each participant receivedan allowance of 10 € per hour.

fMRI Data Acquisition

Scanning was performed on a 1.5 Tesla (T)Avanto Scanner (Siemens, Erlangen, Germany)using a standard eight channel head coil. Scanparameters were: number of slices: 33; slicethickness: 3 mm; matrix size: 64 � 64; field ofview: 192 mm; interleaved slice acquisition;echo time (TE): 50 ms; repetition time (TR):2.91 s. The task was presented via video gog-gles (Nordic NeuroLab, Bergen, Norway) usingPresentation software (NeuroBehavioral Sys-tems Inc.).

fMRI Data Analysis

fMRI data analysis was done using StatisticalParametric Mapping 5 (SPM5, www.fil.ion.ucl.ac.uk/spm/). Preprocessing included realign-ment with slice timing, unwarping, normaliza-tion to an EPI-template and smoothing with an8 mm Gaussian kernel. The hemodynamic re-sponse to each event (each voting decision) wasmodeled by a canonical hemodynamic responsefunction and its temporal derivative.

For the nonparametric first-level analysis, 18vectors were constructed (YES-/NO-vote foreach X and each Q), using the stimulus onsetsof the proposal. Parameter images for the re-spective contrasts of interest were generated foreach subject and then subjected to two differentsecond-level random effects analyses. Pre-defined linear combinations of the group con-

9 Based on the Self-Report Altruism Scale (SRAS), thePAL is more applicable to a sample of young adults from auniversity surrounding. Tankersley and colleagues report ahigh correlation between PAL and SRAS. Because ourquestionnaire already demands a great deal of concentra-tion, we therefore decided to use only the PAL (instead ofthe SRAS) for keeping the total amount of questions as lowas possible.

10 Note that we asked the questions concerning the ex-pected behavior of fellow-subjects for X � 5 and X � 20only. These amounted to 84 questions already. We wereconcerned that an additional 42 questions for X � 10 mightcause subjects to get bored or tired by the repetition andthus, lead them to give increasingly similar answers for allthree values of X. To avoid this, we restricted the questionsto the situations where X � 5 and X � 20 because weexpected significant differences especially between theseextreme cases. The analysis to follow is restricted to thoseobservations where we had the necessary information toclassify them.

As the expectations concerning other subjects’ behaviorwere elicited after the experiment, we decided to use a veryrough set of categories in the questions concerning theseexpectations. For every institution and quorum and for X �5 and X � 20 we asked subjects only two questions:

a) What do you expect the other subjects to do? (pleasepick one)

i) A clear majority of subjects will approve’.ii) A clear majority will reject.iii) There will be no clear majority in favor of approval

or rejection.b) Do you expect it to be a close decision? (Yes/No)

Based on this information, we can classify a large number ofdecisions to belong to one of the three cases. For instance:If the quorum is 10%, answers ai) and aiii) lead to theclassification sure donation case. If the quorum is 50%,answer ai) leads to the classification sure donation case andanswer aiii) leads to the classification sure private moneycase. For a quorum of 80%, answer aii) and aiii) lead to theclassification sure private money case. The close decisioncase is assumed whenever subjects chose bi). Of course, wecannot classify all decisions made by using this scheme.However, we are able to classify most cases and thus,establish a sufficiently large number of observations to runour essential tests. The major advantage of our classificationscheme is that the use of a very rough set of categoriesreduces the risk that the subjects report an expectation thatdiffers to the one that drove their decision during the ex-periment. In addition, we wanted to prevent difficulties thatsubjects may have to report metric estimates for the ap-proval rates to influence our results and create an illusion ofaccuracy that does not exist.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

trast images were at first tested with two-samplet tests. In another second-level random effectsanalysis for the sure-private-money case, weused a full-factorial design with the factors“type” (two levels: altruist/nonaltruist) and“vote” (two levels: yes/no). Statistical thresholdin both second-level analyses was set at a pvalue of .001 voxelwise (uncorrected for multi-ple comparisons) with a cluster size threshold of10 voxels.

Results

We used a postexperimental questionnaire todifferentiate between altruists and nonaltruists(see Table 1). Any participant who scored anabove median score (30) on the altruism-score(toward strangers) proposed by Tankersley et al.(2007) was hereafter classified as an altruist, allother participants were classified nonaltruists.Thereby, we arrived at 13 altruists and 11 non-altruists. No subject consistently approved orrejected the proposal in 100% of the trials butaltruists showed significantly higher approvalrates than nonaltruists for all three stakes (Ber-noulli test, p � .05). The approval rate across all24 participants was consistently declining withhigher stakes and was higher for Q � 10 thanfor Q � 80 (Bernoulli test, p � .05). The samepattern was observed in the subsamples of al-truists and nonaltruists.

We elicited subjects’ expectations concern-ing the voting behavior of their fellow-subjectsfor the stakes 5 € and 20 € and all quora in thepostexperimental questionnaire. Based on theanswers, we classified whether subjects ex-pected to be in a sure-donation, sure-private-money, or close-decision case when decidingabout a certain stake X at a given quorum Q.

This classification was conducted separately forevery subject and every X-Q-combination. Ex-pectations varied substantially across subjects.Among altruists and nonaltruists, the expectedapproval rate for a quorum of 10% was higherthan for a quorum of 80% (see Table 2). Inaddition, expected approval rates declined withstakes.11

Instrumental Motives–fMRI Results

As hypothesized, nonaltruists showed signif-icantly higher activation in the striatum (leftputamen) when approving the proposal in thesure-private-money case in contrast to approv-ing in the sure-donation case (H1: NPY �NDY, two-sample t test, p � .001 unc, k � 10)(see Figure 3).

For altruists, two-sample t tests confirmedhigher activation in the left striatum (caudateand putamen) when approving in the sure-donation case in contrast to approving in thesure-private-money case (Figure 4a). This resultsupports H2 (ADY � APY, p � .001 unc, k �10). We also found additional activation in theright parahippocampus (Figure 4b), left anteriorrostral MFC (Figure 4c) and right pSTC (Figure

11 Finally, we compared the approval rate in situationswhen subjects expected a clear majority of fellow-subjectsto vote YES to the approval rate in situations when subjectsexpected the majority to vote NO. The average approval ratein the first case was significantly higher in the first situationthan in the second one (Bernoulli test, � .01). This resultis in line with Tyran’s observation that subjects’ inclinationto vote YES is positively correlated with their expectationswith respect to the approval rate among fellow-states (e.g.,Bischoff & Egbert, 2013; Tyran, 2004). Therefore, we alsotest for these motives by contrasting the neural activities for�DY to the activities for �PY, both at the quorum of 50%. Ifbandwagon motives are present, the reward-related activa-tion is expected to be higher in the �PY than in �DYconstellations. However, we find no such pattern. One mightargue that the results support the notion that bandwagonmotives were not relevant in our study. In Bischoff andEgbert (2013) who found evidence for bandwagon behavior,subjects took their decision simultaneously and saw eachother while deciding. This type of set-up is more likely toevoke bandwagon behavior than our current set-up wheresubjects decide in an extremely isolated situation. Alterna-tively, the behavioral regularity reported in the first sentenceof this footnote may result from a false-consensus effectrather than from bandwagon motives in the first place. Inany case, it is not clear how to interpret the lack of signif-icant activations in the contrasts �PY versus �DY. Giventhat our major focus rests on the distinction between instru-mental and expressive motives, we do not report in detail onthis contrast.

Table 1Share of Participants That Expected a ClearMajority [%]

Qj X � 5 € X � 20 €

Altruists (n � 13)10% 72.3 83.150% 72.3 33.780% 56.0 35.1

Nonaltruists (n � 11)10% 48.4 40.350% 34.1 5.2080% 29.7 19.5


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

4d). The opposite contrast (H3: APY � ADY,p � .001 unc.; k � 10) exhibited activation inthe left insula (Figure 5a) as well as left caudatenucleus (Figure 5b) and middle prefrontal cor-tex (Figure 5c).

However, the observed striatal activation inthe two opposing contrasts (ADY � APY andAPY � ADY) originated in different processes.Plots of the contrast estimates reveal the influ-ence of each factor of interest on the activationwithin the left putamen and left caudate (F-testfor �AY, p � .001 (FDR), k � 10). As Figure 6aand 6b show, the striatal activity in contrastADY � APY resulted from a negative signalchange in the left putamen as well as in leftcaudate, so these regions were less activated inboth cases. Because the negative signal changeswere higher in the private-money case, the con-trasts showed stronger signals in the sure-donation case (Figure 4a), although the effectsize was negative in both cases. Contrary tothat, the detected caudate activation in the con-trast APY � ADY follows from a positivesignal change in both cases. As Figure 6bshows, the signal change in the sure-private-money case was much higher than in the sure-donation case, therefore we observed a higheractivation in the left caudate nucleus for sure-private-money than sure-donation in the con-trast APY � ADY (Figure 5a).

Expressive Motives–fMRI Results

As expressive motives may apply to bothaltruists and nonaltruists alike, we did not ini-tially distinguish between altruists and nonaltru-ists here. Instead, we contrasted approvals and

rejections of all subjects in the sure-private-money case (H4: �PY � �PN, p � .001 unc.;k � 10) and found no significant activation.With lowering the level of significance to p �.005 unc.; k � 10, two-sample t tests confirmed(among others) higher activation in the left cau-date and left putamen when approving in con-trast to rejecting the proposal in the sure-private-money case (see Figure 7). In a nextstep, we were interested to find out how muchaltruists and nonaltruists each contribute to theobserved neural activity in �PY � �PN. There-fore, we conducted an analysis with a full-factorial-design for the sure-private-money casewith the factors “type” (altruist/nonaltruist) and“vote” (yes/no). We conduct a conjunction anal-ysis (p � .005) for the contrasts APY � APNand NPY � NPN and observed no voxels acti-vated in both contrasts. We mask the contrast�PY � �PN inclusively with the contrast APY �NPY. This approach shows all voxel which aresignificant at .005 unc. with k � 10 across allsubjects and at .05 unc. for altruists. None of theactivations “survived” this masking. We theninclusively mask the contrast �PY � �PN withthe contrast NPY � NPN on the same signifi-cance level and found all activations describedfor �PY � �PN. Thus, the observed neural ac-tivity in �PY � �PN seemed to be solely elicitedby nonaltruists and not by altruists. We thenplotted the contrast estimates for each regionand found further support for this conclusion.

Table 2Average Approval Rates [%]

Qj X � 5 € X � 10 € X � 20 €

All participants (n � 24)10% 68.0 53.8 47.250% 70.2 48.3 31.780% 60.4 38.7 28.92

Altruists (n � 13)10% 75.2 67.0 56.750% 81.8 66.8 44.080% 73.0 53.2 41.1

Nonaltruists (n � 11)10% 60.2 39.5 36.850% 56.6 26.5 17.180% 45.5 21.6 13.0

Figure 3. NPY � NDY: Nonaltruists exhibit higher acti-vation in left striatum when approving the proposal in thesure-private-money case in contrast to the sure-donationcase (p � .001 uncorrected; k � 10).


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

For the left caudate, APY elicited a small andNPY a large positive signal change, whereasAPN showed no effect and NPN a negativesignal change (see Figure 8). We also conducteda separate test for nonaltruists which yielded(among others) significant striatal activation(NPY � NPN, two-sample t test, p � .005 unc,k � 10). Plots of the contrast estimates revealeda positive signal change in the striatum forapprovals and a negative signal change for re-jections of the proposal (see Figure 9). Foraltruists, the same test revealed no significantactivation (APY � APN, two-sample t test, p �.005 unc, k � 10) (Table 3).

Discussion

The purpose of this study was to achieve abetter understanding of the motives in voting byusing the neuroscientific method of fMRI. Wewanted to learn more about rewarding aspectsof the act of voting by investigating the activityof reward-related brain areas, especially thestriatum.12 We explored differences in the stri-atal activation between altruists and nonaltruistsbut our special focus rests on the importance ofexpressive motives in voting. We start by dis-

cussing the evidence on instrumental votingmotives. Here, we expected altruists and nonal-truists to show distinctly different reward-related activation when confronted with theprospect to have to donate money (sure-donation case) as well as when expecting tokeep the money for private use (sure-private-money case).

Instrumental Motives, Nonaltruistic Voters

We hypothesized that a nonaltruistic voterexhibits higher striatal activation when votingYES in the sure-private-money case than in thesure-donation case (H1: NPY � NDY). He ap-proves the proposal in both cases presumably toattain the expressive utility (because �i � pixi)knowing that—with a very high probability,that his monetary payoff do not depend on hisdecision (i.e., pi � 0). The fMRI-results showthat keeping the money in the sure-privatemoney case (NPY) revealed an increased activ-ity in the left putamen, whereas giving to char-

12 The complete list of brain activation is available by theauthors upon request.

Figure 4. ADY � APY: Altruists show higher activation in (a) left striatum (putamen), (b)right parahippocampus, c) left anterior rostral medial frontal cortex (arMFC), and d) rightposterior superior temporal cortex (pSTC) when approving the proposal in the sure-donationcase in contrast to the sure-private-money case (p � .001 uncorrected.; k � 10).


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

ity in the sure-donation case (NDY) showed adecreased activity in the same region. Thus, thenonaltruist showed the hypothesized neural ac-tivation consistent with materialistic and egois-tic preferences.

Instrumental Motives, Altruistic Voters

An altruistic voter wants the proposal to bepursued when he approves it because it is in hisinstrumental interest. However, the donation isonly pursued if the overall approval rate is suf-ficiently high. Thus, we hypothesized that thealtruist shows higher reward-related activationwhen voting YES in the sure-donation case thanin the sure-private-money case (H3: APY vs.ADY). The observed activation in the left stria-tum (caudate and putamen) supports this hy-pothesis at first sight, because these regions areknown to represent monetary and social reward.However, for ADY as well as for APY the leftputamen and caudate were deactivated, only thedeactivation for APY was much higher, so con-trasting both cases, the striatal activity was lessnegative for ADY. This pattern does not fit thelogic underlying hypothesis H2. Given this re-sult, we focus on the other activated regions(left arMFC, right parahippocampus and rightpSTC) in the contrast ADY versus APY for apossible explanation. The additionally activatedanterior rostral MFC (arMFC) is supposed toplay a major role in social cognition (for anoverview: Amodio & Frith, 2006). It is found tobe associated with considering others’ mentalstates and intentions, including reflections aboutwhat we want others to think about ourselves.

However, it is also referred to thinking aboutthe self and processing of what feels like theright thing to do in a moral dilemma (Amodio &Frith, 2006). We also found a higher activity ofthe right pSTC that is associated with self-reported altruism (Tankersley et al., 2007) andwillingness to give (Hare et al., 2010). Consid-ering these additional activations, we might ob-serve in this contrast (ADY � APY) not a clearreward processing but thinking about what isthe right thing to do against the background oftheir own altruistic standards and the standardsof others.

We further hypothesized that altruistsshould exhibit neural activation in areaswhich are related to “frustration” in the sure-private-money case in contrast to the sure-donation case, because they expect the major-ity to reject the proposal which they want tosee pursued (H3). Contrasting APY � ADY,we found stronger activation in the dorsalstriatum (caudate nucleus), left insula andmiddle prefrontal cortex. We found the hy-pothesized activation in the left anterior in-sula. The anterior insula is part of the painprocessing system and also associated withunfairness in social interactions, frustrationand a feeling of discomfort (Abler et al.,2005; Meyer–Lindenberg, 2008). Meyer–Lindenberg even links the anterior insula sig-nal to “gut feeling.” However, the insula isalso associated with the processing of rewardmagnitude and showed a higher activity for ahigher reward magnitude in former studies(Smith et al., 2009). The next activated region

Figure 5. APY � ADY: Altruists exhibit activation (a) in the left insula, (b) in the dorsalstriatum (left caudate nucleus), and c) in the middle prefrontal cortex when they approve theproposal in the sure private-money case (p � .001 uncorrected.; k � 10).


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Figure 6. ADY � APY: The bars represent the signal change in (a) left putamen [MNI:�21, 18, 9] and (b) left caudate [MNI: �18, 19, 9] for altruists approving the proposal in thesure-donation case (left bars) and sure-private-money case (right bars). (c) Shows the signalchange in the left caudate [MNI: �9, 3, 6] in the opposite contrast APY � ADY.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

(the caudate nucleus) has been associatedwith social processes like trust (Baumgartner,Heinrichs, Vonlanthen, Fischbacher, & Fehr,2008). King–Casas et al. (2005) suppose thecaudate nucleus to compute “informationabout the fairness of a social partner’s deci-sion.” However, just like the insula, the cau-date nucleus, too, is also involved in antici-pation and reception of reward (Valentin &O’Doherty, 2009; Dreher, Kohn, Kolachana,Weinberger, & Berman, 2009; Pizzagalli etal., 2009). The third activation in this contrastwas in a middle prefrontal region (mPFC),

that is known to be related to expected rewardvalue and variance (Tobler, O’Doherty,Dolan, & Schultz, 2007). Hence, on the onehand all three regions (insula, caudate, andmPFC) may indicate a more reward-relatedactivity for APY than for ADY. On the otherhand, insula and caudate can also be related tounfairness. While the nonaltruists revealedclear reward-related processes when theyvoted YES but expected not having to givethe money away, the altruists may feeluncomfortable: Not to pursue the proposalmay appear unfair to them, causing an un-

Figure 7. �PY � �PN: Without distinguishing between altruists and nonaltruists, activationfor approvals in the sure-private-money case is observed in (a) left caudate and (b) leftputamen (p � .005 uncorrected.; k � 10).

Figure 8. �PY � �PN: Plots of contrast estimates in the peak voxel of the left caudate(full-factorial p � .005, k � 10 voxel, MNI: �24, �18, 21). The difference in signal changeis largest for NPY � NPN.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

comfortable “gut feeling.” This interpretationsupports the bottom line of hypothesis H3with respect to the altruists motivation whenvoting YES in the sure-private-moneycase.

In a first general view, we observed distinctdifferences between altruists and nonaltruists intheir neural representation of instrumental mo-tives. While the nonaltruists revealed a clearreward-related brain activity in the sure-private-money case (H1), the neural activation patternof altruists was less clear. Donating the moneyin the sure-donation case (ADY) elicited activ-

ity in regions known to be involved in socialcognition, like processes about agency, altru-ism, willingness to give and consideration aboutthe right thing to do (arMFC and pSTC). Do-nating the money in the sure-private-moneycase (APY) activated reward-related brain re-gions but as well regions that are associatedwith a feeling of discomfort and unfairness. Ifthe altruists really wanted the proposal to bepursued—as an investment and consumptionalike—reward-related activity should have beenmore prominent in ADY. Therefore, we couldnot find clear support for H2.

Table 3MNI-Coordinates and Statistic Values for the Significant Activations

Brain region

MNI-coordinates

z-scoreax y Z

Nonparametric analysis: p � .001 (uncorrected for multiple comparisons, extent threshold of 10 voxels)Nonaltruists Putamen (L) �21 18 �3 4,0

NPY � NDYAltruists Caudate (L) �18 18 9 3,76

ADY � APY Putamen (L) �21 18 9 4,07Parahippocampus (R) 18 �12 �21 5,98Anterior rostral MFC (L) �12 60 9 3,35pSTC (R) 54 �60 24 3,59

Altruists Insula (L) �42 18 3 4,52APY � ADY Caudate (L) �9 3 6 4,53

Middle prefrontal cortex 33 48 21 3,92Nonparametric analysis: p � .005 (uncorrected for multiple comparisons, extent threshold of 10 voxels)

All voters Caudate (L) �24 �18 21 3,98�PY � �PN Putamen (L) �21 18 �3 3,16

a z � z-score for the peak activation voxel.

Figure 9. NPY � NPN: Plots of the contrast estimates for nonaltruists reveal a positivesignal change in the left putamen [MNI: �21, 18, �3] when voting YES (left bar) and anegative signal change when voting NO (right bar). In both cases, the voters know that theycan keep the money for themselves.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Expressive Motives, Altruistic andNonaltruistic Voters

Our final yet most important hypothesisaimed at testing for expressive motives in vot-ing. Assuming that these motives are present,we expect all voters—no matter if they arealtruists or nonaltruists—to perceive approvalsin the sure-private-money case as more reward-ing than rejections in this case (H4: �PY ��PN). The observed activation in the dorsalstriatum (left caudate nucleus and left putamen)supports this hypothesis at first sight. However,further analyses revealed that the observed neu-ral activity was nearly completely attributableto nonaltruists. Altruistic voters did not revealthis activation pattern. This finding is consistentwith all other results of this experiment. Onlythe nonaltruists revealed a distinct reward-related neural activity namely when they ap-proved the proposal in the sure-private-moneycase (NPY). The plots of the contrast estimatesfor the left caudate (see Figure 8) support this.Nonaltruists revealed a stronger positive signal-change than altruists when they approved the pro-posal in the sure-private-money case. Rejectingthe proposal elicited a negative signal-change fornonaltruists and nearly no signal-change for altru-ists. Thus, the immediate expressive utility fromthe act of voting seemed to be less distinctive foraltruists.

Despite its interesting first results, our studyhas a number of limitations. First, the number ofsubjects in each group was rather small, whichmight account for partly less conclusive resultsand the need for lowering the level of signifi-cance to p � .005, respectively. Second, wewere unable to analyze voting behavior in close-decision cases because they were not suffi-ciently frequent to allow for a meaningful sta-tistical analysis. The extreme rarity of closedecisions is typical for majority voting andtherefore limits behavioral experiments as wellas field studies on expressive behavior alike. Itcan be avoided if the experimenters appliestreatments in which they explicitly confrontsubjects with situations in which their decisionis likely (or even certain) to tip the scales andother treatments where they are (almost cer-tainly) not pivotal. However, these treatmentscannot be framed as voting decisions in sizablegroups but have to be presented as voting deci-sions in a group of three to five subjects. There-

fore, subjects are likely to perceive these treat-ments with large and small groups to differ inmore than just the probability of being pivotal.Nevertheless, it may be interesting to observethe neural patterns of subject in such treatments.

To keep the task in the fMRI simple, we didnot elicit subjects’ expectations with respect totheir fellow-voters’ behavior at the point in timewhen they make their decision but only after theexperiment. This bears the danger that the an-swers in the postexperimental questionnairemay deviate from the expectations that subjectsentertained when performing the task in thefMRI. However, our results do not rely on thecomparison of close and nonclose decisions. Inour opinion, it seems unlikely that a voter ex-pects a proposal to pass comfortably ex postwhile he expected a clear rejection while in thescanner. Thus, the time-lag between voting de-cisions and answers on expected behavior offellow-subjects is unlikely to jeopardize ourfindings. Finally, we did not generate a suffi-cient number of cases in which nonaltruistsreject the proposals and therefore we are unableto test the mirror-inverted versions of our hy-potheses (e.g., H1=: Higher reward-related neu-ral activities for NPN than for NDN). This is theprice we pay for using charitable organizationsthat produced a high emotional involvementamong subjects. However, we have deliberatelychosen these organizations to make sure thatexpressive motives—if existent at all—arelikely to have an impact on subjects’ behavior.In a future experiment, it may be interesting touse higher stakes and/or emotionally less in-volving organizations and thereby produce asufficient mix of rejections and approvals.

Conclusion

In this article, we provide a first fMRI-studyon voting behavior. We follow Tyran (2004) inour experimental set-up and our primary focuson expressive motives. Furthermore, we use thealtruism-score proposed by Tankersley et al.(2007) to differentiate between altruists andnonaltruists and investigate differences in stria-tal activation between them in the context ofvoting. Our results are in line with former ex-periments on social decision-making and char-itable giving (Harbaugh et al., 2007; Izuma etal., 2010; Moll et al., 2006) but also shed further


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

light on the motivational basis of voting deci-sions.

Summarizing the first results generated byour study, we found different patterns of re-ward-related activity for altruistic and nonaltru-istic voters. For nonaltruistic voters, we ob-served striatal activity when voting YES in thesure-private-money case. Keeping the money isrewarding for nonaltruists. This result shows theimportance of investment-related, instrumentalmotives in voting. Though not surprising assuch, the result indicates the validity of ourset-up. However, for nonaltruistic voters, wefurthermore found approving the proposal in thesure-private-money case to be more rewardingthan a rejection. This result is noteworthy be-cause it provides support for the existence ofexpressive motives among nonaltruists. This in-terpretation further strengthens the notion ofexpressive voting according to which votes—ifnot pivotal—do not refer to the policy outcomebut to the immediate utility from the act ofvoting. For altruists, we observed a neural acti-vation pattern that can be associated with in-vestment (as well as consumption) related mo-tives: In the sure-private-money case, altruistsshowed activation either indicating toward un-fairness because the money will not be given tocharity or indicating toward reward perceptionwith an uncomfortable “gut feeling.” Approvingthe proposal in the sure-donation case activatedbrain regions associated with altruism, willing-ness to give and thinking about the right thing todo. At the same time, we did not observe ex-pressive motives in nonaltruistic voters.

Like Harbaugh et al. (2007) found for chari-table giving and taxation, we found that bothinvestment-related (instrumental: altruism oregoism) and consumption-related (expressive)motives play a role in voting decisions. Previ-ous studies using traditional economic methodshad to rely on behavioral observations andfound only limited support for expressive mo-tives. Using neuroscientific methods, we pro-vide evidence for a neural correlate of expres-sive utility. At this point, our conclusions arestill partly interpretative and need further inves-tigation in follow-up experiments.

References

Abler, B., Walter, H., & Erk, S. (2005). Neural cor-relates of frustration. Brain Imaging, 16, 669–672.

Amodio, D. M., & Frith, C. D. (2006). Meeting ofminds: The medial frontal cortex and social cog-nition. Nature Reviews Neuroscience, 7, 268–277.doi:10.1038/nrn1884

Amodio, D. M., Jost, J. T., Master, S. L., & Yee,C. M. (2007). Neurocognitive correlates of liber-alism and conservatism. Nature Neuroscience, 10,1246–1247. doi:10.1038/nn1979

Andreoni, J. (1989). Giving with impure altruism:Applications to charity and Ricardian equivalence.Journal of Political Economy, 97, 1447–1458. doi:10.1086/261662

Battaglini, M., Morton, R. B., & Palfrey, T. R. Theswing voter’s curse in the laboratory. Review ofEconomic Studies, 77, 61–89, 2010. doi:10.1111/j.1467-937X.2009.00569.x

Baumgartner, T., Heinrichs, M., Vonlanthen, A., Fis-chbacher, U., & Fehr, E. (2008). Oxytocin shapesthe neural circuitry of trust and trust adaptation inhumans. Neuron, 58, 639 – 650. doi:10.1016/j.neuron.2008.04.009

Bischoff, I., & Egbert, H. (2013). Social informationand bandwagon behaviour in voting: An economicexperiment. Journal of Economic Psychology, 34,270–284. doi:10.1016/j.joep.2012.10.009

Brennan, G., & Hamlin, A. (1998). Expressive votingand electoral outcome. Public Choice, 95, 149–175. doi:10.1023/A:1004936203144

Camerer, C., Loewenstein, G., & Prelec, D. (2005).Neuroeconomics: How neuroscience can informeconomics. Journal of Economic Literature, 43,9–64. doi:10.1257/0022051053737843

Carter, J. R., & Guerette, S. D. (1992). An experi-mental study of expressive voting. Public Choice,73, 251–260. doi:10.1007/BF00140921

Chorvat, T. (2007). Tax compliance and the neuro-economics of intertemporal substitution. NationalTax Journal, 60, 577–588.

Coffman, L. C. (2010). Essays in experimental eco-nomics. Cambridge, MA: Harvard University.

Delgado, M. R. (2007). Reward-related responses inthe human striatum. Annals of the New York Acad-emy of Sciences, 1104, 70–88. doi:10.1196/annals.1390.002

Downs, A. (1957). An economic theory of politicalaction in a democarcy. The Journal of PoliticalEconomy, 65 (2), 135–150.

Dreher, J.-C., Kohn, P., Kolachana, B., Weinberger,D. R., & Berman, K. F. (2009). Variation in do-pamine genes influences responsivity of the humanreward system. PNAS, 106, 617–622. doi:10.1073/pnas.0805517106

Farmer, D. (2007). Neuro-Gov: Neuroscience as cat-alyst. Annals of the New York Academy of Sci-ences, forthcoming, (ID: Annals-1412–002).

Feddersen, T. J., & Pesendorfer, W. (1996). Theswing voter’s curse. American Economic Review,86, 404–424.


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Feddersen, T., Gailmard, S., & Sandroni, A. (2009). Moralbias in large elections: Theory and experimental evi-dence. American Political Science Review, 103, 175–192. doi:10.1017/S0003055409090224

Fields, J. M., & Schuman, H. (1976). Public beliefsabout the beliefs of the public. Public OpinionQuarterly, 40, 427–448. doi:10.1086/268330

Fiorina, M. P. (1976). The voting decision: Instru-mental and expressive aspects. Journal of Politics,38, 390–415. doi:10.2307/2129541

Fischer, A. J. (1996). A further experimental study ofexpressive voting. Public Choice, 88, 171–184.doi:10.1007/BF00130417

Fliessbach, K., Weber, B., Trautner, P., Dohmen, T.,Sunde, U., Elger, C. E., & Falk, A. (2007). Socialcomparison affects reward-related brain activity inthe human ventral striatum. Science, 318, 1305–1308.

Harbaugh, W. T., Mayr, U., & Burghart, D. R.(2007). Neural responses to taxation and voluntarygiving reveal motives for charitable donations. Sci-ence, 316, 1622–1625. doi:10.1126/science.1140738

Hare, T. A., Camerer, C. F., Knoepfle, D. T.,O’Doherty, J. P., & Rangel, A. (2010). Valuecomputations in ventral medial prefrontal cortexduring charitable decision making incorporate in-put from regions involved in social cognition. TheJournal of Neuroscience: The Official Journal ofthe Society for Neuroscience, 30, 583–590.

Hillman, A. L. (2010). Expressive behavior in eco-nomics and politics. European Journal of PoliticalEconomy, 26, 403– 418. doi:10.1016/j.ejpoleco.2010.06.004

Izuma, K., Saito, D. N., & Sadato, N. (2010). Pro-cessing of the incentive for social approval in theventral striatum during charitable donation. Jour-nal of Cognitive Neuroscience, 22, 621–631. doi:10.1162/jocn.2009.21228

Kable, J. W., & Glimcher, P. W. (2009). The neuro-biology of decision: Consensus and controversy.Neuron, 63, 733–745. doi:10.1016/j.neuron.2009.09.003

Kamenica, E., & Egan, L. (2011). Voters, dictators,and peons: Expressive voting and pivotality. Uni-versity of Chicago Booth School of BusinessWorking Paper 2011.

King–Casas, B., Tomlin, D., Anen, C., Camerer,C. F., Quartz, S. R., & Montague, P. R. (2005).Getting to know you: Reputation and trust in a twoperson economic exchange. Science, 308, 78–83.doi:10.1126/science.1108062

Kuss, K., Falk, A., Trautner, P., Elger, C. E., Weber,B., & Fliessbach, K. (2013). A reward predictionerror for charitable donations reveals outcome ori-entation of donators. Social Cognitive and AffectNeuroscience, 8, 216 –223. doi:10.1093/scan/nsr088

Loewenstein, G., Rick, S., & Cohen, J. (2008). Neu-roeconomics. Annual Review of Psychology, 59,18.1–18.26.

Meyer–Lindenberg, A. (2008). Psychology: Trust meon this. Science, 321, 778 –780. doi:10.1126/science.1162908

Moll, J., Krueger, F., Zahn, R., Pardini, M., de Ol-iveira–Souza, R., & Grafman, J. (2006). Humanfronto-mesolimbic networks guide decisions aboutcharitable donation PNAS Proceedings of the Na-tional Academy of Sciences, USA of the UnitedStates of America, 103, 15623–15628. doi:10.1073/pnas.0604475103

Pizzagalli, D. A., Holmes, A. J., Dillon, D. G., Goetz,E. L., Birk, J. L., Bogdan, R., . . . Fava, M. (2009).Reduced caudate and nucleus accumbens responseto rewards in unmedicated individuals with majordepressive disorder. The American Journal of Psy-chiatry, 166, 702–710. doi:10.1176/appi.ajp.2008.08081201

Rule, N. O., Freeman, J. B., Moran, J. M., Gabrieli,J. D. E., Adams, R. B. Jr., & Ambady, N. (2010).Voting behavior is reflected in amygdale responseacross culture. Social Cognitive and Affective Neu-roscience, 5, 349–355. doi:10.1093/scan/nsp046

Sanfey, A. G., Loewenstein, G., McClure, S. M., &Cohen, J. D. (2006). Neuroeconomics: Cross-currents in research on decision-making. Trends inCognitive Sciences, 10, 108–116. doi:10.1016/j.tics.2006.01.009

Saxe, R., & Haushofer, J. (2008). For love or money:A common neural currency for social and mone-tary reward. Neuron, 58, 164–165. doi:10.1016/j.neuron.2008.04.005

Seymour, B., Singer, T., & Dolan, R. (2007). Theneurobiology of punishment. Nature Reviews Neu-roscience, 8, 300–311. doi:10.1038/nrn2119

Shayo, M., & Harel, A. (2012). Non-consequentialistvoting. Journal of Economic Behavior & Organi-zation, 81, 299–313. doi:10.1016/j.jebo.2011.10.021

Siegrist, J., Menrath, I., Stöcker, T., Klein, M.,Kellermann, T., Shan, N. J., Zilles, K., & Sch-neider, F. (2005). Differential brain activation ac-cording to chronic social reward frustration. Neu-roreport, 16, 1899 –1903. doi:10.1097/01.wnr.0000186601.50996.f7

Smith, B. W., Mitchell, D. G. V., Hardin, M. G.,Jazbec, S., Fridberg, D., Blair, R. J. R., & Ernst, M.(2009). Neural substrates of reward magnitude,probability, and risk during a wheel of fortunedecision-making task. NeuroImage, 44, 600–609.doi:10.1016/j.neuroimage.2008.08.016

Tankersley, D., Stowe, C. J., & Huettel, S. A. (2007).Altruism is associated with an increased neuralresponse to agency. Nature Neuroscience, 10,150–151. doi:10.1038/nn1833


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Tingley, D. (2006). Neurological imaging as evi-dence in political science: A review, critique, andguiding assessment. Social Science Information,45, 5–33. doi:10.1177/0539018406061100

Tobler, P. N., O’Doherty, J. P., Dolan, R. J., &Schultz, W. (2007). Reward value coding distinctfrom risk attitude-related uncertainty coding in hu-man reward systems. Journal of Neurophysiology,97, 1621–1632. doi:10.1152/jn.00745.2006

Tullock, G. (1971). The charity of the uncharitable.Western Economic Journal, 9, 379–392.

Tyran, J.-R. (2004). Voting when money and moralsconflict: An experimental test of expressive voting.Journal of Public Economics, 88, 1645–1664. doi:10.1016/S0047-2727(03)00016-1

Valentin, V. V., & O’Doherty, P. (2009). Overlap-ping prediction errors in dorsal striatum during

instrumental learning with juice and money rewardin the human brain. Journal of Neurophysiology,102, 3384–3391. doi:10.1152/jn.91195.2008

Westen, D., Blagov, P. S., Harenski, K., Kilts, C., &Hamann, S. (2006). Neural bases of motivatedreasoning: An fMRI study of emotional constraintson partisan political judgment in the 2004 USPresidential election. Journal of Cognitive Neuro-science, 18, 1947–1958. doi:10.1162/jocn.2006.18.11.1947

Received December 6, 2012Revision received September 2, 2013

Accepted September 8, 2013 �

E-Mail Notification of Your Latest Issue Online!

Would you like to know when the next issue of your favorite APA journal will be availableonline? This service is now available to you. Sign up at http://notify.apa.org/ and you will benotified by e-mail when issues of interest to you become available!


Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Documents

The Neuroeconomics of Voting: Neural Evidence of Different