social preferences and political participation

7/28/2019 Social Preferences and Political Participation

1/13

Social Preferences and Political Participation

Christopher T. Dawes University of CaliforniaPeter John Loewen University of TorontoJames H. Fowler University of California

Models of political participation have begun to incorporate actors who possess social preferences. However, these

models have failed to take into account the potentially incongruent political goals of different social preferencetypes. These goals are likely to play an important role in shaping political behavior. To examine the effect of distinctsocial preferences on political activity we conducted an experiment in which participants played five rounds of a

modified dictator game (Andreoni and Miller 2002). We used the decisions in these games to determine theirpreference type and mapped these types to reported political activity. Our results show that subjects who were mostinterested in increasing total welfare in the dictator game were more likely to participate in politics than subjectswith selfish preferences, whereas subjects most interested in reducing the difference between their own well-beingand the well-being of others were no more likely to participate than subjects with selfish preferences.

Rational choice models of political participa-tion based on the assumption that actors arepurely self-interested typically fail to explain

why so much participation occurs (Fiorina 1990).1 Inthese models the cost of participation, even if small,typically outweighs the expected personal benefit ofparticipation. If this is the case, we should observevery little political participation, but in fact largenumbers of citizens are politically active. Alternativemodels have begun to incorporate actors who care

not only about themselves, but also about the well-being of others. In particular, formal models based onan assumption that actors have social preferenceshave been constructed to explain voting behavior(Edlin, Gelman, and Kaplan et al. 2007; Fowler 2006;Jankowski 2002, 2007; Feddersen and Sandroni 2006)and political participation more broadly (Fowler andKam 2006). They have also been tested empirically(Fowler 2006; Fowler and Kam 2006; Loewen 2010).This approach is motivated in part by a vast literaturewithin behavioral economics which shows that in anexperimental setting many individuals are willing tohelp others, often anonymous to them, even whenit is personally costly to do so. In models assumingpolitical actors with social preferences, individuals

consider the benefits to themselves and to others theycare about and weigh them against the cost ofparticipation. This decision-making calculus is ra-tional in the sense that individuals have completeand coherent preferences, but these preferences arenot purely selfish (Charness and Rabin 2002).

Thus far, the political science literature has treatedsocial preferences in nearly monolithic fashiontheprincipal type of social preference that has beenconsidered is one in which there is a direct corre-

spondence between how concerned actors are forothers and how much they participate (Edlin, Gelman,and Kaplan 2007; Fowler 2006; Jankowski 2002,2007). The only exception to this has been allowingactors to express differing degrees of concern fordifferent groups of citizens (Fowler and Kam 2006;Loewen 2010). While behavioral economists havemodeled, and observed in the laboratory, severaldistinct forms of concern for the payoffs of others,thus far no studies have empirically related differenttypes of social preferences to political participation.Our contribution is to demonstrate the differingrelationships between different types of social prefer-ences and a composite measure of political partic-ipation. By doing so, we move beyond both the more

The Journal of Politics, Vol. 73, No. 3, July 2011, Pp. 845856 doi:10.1017/S0022381611000508

Southern Political Science Association, 2011 ISSN 0022-3816

1Loewen acknowledges the support of the Killam Trusts and the Social Science and Humanities Research Council of Canada. Anonline appendix for this article is available at http://jhfowler.ucsd.edu and http://www.journals.cambridge.org/jop. Data and supportingmaterials necessary to reproduce the numerical results in the paper will be made available at http://jhfowler.ucsd.edu and http://www.journals.cambridge.org/jop upon publication.

845


2/13

narrow consideration of uniform social preferencesand the act of voting. Our work thus speaks to both asocial preferences literature and to the much broaderempirical literature concerned with the individual-level differences that motivate political participation,whether senses of duty (Blais 2000), race and gender(Leighley and Nagler 1992a,b), information (Lassen2005), or socioeconomic status (Verba, Schlozman,and Brady 1995).

Two prominent types of social preferences thathave been identified are Rawlsian and utilitarianpreferences (Andreoni and Miller 2002).2 Both ofthese types extend utility beyond narrow self-interest,but they differ in important ways. Rawlsians caremore about reducing the difference between theirown well-being and the well-being of others, whereasutilitarians care more about increasing everyonesaverage well-being. This difference is important for

applications of social preferences to political participa-tion. Utilitarians might participate more than thosewho are purely selfish, but only if political outcomeshave an effect on total social welfare. Similarly, Rawl-sians might participate more than those who are purelyselfish, but only if political outcomes can be used toredistribute resources to those most in need. In reality,politics probably influences both welfare and the dis-tribution of resources, but it is not clear which effectdominates or how much they affect participationamong those with other-regarding preferences.

This issue poses a potential limitation for em-

pirical studies linking social preferences to individualpolitical behavior. Individuals with Rawlsian prefer-ences and individuals with utilitarian preferencesboth behave altruistically, or express a desire foraltruism, rendering them observationally equivalent ifnot examined carefully. In the parlance of thedictator game, both types of subjects may give awaymoney to others, though they may have differentunderlying motivations for this action. If these twogroups do have different views on politics as a meansfor benefitting others, the more complex underlyingrelationship between social preferences and political

participation may be clouded by lumping the twogroups together. What is required is a behavioralmeasure which can reveal distinct motivations andthen map these motivations to political participation.

To examine the effect of distinct social prefer-ences on political activity we designed a two-partstudy. First, we instructed participants to play five

rounds of a modified dictator game and used thedecisions in these games to determine their prefer-ence type (Andreoni and Miller 2002). Second,participants completed a detailed survey of theirrecent political activity. This research design allowsus to map revealed preferences to a large variety ofpolitical activities. To preview our results, we findthat subjects who were most interested in increasingtotal welfare in the dictator game were more likely toparticipate in politics than subjects with selfishpreferences, whereas subjects most interested in re-ducing the difference between their own well-beingand the well-being of others were no more likely toparticipate in politics than subjects with selfishpreferences. These results highlight the importanceof accounting for heterogeneity when studying thelink between social preferences and political behavior.Indeed, the results that follow suggest that different

social preferences are related to distinctly differentrates of political participation.

Social Preferences and PoliticalParticipation

Utilitarianism, in the words of Harsanyi, 116 isconcerned with maximizing social utility...definedas the sum, or arithmetic mean, of all utilities (1980,116). Harsanyi (1980) theorized that the paradox of

voting could be overcome by individuals guided byutilitarian principles. Voters deciding the fate of somepolicy measure they view as socially desirable, heargued, would go to the polls, even though it wascostly to do so, because voting would yield highersocial utility than abstaining. Edlin, Gelman, andKaplans (2007) social benefit model of voting, forexample, is an extension of this thinking. The modelassumes that the benefit associated with participationis B

s+ aNBo, where N is the number of people

affected by the political outcome, Bs

is the benefit tooneself and B

ois the benefit to the average voter if the

preferred outcome occurs, and a is a scalar measure,between zero and one, of an individuals level of al-truism. The expected value of participation is P(B

s+

aNBo) Cwhere Pis the probability that participationgenerates the desired political outcome. If P 1

N(as

many scholars have arguedsee Gelman, King, andBoscardin 1998; Gelman, Katz, and Bafumi 2004;Mulligan and Hunter 2003), this calculus of votingreduces to a decision to vote when aBo . C. Thismodel is largely similar to those proposed by others(Fowler 2006; Jankowski 2002, 2007; Feddersen and

2Andreoni and Miller also refer to these as Leontief andsubstitute preferences, respectively.

846 christopher t. dawes, peter john loewen, and james h. fowler


3/13

Sandroni 2006). We note that its logic extends to anyact of political participation which resembles a collec-tive action problem and not only the act of voting.

Rawlsian preferences, on the other hand, havenot been explicitly linked to political participation inany formal or empirical model. Harsanyi (1975), in acriticism of John Rawls maximin principle, offers apossible explanation for why a desire to reduce thedifference between ones own well-being and thewell-being of others may not translate into increasedpolitical participation. Harsanyi argues that focusingon this difference would require us to give absolutepriority to the interests of the worst-off individual, nomatter what, even under the most extreme condi-tions. Even if his interests were affected only in a veryminor way, and all other individuals in society hadopposite interests of the greatest importance, hisinterests would always override anybody elses(596).

Political outcomes are often redistributive, favoringone group over another, but individuals with Rawlsianpreferences would only support redistributions thatstrictly go to the worst-off. Because politics rarely isthe venue for this type of redistribution, individualswith Rawlsian preferences may devote their altruism toother outlets better capable of achieving their distri-butional goals. For example, Andreoni and Vesterlund(2001) provide evidence of a possible link betweenRawlsian preferences and charitable giving. In thecontext of voting and other means of political engage-ment then, we should expect utilitarians to exhibit

greater participation than Rawlsians. Accordingly, ourwork examines this link between different socialpreferences and political participation across a numberof acts of political engagement (Verba, Schlozman, andBrady 1995).

Research Design

Our study was conducted at the University ofCalifornia, Davis in the spring of 2005. Subjects wererecruited from various social science majors and were

offered course credit for participating in the study. Atotal of 234 subjects participated in the completestudy. The study had two parts. First, subjects wereasked to make five decisions in what is called thedictator game. Upon completion of the first taskthey were asked to complete a questionnaire elicitinginformation about their political behavior and views,age, sex, and socioeconomic background. Subjects inthe study ranged in age from 19 to 49 with an averageage of 20.6 years. The subject pool was 41% womenand 56% of subjects identified themselves as white.

The dictator game involves two players. Thefirst player determines a split of some endowment,usually a sum of money, between herself and theother player. The second player simply receiveswhatever the first player has allocated to her andthus has no strategic role in the game. If the firstindividual were only concerned with her own well-being, she would keep the entire endowment and passnothing. Therefore, any positive allocation to thesecond player is viewed as altruistic behavior. For areview of the literature on the dictator game, seeCamerer (2003).

Subjects in our study were asked to play fivemodified dictator games, modeled after Andreoniand Miller (2002). In each dictator game, subjectswere asked to split 10 lottery tickets between them-selves and an anonymous recipient. Subjects wereinformed that every ticket they contributed would

give the recipient either 1, 2, or 3 chances to win theprize and every ticket they kept would give them-selves either 1, 2, or 3 chances to win the prize,depending on the game.3 The order of the games wasrandomized. This design serves to vary the relativeprice of giving in order to check decisions againstrationality conditions and determine the social pref-erence type of the subject.4

Subjects were informed that their participationmade them eligible to win $100. They were told that anumber corresponding to a ticket would be randomlydrawn and matched with the subject holding it. As

noted above, the drawing was weighted according tothe type of ticket held, with some giving the holderone chance, some giving them two chances, and somegiving them three chances. At the conclusion of thestudy, the winning number was announced by emailand a winner came forward with the winning ticket toclaim the $100 prize. Fowler (2006) has shown thatthis type of experimental design yields results con-sistent with those reported in experiments wheresubjects were asked to split a cash endowment. Ourresults are also very similar to Andreoni and Miller(2002) leading us to conclude the lottery prompts

subjects to treat the experiment as if they were beingpaid for a randomly chosen decision.

Though economists sometimes criticize low-stakes experiments like this one, a survey of theexperimental economics literature by Camerer and

3This thus gives us five payoff conditions, 3:1, 2:1, 1:1, 1:2, 1:3.

4The relative price of giving is simply the value of keeping theticket divided by the value of the ticket to the other player ifpassed. For example, if a ticket is worth 1 chance if kept and 3chances to the other player if passed, the relative price of giving is0.33.

social preferences and political participation 847


4/13

Hogarth (1999) shows that stake size has only a smalleffect on average behavior and the biggest effect ofstakes on behavior is changing from zero to positivestakes. Furthermore, Carpenter, Verhoogen, andBurks (2005) show specifically for the dictator gamethat changing from low stakes to high stakes has noeffect on mean allocations.

Finally, compared to verbal questions in which itis costless for a subject to succumb to social desir-ability, behavioral measures from choices made in thedictator game require subjects to pay a cost to revealfalse social preferences in the name of social desir-ability. By comparison, it is never financially costly inconventional survey research for a subject to lie abouttheir preferences. Accordingly, on the preferences side,at least, our design works against social desirabilitybetter than conventional survey questions. Moreover,we have no reason to believe that those who succumb

to social desirability would be more likely to revealdifferent kinds of social preferences. For example, it isnot clear whether egalitarian motives or motives tomaximize group welfare represent the most desirableoutcome perceived by subjects. As a result, desirabilityshould not bias efforts to classify subjects who expresssocial preferences.

Preferences

The modified dictator game provides two critical piecesof information. First, allocation choices (subject to thebudget constraint) in each relative price condition canbe analyzed to see if they meet the conditions requiredfor rationality. Over 94% of our subjects made con-sistent choices across the five games.5 Second, thesechoices can be used to determine each subjectspreference type. A set of rational choices can berepresented by a utility function and thus key param-eters of interest can be estimated econometrically inorder to provide more information about preferences.

ClassificationSubjects were classified into three utility types: selfish,Rawlsian, and utilitarian. Strong selfish types choose

according to the utility function Us(p

s, p

o) 5 ps,

strong Rawlsians6 choose according to Us(p

s, p

o) 5

min{ps, p

o}, and strong utilitarians choose according

to Us(p

s, p

o) 5 ps + po where ps is the subjects own

payoff in the dictator game, po

is the payoff to theother player, and U

sis the subjects utility. Optimal

utility for each player varies by type. These optimalutilities for each game are presented graphically inFigure 1. Notice that Rawlsians achieve optimalutility by giving progressively less as it becomes lessexpensive to give.7 Meanwhile, utilitarians shouldkeep everything when it is expensive and give every-thing away when it is cheap. In other words,Rawlsians achieve maximum utility by ensuring equalchances between themselves and the other player ineach game, while Utilitarians maximize utility byensuring the greatest number of total chances be-tween themselves and other player.

Those subjects not perfectly falling into one ofthese three categories are grouped into the categorythey most resemble based on the Euclidean distanceof their allocation choices from those of the strongutilitarian type.8 The breakdown of subjects intocategories is presented in Table 1. Pure subjects arethose who gave all of their endowment away in eachof the five games.

Our experiment yielded fewer subjects withutilitarian preference types and more selfish andRawlsian types than Andreoni and Miller (2002).They find 47.2% of their sample made up of selfish

utility-types, 30.4% Rawlsian, and 22.4% utilitarian.The discrepancy between our results and those ofAndreoni and Miller may be explained by samplingdifferences. Our study recruited subjects from avariety of social science majors, whereas Andreoni

5For details on the Weak Axiom of Revealed Preference (WARP)see Varian (1992). In short, subjects choices are in violation ofWARP if they choose the most expensive Rawlsian (or utilitarian)allocation while not choosing the less expensive Rawlsian (orutilitarian) allocation. Three subjects violated WARP in two ormore of their five decisions. Ten subjects violated WARP in oneof five decisions. Our results do not change when we excludeindividuals who violated WARP.

6Single tickets are not divisible. As a result, we cannot classify anindividual as a strong Rawlsian. This is because a strong Rawlsianwould keep 6.67 tickets when the value of given tickets was worthdouble and 7.5 when they were worth triple. She would give 3.33tickets when the value of kept tickets was worth double and 2.5when they were worth triple. There is no way of knowing if astrong Rawlsian subject is choosing to round up or down.

7This point is best illustrated using a numerical example. Givingto the other player is most expensive in our game when thefraction of the endowment kept is tripled and least expensivewhen the fraction sent is tripled. When giving is most expensive,a Rawlsian subject maximizes her utility when she keeps 2.5(assuming a perfectly divisible endowment), which is tripled, andpasses 7.5 (min{2.5 3 3, 7.5} 5 7.5). When giving is leastexpensive, a Rawlsian maximizes her utility when she keeps 7.5and passes 2.5, which is tripled (min{7.5, 2.5 3 3} 5 7.5).

8Because strong utilitarians may choose any allocation when therelative price is one, the Euclidean distance is calculated for avector of four allocation decisions in which the relative price isnot equal to one.



5/13

and Miller used only upper-level economics students.Fehr, Naef, and Schmidt (2006), in an experimentstudying preferences for efficiency versus equity, foundthat a majority of economics students chose efficientallocations over egalitarian ones, whereas a majority ofnoneconomics social science majors in the subjectpool favored allocations consistent with a desire toavoid inequality. It is likely that the significantly highernumber of utilitarians Andreoni and Miller found ispartially an artifact of their sampling technique.

We perform two further tests of our estimates of

subjects preferences, namely by econometrically es-timating a demand function for each utility type andby examining how each utility type is related to statedpreferences for equality. These further tests supportour classifications and can be found in the first twosections of the online appendix.

Political Participation

Subjects in our study were asked a series of yes/noquestions about what type of political activity they

engage in, as well as more detailed questions about

their political attitudes and interest in politics. Wefollow Verba, Schlozman, and Bradys (1995) con-struction of an index of overall political activity as theunweighted sums of yes/no responses. The compo-nents of the index are volunteering for a candidatescampaign, volunteering for a local government boardor council or attending regularly, initiating contactwith a government official, engaging in informalcommunity work, participating in a protest, contri-buting money to a candidate, being a member of orgiving money to an organization that takes stands onpolitical issues, and voting in the 2004 generalelection. Therefore, there is maximum of eight totalacts of political participation. We note that this indexcovers a range of conventional and nonconventional(or formal and informal) activities and is widely used(e.g., Burns, Schlozman, and Verba 2001). Moreover,it resembles earlier measures of participation (e.g.,

Verba and Nie 1972). It displays a respectably highinternal consistency within our sample (Cronbachsa 5 0.67).9

The study averages are: 28% of subjects volun-teered for a campaign, 30% volunteered for a localgovernment or council, 47% initiated contact with agovernment official, 39% did informal communitywork, and 44% participated in a protest, 23%contributed money to a candidate, 48% were mem-bers of or gave money to an organization that takesstands on political issues, and 75% voted in the 2004general election. Subjects participated in 3.35 activ-

ities on average. This compares closely with Verba,Schlozman, and Bradys average estimate of three actsof participation for college educated respondents(1995, 207).

A simple examination of the raw data, presented inTable 2, suggests that utilitarian participation is sig-nificantly greater than that of Rawlsian (p 5 0.04 one-tailed test), pure (p 5 0.02 one-tailed test), and selfish(p 5 0.005 one-tailed test) subjects. This analysis isonly suggestive of a relationship between social prefer-ences and political participation. An appropriate testmust include controls for other variables known to

influence political participation.

TABLE 1 Subject Types

Strong Weak Total Percent

Selfish 60 68 128 55.4Rawlsian 73 73 31.6Utilitarian 3 15 18 7.8Pure 12 12 5.2

Note: Subjects are classified according to the procedure describedin the appendix.

FIGURE 1 Optimal allocation to the other playerin each dictator game

9We also conducted a maximum-likelihood factor analysis onour eight measures of participation. This analysis revealed onlyone factor with an eigenvalue greater than one (1.64). Factorloadings of the scale components ranged from 0.26 (for vote) to0.54 (for campaign volunteering). Our results do not changesignificantly if we drop the weakest loading factor. Likewise, ourresults do not change significantly if we consider political andnon-political factors separately.



6/13

Regression Analysis

The first regression presented in Table 3 is a simplemodel with dichotomous variables for Rawlsian andutilitarian preferences, as revealed in decisions made inthe dictator game. We also include a dichotomous

variable for pure subjects, i.e., those who gave awaytheir entire endowment in all five games. The basecategory, not included in the model, is the selfish

utility type. Therefore, the participation among Rawl-sian and utilitarian preferences types is being com-pared to participation of selfish types. In other words,we are asking if Rawlsians and utilitarians participatesignificantly more than selfish preference types. Thesimple model shows that only the utilitarian prefer-ence type is significant (at the 5% level). We note thatthis significant result is despite the small number ofutilitarians present within our sample. Utilitarians alsoappear more likely to participate than Rawlsians(F(1,231) 5 2.95, p 5 0.04 one-tailed).

In order to control for other political, socioeco-nomic, and nonpolitical institutional affiliation varia-bles that may influence levels of participation, weinclude several variables studied by Verba, Schlozman,and Brady (1995; their book provides an exhaustivereview of these variables). These variables includepolitical controls for the respondents level of interest

in politics, how politically informed they are, theirfeelings about whether or not they have a say in whatgovernment does, their party identification and howstrongly they identify with the party; socioeconomiccontrols for whether or not their family income is at orabove the sample median, gender, ethnicity, and age;and religious attendance and contributions to controlfor nonpolitical activity. Even with these controlsadded and even given our small number of utilitarians,the coefficient on utilitarian preferences remains sig-nificant (p 5 0.03). However, we note that in thisspecification, the difference in participation between

utilitarians and Rawlsians is less clear (F(1,218)5 1.80,p 5 0.09, one-tailed). As was found by Verba,Schlozman, and Brady (1995), political interest exertsa strong (significant) influence on participation,whereas efficacy and strength of partisanship do not.Our results differ from Verba, Schlozman, and Bradys(1995) in that we do not find political information andfamily income to be significant. This difference is mostlikely due to the lack of variation present in these twovariables, given our convenience sample. Finally, wefind that that partisan strength is not related toparticipation, but that partisanship is related. Repub-

lican respondents are less likely to report engaging inpolitical participation than Democratic respondents.

A potential criticism of our analysis is that thereare only 18 utilitarian subjects in our sample. Alimited number of utilitarians makes it more difficultto detect differences in participation by preferencetype, due to a potential lack of power. But, as ourresults show, we do not encounter this problem. Wefind a significant difference between utilitarian andselfish voters even with a limited number of utili-tarians in our sample.

TABLE 3 Dependant Variable: Political ActivityIndex

Variable

Simple Model With Controls

Coef S.E. p Coef S.E. p

Social Preference Type

Rawlsian 0.53 0.32 0.10 0.47 0.29 0.10Utilitarian 1.52 0.55 0.01 1.14 0.48 0.02

Pure 0.08 0.66 0.90 0.27 0.59 0.64Political Variables

Political Interest 3.50 0.55 0.00High Political Info 0.08 0.31 0.80Internal Efficacy 0.44 0.48 0.35External Efficacy 0.53 0.61 0.39Partisan Strength 0.62 0.41 0.14Partisan ID 20.91 0.40 0.02

Citizen 0.54 0.64 0.40Socioeconomic Status

High Income 0.34 0.26 0.19Female 0.45 0.28 0.11White 0.56 0.26 0.03

Age 0.92 0.53 0.08Religious Activities

Attendance 0.93 0.52 0.08Contribution 0.02 0.30 0.94

(Intercept) 3.00 0.19 0.00 21.79 0.83 0.03

LogLik (model) 2496.4 2456.2

LogLik (null) 2500.7 2500.7

N 5 234

Note: The models are estimated using tobit regression. Allvariables are rescaled to [0,1] for ease of interpretation. SocialPreference Types are measured dichotomously.

TABLE 2 Participation Raw Averages

Number of Activities S.E.

Selfish 3.08 0.17Rawlsian 3.56 0.25Utilitarian 4.56 0.49Pure 3.08 0.51

Experiment Total 3.35 0.13

Note: Cell entries indicate the average number of politicalactivities undertaken by each subject type and their standarderrors.



7/13

Another objection is that valuable information isbeing lost by using dichotomous variables for eachutility type and that our results may be driven by oursmall number of pure utilitarians. To address thisobjection, we have reestimated our models usingcontinuous measures of each utility type. For eachmeasure, we calculate the Euclidean distance betweena purely selfish or altruistic allocation type and astrong Rawlsian or utilitarian type. According tothese measures, an individual with a 1 on Rawlsian-ism is the maximum possible distance from another

pure type, while an individual with a 0 on Rawlsian-ism is the maximum possible distance from the pureRawlsian type. The same logic applies for an indi-vidual with a 1 or 0 on utilitarianism. We presentthese results in Table 4. As can be seen, our results arerobust to this other specification.

The simple model and the model with controlsare very similar to those presented in Tables 2 and 3.The continuous measure of utilitarian preferences issignificant with and without controls but the measureof Rawlsian preferences is not in either case. A

likelihood ratio test also rejects adding a measure ofselfish preferences to the model (the null hypothesisthat the coefficient is zero fails to be rejected, p 50.17). Once again, utilitarians appear more likely toparticipate than Rawlsians (F(1,232)-2.83, p 5 0.04one-tailed) in the simple model. They also appearmore likely to participate in the model with controls(F(1,219) p 5 0.04 one-tailed). All of the controlvariables also maintain their signs and similar mag-nitudes and significance. The interpretation of theutilitarian coefficient is that as one gets closer tohaving pure utilitarian preferences, the more politi-cally active one becomes, controlling for other po-tentially significant influences.

Finally, we note that we present a number ofrobustness checks in the online appendix. First, wereplaced our revealed measures for utilitarian andRawlsian preferences with responses from the ques-

tion used to probe expressed preferences for equality.When added to the full model with controls, thecoefficient on this variable is insignificant (p 5 0.18),suggesting that the dictator game measures moreprecisely capture the components of altruism thatcontribute to political activity better than simpleexpressed preferences for equality.10 Second, wepresent Huber regressions to account for the effectsof possible outliers on our estimations. Third, wepresent regressions which consider each participatoryact separately. Finally, we present results with purealtruists excluded, to demonstrate that our results are

not driven by this classification. Our results holdacross each of these specifications.

Conclusion

Our research design allows us to map individualsrevealed social preferences to political activity, and weare able to show that utilitarian social preferencesmake one more likely to be politically active. Thissuggests that utilitarians view political activity as a

means to increase social welfare. Individuals withRawlsian preferences, on the other hand, are no morelikely than selfish individuals to participate in poli-tics. This is potentially because they do not viewpolitics as a means to help those least fortunate. Anarea for future research is to explore the nonpolitical

TABLE 4 Dependant Variable: Political ActivityIndex

Variable

Simple Model With Controls

Coef S.E. p Coef S.E. p

Rawlsianism 0.42 0.58 0.47 0.30 0.52 0.57

Utilitarianism 2.32 0.97 0.02 1.98 0.85 0.02Political Variables

Political Interest 3.60 0.55 0.00High Political Info 0.08 0.31 0.79Internal Efficacy 0.39 0.47 0.41External Efficacy 0.49 0.61 0.42Partisan Strength 0.51 0.41 0.21Partisan ID 20.99 0.40 0.01

Citizen 0.37 0.65 0.56Socioeconomic Status

High Income 0.37 0.65 0.56Female 0.49 0.28 0.08White 0.57 0.26 0.03

Age 0.97 0.52 0.06Religious Activities

Attendance 0.96 0.52 0.07Contribution 0.06 0.30 0.84

(Intercept) 2.39 0.42 0.00 22.11 0.86 0.02

LogLik (model) 2497.7 2456.7

LogLik (null) 2495.0 2495.0

N 5 234

Note: The models are estimated using tobit regression. Allvariables are rescaled to [0,1] for ease of interpretation. SocialPreference Types are measured continuously.

10Furthermore, the failure of this variable to significantly predictgreater participation suggests that our subjects are not succumb-ing to social desirability on either our behavioral measures ofsocial preferences or our dependent variable.



8/13

activities of utilitarians and Rawlsians, such as reli-gious or charitable contributions and volunteering, todetermine if this is indeed the case.

Our work has two clear limitations. First, we haveelicited just two types of social preferences, whileothers exist. Future work could distinguish furtherother variants, such as paternalist or nosy preferences(Sen 1976) or spitefulness (van Lange 1999). Second,our experiment was conducted on a conveniencesample of university students. The relationships wehave uncovered may exist only among similar pop-ulations. Future replications of our experiment onmore general populations would test the general-izability of our findings.

Our current results speak directly to existingwork on social preferences and political participation,especially given their conflicting results. For example,Fowler (2006) finds that altruism, as measured in a

single dictator game, has a positive but not significantdirect effect on the tendency to vote, while Fowlerand Kam (2006) find that general altruism andaltruism particular to certain groups, as revealed indictator games, both predict political activity. Finally,Loewen (2010) finds that general altruism, as meas-ured in dictator games, does not predict the decisionto vote, but altruism towards certain groups does. Healso finds that turnout increases as the size of groupsincreases. Our results qualify this previous work byhighlighting the importance of different types ofsocial preferences. Such differences cannot be uncov-

ered in a conventional dictator game. Since Rawlsianstend to split their allocation 50-50 in a normaldictator game, utilitarians might do the same sincethe total amount divided remains the same for anysplit. Thus, both the pool of self-interested subjects(those who give nothing) and the pool of altruisticsubjects (those who give something) might containutilitarians. Moreover, the pool of altruists containsRawlsians who apparently do not participate inpolitics any more than self-interested subjects do.The inability to distinguish these disparate motiva-tions means that a simple attempt to correlate giving

in a single dictator game with political participationis likely to show only a weak (if any) correlation.

Our finding that utilitarians participate morethan selfish individuals and Rawlsians do not alsocarries important implications for political theory,formal models of participation, cross-national em-pirical accounts of participation, and individual-levelaccounts of turnout. First, it supports Harsanyis twinarguments about the effect of redistributive andwelfare-maximizing social preferences on participa-tion. He noted that since utilitarians care less about

their own costs and more about policies that makeeveryone better off, they would be more likely to votethan purely self-interested individuals. He also ar-gued that Rawlsians would find politics a much lesssuitable domain for action than utilitarians because itis difficult to use sweeping policies to target redis-tribution to the least fortunate. As a result, they maynot participate because they view time spent onpolitical activity as a substitute for time spent onnonpolitical activities aimed at directly helping thepoor. The empirical evidence in this article supportsboth of these claims.

Second, the evidence also supports assumptionsmade in a recent game theoretic analysis of other-regarding preferences by Feddersen and Sandroni(2006). They assume that individuals are motivatedprimarily by aggregate welfare and they show that alarge fraction of a population of such individuals will

choose to participate. However, our results alsosuggest that the population is heterogeneous, withmost people exhibiting selfish preferences and othersexhibiting nonutilitarian other-regarding preferences.Although the Feddersen and Sandroni model predictsthat many utilitarians will participate, these individualsmay make up only a small fraction of the population.Our results may not be representative as they are basedon a laboratory experiment with college students, butthey do suggest that Rawlsians and self-interestedindividuals may make up a large and relatively non-participatory part of the population.

Third, our findings speak to work examining andexplaining cross-national variation in political par-ticipation, especially institutional explanations (e.g.,Jackman 1987; Powell 1982). If such work overlooksdifferences in the distribution of social preferenceswithin various populations, or does not take accountof important national differences in redistributivepolitics, then it risks overstating the importance offormal institutions in the explanation of varying ratesof political participation (Blais 2006). Our worksuggests that future analysis should pay greaterattention to the degree and type of redistribution

present in different countries.Fourth, our findings add to an expansive liter-

ature on individual-level determinants of politicalparticipation (e.g., Blais 2000; Finkel 1985; Hether-ington 1999; Leighley and Nagler 1992a, 1992b;Verba, Schlozman, and Brady 1995). This literaturehas catalogued an exhaustive list of individual-leveldifferences that correlate with voter turnout, such thatwe know the typical effects of income, education, age,gender, martial status, political interest, information,trust, religious attendence, and scores of other



9/13

variables. We add to this work by demonstrating thatother individual-level differences, namely differentsocial preferences, also account for variation in polit-ical participation. They do not stand in for these otherdeterminants, as the effects of control variables show.Instead, they explain an element of the decision to votelargely uncovered by this empirical literature. More-over, our article suggests an experimental frameworkmay be best suited for uncovering such additionaldeterminants.

An area of future research is to explore theproperties of social preferences, including their originsand malleability. For example, are social preferencesfixed or able be changed under certain conditions andover the course of the life cycle (see Fehr, Bernhard,and Rockenbach 2008)? Also, are different socialpreference types learned or do they have a morefundamental basis? Research in behavioral genetics,

using both twin study (e.g., Cesarini et al. 2009), andgenetic association designs (Knafo et al. 2008), hasshown a genetic component to the other regardingbehavior revealed in economic games. Searching for agenetic basis for different social preferences is apromising avenue of research and, based on the resultswe present here, has important implications for thestudy of political behavior.

Accounting for actors with social preferences hasdone much to advance our understanding of thephenomenon of widespread political participation.However, as our results have shown, there is theo-

retically and empirically important variation in socialpreferences. The puzzle of participation may persistuntil we can understand how people with differentkinds of preferences are motivated to take part inpolitical life. What is more certain is that suchpreferences should form a starting point for theexplanation of political participation.

Appendix

Political Activity Index

The Political Activity Index is comprised of eight actsof participation. They include: Vote: In talking topeople about elections, we find that they are some-times not able to vote because they are not registered,they do not have the time, or they have difficultygetting to the polls. Did you happen to vote in the2004 general election in November? Volunteer Can-didate: Have you ever as a volunteerthat is, for nopay at all or for only a token amountfor a

candidate running for national, state, or local office?Contribute to a Candidate: Have you ever contrib-uted money to an individual candidate, a partygroup, a political action committee, or any otherorganization that supports candidates in elections?

Join a Political Organization or Donate to a PoliticalOrganization: Are you a member of any organiza-tions that take stands on any public issues, eitherlocally or nationally? or Not counting membershipdues, have you given money to any organizations thattake stands on any public issues, either locally ornationally? Attend Local Board Meetings or VolunteerLocal Board: Have you ever regularly attendedmeetings of an official local government board orcouncil? or Have you ever served in a voluntarycapacity, that is, for no pay at all or for only a tokenamount, on any local governmental board or councilthat deals with community problems and issues such

as a town council, a school board, a zoning board, aplanning board, or the like? Informal CommunityWork: Aside from membership on a board orcouncil or attendance at meetings, have you everinformally gotten together with or worked withothers in your community or neighborhood to tryto deal with some community problem? GovernmentContact: Aside from contacts made as a regular partof your job, have you ever initiated any contacts witha government official or someone on the staff of suchan official, such as someone in the White House, aCongressional or Senate Office, or a federal agency

like the EPA or IRS? Protest: Have you ever takenpart in a protest, march, or demonstration on somenational or local issue (other than a strike against

your employer)? (0 5 No, 1 5 Yes).

TABLE A1 Experiment means for components ofthe political activity index by

preference classification

Selfish Rawlsian Utilitarian Pure

Vote 0.76 0.79 0.83 0.42Board 0.28 0.29 0.50 0.08Volunteer 0.26 0.27 0.56 0.17Contact 0.48 0.41 0.61 0.50Informal 0.36 0.40 0.61 0.33Protest 0.34 0.52 0.61 0.67

Organization 0.44 0.56 0.50 0.58Contribute 0.16 0.32 0.33 0.33



10/13

TABLE A2 N 5234. Components of the Political Activity Index are estimated using logit regression

Vote Board Volunteer Contact Informal Protest

Coeff p Coeff p Coeff p Coeff p Coeff p Coeff p

Rawlsian 0.46 0.31 0.00 1.00 0.12 0.77 20.25 0.46 0.08 0.80 0.80 0.02

Utilitarian 0.17 0.82 0.84 0.14 1.17 0.04 0.39 0.49 0.97 0.07 1.14 0.04Pure 21.98 0.03 22.20 0.06 20.47 0.61 0.44 0.56 20.25 0.72 1.49 0.04Political

Interest 1.63 0.07 2.45 0.00 3.14 0.00 1.68 0.01 1.76 0.01 2.00 0.00Info 0.13 0.79 20.43 0.28 0.36 0.43 0.49 0.18 20.24 0.49 0.33 0.35Int Eff 0.70 0.34 0.38 0.55 0.26 0.70 0.01 0.98 0.05 0.93 0.24 0.67Ext Eff 2.29 0.01 1.09 0.18 1.50 0.10 0.38 0.59 20.23 0.75 20.55 0.43Strength 0.63 0.30 20.38 0.48 1.42 0.02 0.69 0.15 20.79 0.10 20.21 0.67

ID2

0.70 0.292

1.54 0.01 0.56 0.272

0.01 0.982

1.18 0.022

0.93 0.06Citizen 18.94 0.99 20.26 0.80 22.15 0.02 15.85 0.99 0.31 0.74 21.36 0.15SESIncome 0.66 0.10 20.12 0.72 0.38 0.28 0.44 0.14 0.10 0.73 20.24 0.43

Female 1.09 0.02 20.22 0.54 0.97 0.01 0.29 0.38 0.25 0.44 20.02 0.96White 0.98 0.02 0.30 0.38 0.14 0.70 0.35 0.25 0.52 0.09 0.02 0.96Age 0.95 0.24 0.79 0.24 0.84 0.25 0.65 0.29 20.77 0.21 0.49 0.43ReligiousAttend 1.36 0.11 1.47 0.03 21.71 0.02 0.45 0.47 1.18 0.06 0.37 0.55Contrib 21.19 0.01 0.19 0.62 0.81 0.05 20.12 0.73 0.05 0.89 20.17 0.63(Intercept) 222.23 0.99 22.98 0.02 24.67 0.00 218.98 0.98 21.07 0.35 20.33 0.77

LL (model) 289.05 2119.77 2110.44 2140.98 2141.55 2141.24

LL (null)2

127.942

139.112

137.272

159.632

154.442

158.29

All variables are rescaled to [0,1] for ease of interpretation.


11/13

References

Andreoni, J., and J. Miller. 2002. Giving According to Garp: AnExperimental Test of the Consistency of Preferences forAltruism. Econometrica 70 (2): 73753.

Andreoni, J., and L. Vesterlund. 2001. Which is the Fair Sex?Gender Differences in Altruism. Staff General ResearchPapers 1951. Iowa State University.

Blais, A. 2000. To Vote or Not Vote: The Merits and Limits ofRational Choice Theory. Pittsburgh, PA: University of Pitts-burgh Press.

Blais, A. 2006. What Affects Voter Turnout? Annual Review ofPolitical Science 9: 11125.

Burns, N., K. Schlozman, and S. Verba. 2001. The Private Roots ofPublic Action. Cambridge, MA: Harvard University Press.

Camerer, C. 2003. Behavioral Game Theory: Experiments inStrategic Interaction. Princeton, NJ: Princeton University Press.

Camerer, C., and R. Hogarth. 1999. The Effects of FinancialIncentives in Experiments: A Review and Capital-LaborProduction Framework. Journal of Risk and Uncertainty 19(13): 742.

Carpenter, J., E. Verhoogen, and S. Burks. 2005. The Effects ofStakes in Distribution Experiments. Economic Letters 86 (2):15262.

Cesarini, D., C. Dawes, M. Johannesson, P. Lichtenstein, andB. Wallace. 2009. Genetic Variation in Preferences for Givingand Risk Taking. Quarterly Journal of Economics 124 (2): 80942.

Charness, G., and M. Rabin. 2002. Understanding SocialPreferences with Simple Tests. Quarterly Journal ofEconomics 117 (3): 81769.

Craig, S., R. Niemi, and G. Silver. 1990. Political Efficacy andTrust: A Report on the NES Pilot Study Items. PoliticalBehavior 12 (3): 289314.

Edlin, A., A. Gelman, and N. Kaplan. 2007. Voting as a RationalChoice: Why and How People Improve the Well-Being ofOthers. Rationality and Society 19 (3): 293314.

Feddersen, T., and A. Sandroni. 2006. A Theory of Participationin Elections. American Economic Review 96 (4): 127182.

Fehr, E., H. Berhard, and B. Rockenbach. 2008. Egalitarianismin Young Children. Nature 454: 107983.

Fehr, E., M. Naef, and K. Schmidt. 2006. Inequality Aversion,Efficiency, and Maximin Preferences in Simple DistributionExperiments: Comment. American Economic Review 96 (5):191217.

Finkel, S. E. 1985. Reciprocal Effects of Participaton andPolitical Efficacy: A Panel Analysis. American Journal ofPolitical Science 29 (4): 891913.

Fiorina, M. 1990. Information and Rationality in Elections. InInformation and Democratic Processes eds. J. Ferejohn andJ. Kuklinski. Urbana-Champaign: University of Illinois Press,32942.

Fisman, R., S. Kariv, and D. Markovits. 2007. IndividualPreferences for Giving. American Economic Review 97 (5):185876.

Fowler, J. 2006. Altruism and Turnout. Journal of Politics68 (3): 67483.

Fowler, J., and C. Kam. 2006. Beyond the Self: Social Identity,Altruism, and Political Participation. Journal of Politics69 (3): 81327.

Gelman, A., J. Katz, and J. Bafumi. 2004. Standard VotingPower Indexes Dont Work: An Empirical Analysis. British

Journal of Political Science 34 (4): 65774.

Gelman, A., G. King, and W. Boscardin. 1998. Estimating theProbability of Events That Have Never Occurred: When isYour Vote Decisive? Journal of American Statistical Associa-tion 93: 19.

Harsanyi, J. 1975. Can the Maximin Principle Serve as a Basisfor Morality? A Critique of John Rawlss Theory. American

Political Science Review 69 (2): 594606.Harsanyi, J. 1980. Rule Utilitarianism, Rights, Obligations and

the Theory of Rational Behavior. Theory and Decision 12 (2):11533.

Hetherington, M. J. 1999. The Effect of Political Trust on thePresidential Vote, 1968-96. The American Political ScienceReview 94 (2): 31126.

Jackman, R. 1987. Political Institutions and Voter Turnout inIndustrial Democracies. American Political Science Review81 (2): 405424.

Jankowski, R. 2002. Buying a Lottery Ticket to Help the Poor:Altruism, Civic Duty, and Self-Interest in the Decision toVote. Rationality and Society 14 (1): 5577.

TABLE A3 Simulated first differences and 95% confidence intervals based on the regressions in AppendixTable 2

Rawlsian Utilitarian

First Difference 2.5% 97.5% First Difference 2.5% 97.5%

Vote 0.00 0.00 0.05 0.00 20.01 0.01

Board 0.012

0.12 0.15 0.172

0.07 0.43Volunteer 0.00 20.12 0.14 0.25 0.01 0.53Contact 20.03 20.17 0.11 0.08 20.16 0.32Informal 0.01 20.14 0.15 0.20 20.03 0.43Protest 0.18 0.02 0.32 0.24 0.02 0.49Organization 0.10 20.05 0.26 20.01 20.22 0.23Contribute 0.13 0.02 0.29 0.12 20.03 0.38

The first difference is the change in dependent variable resulting from a switch from selfish preferences to either Rawlsian (first threecolumns) or Utilitarian (second three columns) preferences. All other variables are held at their means. The reported results are basedon 1000 simulations.



12/13

Jankowski, R. 2007. Altruism and the Decision to Vote:Explaining and Testing High Voter Turnout. Rationalityand Society 19 (1): 534.

Knafo, A., S. Israel, A. Darvasi, R. Bachner-Melman, F. Uzefovsky,L. Cohen, E. Feldman, E. Lerer, E. Laiba, Y. Raz, L. Nemanov,I. Gritsenko, C. Dina, G. Agam, B. Dean, G. Bornstein, andR. Ebstein. 2008. Individual Differences in Allocation of Fundsin the Dictator Game Associated with Length of the Arginine

Vasopressin 1A Receptor RS3 Promoter Region and CorrelationBetween RS3 Length and Hippocampal MRNA. Genes, Brainand Behavior 7 (3): 26675.

Lassen, D. D. 2005. The Effect of Information on VoterTurnout: Evidence from a Natural Experiment. American

Journal of Political Science 49 (1): 10318.

Leighley, J. E, and J. Nagler. 1992a. Individual and SystematicInfluences on Turnout: Who Votes? 1984. Journal of Politics54 (3): 71840.

Leighley, J. E., and J. Nagler. 1992b. Socioeconomic Class Bias inTurnout, 1964--1988: The Voters Remain the Same. Ameri-can Political Science Review 86 (3): 72536.

Loewen, P. J. 2010. Affinity, Antipathy and Political Participa-

tion: How Our Concern for Others Makes Us Vote.Canadian Journal of Political Science 43 (3): 66187.

Mulligan, C., and C. Hunter. 2003. The Empirical Frequency ofa Pivotal Vote. Public Choice 116 (12): 3154.

Niemi, C., S. Craig, and F. Mattei. 1991. Measuring InternalPolitical Efficacy in the 1988 National Election Study.

American Political Science Review 85 (4): 140712.

Powell, G. B. 1982. Comparative Democracies: Participation,Stability and Violence. Cambridge, MA: Harvard UniversityPress.

Sen, A. 1976. Liberty, Unanimity and Rights. Economica 43(171): 21745.

Timpone, R. 1998. Structure, Behavior, and Voter Turnout inthe United States. American Political Science Review 92 (1):14558.

Van Lange, P. 1999. The Pursuit of Joint Outcomes and Equalityin Outcomes: An Integrative Model of Social Value Orienta-tion. Journal of Personality and Social Psychology 77 (2):33749.

Varian, H. 1992. Microeconomic Analysis. New York: W. W.Norton & Company.

Verba, S., and N. Nie. 1972. Participation in America. New York:Harper and Row.

Verba, S., K. Schlozman, and H. Brady. 1995. Voice and Equality:Civic Volunteerism in American Politics. Cambridge, MA:Harvard University Press.

Chris Dawes is a Ph.D. candidate in the Depart-

ment of Political Science at University of CaliforniaSan Diego, La Jolla CA 92093.

Peter John Loewen is an Assistant Professor ofPolitical Science at the University of Toronto atMississauga, Mississauga, Ontario L6L 1C6.

James H. Fowler is a Professor of MedicalGenetics and Political Science at the University ofCalifornia, San Diego, La Jolla CA 92093.



13/13

Copyright of Journal of Politics is the property of Cambridge University Press and its content may not be copied

or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.

However, users may print, download, or email articles for individual use.

social preferences and political participation

Documents