Beyond the Self: Social Identity, Altruism, and Political Participation

15 Pages·2007·0.34 MB·English

Beyond the Self: Social Identity, Altruism, and
Political Participation
James H. Fowler University of California, San Diego
Cindy D. Kam University of California, Davis
Scholarshaverecentlyextendedthetraditionalcalculusofparticipationmodelbyaddingatermforbenefitstoothers.
Weadvancethisworkbydistinguishingtheoreticallyaconcernforothersingeneral(altruism)fromaconcernfor
othersincertaingroups(socialidentification).Wepositthatbothconcernsgenerateincreasedbenefitsfrompartici-
pation.Totestthesetheories,weuseallocationsindictatorgamestowardsanunidentifiedanonymousrecipientand
two recipients identified only as a registered Democrat or a registered Republican. These allocations permit a
distinctionbetweenaltruismandsocialidentification.Theresultsshowthatbothaltruismandsocialidentification
significantlyincreasepoliticalparticipation.Theresultsalsodemonstratetheusefulnessofincorporatingbenefitsthat
stemfromsourcesbeyondmaterialself-interestintorationalchoicemodelsofparticipation.
“Avarice, or the desire for gain, is a universal passion viduals consider benefits to others, beyond the self,
which operates at all times, in all places, and upon all when deciding whether or not to participate. Even
persons”(Hume,OftheRiseandProgressoftheArtsand
Downs,sooftenportrayedasthearchetypalchampion
Sciences,[1742]1991,113)
of self-interest as a motivating factor of political
choice,states that a concern for the welfare of others
A
large body of work, ancient and modern, might influence political attitudes and behaviors:“In
posits that self-interest is the primal force for reality, men are not always selfish, even in politics.
politicalattitudesandbehaviors.Althoughthis They frequently do what appears to be individually
parsimonious assumption explains many observed irrationalbecausetheybelieveitissociallyrational—
politicalphenomena,itfailstoilluminatesomeof the i.e., it benefits others even though it harms them
most important features of political life (Citrin and socially”([1957] 1985,27).Extending the foundation
Green1990;Mansbridge1990;SearsandFunk1991). forpoliticalchoicebeyondtheself isnotaneasytask.
For example, rational choice scholars have typically It forces analysts to confront an important question:
approached the problem of political participation by When individuals decide whether and how to act, to
using models based on pure self-interest (Aldrich whom do they refer? On whose benefit will they act?
1993;Downs[1957]1985;FeddersenandPesendorfer Inthisarticle,wedistinguishtheoreticallytwodif-
1996; Ledyard 1982; Palfrey and Rosenthal 1985). ferent kinds of other-regarding considerations that
These models encounter a well-known difficulty: influence political participation. Some people are
although an individual may derive personal benefits motivated by social identification, which creates a
fromacertainpoliticaloutcome,theprobabilitythata desire to improve the welfare of certain groups in
single act of participation will significantly affect the society,possiblyattheexpenseofothergroups.These
outcomeisverysmallinlargepopulations.Thisgives individuals will likely participate when they believe
individuals an incentive to avoid the costs of partici- that their actions will give them an opportunity to
pation and free ride on the efforts of others,produc- help their preferred group(s).Other people are moti-
ing the well-known paradox of participation. vatedbyaltruism,awillingnesstopayapersonalcost
If self-interestdoesnotmotivatepoliticalpartici- to provide benefits to others in general, regardless of
pation, then what does? One possibility is that indi- theidentityofthebeneficiaries.Theseindividualswill
TheJournalofPolitics,Vol.69,No.3,August2007,pp.813–827
©2007SouthernPoliticalScienceAssociation ISSN0022-3816

  .    . 
likely participate when they believe that their actions nationoftheimpactofpartisanshipondictatorgame
willgivethemanopportunitytomakeeveryonebetter allocations, this work should be of interest to behav-
off. Our core expectation is that altruists and social ioralandexperimentaleconomists.Itshouldalsobeof
identifiers will participate more than egoists—that is, interest to psychologists and sociologists, since our
individuals who are primarily self-interested. uniquely designed dictator game provides a novel
Wetestthesocialidentifierandaltruismtheoriesof meansoftappingsocialidentity.Mostexistingworkon
participationusingauniqueexperimentaldesign.Sub- socialidentificationdoesnotforceindividualstosac-
jectsareaskedanumberofstandardquestionsregard- rificetheirownmaterialwell-beinginordertoaffirm
ingtheirsocioeconomicstatus,politicalattitudes,and support for their ingroups, but in the dictator game,
participationbehavior.Theythenplaythree“dictator” socialidentifiersmustdeliberatelydeprivethemselves
games(Forsytheetal.1994),inwhichtheydivideaset ofpersonalrewardssothattheycanaffirmtheposition
of lottery tickets between themselves and an anony- ofsomeoneelseintheirgroup.Inourdesign,affirming
mous individual. The recipient is completely anony- social identification has a cost. We demonstrate that
mousinallthreegames.However,intwoofthegames, subjects are in fact willing to bear this cost, and we
subjectsareinformedthattherecipientisaregistered demonstratethepoliticalconsequencesof thisbehav-
Democrat or a registered Republican. We use these ior.Finally,ourworkshouldbeof interesttopolitical
dictator games to uncover the degree to which each scientists,sincewenotonlyintroduceaninnovationin
subjectisgenerallyconcernedaboutthewell-beingof the measurement of dispositions towards groups and
others,asevidencedbyallocationstotheunidentified others in general, but we also identify the political
anonymous recipient, and the degree to which each implications of these dispositions by using them to
subject socially identifies with the political parties,as predict political participation. Our work therefore
shownbyallocationstotheDemocratandRepublican. allowsustoaddresstheliteratureonrationalchoiceby
These experiments yield several novel results for demonstratingthatthecoremotivationalelementsof
behavior in the dictator game and its relationship to rationalchoicetheoryneednotrestentirelyorsolelyon
politicalparticipation.First,weshowthatbehaviorsin self-interest, that other-regarding behavior can and
these dictator games reveal a key characteristic of shouldbetakenintoaccount,andthatrationalityinno
socialidentification:apreferencefortheingroupversus obviousornecessarywayrequiresmaterialself-interest
the outgroup. Democrats and Republicans both give tobeprivilegedastheprimarymotivatorinmodelsof
more to the recipient from their own party than the politicalbehavior.
opposingparty;independentsgivemoretotheanony-
mousrecipientthanthepartisanrecipients,whilepar-
Social Identity,Altruism,
tisansdojusttheopposite.Second,behaviorsinthese
and Participation
dictator games reveal that strength of social identity
magnifies preferences for the ingroup. Subjects who
identify themselves as strong Democrats and strong Traditional rational models of participation based on
Republicans tend to give much less to the recipient self-interest posit that individuals receive a benefit B
from the opposing party than partisans identifying fromsomepoliticalactivityiftheirpreferredoutcome
with weaker affiliations. Third, we uncover a bias occurs.However,the participatory acts that yield this
againstRepublicans.TheRepublicanrecipienttendsto outcome are individually costly (e.g., Aldrich 1993;
receive less than the Democrat or the unidentified Downs[1957]1985;FeddersenandPesendorfer1996;
anonymous recipient, even when the donor is a Ledyard1982;PalfreyandRosenthal1985).Thestick-
Republican. Finally, both altruism and social identity ing point for these models is that a single act of par-
increase political participation.People who share with ticipationusuallyhasonlyaverysmallprobabilityPof
ananonymousindividualinthedictatorgamepartici- affecting some political outcome. For example, if the
pate in politics more than those who do not share. participatoryactisvoting,thentheoutcomecanonly
People who vary the amount they give depending on bechangedwhenthereisanexacttie,orwhenthevote
thepartisanaffiliationoftherecipientalsoparticipate can create a tie.If the participatory act is a contribu-
more than those who give (or withhold) the same tionofmoneyortimetoacandidateorpoliticalorga-
amountto(from)everyone.Theseresultssuggestthat nization, it may be just one of thousands or even
other-regarding behavior plays an important role in millions of other contributions. Thus, the expected
the decision to participate. benefit of participation PB is typically less than the
Our work has broad implications for existing cost C, even when populations are not too large and
scholarshipinseveralfields.Sinceitisthefirstexami- even when the cost of participation is very low.
  :  , ,    
Riker and Ordeshook’s (1968) D term seems to actof participatingwillhavenoinfluenceonbenefits
offer one solution to the paradox of voting. The D derived from policy outcomes. In contrast, we argue
termsuggeststhatindividualswhoparticipateinpoli- thataltruismandsocialidentitywillencouragepoliti-
tics derive a benefit associated with the act of voting, cal action in order to benefit others,generally or spe-
resulting from satisfying a sense of citizen obligation, cifically;altruism and social identity affect B.
affirming their allegiance to the political system or According to social identity theory, individuals
reinforcing their own sense of efficacy. This benefit yearn to acquire and maintain a positive self-identity
associated with completing the act of voting is (Tajfel 1981). This sense of self is derived in large
orthogonal to the benefits derived from the policy part from formal membership with or psychological
outcomeofthepoliticalaction.Thus,politicalpartici- attachment to social groupings. In contrast with a
pation is an expressive act in which the desired policy theory based purely on self-interest, social identity
outcomesareessentiallyirrelevantintheparticipation theorysuggeststhatindividualsgainutilityfromaffili-
calculus,given how small P and B are.1 ating with social groups, from bestowing benefits
TheDtermprovidesoneanswertotheparadoxof upon the ingroup, and from withholding benefits
voting, but it is not the only answer. We argue that from the outgroup. Social identity theory resonates
citizenscanconsiderpoliticalactiontobeinstrumen- with Converse’s (1964) observation that the funda-
tal not only for themselves but for others as well. mental way in which many citizens understand poli-
Empiricalresearchsuggeststhistobethecase:activists tics is through groups. Social identity theory implies
frequently participate in politics in order to enact that individuals will make political choices by using
changes in public policy—that is,they act for instru- specificgroupsratherthantheselfasareferencepoint.
mental reasons—no matter how “irrational” this Social identity predicts policy preferences (Campbell
motivation seems (Schlozman, Verba, and Brady etal. [1960] 1980; Kinder and Winter 2001; Price
1995).Further,activistsfrequentlynotethatthepoliti- 1989), and under some conditions, social identity
cal stakes of participation affect individuals beyond spurscollectiveaction(forareview,seeHuddy2003).
themselves and their families (Schlozman,Verba,and So far, however, the literature has not linked social
Brady 1995).That is,they act instrumentally,not just identitywiththepolicy-orientedbenefitsofparticipa-
fortheirownbenefit,butforthebenefitof others.As tion in an attempt to address the paradox of
such, the policy outcomes of political actions should participation.2
affect individual decision making. We explicitly We argue that social identifiers may be spurred
address the possibility that an individual may care into political action when they believe that political
about the impact of policies as these policies apply outcomes will positively affect members of their
beyondtheself.Wedosobyincorporatingsocialiden- group. When individuals perceive political outcomes
tity and altruism into the calculus of participation. asdistributive—asopportunitiestotransferresources
Note that the benefits associated with altruism and from outgroups to their ingroup—social identifiers
social identification are distinct from those captured should be more likely to participate than individuals
by the D term. The D term can be conceived of as who are self-interested. Moreover, as people identify
systemaffirmationorfulfillmentofamoralobligation more strongly with their ingroup or more strongly
toparticipate.Moreover,theDtermisindependentof againstsomeoutgroup,theyshouldexperiencegreater
political outcomes—people with a strong sense of benefits from distributive politics and thus be more
socialobligationwillparticipateeveniftheythinkthe likely to participate.
While social identity theory suggests that indi-
viduals partition the world into ingroups and out-
groups, in a wide range of contexts, human beings
1RikerandOrdeshook’s(1968)approachisdecision-theoreticand
basedontheassumptionthattheDtermisexogenous.However,
tworecentattemptstoendogenizetheDterminagametheoretic
modelshowthat“ethical”preferencescanhelptoexplainturnout
evenwhenvotersarewellinformedandfullystrategic(Coateand 2One exception worth noting is Uhlaner’s (1989) treatment of
Conlin 2004; Feddersen and Sandroni 2006a, 2006b). These group members, group leaders, and candidates in her formal
modelssuggestthatvotersactassocialplannersbytryingtomaxi- modelof turnout.Shearguesthatgroupleaderscanmanipulate
mizesocialwelfare,andtheygainutilityfrom“doingtheirpart.” thecostsandbenefitsofvoting,e.g.,throughostracismofabstain-
However,bothofthesemodelsassumethatvoterspreferthelowest ers or social invitations directed at compliers. This approach
turnoutpossible,andneitherofthesemodelsconsidersthepossi- differs from ours because group leaders manipulate benefits
bilitythatvotersmightcareaboutthedistributiveimplicationsof obtained from the act of voting—where the act of voting is still
politicaloutcomesthatprovidebenefitstosomegroupsatacostto expressiveandnotinstrumental,andtheinstrumentalfunctionsof
others. votingremainuntouchedbytheactionsofgroupleaders.
  .    . 
have been observed to be motivated by the welfare of Finding Altruists and Social
othersingeneral(FehrandFischbacher2003;Monroe Identifiers among Dictators
1996;Piliavin and Charng 1990).They engage in acts
of altruism,or“behavior intended to benefit another,
Our study contributes to existing empirical work by
evenwhenthisriskspossiblesacrificetothewelfareof
adoptinganinnovativemeasureofaltruismandsocial
the actor” (Monroe 1996, 6). In contrast with social
identity. Previous attempts to examine the relation-
identifiers,altruistsdonottypicallytargetindividuals
shipbetweenother-regardingbehaviorandparticipa-
from certain groups for benefits. Monroe (1996)
tionhavereliedonquestionsintheNationalElection
explains thatindividualswhoarewillingtoengagein
Study (NES) pilots.Knack (1992) creates an index of
uncommonactsofaltruismexpressasenseofuniver-
“socialaltruism”fromquestionsaboutcharity,volun-
salism in viewing the human condition. Instead of
teer work, and community involvement on the 1991
viewing an individual (and the self, in particular) as
NES Pilot Study and finds a positive relationship
tiedtospecificsocialgroupings,altruists“shareaview
between the index and voter turnout. However, the
of theworldinwhichallpeopleareone”(198).Thus,
questionsusedintheindexareveryclosetothoseused
while social identifiers are more likely to help
by scholars who argue that organizational involve-
membersoftheiringroup,altruistsareunlikelytodis-
ment (not altruism) enhances political participation
criminate in whom they help.3
(Verba, Schlozman, and Brady 1995). Jankowski
Scholarshaverecentlyincorporatedaltruisminto
(2007)findsarelationshipbetweenvoterturnoutand
the traditional calculus of participation model by
“humanitarian”norms(i.e.,agreementwiththestate-
assumingthateachcitizenalsocaresaboutthebenefits
ment that“One of the problems of today’s society is
thatotherssecurefromthepreferredoutcome(Edlin,
that people are often not kind enough to others.”).
Gelman, and Kaplan 2007; Fowler 2006; Jankowski
These questions certainly reflect expectations about
2002, 2007). Although a single participatory act may
thealtruismofothers,butitisnotclearhowtheyrelate
havelittleeffectonapoliticaloutcome,thenumberof
to the respondent’s own willingness to bear costs to
people who benefit may be quite large. Thus, those
provide benefits to others. Typical measures of social
who exhibit a sufficient degree of concern for the
identification rely upon self-reports (Kinder and
welfare of others will be willing to engage in costly
Winter 2001) or are based on group membership
political participation. Moreover, as people become
(Price 1989).
moreconcernedforthewelfareof others,theyshould
The above studies rely on respondents’ expressed
experience greater benefits when political outcomes
preferences for helping others generally or for identi-
portend improvements for the welfare of others gen-
fying with a group. In neither case do respondents
erally.Thus,altruistswillbemorelikelytoparticipate
actuallyexperienceacostinordertogiveabenefitto
than individuals who are self-interested.4
someone else. In contrast, preferences for helping
Political outcomes might be construed by indi-
others are revealed in what experimental economists
viduals as improving the general welfare and/or as
call the“dictator game”(Forsythe etal.1994).In this
favoringparticularsocialandpoliticalgroups.Conse-
game, the experimenter gives player 1 a certain
quently,the decision to participate in politics may be
amountof moneyandthenasksthesubjecttodivide
motivated by both a desire to make things better for
thatmoneybetweenherselfandplayer2.5Ifplayer1is
everyone(altruism)andadesirespecificallytoacquire
motivatedonlybyherowneconomicgain,sheshould
as many benefits as possible for the ingroup (social
keepallthemoneyforherself andallocatenothingto
identification). Thus the benefit from participation
player 2. However, this is not what players normally
may be derived by some combination of self-interest,
do. In a survey of dictator game results, Camerer
altruism,and social identity.
(2003) shows that the mean allocation to player 2
rangesfrom10%to52%.Anonymityconditionstend
todecreasethemeanallocation,buteveninthemost
anonymous treatments (Hoffman etal. 1994) about
3Notethatwedonotaddresstheevolutionaryorsocialoriginsof
40% of the allocations still exceed 0.
altruism and other forms of nonself-interested behavior as has
beendoneelsewhere(e.g.,Samuelson1993).Ourprimaryinterest
hereisinhowvariationinaltruismandsocialidentificationpre-
dictspoliticalparticipation.
5Unlike the ultimatum game (cf.Hibbing and Alford 2004),the
4Foraformalizedsketchofhowaltruismandsocialidentitymight dictatorgamedoesnotgiveplayer2anopportunitytoacceptor
beinsertedintotheclassicparadoxofvotingmodel,seetheonline rejecttheoffer.Inthedictatorgame,player2simplypocketsthe
appendixathttp://journalofpolitics.org/articles.html. moneythatplayer1allocatestoherandthegameisover.
  :  , ,    
Interpretations of Excess Giving in the Inaningeniousdesignthatcomparesthealtruism
Dictator Game and fairness interpretations, Andreoni and Miller
(2002) examine choices in a series of dictator games
Excessgivingindictatorgamesisareplicableempiri- with different payoffs.In some treatments,player 2 is
cal regularity. Scholars offer several explanations for given $.20 or $.30 for every $.10 player 1 allocates.In
thisexcess.Weadoptthemostprominentexplanation: other treatments, player 1 must allocate $.20 or $.30
altruism—that individuals engage in “other- forevery$.10player2receives.Byvaryingthepayoffs,
regarding” behavior. The altruism explanation sug- Andreoni and Miller are able to distinguish between
geststhatdictatorsgivetoothersbecausetheywantto individuals who give in order to equalize payoffs
improve the well-being of other individuals, even (whom they call“Rawlsians”) and those who give in
when doing so impinges on their own material inter- ordertomaximizetotalpayoffstobothplayers(whom
ests. In his thorough overview of dictator games, they call “utilitarians”). The results show that about
Camerernotesthat“thereissomepurealtruism”that two-thirds of those who incorporate the recipient’s
explainsexcessgiving(2003,56).Forexample,intheir utility in their decision can be described as
studyofaltruismanddictatorgames,EckelandGross- “utilitarians.” Thus, while a concern for fairness
man (1996) manipulate the target of the giving; they undoubtedly plays an important role, altruism, or
find that subjects are much more likely to give when consideration of others’ welfare, appears to be the
the target is the Red Cross.Eckel and Grossman con- dominant motivation behind giving in the dictator
clude that“altruism is a motivating factor in human game.
behavior in general and in dictator games in particu- Another explanation for excess giving rests on
lar”(1996,182). the notion of reciprocity. Hoffman etal. (1994) and
A companion explanation for excess giving is a Hoffman, McCabe, and Smith (1996a) argue that
taste for fairness;this fairness hypothesis is often dis- excess giving occurs in order to satisfy norms of reci-
cussed interchangeably with altruism, but they are procity. Dictators give to others because future
distinguishable from each other. The altruism expla- rewards are contingent upon the individual’s “social
nation hinges upon the idea that individuals care reputation as a cooperative other-regarding person”
about others’ welfare. The fairness explanation in its (Smith2000,84).Dictatorsthusgivemorethanwould
simplest formulation is standards-oriented: an indi- be expected because they are concerned, in the short
vidual cares that the division of goods satisfies some run, that appearing “greedy” will decrease the likeli-
standard of equity (typically,in the standard dictator hoodthattheywouldbeinvitedbackformoreexperi-
game, one-half). Further elaborations of the fairness ments,ortheyareconcernedinthelongrunof other
explanation tilt the balance even more towards the negative consequences for themselves. To dispute
self: Fehr and Schmidt (1999) offer an extension of this reciprocity argument, Johannesson and Persson
this line of reasoning in suggesting that individuals (2000) manipulate the target recipient in a dictator
care about not just equity in outcomes across indi- game,specifyingthattherecipientisoneof theother
vidualsbutalsoabouttheabsolutedifferencebetween subjectsrecruitedforthestudyorarandomlyselected
anindividual’sallocationsvis-à-visthatofotherindi-
viduals. They specify an asymmetric utility function,
where individuals receive the most utility when
andfromtheirrelativestandingcomparedwithothers.Thecore
payoffs are equal, slight (and increasing) disutility
argument,thus,isthatindividualsdonotreallycareaboutmaking
from being advantaged when compared to others
others better off; instead, they take their own standing and the
(“guilt,” per Camerer 2003, 102), and sharper (and relative standing of others into account. However, the ERC still
increasingly sharper) disutility from being disadvan- includesthenotionofa“socialreferencepoint,”astandardagainst
whichdecisionsaremeasured.Indictatorgames,thissocialrefer-
taged compared to others (“envy,”per Camerer 2003,
ence point is an equal division of the payouts. The notion of
102).Yet,even after incorporating a sense of fairness, equity,orfairness,hasapivotalplaceinthisformulation.Asubtle
guilt, and envy into account, Fehr and Schmidt still distinctionbetweentheFehrandSchmidt(1999)andBoltonand
Ockenfels (2000) models is that in the former, individuals care
notethat“Altruismisconsistentwithvoluntarygiving
about the absolute level of difference between themselves and
in dictator and other public good games” (1999, others,whereasinthelatter,individualscareabouttheirrelative
854).6,7 sharesintheallocationsratherthanabsolutedifferencesinthese
shares(Camerer2003,104).
7Fehr and Schmidt (1999) note that it is harder to account for
6Using a similar approach, the Bolton and Ockenfels (2000) behaviors in other games using altruism. This is not a central
Equity,Reciprocity,andCompetition(ERC)Modelspecifiesthat concernofoursinthispaper,aswedonotbelievethatothergames
individualsreceiveutilityfromtheirownlevelofmaterialstanding provideasappropriateameansoftappingaltruism.
  .    . 
individual from the general population. They argue increased giving suggests that, “the more we know,
that “If donations in dictator games are motivated the more we care” (citation from Schelling 1968).
solelybyreciprocity,donationsshouldthereforedrop Camerer notes that the“identification effect is target
tozerowiththisexperimentaltreatment”(2000,138). specific and is not the result of general sympathy
JohannessonandPerssonareunabletorejectthenull toward others”(2003,76).
hypothesis of no difference between the two groups, Our design enables us to capture this distinction
whichsuggeststhatexcessgivinginthedictatorgame between concern for others, generally, and a concern
cannot be ascribed to reciprocity on its own. forspecificgroups.Weareinterestedinwhetherornot
A final explanation for excess giving is that sub- people give more to members of one political group
jectsdonotunderstandthegameandarejustmaking thananotherorwhethertheygivethesameamountto
random allocations. Andreoni and Miller (2002) ad- ananonymousindividualversusindividualsaffiliated
dress this concern by examining within-subject pat- withgroups.Byvaryinginformationaboutthepoliti-
terns of choices in their series of dictator games with calgrouptowhichthetargetrecipientbelongs,wecan
different payoffs. They find that 98% of the subjects uncovertheextenttowhichsocialidentitymightdrive
make choices that are consistent with the general allocation decisions. In contrast to Bohnet and Frey
axiomofrevealedpreferencesacrosseighttreatments, (1999), we compare giving in the anonymous game
suggesting that most of them understand the game with giving to anonymous individuals affiliated with
and are not choosing randomly. political parties. As a result, we find that there is
These results from the literature on giving in the an important qualification to the observation that
dictator game suggest that while there are several decreasingsocialdistanceincreasesgiving:itisnotjust
factors that might explain giving, dictator game allo- themoreyouknow,butboththemoreyouknowplus
cations may be a good proxy for an individual’s how you feel about the target. Dictators can, as
concernforthewell-beingofothers.Thewell-beingof Camerer notes, show “empathy or contempt” (2003,
others is probably more important to a person who 76).
choosestoallocate20%thanonewhoallocates0%.In We select individuals from partisan groups as
fact,the utility function used in Andreoni and Miller target recipients because political parties are among
(2002)toexplainbehaviorinthedictatorgameyields themostrelevantgroupingsinpoliticallife.AsCamp-
a monotonic relationship between the equilibrium bell etal. argue,“the strength and direction of party
allocationinthedictatorgameandtheweightaplayer identification are facts of central importance in
placesontheotherplayer’sutility.Inotherwords,the accounting for attitude and behavior” ([1960] 1980,
moreaplayercaresaboutthewell-beingofothers,the 121).Identification with parties is typically measured
moreshewillallocatetotheotherplayerinthedicta- witha7-pointLikertscaleofsubjectiveidentification,
tor game. although it has also been measured with self-reports
on closeness to parties and implicit associations
TappingAltruism and Social Identity in a (Huddy2003).Thesemeasuresofsocialidentification
Unique Dictator Game allow individuals to claim allegiance or closeness to
groups,buttheydonotrequireindividualstosacrifice
Behavior in dictator games can reveal other- anything personally in making such a claim. Using
regardingnessatagenerallevel(thatis,altruism).We dictatorgameallocationsasameasureof socialiden-
can also use dictator games to investigate whether or tificationisamethodologicalinnovation.Thedictator
not individuals exhibit politically relevant group- game enables us to measure an individual’s willing-
based preferences. Past experiments have varied the nessnotonlytoclaimallegiancetoapartybutalsoto
characteristicsof theanonymousrecipientwithsome affirmthatallegiancebywithholdingmaterialbenefits
interesting results. People are more willing to give to fromtheselfinordertotransferbenefitstoadifferent
charities than an anonymous individual (Eckel and individual who happens to be a member of the
Grossman1996),towomen(SaadandGill2001),and ingroup. Further, the nature of political competition
topeoplewhohavebeenintroducedtothem(Bohnet makes it more socially acceptable for individuals to
andFrey1999).Intheirexperiment,BohnetandFrey confer benefits to in-partisans and deny benefits to
manipulate the amount of information provided out-partisans (as opposed, to, say, racial groupings
about the target to the dictator. They find that more whichmightinvokesocialdesirabilityconcerns).This
information“transforms anonymous,faceless entities socialacceptabilitythusimprovesourabilitytodistin-
into visible, specified human beings, i.e., identifiable guish between self-interested, social-identity-based,
victims” (1999, 339). They argue that this pattern of and altruistic behavior.
  :  , ,    
In December 2004, about 350 subjects were mailbox under their computer, and then keep the
recruited from undergraduate political science and other envelope for themselves.They then type on the
sociology courses at a largeWestern public university computer the number of tickets they kept for them-
to participate in a computer-based survey.8 Subjects selves.11 Computers and the locked mailboxes are
wereofferedcredittowardstheircoursegradetopar- separated by partitions to protect the anonymity of
ticipateinthestudy;306(about85%)ofthemelected choices each subject makes. After the study a ticket
todoso.Eachindividualansweredanumberof stan- numberforeachofthethreedictatorgameprizeswas
dard questions regarding their socioeconomic status, drawn and announced by email to participants.12 All
political attitudes, and participation behavior and three prizes were claimed.
then played three dictator games.9 In one game, sub-
jects are told,“You know nothing about this anony-
mousindividual.”Intheothertwogamessubjectsare Partisanship and Dictator
told,“Theonlythingyouknowaboutthisindividualis Game Allocations
thatheorsheisaregisteredRepublican[Democrat].”
The order of these treatments is randomized, and a
We begin with mean allocations for each of the three
variableindicatingtheorderisincludedintheanalysis
kinds of recipients.In general,results from the dicta-
below.Allocations to the anonymous recipient reveal
tor game in this experiment appear to be similar to
thedegreetowhicheachsubjectcaresaboutthewell-
those of other researchers.Forsythe etal.(1994) spe-
being of others generally, while allocations to the
cifically compare“with pay”dictator games in which
Democrat and Republican reveal the extent to which
subjects are given $5 or $10 to divide and “without
subjects are motivated by social identity.
pay” dictator games in which subjects are asked to
In a typical dictator game, subjects are given a
make hypothetical choices. They find that more
smallamountofmoney($5to$10)andtheythengive
people keep everything for themselves in the “with
backtheportionofthemoneytheychoosetoallocate
pay”treatment(30.4%vs.13.0%)andthemeanallo-
to the other player.This procedure can be very costly
cation is lower (22.6% vs. 38.7%). By comparison,
forlargersamples,soweemployadifferenttechnique.
subjects in this experiment were even more likely
Subjects are given 10 lottery tickets that each have an
(38.0%)tokeepeverythingforthemselvesthanthose
equal chance of winning a prize of $100.10 They are
in both treatments. However, the mean allocation
then given two identical opaque envelopes. They are
(29.9%) falls between the two treatments. This sug-
asked to place the tickets they wish to keep for them-
geststhatthelotterymechanismusedinourdesignis
selves in one envelope and the tickets they wish to
replicating at least some of the incentives from dicta-
share with an anonymous individual in the other
tor games that use cash stakes.13
envelope.Theysealbothenvelopes,placetheenvelope
designated for the anonymous individual in a locked
11Achi-squaretestofthedistributionofcomputerresponsesand
thedistributionofticketsthatwerephysicallyplacedinthemail-
8Subjectsrangeinagefrom18to43years;theaverageageis21. boxes suggests that these two distributions are not statistically
Thesampleconsistsof56%womenand43%minority;itisquite different.
similar to the undergraduate body from which it is drawn (the
12Inmanydictatorgamestherecipientsarealsosubjects.Thiswas
undergraduate body is 56% female and 51% minority). The
nottrueinourexperiment—recipientsaredrawnrandomlyfrom
medianfamilyincomeisabout$80,000ayear.Theaveragesubject
the U.S. population (this is also the case in Johannesson and
leans left and Democratic—the modal response to the liberal-
Persson2000,whosendtheallocationstoarandomlydrawnindi-
conservative7-pointscaleisa“2,”or“liberal”(30%ofthesample),
vidualintheSwedishpopulation).Increasingthesocialdistance
and57%ofsubjectsidentifyasDemocratic.
betweenthedictatorandtherecipientshouldminimizethepoten-
9For a summary and exact question wording, see the online tialeffectofreciprocityandthusmakealtruismamorecompelling
appendix. explanationforexcessgiving(JohannessonandPersson2000).We
didnothandalltheenvelopeswithdonatedticketstorandomly
10Oneimportantdifferencebetweenourmethodandthetypical
chosenindividuals.Instead,wewaitedtoseeifadonorclaimedthe
dictatorgameisthestakesize.Notethattheexpectedvalueofthe
prize for a given dictator game. If they did not, then we used
prizeisonly$100/Nª$.33.Thougheconomistssometimescriti-
random digit dialing to locate an individual and request their
cizelow-stakesexperimentslikethisone,asurveyof theexperi-
name and address (and partisanship for the Republican and
mental economics literature by Camerer and Hogarth (1999)
Democrattreatments)andmailthemtheprize.
showsthatstakesizehasonlyasmalleffectonaveragebehavior
andthebiggesteffectofstakesonbehaviorischangingfromzero 13Another way to compare the results of this experiment to the
topositivestakes.Furthermore,Forsytheetal.(1994)andCarpen- existingliteratureisbyexaminingtherelationshipbetweendicta-
ter,Verhoogen,andBurks(2005)showspecificallyforthedictator torgameallocationsanddemographicvariables.Camerer(2003)
gamethatchangingfromlowstakestohighstakeshasnoeffecton notesthatmostdemographicfactorshavelittleeffectondictator
meanallocations. gameallocations,buttherearetwonotableexceptions.Carpenter,
  .    . 
TABLE 1 PartisanshipandAllocationsintheDictatorGame
Donor AnonymousRecipient DemocratRecipient RepublicanRecipient N
All 29.9% 30.1 27.3 306
Democrat 29.6 31.5 26.0 173
Republican 29.2 29.6 32.7 78
Difference .4 1.9 -6.7
p-value .41 .23 .02
Note: p-valuesreflectprobabilitythattruerelationshipisoppositetothesignofthedifference(Wilcoxonsignedranktest).
Beforemovingtotestsoftherelationshipbetween RepublicantargetthanweakRepublicanswould.Con-
socialidentity,altruism,andparticipation,wediscuss versely, strong Democrats will likely withhold more
how social identity manifests itself in the dictator from a Republican target than weak Democrats will;
game results. No previous studies have examined the strongRepublicanswillwithholdmorefromaDemo-
effectof partisanshipondictatorgameallocations,so cratic target than weak Republicans will.
weprobethisrelationshipinsomedetail.Ourexperi- Bias against Republicans. Experimental work
mentaldesignprovidesuswithauniqueopportunity suggests that individuals may discriminate against
to test several hypotheses about social identification members of different groups when they are choosing
behavior as it applies to partisanship: whether or not to bear a personal cost to help them.
Preferencefortheingroupovertheoutgroup.Parti- Additionally, considerations of deservingness enter
san identifiers will be more generous when asked to into dictators’ decisions. For example, Eckel and
allocaterewardsbetweenthemselvesandamemberof Grossman (1996) note that altruism increases when
their own party compared with a member of the the recipient appears to be more “deserving” or in
oppositionparty.Partisanidentifierswillalsobemore need of resources (for more on deservingness, see
generous to an ingroup member compared with Hoffman, McCabe, and Smith 1996b; Burrows and
someone not in the ingroup (the anonymous indi- Loomes 1994). The Republican Party has typically
vidual). This implies that Democratic identifiers will been associated with business interests and the
give more to a Democratic target than a Republican wealthy, whereas the Democratic Party has typically
target,and Democratic identifiers will give more to a been associated with the working class and the less
Democratic target than an anonymous individual. advantaged(BastedoandLodge1980;Campbelletal.
Likewise, Republican identifiers will give more to a [1960] 1980; Miller, Wlezien, and Hildreth 1991).
Republican target than a Democratic target, and Theseassociationsimplythat,onaverage,individuals
Republican identifiers will give more to a Republican might be less generous towards a Republican target
target than an anonymous individual. Independents compared with a Democratic target.
willgivelesstoaDemocratictargetandaRepublican Table1 shows mean allocations in the dictator
target than to the anonymous individual,since parti- game by partisanship of the donor and recipient.
santargetsaremoreobviouslyan“outgroup”toinde- NoticefirstthattheRepublicanrecipientreceives2.8%
pendents than an anonymous individual would be. less on average than the Democrat from all donors.
Strength of social identity. The stronger the parti- However, this difference may be due to the larger
san attachment, the more the ingroup should be number of Democrats in the sample. When we take
rewarded and the more the outgroup should be intoaccountthepartisanshipofthedonor,meanallo-
deprived.Assuch,wewouldexpectstrongDemocrats cationstendtodivergealongpartylines.Subjectswho
togivemoretoaDemocratictargetthanweakDemo- identify themselves as Democrats and Republicans
cratswouldandstrongRepublicanstogivemoretoa both give about the same amount to the anonymous
recipient, but they tend to give more to the recipient
from their own party, suggesting ingroup favoritism
Verhoogen,andBurks(2005)findthatsubjectswithhigherfamily
occurs. Notice that the Republican recipient inspires
incomestendtogiveless,whileEckelandGrossman(1998)find
the largest divergence, receiving 6.7% more from
thatwomentendtogivemore.Ourresultsreplicatebothfindings.
Consistentwithresultsfromotherdictatorgames,subjectsfrom RepublicandonorsthanDemocraticdonors,orabout
families with low incomes (below the median) give 6.4% more a fifth of the mean allocation.
thansubjectsfromfamilieswithhighincomestotheanonymous
Table2indicatesthatbothdirectionandstrength
recipient. Further, in this experiment, women give away 6.1%
moreticketsthanmen. of partisanship are significantly related to dictator
  :  , ,    
TABLE 2 StrengthofPartisanshipandAllocationsintheDictatorGame
PartisanRecipients
Donor AnonymousRecipient In-Party Out-Party N
StrongPartisan 24.4 31.7 23.3 127
WeakPartisan 34.6 32.0 31.0 124
Independents 32.0 26.7 55
TABLE 3 Within-SubjectDifferenceinGivingtoAnonymousandPartisanRecipients
AmountGivento AmountGivento AmountGivento
RepublicanMinusAmount RepublicanMinusAmount DemocratMinusAmount
GiventoDemocrat GiventoAnonymous GiventoAnonymous
Donor Mean p-value Mean p-value Mean p-value
All -2.8% .01 -2.2 .09 .3 .28
Democrat -5.5 .00 -3.6 .03 1.9 .19
Republican 2.8 .16 4.1 .10 1.2 .33
Independent -2.1 .32 -6.2 .19 -6.2 .05
Note: p-valuesreflectprobabilitythattruerelationshipisoppositetothesignofthedifference(Wilcoxonrankedsigntest).
game allocations. Strong partisans give most to within-subjecttendencytogivemoretosomepolitical
in-party targets,and they give significantly less to the groupsandlesstoothers.About61.7%ofthesubjects
anonymous recipient (Wilcoxon signed rank test, gaveexactlythesameamounttotheregisteredDemo-
p=.007)andtotheout-partytarget(p=.001).Weak crat and registered Republican that they gave to the
partisans show about the same degree of favoritism anonymousrecipient.Theremaining38.3%ofthesub-
towards the in-party target, but they show much less jectsdiscriminatedacrosstargets,changingtheirallo-
hostility towards the out-party target,compared with cation in at least one of the games based solely on
strongpartisans.Weakpartisansalsoshowmuchmore information about the partisanship of the recipient.
generosity towards the anonymous recipient com- Table3showseachofthethreepossiblecombinations
pared with strong partisans. This evidence suggests ofwithin-subjectdifferencesintheamountgiventothe
that strength of partisanship does not necessarily anonymous and partisan recipients and how this
affect generosity towards the ingroup, but it does breaks down by partisanship of the donor.First,note
affect punishment of the outgroup. The stronger the that the average subject gave somewhat less to the
partisanship, the greater the propensity to withhold RepublicanthantheDemocratoranonymousdonor,
benefits from those not explicitly affiliated with the yielding additional evidence for an anti-Republican
ingroup.Wealsoseethatindependentsmakedistinc- biasingiving.Thepartisanidentityof thedonoralso
tionsaswell—butdifferentlyfrompartisans.Indepen- seemstohaveaneffectonallocations.Democratsgive
dentsaremuchmoreinclinedtobegeneroustowards significantlylesstotheRepublicanthantotheDemo-
an anonymous individual compared with a partisan cratoranonymousrecipient.Republicansgivemoreto
identifier.Thisisconsistentwithourexpectationthat the Republican than to the Democrat or anonymous
independentsseethemselvesasagroupseparatefrom recipient,butthesignificanceofthedifferenceisweak.
the major political parties and thus are less generous Onceagain,therawdataappearstosuggestanin-party
towards these explicit outgroups compared with the bias,withthestrongestdifferenceinbehaviorexhibited
anonymous individual. by Democrats towards Republicans. Finally, people
Recall that a single subject participates in three whodidnotidentifythemselvesaseitheraDemocrator
dictatorgames,soourdesignenablesustodetermine RepublicantendtogivelesstoboththeRepublicanand
howthepartisanshipofthetargetrecipientaffectsthe Democratic recipient.In fact,the mean difference for
  .    . 
TABLE 4 DeterminantsofWithin-SubjectDifferencesinGivinginDictatorGameswithAnonymousand
PartisanRecipients
AmountGivento AmountGivento AmountGivento
RepublicanMinusAmount RepublicanMinusAmount DemocratMinusAmount
GiventoDemocrat GiventoAnonymous GiventoAnonymous
Coef. S.E. p Coef. S.E. p Coef. S.E. p
DonorCharacteristics:
Partisan .11 (.05) .01 .12 (.05) .01 .01 (.04) .42
Identification
PartisanStrength -.03 (.06) .32 .14 (.06) .02 .16 (.06) .00
HighIncome .01 (.03) .41 .04 (.04) .11 .04 (.03) .13
Female -.05 (.03) .08 -.07 (.03) .03 -.02 (.03) .24
White .01 (.04) .37 -.02 (.04) .28 -.03 (.03) .15
OrderVariables:
RepublicanFirst -.04 (.04) .15 .00 (.04) .46 .04 (.04) .10
DemocratFirst -.04 (.04) .15 -.04 (.04) .14 .00 (.04) .47
Constant .00 (.06) .48 -.12 (.06) .03 -.12 (.06) .01
Logscalevariable -1.28 (.04) .00 -1.24 (.04) .00 -1.38 (.04) .00
Deviance/Null 43/57 108/123 23/35
Deviance
Note: N=300.Intervalregression,wheredependentvariableiswithin-subjectdifferenceinallocationinthedictatorgame.Allindepen-
dentvariablesaredichotomousexceptthepartisanvariableswhicharescaledfrom0to1.Ordervariablesindicatewhichdictatorgame
subjectplayedfirst.
bothisexactlythesameat6.2%,oraboutone-fifthof anonymous recipient. Given that there is no such
the mean allocation. Although these differences are effect for the difference in giving between the Demo-
onlyweaklysignificant,theylendqualifiedsupportto crat and the anonymous recipient,these two findings
the strength of social identification observed above. suggest that subjects give less to Republicans than
Partisans tend to receive less from nonpartisans and other kinds of recipients and an anti-Republican bias
viceversa. exists.
Tofurtherassesstheeffectof partisanshipondif- The regressions reveal partisan strength bias:
ferencesindictatorgameallocations,Table4presents strong partisans give 14% and 16% more than
results from three sets of multiple regressions that independents do to the Republican and Democratic
also control for demographic factors.14 The first set targets, respectively. These results suggest that strong
of results analyzes partisan discrimination:the extent partisans see themselves as part of two ingroups,
to which subjects make distinctions between the rewarding members of their own party at the
Republican and Democratic targets.We see that par- expense of the opposing party, and rewarding
tisan identification has a positive and significant members of any party over those who do not affiliate
effect on the difference in the amount given to the with a party.
Republican versus the Democrat, providing addi-
tional evidence for ingroup preference and outgroup
hostility. The direction of partisan identification also
Altruism, Social Identity, and
has a positive and significant effect on the difference
in the amount allocated to the Republican versus the Political Participation
Ourmainexpectationisthatthosewhoaremotivated
14InTables4and5weuseintervalregressionbecausethedepen-
dent variable is truncated at its minimum and maximum value. by altruism and by social identity will participate in
Thisestimationmethodiscommonintheliteratureondictator politics more than those who are motivated by mate-
games (e.g., see Carpenter, Verhoogen, and Burks 2005) and is
rial self-interest. To test this expectation, we create a
conductedusingmaximumlikelihood.Wescaleallvariablesfrom
0to1foreaseofcomparisonacrosscoefficients,andwereportthe 7-point scale of participatory acts, including voting,
residualdevianceofthemodelandcompareittothenulldeviance contributingtoacandidate,joiningapoliticalorgani-
ofamodelwithaconstant.