Birds of Different Feathers Cooperate Together: No Evidence for Altruism Homophily in Networks

23 Pages·2015·0.47 MB·English

Birds of Different Feathers Cooperate Together:
No Evidence for Altruism Homophily in Networks
Brent Simpson,a Matthew Brashears,b Eric Gladstone,b Ashley Harrella
a)UniversityofSouthCarolina,(b)CornellUniversity
Abstract:Manyevolutionarymodelsofcooperationassumethataltruistspossesstelltalesignsofdispositionthatthey
usetofindandselectivelyassociatewitheachother.Priorresearchfindsthatpeoplecandetectthesesignsofaltruism
instrangers,butwedonotyetknowwhetherthisresultsinaltruismhomophily. Wearguethatdispositionsshould
matterlessinrepeatedinteractions,wherebehaviorisbasedonreciprocity.Asaconsequence,weshouldnotexpect
peopletohaveaccurateinsightintothedispositions(egoismvs.altruism)oftheirfriends,norshouldweexpectthese
relationstobecharacterizedbyaltruismhomophily.Threestudies,employingdiversemethodologiesandmeasures,find
noevidenceofaltruismhomophily.Moreover,wefindthatpeoplehavepoorinsightintotheirfriends’altruism.Wediscuss
theimplicationsofthesefindingsfortheemergenceofaltruismandtheroleofembeddedinteractionsinsustaining
humancooperation.
Keywords:relations;friends;altruism;prosocialbehavior;trust;trustworthiness
Editor(s):JesperSørensen,GabrielRossman;Received:September4,2014;Accepted:October11,2014;Published:December22,2014
Citation:Simpson,Brent,MatthewBrashears,EricGladstone,andAshleyHarrell.2014.“BirdsofDifferentFeathersCooperateTogether:NoEvidence
forAltruismHomophilyinNetworks.”SociologicalScience1:542-564.DOI:10.15195/v1.a30
Copyright:(cid:13)c 2014Simpson,Brashears,Gladstone,andHarrell.Thisopen-accessarticlehasbeenpublishedanddistributedunderaCreativeCommons
AttributionLicense,whichallowsunrestricteduse,distributionandreproduction,inanyform,aslongastheoriginalauthorandsourcehavebeen
credited.cb
How does altruism persist in the face of in- tirelyfromstudiesexaminingwhetherpeoplecan
centives to exploit it? This question has re- predict the prosociality of strangers in one-off
ceived increased attention with growing evidence interactions (e.g., Frank, Gilovich, and Regan
of heterogeneity in altruism, or other-regarding 1993; Brown, Palameta, and Moore 2003; Yam-
preferences, in humans (van Lange et al. 1997; agishi et al. 2003; Oda et al. 2009; Fetchenhauer,
Fehr and Gintis 2007; Simpson and Willer 2008). Groothuis, and Pradel 2010; Feinberg, Willer,
Oneanswerfromevolutionarymodelsofcoopera- and Keltner 2012). As a result, there is scant evi-
tionisthataltruists(whopossessother-regarding denceforaltruism homophily,orthetendencyfor
preferences and act prosocially even in the ab- altruists to be friends with each other at higher
senceofstrategicincentives)possesstelltalesigns than chance levels. Thus, while people are able
of character that identify them as altruists to to read telltale signs of others’ dispositions with
prospective friends and exchange partners. Ego- better than chance accuracy, we do not know
ists (who are less apt to act prosocially in the whether this ability plays a role in relationship
absence of reputational or other strategic incen- formation. This paper draws on theories of reci-
tives) lack these characteristics and are avoided. procity (Gouldner 1960; Trivers 1971; Axelrod
As a result, we should expect social networks to 1984)toarguethatweshouldnotexpectaltruism
exhibitaltruismhomophily,withaltruistssorting homophily.
together and reaping benefits from social cooper- Inadditiontoaltruismhomophily,wealsoad-
ation, leaving egoists to either pair up together, dress whether people have accurate insight into
or go it alone (Wilson and Dugatkin 1997; Frank their friends’ dispositions. On one hand, the mo-
1988;MacyandSkvoretz1998;SoberandWilson tivation and ability to read telltale signs is of
1998; see also Rand and Nowak 2013). paramountimportancetoaltruismhomophilyhy-
There is evidence that people can distinguish potheses, and relationship formation and contin-
altruists from egoists, but it derives almost en- uationshouldbelargelybasedonassessmentofa
sociologicalscience|www.sociologicalscience.com 542 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
prospective partner or friend’s disposition. Thus, has assumed that individuals prefer to associate
friends should be able to predict each other’s withthosewhohavesimilarknowledgeorcultural
prosociality at a better than chance level. From tastes (e.g., Carley 1991; Mayhew et al. 1995).
a reciprocity perspective, on the other hand, the Here we expand the literature on homophily by
“shadow of the future” that characterizes friend- asking a more fundamental question: Do humans
ships implies that both egoists and altruists will prefer to associate with those whose basic values
tend to act prosocially toward their friends. As a (e.g., altruism and egoism) are like their own?
result, we are likely to have limited insight into In what follows, we introduce necessary ter-
the extent to which our friends are altruistic be- minology and briefly review arguments linking
cause the structure of the relationship, rather the ability to read telltale signs of character with
than individual dispositions, acts to guarantee altruism homophily and thus the emergence of
their benevolence. altruism and cooperation. We then argue why
Addressing these competing predictions is im- we should not necessarily expect detection abili-
portant for a number of reasons. Most broadly, ties to translate into altruism homophily and ex-
many evolutionary models of altruism, or other- tend these arguments to the question of whether
regarding behavior, hinge on humans’ ability to people have knowledge of their friends’ disposi-
read telltale signs of disposition and to form tions. Wethenintroducethreestudiestotestthe
friendships based on those abilities. But while homophily predictions, including a Respondent
prior work suggests that people can read these Driven Sampling survey of ties between students
signs, we don’t know that these abilities lead at a large public university (Study 1), a study of
to altruism homophily. If the ability to discern friendship ties within a sorority chapter (Study
dispositions does not lead altruists to selectively 2), and a laboratory experiment (Study 3). The
exclude egoists from productive ongoing relation- laboratory experiment also allows us to test pre-
ships,itislessclearhowtheabilitytoreaddispo- dictionsaboutpeople’sknowledgeoftheirfriends’
sitions can account for altruism homophily and altruism. Across all studies, using a variety of
thus the emergence and persistence of altruism. dataandanalytictechniques,wefindvirtuallyno
Further, altruism homophily approaches sug- evidenceofaltruismhomophily. Wealsofindthat
gest that relationships between altruists will pro- friendshavesurprisinglylimitedinsightintoeach
duce more value than those between egoists, other’s dispositional altruism. Together, results
whereas reciprocity approaches suggest that ben- fromthethreestudiesofferconsistentsupportfor
efits derive from the fact that two individuals are our argument that there is little reason to expect
connected, rather than from the combinations of altruism homophily in networks.
their dispositions. Thus, beyond shedding light
on evolutionary explanations of altruism, a bet-
Altruism and Telltale Signs
ter understanding of altruism homophily should
yield clearer insight into the costs and benefits of Disposition
of relationships. An investigation of the social
processesunderlyingtheformationofcooperative We define prosocial behavior as any behavior
relationships thus has implications both at the that benefits another person, regardless of the
micro level of dyadic interaction and the macro underlying motivation for the act (Piliavin and
level of societal integration. Charng 1990). Thus, prosocial behavior is an
Finally, homophily, or the tendency for indi- outcome, whereas altruism and egoism are moti-
viduals to associate with those like themselves vational precursors to behavior. Recent research
(Lazarsfeld and Merton 1954), is one of the most hasestablishedstrongbehavioral(e.g,. vanLange
robust of all social science findings (e.g., McPher- 1999; Simpson and Willer 2008) and neurological
son et al. 2001; Smith et al. 2014). Homophily (Emondsetal. 2011)evidencefortheexistenceof
largely depends on the presence of similar others heterogeneous social preferences, as well as ego-
in foci (Feld 1981) frequented by the individual, ists’ and altruists’ divergent responses to various
but some doubtless derives from preferences for incentives for prosocial behavior.
associating with those like oneself (e.g., Huston Thenotionthataltruistspossesstelltalesigns
and Levinger 1978). However, research to date that allow them to identify each other informs
sociologicalscience|www.sociologicalscience.com 543 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
much of the work on selective assortation (Eshel This prior work shows that people can read
and Cavalli-Sforza 1982; Wilson and Dugatkin telltale signs of disposition, but it does not tell
1997; Sober and Wilson 1998; see also Macy and us whether, as a result, they form relationships
Skvoretz 1998), but it has arguably received its characterized by dispositional homophily. To
most extensive treatment in Frank’s (1988) com- our knowledge, only two studies have addressed
mitmentmodel. Frankarguesthatindividualdif- whether altruists attract and associate with each
ferences in moral sentiments, such as sympathy, other. First, Sheldon et al. (2000) looked at
compel individual differences in moral behavior. the “prosocial values” of friends and found a
If these moral sentiments are revealed to others marginally significant correlation. However, it
by telltale signs of character, such as emotional is not clear that their measure of prosocial val-
displays,theargumentcontinues,altruistswillbe ues was related to altruism or prosociality as the
preferentially chosen as exchange partners. Be- terms are used in the literature.1
cause altruists will seek to associate with others A second study by Pradel and colleagues
whoalsodisplaysignsofaltruismandavoidthose (2009) measured generosity in six relatively small
who do not, altruistic types will tend to pair up secondary school classrooms (14 to 29 students)
with one another. Egoists are thus left to go using a “modified dictator game.” Each student
it alone (and miss out on the benefits of social was asked to imagine that they had been given
cooperation) or make do with one another. Such a monetary endowment and to decide how much
a process allows altruism to compete successfully of the endowment to pass on to an unspecified
with egoism. classmate. Participants were told that “dictators”
If telltale signs of disposition create an ad- and“recipients” wouldbematchedapproximately
vantage for altruists, they create an even greater one week later. Pradel and colleagues observed
benefit to egoists who are able to fake the signal modest correlations between the friends’ generos-
andreapthebenefitsofhavingaltruisticpartners ity. Further, a series of follow-up questions found
withoutpayingthecostsofaltruism(Trivers1971; that students were able to predict the generos-
Dawkins 1976; Macy and Skvoretz 1998). While ity of their friends and enemies (but not other
acknowledging the potential for egoists to feign classmates) with greater than chance accuracy.
altruists’ signs, Frank argues that the emotional While suggestive, several important issues
cues are costly and difficult to fake for those who limittheconclusionswecandrawfromthePradel
do not actually experience the underlying emo- et al. study. First, because students were told
tion. The incentive for egoists to mimic altruism, that their dictator decision would affect an un-
andtheconcomitantneedforaltruiststobewary specified member of the class, the measure may
ofmimics,ispredictedtoresultinanequilibrium have captured the dictator’s general orientation
mix of egoists skilled at deception and vigilant toward the class as a whole, rather than altru-
altruists continually on the lookout for wolves in ism. Thus, we cannot know whether the method
sheep’s clothing (Frank 1988). trackedaltruismhomophilyorwhetherthosewith
The empirical evidence supports the ability similar sentiments toward the class were simply
of people to read telltale signs of dispositional more likely to be friends.2 Alternatively, if the
altruism. Frank et al. (1993) found that partici-
1Although presented as a test of whether altruists
pants in a laboratory experiment could predict
sort together, Sheldon and colleagues measured values
cooperation and defection in interaction part- using an Aspirations Index that distinguished intrinsic
ners at better than chance levels (see also Brosig values(e.g.,“IwillknowandacceptwhoIreallyam”)and
2002). Other work shows that people can distin- extrinsicvalues(“Iwillhavepeoplecommentoftenabout
how attractive I look”). Because it is not clear to what
guish altruists from egoists based on discussion
extent, if at all, these values overlap with dispositional
of unrelated issues (Yamagishi et al. 1999), brief altruismoregoism,itisnotclearwhattheSheldonetal.
video exposure (Brown, Palameta, and Moore findingstellusaboutaltruismhomophily.
2Forinstance,HarbaughandKrause(2000)foundthat
2003; Oda et al. 2009; Fetchenhauer et al. 2010;
the greater proportion of a child’s education that had
Feinberg, Willer, and Keltner 2012) or even a
beenspentamongcurrentclassmates,themoregenerous
photograph (Yamagishi et al. 2003; Verplaetse et sheorhewastotherestoftheclass. Totheextentthat
al. 2007). those who have been at a given school longer are more
apttobefriends,thiscouldresultintheappearanceof
altruismhomophily.
sociologicalscience|www.sociologicalscience.com 544 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
measure tracked generosity to a person, rather tinue can eliminate the otherwise strong behav-
than orientation to the class as a whole, it isn’t ioral differences we observe between egoists and
clear to what extent it tracked generosity toward altruists in one-off interactions (van Lange et
a close other versus a distant other. This would al. 2011). Further, egoists and altruists show
depend, among other things, on the assumptions no difference in willingness to sacrifice for their
the dictator made about who would ultimately romantic partners (e.g., van Lange et al. 1997).
benefit from her decision. Fromtheperspectiveofreciprocityapproaches,
Second, as the authors note (p. 109), it isn’t treatmentoflongtermpartnersismoreinfluenced
clear to what extent the findings were driven by by structures of interdependence in a given rela-
generosity versus reputation management: “Our tion (e.g., the dependence of each person on the
experimental setting, in which we seated partici- otherandtheavailabilityofsuitablealternatives)
pantsinacircletoenablethemtolinkclassmates than either partner’s dispositional altruism or
with subject numbers, might have created a con- egoism (see also Kelley and Thibaut 1978; Rus-
text that students perceived as partly public.... bult and van Lange 2003). Repeated interactions
Observed altruistic behavior ...therefore, cannot lead to a convergence of strategies within those
unconditionally be interpreted as ‘pure’ altruism relations that make altruists and egoists behav-
inthesensethatitwasindependentofreputation iorally indistinguishable.
management efforts.” While prior work establishes that the mere
possibility of repeated interaction moderates the
impact of dispositions on behavior, we do not
Reciprocity and Social know whether people actually anticipate this ef-
fect. Ifpeopledonotintuitthatongoingrelation-
Relations
ships attenuate the effects of others’ dispositions,
we would expect people to attach similar value
Whereas the arguments just reviewed assume to information about a potential partner’s dis-
that prosociality emerges out of individuals’ dis- position whether they anticipated one-off or re-
positions, reciprocityapproaches(Gouldner1960; peated interactions. As a consequence, we would
Trivers1971;Axelrod1984)viewprosocialityand expect to observe a tendency towards altruism
cooperation as stemming from relational induce- homophily even though dispositions matter little
ments. We build on this perspective to address once relations are established.
why individual differences in dispositional altru- However, we expect that people understand
ism need not factor into longterm social relations thatrelationshipsandrepeatedinteractionsmod-
and thus why we should not expect to observe eratetheimpactofdispositions. Ifthisisthecase,
altruism homophily. they will invest less in gaining information about
In a reciprocity approach, not only is coop- others’ character when they expect repeated ver-
eration the most effective way for altruists to sus one-off interactions. Such a pattern would
realize their goals, assuming others cooperate; provide a stronger theoretical foundation for pre-
egoists can also maximize their self-interest in dicting the absence of altruism homophily. In
longterm social relations by engaging in tit-for- a series of ancillary experiments, available upon
tat exchanges, or “reciprocal altruism” (Kollock request from the first author, we found empirical
1998; van Lange et al. 2011). As a result, dis- evidence for precisely this pattern.
positional differences that have powerful effects To summarize, individual differences in al-
on behavior in interactions with strangers have truism that emerge in one-off interactions with
limited impact in repeated interactions. strangers lose their predictive power in repeated
Empirical evidence supports the prediction interactions. Given that social relations such as
that the behaviors of egoists and altruists, while friendships are repeated interactions, there is no
highly divergent in one-off interactions, tend to reason to expect altruism homophily.This stands
matterlessinlongertermrelations(e.g.,Kuhlman in marked contrast to the evolutionary models
and Marshello 1975; Parks and Rumble 2001). described earlier that explain the emergence of
Even the possibility that an interaction may con- altruism and cooperation via altruism homophily.
sociologicalscience|www.sociologicalscience.com 545 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
Perceived Altruism Homophily experiment (Study 3). Second, we assess compet-
ing predictions about whether people can intuit
Theabilityandmotivationtoreadtelltalesignsof
theirfriends’prosociality. Altruismhomophilyar-
disposition is fundamental to altruismhomophily
gumentspredictthatfriendswillhaveinsightinto
approaches. Further, these approaches suggest
the extent to which their friends will act proso-
that it is precisely in longterm relations such
cially toward strangers. Our arguments, however,
as friendships that we should observe evidence
lead us to expect that friends will be poor at pre-
that people have insight into their partners’ dis-
dictingeachother’sprosocialitytowardstrangers.
position (e.g., Frank 2005), since these are the
Further, we expect that this inaccuracy will be
relationships that matter most and where there
driven in part by perceived altruism homophily—
would be the greatest opportunity to detect sub-
i.e., the tendency to (inaccurately) perceive one’s
tle telltale signs. But here we instead extend our
friend as being as altruistic or egoistic as one-
application of reciprocity theories to explain why
self. We test these predictions in the laboratory
friends’ altruism should be relatively inaccessible.
experimentusingseveralmeasuresofprosociality.
Peopleoftenhavelimitedinformationabouta
friend’sprosocialityoutsidethefriendship. What
information is available may be distorted by the Study 1: Homophily in a
friend’saccounts(ScottandLyman1968)orother
Respondent Driven Sample
impression management strategies. Additionally,
the fact that both altruism and egoism (through
OurfirsttestusesdatafromaRespondentDriven
the possibility of reputational gains, for instance)
Sampling (RDS) study. RDS samples begin with
can motivate prosociality allows for varied inter-
a modest number of initial respondents who re-
pretations of a single prosocial act by the friend.
cruit the second wave of respondents through
Prior work suggests that altruistic versus egois-
their network connections, and so on for subse-
tic dispositions influence the attributions people
quentwaves. Givenasufficientnumberofrecruit-
make about others’ prosociality (see Simpson
mentwaves,thecharacteristicsofanRDSsample
and Willer 2008), such that altruists and egoists
converge on those of a random sample, thereby
project their own values and motives onto others’
permittingstatisticalinference(Heckathorn2007).
behaviors (Orbell and Dawes 1993). This implies
Respondents were students at a large south-
that altruism homophily will be more perceived
eastern university and could earn up to $25 for
thanreal,aspeopleinterpretfriends’behaviorsin
completing a web-based survey and recruiting as
light of their own dispositions. That friendships
many as three other respondents. The survey
will be characterized by perceived dispositional
was part of a larger project (Wejnert 2010) and
homophily is in line with the broader homophily
includedbasicdemographicquestions,astandard
literature, which shows that people tend to think
measure of dispositional altruism, and questions
that their friends are more similar to them with
about student life (e.g., whether the respondent
respect to attitudes and beliefs than is actually
was a member of a Greek organization, lived on
the case (McPherson, Smith-Lovin, and Cook
or off campus, etc.) and personal networks.
2001; Goel, Mason, and Watts 2010).
Procedures
Hypotheses and Overview
The sample began with nine respondents who
of Studies received an email explaining the purpose of the
project and details on compensation. If a respon-
Our first empirical objective is to assess whether dent agreed to participate, he or she clicked on a
friendships will exhibit altruism homophily (as link to the online survey. After completing the
suggested by evolutionary models of altruism re- survey,therespondentreceivedthreerecruitment
viewed above) or whether, as we expect, friend- emails that they were asked to forward to three
ships will be random with respect to dispositions. potential recruits who were students at the uni-
We test these competing predictions in two net- versity and with whom they had regular contact.
work surveys (Studies 1 and 2) and a laboratory These later recruits followed the same process
sociologicalscience|www.sociologicalscience.com 546 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
as the initial seeds. Each participant was paid did not make at least six choices consistent with
$10 for completing the 20-minute survey and an a given SVO and thus could not be classified.
additional $5 for each recruit who did so, for a
total individual maximum compensation of $25 Homophily
and a mean of $15.
We outline the results of four logistic regression
models. Thefirsttwomodelspredictwhetherthe
Altruism Measure
recruiterisanaltruistoranegoistasafunctionof
We assessed altruism using the standard nine- the respondent’s SVO. These models allow us to
item measure of social value orientation (SVO), determine whether there is a general preference
arguably the most widely used survey measure of forassociationwithaltruists, aswemightexpect,
dispositional altruism (van Lange 1999; Feinberg as well as permit us to detect dispositional ho-
et al. 2012). Social value orientations are sta- mophily. The third model predicts whether the
ble preferences for how outcomes are distributed recruiter is a female based on the respondent’s
between self and others (Messick and McClin- sex, permitting us to detect gender homophily.
tock 1968). Our measure presented respondents Investigationofgenderhomophilyallowsustoas-
withaseriesofninedecomposedgames,eachcon- sesswhetherourmethodstrackhomophilywhere
sisting of three different distributions of points itexists.3 Finally, wemightexpectthatrelations
for self and another (unidentified) person. The characterized by more frequent interactions or
other was presented to respondents as “someone closer ties would be more apt to exhibit altruism
you do not know and that you will not know- homophily than weaker relations (Frank 2005;
ingly meet in the future.” Individualist choices Pradel et al. 2009). The fourth model thus ad-
maximize payoffs to self, without regard to the dresseswhethertiestrengthimpactsdispositional
other’s payoff; competitive choices maximize the homophily and uses a new dependent variable
differencebetweenthepayofftoselfandother;al- that distinguishes SVO-homogeneous dyads from
truistic choices maximize the aggregate payoff to SVO-heterogeneous dyads without regard to the
self and other and minimize inequality in payoffs. SVO of the recruiter.
Previous research has established the temporal AllfourmodelsaregiveninTable1. Thefirst
stability (van Lange 1999) and overall predictive model predicts whether a respondent’s friend is
power (Liebrand 1986) of such measures for a an altruist, the second predicts whether a re-
wide range of behaviors, including contributions spondent’s friend is an egoist, the third predicts
to real world charities (van Lange et al. 2007). whether the respondent’s friend is female, and
Following prior work, we classified partici- the fourth predicts whether the respondent and
pants as a given SVO (altruist, individualist, or his or her recruiter have dispositional homophily.
competitor) only if their responses to at least Givenourfocusonhomophily, ineachofthefirst
six of the survey items were consistent with that three models the main independent variable is a
orientation. Most prior work also collapses re- dummy variable indicating whether the respon-
sponses from individualists and competitors into dent is an altruist (model 1), an egoist (model
one“egoist” category. Asanalysesusingagreater 2), or female (model 3). The fourth model con-
level of specificity did not affect our conclusions, tains two additional dummy variables measuring
we follow this prior work. Thus, our key compar- tie strength: whether the respondent reported
ison centers on the distinction between altruists interacting with the recruiter at least once per
and egoists.
3Specifically, if we do not observe dispositional ho-
mophily or gender homophily, it will suggest that our
datamaybecompromised,giventhatgenderhomophily
Study 1 Results isarobustphenomenoninsocialnetworks(McPherson
et al. 2001; Smith et al. 2014). If we observe disposi-
tional homophily and gender homophily, it will suggest
A total of 310 respondents (145 females) par- thatfriendsaremorelikelythanchancetohavethesame
ticipated in Study 1. Among these, 147 were disposition. Finally,ifweobservegenderhomophilybut
notdispositionalhomophily,itwillsuggestthatourdata
classified as altruists and 126 egoists (84 individ-
aresufficienttoidentifyhomophily,butthatfriendsare
ualists and 42 competitors). The remaining (46) notmorelikelythanchancetohavethesamedisposition.
sociologicalscience|www.sociologicalscience.com 547 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
Table1:Logistic Regression Estimates for Models Predicting Altruism Homophily in a Respondent
Driven Sample (Study 1)
Altruistic Egoistic Female SVO
Friend Friend Friend Homophily
Altruistic Respondent 0.33 −0.12
(0.27) (0.31)
Egoistic Respondent 0.41 −0.32
(0.27) (0.38)
Unclassified Respondent −0.11 0.37 −0.50 −1.98†
(0.36) (0.32) (0.39) (0.75)
Altruistic Referrer −0.01
(0.41)
Egoistic Referrer −0.32
(0.27)
Unclassified Referrer 0.15 −1.63∗
(0.65) (0.78)
Female Respondent −0.25 0.37 2.01† −0.22
(0.35) (0.34) (0.30) (0.33)
Female Referrer −0.23 0.18 −0.22
(0.53) (0.54) (0.42)
Female Respondent × Female Referrer 0.33 −0.36 0.20
(0.61) (0.61) (0.53)
Interacts Weekly or More 0.16
(0.49)
Discusses Important Matters 0.11
(0.32)
Constant −0.11 −0.73∗ −1.36† 0.28
(0.30) (0.29) (0.42) (0.49)
N 310 310 310 307
†p<0.01; ∗ p<0.05.
week, and whether the respondent reported dis- Likewise, egoists are no more likely to be associ-
cussing important matters with the recruiter. In atedwithegoiststhanaltruists(model2). Incom-
allmodelswecontrolforhomophilyofothertypes bination, these two models fail to show evidence
(models 1, 2, and 4 control for gender homophily; for dispositional homophily. In contrast, we ob-
model3controlsforaltruismhomophily). Weran serve substantial homophily by gender (model 3).
other models with additional control variables, As expected, female respondents are more likely
including controls for tie strength in models 1, to be recruited by female friends than are male
2, and 3. The alternative models yielded sub- respondents. Since respondents are recruited via
stantively identical conclusions to those reported existing social network ties, this result indicates
below. Because the same individual could recruit that interpersonal ties do exhibit homophily by
up to three others, dyads are not statistically gender, as we would expect from prior work, but
independent, and we adjust our models for this do not exhibit homophily by disposition. Finally,
clustering of observations on recruiters. (Omit- model 4 shows that stronger ties are no more
ting this adjustment does not alter our results.) likely than weaker ties to exhibit dispositional
On the whole, Table 1 shows no evidence of homophily. In short, these data do not support
dispositionalhomophily. First,asshowninmodel the altruism homophily hypothesis.
1, wefindthataltruisticrespondentsarenomore In estimating the logit models, we treated
likelytohaveanaltruisticfriendthanareegoists. the RDS data as though they were derived from
sociologicalscience|www.sociologicalscience.com 548 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
a standard survey and did not weight to ad- is homogeneous: all members are female, they
just for the non-random nature of respondent are similar in age and socioeconomic background,
recruitment, although we did adjust for the clus- and the vast majority are white. Members re-
tering of observations. This decision stems in ceived an email, sent with permission over the
part from the fact that multivariate models rely chapter listserv, with a link to an online survey.
on correlations between variables that are rarely Thesurveyincludedthemeasureofaltruismused
impacted by weighting procedures. Thus, in- in Study 1 and a number of network measures.
cluding weights would not alter our conclusions
(Heckathorn 2007). As a check, we conducted Altruism and Association Measures
ancillary analyses employing methods specific to
RDSthatincorporatecorrectionsfornon-random In addition to completing the altruism measure,
selection of respondents (Heckathorn 2007; Wejn- respondentsviewedacompletechapterrosterand
ert 2010). These models, available upon request, were asked to select the person they considered
yielded the same conclusions given above. The their best friend in the sorority. We also used
ancillary analyses indicate the presence of ho- several other measures that have been used in
mophily by gender, age, and class, as well as prior work on friendship and close relations (e.g.,
on a various aspects of university life, such as Paxton and Moody 2003): each respondent was
membership in Greek organizations, living on- asked to select “up to five [chapter] sisters with
or off-campus, and participation in intramural whom you regularly discuss important matters;”
sports. But they do not suggest any evidence of “up to five sisters with whom you regularly go
dispositional homophily. out socially;” and “up to five sisters in whom you
regularly confide and/or go to for advice.”
In addition to the measures of social ties, re-
Discussion of Study 1
spondentswereaskedtoindicate“uptofivesisters
The results of the RDS study show no evidence who you feel sacrifice most on behalf of the soror-
of altruism homophily. Study 1, however, relied ity.” Unliketheprecedingfourmeasures,thislast
upon a sample drawn from an entire university questiondoesnotmeasuresocialtiesbetweenthe
population. It is possible that the emergence of respondent and another member of the sorority,
altruism homophily may require a high level of but instead functions as a reputational measure.
mutual interaction so that altruists and egoists This measure captures whether members of the
can be more easily identified. It is also possible sorority can identify those who act more proso-
that altruism homophily is more apt to emerge ciallytowardsthegroup, whetherornottheyare
in groups that are more homogeneous on key more inclined to socialize with them.
sociodemographic dimensions. For instance, peo- Finally, respondents were asked to approxi-
ple might first sort on characteristics such as matewhatpercentageoftheirclosestfriendswere
race,ethnicity,orreligionandonlytheninvestre- chaptermembers. Thisprovidesameasureofthe
sourcesindiscerningtelltalesignsofaltruismand extent to which the sorority is a closed system.
forming social relations based on these insights. Wealsocollectedinformationontherespondent’s
Given these questions, Study 2 investigates altru- pledge (year) group and college major, as well as
ism homophily in a bounded homogeneous group basic demographic questions, including religious
where the conditions of repeated mutual inter- beliefs,politicalvalues,andfather’sandmother’s
actions between people of similar socioeconomic education. The survey took about 25 minutes to
profiles can be met. complete.
Study 2: Homophily in a Study 2 Results
Sorority Chapter
The sample includes 127 members of the soror-
ity, for an approximate 51 percent response rate.
Respondents in Study 2 were members of a tradi-
Given this low response rate, we treat these data
tionalsororitychapter(with247members)atthe
asanegonetworksampleforpurposesofanalysis.
sameuniversitysampledinStudy1. Thechapter
Morecomplexanalysesusingexponentialrandom
sociologicalscience|www.sociologicalscience.com 549 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
graphmodels(e.g.,Lusher,Koskinen,andRobins equals one when the respondent is an altruist
2013) produced substantially similar results, but and the other which equals one when the respon-
as these models are compromised by the large dent has an unclassifiable SVO. Thus, egoist is
amount of missing data, we limit discussion to the reference category, and the dummy variables
the ego network approach. allow us to determine if respondents of a given
Altruists make up 38.6 percent of the sample, disposition are more likely to show dispositional
and individualists and competitors make up 25.2 homophily. Thecontrolvariablesforthesemodels
percent and 16.5 percent, respectively. This dis- are the number of names given for each relation
tribution is comparable to Study 1 and to those andthepercentageofalltherespondent’sfriends
from prior work (e.g., Simpson and Willer 2008). whoderivefromthesorority. Theformercontrols
Again, we group individualists and competitors for the opportunity to be named by the respon-
into one “egoist” category (41.7 percent of the dent (i.e., the more alters a respondent reports,
sample). Those who did not give at least six re- the more likely it is that some will share an SVO
sponses consistent with either altruism or egoism with the respondent). The latter controls for the
make up 19.7 percent of the sample. respondent’s dependence on the sorority for her
Inancillaryanalyses,welookedatthenumber social ties.
of others chosen by each sorority member (outde- Model 5, investigating the disposition of a
gree) and the number of times they were chosen respondent’s best friend in the sorority, includes
by others (indegree) in each of the five relations the same independent variables. The control
(important matters, social activities, advice, sac- variable(numberofnamesgiven)isomittedfrom
rifice, and best friend). These analyses, available these models, however, as respondents could only
upon request, revealed that altruists and ego- provide a single name. In its place a variable is
ists had similar outdegrees and similar indegrees added indicating whether or not the best friend
across all measures with one exception: as dis- reciprocates the nomination.
cussedbelow,altruistsweresomewhatmorelikely Across the models, the only consistent effect
than egoists to be selected by other members as is that the more alters named by the respondent,
sacrificing on behalf of the chapter. We also as- the more homophilous friends a respondent has.
sessed whether there was a general preference Thesoleexceptiontothisgeneraltrendisforthe
for altruists in any of the relationship measures, selectionofthosewhosacrificemostforthesoror-
including whether respondents of a given dispo- ity (model 4). In this case respondents are not
sition are more likely to have large numbers of selectingfriendsbutratherarereportingonthose
altruistic friends. These analyses showed that whom they perceive to be the largest contribu-
altruists count approximately the same number tors to the group, and altruists are more likely
of altruists among their friends as do egoists. to be chosen by everyone. This finding suggests
that members of the sorority can accurately iden-
Is There Dispositional Homophily? tify more altruistic members (via contributions
to the group), but do not exhibit preferences for
We now turn to whether relations in the sorority associating with them.
show dispositional homophily. Table 2 gives a Finally, model 5 suggests that altruists may
series of five regression models, one for each of be somewhat more likely to have altruistic best
the five relation types. Models 1–4 take as their friends. However, this may result from the fact
dependent variables a count of the number of that that there are more altruists than individ-
nominated individuals with the same disposition ualists, competitors, or unclassified respondents,
as the respondent. Model 5 estimates the prob- which could lead them to show dispositional ho-
ability that the respondent and her best friend mophily in random assortment (i.e., due to ran-
share the same disposition. domchance, altruistsaremorelikelytoselectan-
Models1–4includetwoindependentvariables otheraltruistasabestfriend). Achi-squaredtest
(altruist and unclassified) and two control vari- for independence indicates that once marginal ef-
ables (number of names given for that relation fectsaretakenintoaccountthereisnoassociation
and percent of friends in sorority). The two in- between the disposition of the respondent and
dependent variables are dummies, one of which the disposition of her best friend (p=0.12), sug-
sociologicalscience|www.sociologicalscience.com 550 December2014|Volume1
Simpson,Brashears,Gladstone,andHarrell NoEvidenceforAltruismHomophilyinNetworks
Table2:Regression Estimates for Models Predicting Altruism Homophily Across Four Types of Close
Relations in a Sorority (Study 2)
——————— SVO Homophily ———————
Important Social Best
Matters Activities Advice Sacrifice Friend
Altruist −0.17 −0.29 −0.23 1.01† 1.31∗
(0.22) (0.19) (0.21) (0.22) (0.59)
Unclassified −0.91∗ −1.08† −0.97† 0.78 −1.01
(0.37) (0.36) (0.35) (0.42) (1.14)
Number Given 0.44† 0.51† 0.46† 0.55†
(0.08) (0.09) (0.09) (0.08)
Percent Friends in Sorority 0.00 0.00 0.00 −0.00 0.00
(0.01) (0.01) (0.01) (0.01) (0.01)
Best Friend Reciprocity 1.09
(0.59)
Constant −1.42† −1.58† −1.68† −2.01† −2.45†
(0.37) (0.39) (0.39) (0.37) (0.86)
N 127 127 127 127 80
Dependent variable Count Count Count Count 0/1
Model Poisson Poisson Poisson Poisson Logit
†p<0.01; ∗ p<0.05.
gesting that the effect in model 5 is simply a that SVO tracks actual motivational and behav-
result of demographic availability (Blau 1977a). ioral differences comes from the Study 2 finding
Onthewhole,theseresultssuggestthatsoror- that those who made more altruistic responses
ity members are able to identify at least some were more often listed by their fellow sorority
altruists—namely those who sacrifice most for sisters as making greater sacrifices for the group.
the wellbeing of the sorority as a whole—but This result adds to a large body of work demon-
display no general tendency to preferentially as- stratingthatsocialvalueorientationisapowerful
sociate with them. There is, additionally, no predictor of altruistic behavior in a wide range of
evidence that egoists are forced into association settings (see Bogaert, Boone, and Declerck 2008).
with each other as a result of collective exclusion That said, it was costless for the respondents of
byaltruists. Wemustthereforeconcludethatthe Studies1and2togivealtruisticresponsestothe
altruism homophily hypothesis is unsupported. survey measure. Moreover, given that we used
the same measure in both studies, it is possible
Discussion of Study 2 that the absence of altruism homophily resulted
from measurement error. Our next study aims
Findings from Study 2 are highly consistent with to establish greater confidence in our conclusions
those of Study 1. The two studies converge on by expanding the range of altruism measures to
the same basic conclusion, but they also share includecostlybehaviors. Finally,ourresultsthus
a limitation: both studies used the same mea- fardonottelluswhetherpeoplehavepoorinsight
sure of altruism. Like other survey measures into their friends’ dispositions, as we expect, or
meant to tap into socially desirable preferences whether they possess accurate insight but simply
or behaviors, it may be prone to response bias, do not care about—and thus do not sort on the
although recent research suggests this is not the basis of—altruistic disposition. Study 3 allows
case (Willer et al. 2014). Additional evidence us to address this question.
sociologicalscience|www.sociologicalscience.com 551 December2014|Volume1