...

Deconstructing Nepotism

by user

on
Category: Documents
2

views

Report

Comments

Transcript

Deconstructing Nepotism
Deconstructing Nepotism
Sheheryar Banuri, Catherine Eckel, & Rick K. Wilson1
April 2016
Abstract
Nepotism arises when favoritism toward one’s group affects personnel and contract decisions. It
is widely regarded to be welfare reducing, yet it persists. In this paper we address the motives
for engaging in nepotism. Using naturally-occurring groups, we present a laboratory experiment
to test the strength of two motives for engaging in nepotism: beliefs regarding worker
performance within and outside one’s group, and the desire to reward members of one’s group in
the form of favoritism. Nepotism is introduced by allowing subjects to select their partners in a
trust game. The design varies two factors in a 2x2 design: the efficiency of group members and
the ability to select partners. We find beliefs about group member productivity to be the
predominant motive. These beliefs bear out: ingroup members trust each other more, and
reciprocate at higher levels, even when they are less productive. Selecting ingroup partners is
profitable. These results help explain why nepotism persists.
Keywords: Nepotism, Group Identity, Discrimination, Trust, Reciprocity
JEL Classification Codes: C92, D73, M51
1
Banuri: Development Economics Research Group, World Bank, 1818 H St NW, MC 3-356, Washington, DC,
20433 (e-mail: [email protected]); Eckel: Department of Economics, Texas A&M University, 4228 TAMU,
College Station, TX, 77845 (e-mail: [email protected]); Wilson: Department of Political Science, Rice
University, MS 24, Houston, TX, 77251 (e-mail: [email protected]). The authors have no relevant or material financial
interests that relate to the research described in this paper. The findings, interpretations, and conclusions expressed
in this paper are entirely those of the authors and do not necessarily represent the views of the World Bank, its
Executive Directors, or the countries they represent. We are indebted to Klaus Abbink, Rachel Croson, Sherry Xin
Li, Angela de Oliveira, Ngoc Phan and participants of the EITM summer school at Washington University-St.
Louis, NSF Conference on Politics Experiments at the University of Virginia, the NYU Experimental Political
Science Conference, and the Economic Science Association meetings in Tucson, AZ. Funding was provided by the
National Science Foundation (NSF SES-0921884). Any errors remain our own.
0 “Better to dance with the devil you know than the angel you don’t.” – English proverb
INTRODUCTION
Consider a manager who is in a position to hire one of two possible candidates with
identical levels of skill. One candidate has social ties to the manager, while the other candidate
is randomly selected from the general population. Which candidate will the manager select for
the position? Will his answer be the same if the candidate with social ties has a lower level of
skill? If the manager chooses the candidate with social ties, this can be considered nepotism,2 an
act that is widely regarded as inefficient and discriminatory, and yet is pervasive. There is little
agreement about why people engage in nepotism, and whether it is profitable for them. In this
paper, we address both these questions using naturally occurring groups and a novel
experimental design.
From the perspective of traditional economic theory, nepotism can only reduce profit,
since restricting employment to a favored group yields a less qualified candidate, on average,
than an open, full search (Becker 1971). Current empirical research also supports the notion that
nepotism is damaging for firm profitability (Bennedsen et al. 2007, Perez-Gonzalez 2006). Yet,
despite its impact on performance and efficiency, nepotism persists.
We define nepotism as the choice of a partner from one’s own primary group (kin,
friendship, or identity group) in a setting involving trust. Two motives have been offered for
engaging in nepotism: one is based on the claim that nepotism is rewarded - members of the
same group work harder, thereby reciprocating the trust placed in them (McConaugby, et al.,
2001, Kets de Vries 1993, Davis et al. 1997). The second motive (more common in the
literature) is the desire to confer benefits on group members (Vanhanen 1999, Brewer 1999,
Chen and Li, 2009, Brandts and Sola 2010, Belot and van de Ven 2011). While both motives are
plausible, it has been difficult to determine the relative contribution of each in determining
nepotism. This is due, in part, to the inherent difficulty of observing motives underlying
nepotism in the field.
2
Note that we define nepotism in a broader framework than simply kin-based relationships. Nepotism is defined as
“discrimination in favor” of a group member relative to the population (Fershtman et al. 2005, Becker 1971). This is
divergent from traditional biological definitions of nepotism, which stress kin-based relationships.
1 We employ a laboratory experiment to investigate the motives for engaging in nepotism.
Using the familiar trust game (Berg et al. 1995) as our starting point, we introduce two types of
responders: an ingroup member and a non-ingroup member.3 In one treatment responders can
choose which partner they prefer, and in another they are randomly assigned. Each treatment is
further divided into a condition where ingroup members are less efficient than others, and
another where ingroup members are equally efficient. This design allows us to identify motives
for engaging in nepotism and its impact on subsequent trust and reciprocity.
Motives are difficult to uncover in observational studies. Nepotism, for example, is often
either illegal or socially undesirable, and therefore observing it is problematic. Our design
allows us to eliminate social desirability, since all interaction is anonymous, and to manipulate
the efficiency of ingroup members, which is not easily observable in the field. There may be
other factors that contribute to nepotism than those we consider in our experiments (such as labor
market conditions, and reputations within the group). However, by restricting our attention to
key aspects of nepotism, we are able to focus on mechanisms through which it operates.
We address three central questions. First, why do individuals engage in nepotism, when
ingroup members are less efficient? Is it a strategic choice based on expectations of reciprocity
(beliefs), or is it out of concern for the wellbeing of the group (favoritism)? Second, what is the
impact of nepotism on trust and reciprocity? And finally, is engaging in nepotism profitable?
We find that individuals engage in nepotism because of their beliefs about the reciprocity
of ingroup members. We also find that ingroup members trust each other more, and that
nepotism has a positive impact on reciprocity among group members, leading to the primary
reason for the persistence of nepotism: it is profitable. We find that having an ingroup member
as a partner carries an earnings premium. Thus, we find evidence for nepotism as a social
dilemma: it is individually beneficial to engage in nepotism, but may be welfare reducing due to
negative externalities on meritocracy.
3
As opposed to a majority of the literature in this area, we use naturally occurring group to study nepotism: Rice
university’s residential college system, where students are randomly assigned to one of eleven different residential
colleges at the beginning of their freshman year. These residential colleges are the basis of our ingroups.
2 RELATED RESEARCH
A primary theme of research on nepotism is that it leads to inefficient outcomes. These
studies are diverse and include a wide variety of settings. For example, Brick et al. (2005) find
that excess compensation of boards of directors, which they interpret as evidence of cronyism,
predicts future firm underperformance. In a study of the emergence of liberal democracy in
Africa following the demise of colonialism, Englebert (2000) makes a similar argument, noting
that in countries where colonial institutions conflict with historical (formal or informal)
institutions, reversion to ethnicity-based resource allocation decisions is more likely, again
yielding inefficient outcomes. Nepotism also is a widespread phenomenon within professional
groups. To name two examples, Lentz and Laband (1989) show that children of doctors are 14
percent more likely to be admitted to medical school than are comparable other candidates,4 and
Singell and Thornton (1997) find that many dairy farmers in Utah regularly make hiring
decisions based on family and group ties, and that these farms underperform when compared
with farmers that do not.
So why does nepotism persist if it is welfare-reducing? There are two dominant
explanations: (1) beliefs, and (2) favoritism. Previous literature supports both beliefs about
trustworthiness (e.g., Ashraf et al 2006; Barr 2003; Buchan et al. 2008) and favoritism (Falk and
Zehnder 2007) as determinants of the level of trust in the investment game.5 The argument for
beliefs as a motive is based on the assumption that nepotism may be profitable because of the
superior performance of group members. This assumption may stem from the belief that group
members are in fact more capable, on average. Alternatively, higher productivity can arise from
enhanced monitoring due to social ties: Social ties can substitute for incomplete contracts or
weak legal institutions.6 Fearon and Laitin (1996) point out three factors that enhance trust and
cooperation within groups: greater information regarding other members of the group, individual
reputations that are sustainable and credible, and the availability of sanctions from within the
group when defection is observed. All of these factors may serve to enhance productivity and
4
Their research cannot rule out the effects of legacy and donations on college acceptance, and note that
intergenerational human capital transfers may also be a reason for larger acceptance rates.
5
Cox (2004) presents a “triadic” design that carefully explores the relationship between altruism and trust. He
argues that the amount sent in the trust game incorporates both altruism and trust. Thus in our context, trust would
be greater for fellow group members if there is greater altruism toward group members.
6
For example, McConaugby et al. (2001) argue that family-controlled firms are more likely to hire fellow group
members as a solution to the agency problem, and that reduced monitoring costs can yield higher firm valuations.
3 reciprocity. Even if fellow group members are less capable, they may be more likely to engage in
reciprocal behavior, effectively working harder than their more-qualified counterparts. Strong
group identity yields high motivation for reciprocity, and results in a greater preference for
nepotism.
The second dominant explanation for nepotism is that individuals with a strong sense of
group identity are more likely to select in-group members because of favoritism. This can be
either taste-based discrimination (Becker, 1971) or because of strong ties to a social identity
(Akerlof and Kranton, 2000).7 Nepotism is employed to benefit fellow group members, or
because of the higher value placed on interactions within the group. Tajfel and Turner’s (1979)
social identity theory suggests that individuals derive utility from group membership and actively
work towards maintaining ties within the group, culminating in favoritism.8 Behavior favoring
fellow group members is commonplace in these studies (for reviews, see Brewer and Brown,
1998; Messick and Mackie, 1989).9
Nepotism has implications for trust and reciprocity among group members: Slonim and
Garbarino (2008) show that merely providing subjects with the ability to select their partners
(based on gender and age) in the trust game increases trust. Brandts and Sola (2010) find higher
reciprocity among friends in a lab experimental study using the trust game, justifying the
selection of friends as partners, even when their efficiency is, by design, lower. Fiedler et al.
(2011) find a similar result among a sample of second life players (but not among a standard lab
sample of college undergraduates): reciprocity is higher among ingroup members. Fershtman, et
al., (2005) also find nepotism using a unique pool of subjects in Israel and Belgium. Orthodox
Jews trust other Orthodox Jews more than the general population and Belgian subjects are less
trusting of identifiable out-groups (Flemish vs. Walloon). Despite their breadth, these studies do
7
Psychologists have studied the effects of ingroup bias, understood as discriminating in favor of the primary group
of the individual relative to an out-group (Brewer, 1999). Once individuals establish their identities as part of a
particular group, pro-social behavior towards their group members increases based on this linkage. Thus, the
stronger an individual identifies with their group (relative to an outgroup), the greater the instance of pro-social
behavior. See Chen and Li (2009) and Goette et al. (2006) for recent examples.
8
Much of the research in this area utilizes lab experiments, and employs the minimal group paradigm (Billig and
Tajfel 1973), a relatively weak procedure for manipulating group identity in the lab. The procedure creates an
ingroup as well as a complementary outgroup.
9
In contrast, several studies find ingroup denigration. Lewis and Sherman (2003) document two such situations.
They show that individuals are more likely to hire out-group members when both applicants are unqualified (for
qualified candidates, the favoritism result holds), or when a marginally qualified ingroup member might confirm a
negative stereotype about the ingroup.
4 not distinguish between the two motives for nepotistic behavior, because they do not collect
information on expected reciprocity or strength of friendship.
Two studies support the idea that motives for nepotism lie with favoritism. In a field
experiment with children aged 6-8 and 10-12, Belot and van de Ven (2011) demonstrate that
younger children are more likely to select friends as group members regardless of performance.
But for older children performance becomes important. This study also finds that favoritism
improves performance, as group members who are selected exert more effort, consistent with
(accurate) beliefs about the productivity of fellow group members. In a field experiment in a
fruit-picking firm, Bandiera, et al., (2009) provide evidence that managers favor workers who are
socially close to them when it is costless to do so, but when it is costly, favoritism is eliminated.
These studies demonstrate the prevalence of the favoritism motive and suggest that beliefs about
higher performance can also play an important role.
While these studies are informative, they do not assess the relative strengths of beliefs
and favoritism as motives for selecting ingroup members under conditions where ingroup
members are less or equally efficient. Our experiment allows us to make this distinction. We
use laboratory experiments to examine which behavioral factors influence nepotism, and the
impact of nepotism on trust and reciprocity.
EXPERIMENTAL DESIGN
We modify the standard trust game (Berg et al. 1995) by introducing groups and
differences in partner efficiency. Our treatment of groups is somewhat different from prior
studies in that we do not have a true outgroup. We allow the proposer, the first mover in the trust
game, to choose either an ingroup or a non-ingroup member as responder (second mover). A
non-ingroup member is an individual that is “not in” the ingroup. Thus, by design, the
“outgroup” has no identity: it is a random individual from the population, reflecting common
situations where nepotism plays a role. This is an important distinction, since favoritism towards
one’s group is not the same as out-group dislike (Brewer 1999).10 Previous studies of ingroup
favoritism typically include an identifiable outgroup, confounding ingroup favoritism with
10
Brewer (1999) argues that ingroup favoritism and outgroup discrimination are separable phenomena, and thus it
may be unclear whether behavioral variation is driven by a preference for the ingroup or dislike towards the
outgroup. We have no identifiable outgroup, meaning that subjects cannot discriminate against outgroup members.
5 outgroup dislike. By structuring the game in this way, we are able attribute any preferential
treatment shown the ingroup members as favoritism, rather than outgroup dislike. We increase
the external validity of our study by using naturally-occurring groups, as explained below.
The experiment includes two factors with four treatments, in a 2x2 design. The first
factor is nepotism, labeled as “Random match” and “Nepotism” treatments. The difference
between the two is that the latter allows proposers to choose their partners, while the former does
not.11 The other factor varies the efficiency of ingroup members: “Equal efficiency” or “Low
ingroup efficiency.” In the Equal efficiency treatments, the standard trust game multiplier of
three applies to both ingroup members and others. In the Low ingroup efficiency treatments, a
lower multiplier of 2.5 is applied to ingroup members, while transfers to others retain the original
multiplier of three. Each treatment combination is conducted with an independent sample in a
between-subjects design.
Figure 1: Experiment timeline
Figure 1 shows a timeline of events in each session. The experiment begins with a pregame survey (collecting demographic information) followed by three games (Trust, Dictator,
Risk, presented in random order by session) and then a post-game survey (collecting game
specific information).
Nepotism game
Proposers and responders in the trust game are endowed with 20 tokens (with each token
equal to $0.50). In the Nepotism treatments, proposers make three decisions in the trust game;
(1) choose between ingroup and other, (2) choose how many tokens to send to responder, and (3)
11
In the Random match treatment, during the post-game survey we ask the subjects if they could choose, which
group would they choose their responder from, giving us survey based information on partner preferences, which
has no bearing on outcome of the game, and is not incentive compatible.
6 estimate the number of tokens sent back by the responder. All three decisions are incentivized.12
We use the strategy method, meaning that each subject makes a trust decision and a return
estimate for both possible counterparts: the ingroup and the other.13 Responders make two
decisions: (1) estimate how much they will receive, and (2) choose how much to send back for
all possible amounts received (using the strategy method). Both these decisions are incentivized.
We use complete information, meaning that both proposers and responders are aware that
proposers choose partners.14
In the Random match treatment, the setup is identical to above except for decision (1).
Proposers do not choose a responder group, but instead are informed that there is a 50% chance
they will be matched with either group. Both proposers and responders are aware that actual
matching is random. In the post-game survey for this treatment, proposers are asked which
group they prefer to be matched with if they could choose; however, their response has no
bearing on the matching protocol.
In the Low ingroup efficiency treatments, proposers are informed that pairing with an
ingroup member yields a lower multiplier (of 2.5, compared to a multiplier of 3 for a pairing
with the non-ingroup counterpart). In the Equal efficiency treatment the multiplier is 3, and is
the same for ingroup members and others. Both proposers and responders are informed of this.
Preference controls
To measure subject’s key preferences – favoritism and risk aversion – we conduct additional
games. Favoritism is measured by a variation on the standard dictator game. Proposers are
endowed with 20 tokens (each worth $.50 USD) and are asked how much they want to send to an
ingroup responder, and how much they want to send to a non-ingroup responder. They make this
decision simultaneously (on a single screen, with the order randomized). In the Nepotism
12
Belief estimates are rewarded using a binary scoring rule. Subjects receive a 2-token ($1) bonus if they estimate
correctly.
13
As explained in the next section, subjects are informed that there is a chance their first choice responder will not
be available, and are thus asked to take both ingroup and other decisions simultaneously. The matching protocol
(described below) ensures that there is some chance they will be matched with their second choice of responder.
This provides us with appropriate counterfactual data for each subject.
14
The trust game is played once, all participants have fixed positions, and the pairings are anonymous. The game is
computerized using z-tree (Fischbacher, 2007).
7 treatment, proposers select the group (ingroup or other) to which they send tokens.15 The
protocol in the Random match treatment is identical, but the choice of responder group is
removed.
We implement a simple measure of risk aversion as in Eckel and Grossman (2008),
wherein subjects are asked to select one of six possible gambles. Appendix A displays a
screenshot of the gambles viewed by the subjects. Gambles one through five increase in both
expected value and variance. Gamble six increases in variance, but holds the expected value the
same as in gamble five. Each gamble has a 50% chance of paying out a low amount or a high
amount.
In addition, as an additional control variable, we measure the strength of group identity
using a 7-point Likert-scale survey question (“How strongly do you identify with members of
[primary group]?”) and use survey measures of generalized trust and perceptions of generalized
fairness (from the World Values Survey).
EXPERIMENTAL PROCEDURES
We conducted the experiment at Rice University, making use of their Residential College
system. Upon entrance to the university as freshmen, undergraduates are randomly assigned to
one of eleven Residential Colleges. Colleges have their own dining halls, dorms, and faculty
advisors, which cultivates a strong group identity. Furthermore, a week-long orientation period
for freshman and regular competitions among colleges further establish strong group
affiliations.16 These residential colleges serve as the primary group affiliation for undergraduates
on campus.
Subjects were recruited during lunch and dinner hours at the dining hall for each
particular college. The experiments explicitly make reference to the primary college under
15
Again, both decisions are incentive compatible as the matching protocol allows for a chance that the proposer will
not be matched with the responder they selected.
16
We utilize Rice University’s residential college system as the basis for our groups. This is useful as (1) we can
implement the partner choice mechanism with an in-group but no identifiable out-group, and (2) we can conceal the
identity of the partner so as to mitigate post-game play. Furthermore, random assignment assures that potentially
confounding factors are not correlated with treatments, and the possibility of selection bias in group assignment is
avoided. However, one threat to randomization is the possibility of legacy admissions; i.e., undergraduates
requesting to be assigned to a particular college based on previous affiliation. The number of legacy admissions is
relatively small at Rice, and given the relatively small sample of subjects, the probability of legacy students
participating in the study is low. For more information on the residential college system, please see:
http://students.rice.edu/students/Colleges.asp
8 observation in order to establish a basis for engaging in nepotism. All partners are anonymous,
and no identifiable characteristics (other than group membership) are revealed.
Table 1: Study Design
Nepotism
Equal multiplier
N = 78
Low ingroup multiplier
N = 78
Random match
N = 68
N = 72
Table 1 contains the overall design of the study. Sessions were conducted at the
Behavioral Research Lab at Rice University in April and October 2009, and October 2010. A
total of 296 subjects participated in the study. There were a total of 17 sessions, with an average
of 16 subjects in each session. In all cases, the ingroups were labeled in accordance with the
name of the residential college.
As detailed above, the experiment consisted of an initial short entry survey (collecting
demographic information), and the three games (Dictator, Nepotism, and Risk described above),
followed by a post-game survey. Each game started with instructions, two examples, and a short
quiz to test understanding, followed by the game itself.
The overall experiment has a number of games, and well as two types of groups (ingroup
and other), which brings up concerns about order effects. To minimize these concerns, we
randomized a number of things in the experiment. First, we randomized the order of the games:
subjects had an equal chance of starting with the dictator, nepotism, or risk game. Second, we
randomized the order of the groups: each subject made ingroup decisions on the left side of the
screen or the right side (with the other decisions on the adjacent side).17 Third, we randomly
selected one game (dictator, nepotism, or risk) for payment at the end of the session, so as to
induce independence in decisions across games.
Upon arriving at the lab, subjects signed in and were asked to confirm their residential
college name and then promptly seated at a terminal. Instructions referred to ingroup subjects by
the name of their college (for example, “individuals in Baker College”) and others were referred
to as “individuals not in Baker College but from the Rice University population” (emphasis
17
Ingroup and other decisions were made on the same screen to avoid biasing the subjects.
9 added).18 No feedback was provided on earnings between tasks during the experiment. At the
end of the session, the experimenter entered the lab area and asked for a volunteer. The
volunteer rolled a die to determine the game that would be paid for in the session. If the risk
game was selected for payment, subjects were directed to the payment area and rolled a six-sided
die. A roll of 1 through 3 gave them the low amount listed for their chosen gamble, and the roll
of 4 through 6 gave them a payout of the high amount.
Subjects were assigned to one of two roles at the beginning of the session: proposer or a
responder. Subjects kept this role through the entire session. In all sessions, all proposers
belonged to the ingroup, while approximately half of the responders belonged to the ingroup.
The remaining responders belonged to residential colleges other than the ingroup’s. All
participants were aware of this.
Appendix A contains screenshots of the proposer and responder decision screen,
respectively (figures A.2 and A.3). In the Nepotism treatment, proposers had the option to select
the group that their counterpart would be drawn from for each task. In the Random match
treatments, subjects were not given this option, but were told that there would be “approximately
a 50% chance” that they would be matched with a responder from either group (i.e., their own
group or not).
Subjects were paired using a matching algorithm that is a variation on one developed by
Castillo and Petrie (2010) for eliciting preferences for partners in a public goods game. For the
Nepotism treatments, one proposer was selected at random. His preferred group choice was
noted, and then a responder was randomly selected from his preferred group. Next, a second
proposer was randomly selected and given his first choice of group from the remaining candidate
responders. This process continued until each proposer was matched with a responder in the
session. In the event that the pool of responders from any particular group was exhausted, but
still had been requested by a proposer, then the proposer was matched with a responder from the
alternate group. In the Random match treatments, each proposer was matched with a responder
at random. The matching algorithm was triggered once all subjects had completed all tasks and
the surveys. Each proposer was matched with a single responder.
18
Note that the “others” belonged to Rice University, which constitutes another in-group for the subjects, but one
that is not as salient as their own college.
10 RESULTS
In this section we first examine the main treatment effects. We then focus on the motives
for nepotism: favoritism or beliefs. Next, we discuss the impact of nepotism on trust and
reciprocity. Finally, we address the question of whether nepotism is a profitable strategy.
Figure 2 presents preferences for nepotism in all treatments. Note that in both the
Random match and Nepotism treatments, over 80 percent of proposers prefer/choose ingroup
members as their partner when the ingroup member is as efficient as the general population.19
Nepotism, however, is rarely costless, and we find that when it is costly, it is less utilized.
Across both the Random match and Nepotism treatments, when ingroup members are less
efficient than the general population, 44 percent of the proposers choose ingroup members as
their partner, significantly lower than when ingroup members are as efficient (from 85 percent to
44 percent, two-sample z-test of proportions, z=5.19, p<0.01).20 Note, however, that even when
ingroup members are less efficient, a large proportion of subjects still choose them as partners.
19
Please note that in the random match treatment, subjects cannot directly choose their partner, but indicate their
preference in the exit survey. The figure reports this metric for the random match treatment.
20
There are no significant differences between the Random match and Nepotism treatments both when ingroup
member efficiency is identical to or lower than the general population.
11 Figure 2: Percentage of subjects choosing/preferring ingroup members (95% confidence interval)
Motives for nepotism
We now test for the determinants of nepotism. Our primary variables of interest are (1)
favoritism towards ingroup members21 and (2) beliefs regarding the performance of the trustee.22
We also control for risk preferences, measured using the Eckel-Grossman risk elicitation method
(2008).23 Finally, we add controls for gender, group identity,24 and the survey-based attitudinal
measures of trust and fairness.25
21
We measure favoritism by constructing a measure using the dictator game, defined as the amount donated to an
ingroup member less the amount donated to a member of the general population. If the difference in dictator game
giving favors the ingroup member then this variable is positive. Stronger altruism toward group members is likely to
play a role in partner selection. Thus, this variable measures the extent to which the subject is altruistic towards his
ingroup relative to his altruism towards the general population, using the following formula: ! =
    −    ℎ .
22
We measure beliefs in the following way: after subjects make their partner choice (depending on treatment) and
trust decisions, we inform them the total available to their ingroup and other responders. We then ask them how
much of that total they expect back from each responder. The elicitation is incentive compatible in that subjects are
paid a bonus for guessing correctly, and zero otherwise. The difference in beliefs about the performance of ingroup
members and others is our measure, using the following formula:
   
   ℎ
! =
−
   
   ℎ
23
Ben-Ner and Putterman (2001) argue that trust is necessarily a risky decision due to lack of information between
partners (see also Eckel and Wilson 2004, who find little relationship between risk attitudes and trust, and Schechter
2007 who does find a positive relationship between risk-tolerance and trust). The decision to trust is inherently risky
due to the possibility of betrayal (see Bohnet and Zeckhauser 2004). Trust decisions involve uncertainty regarding
12 Our dependent variable is a dummy variable equaling 1 if the subject chose an ingroup
member as partner in the trust game. Therefore, we use a probit model to estimate the
probability of the subject choosing an ingroup member as partner. Furthermore, we restrict our
analysis to the Nepotism treatment, where partner choice is incentive compatible.26 The results
are provided in table 2 (marginal effects are reported).27
Table 2: Probability of selecting ingroup members as partner (marginal effects)
Dependent variable: Partner choice (1 = Chose ingroup member)
Treatment
(1 = Low ingroup efficiency)
Beliefs
(Ingroup less other)
Favoritism
(Ingroup less other)
Risk Preferences
(6 = Risk seeking)
Gender (D)
(1 = Female)
Group Identity
(7 = Strong ingroup identity)
Generalized Trust
(7 = More trusting)
Generalized Fairness
(7 = Others are fair)
I
II
III
IV
V
-0.339***
-0.422***
-0.355***
-0.410***
-0.397***
(0.11)
(0.10)
(0.10)
(0.11)
(0.11)
0.069*
2.471**
(1.09)
0.072
2.580**
(1.10)
0.064
2.881**
(1.11)
0.067
(0.04)
(0.04)
(0.04)
(0.05)
-0.116***
-0.094**
(0.04)
(0.04)
2.655**
(1.11)
0.153
(0.10)
0.011
(0.04)
-0.017
(0.04)
0.013
(0.04)
behavior of the counterpart; this uncertainty is diminished in interactions between individuals with a common social
identity. Individuals choosing between in-group partners and “others” have a shared history with in-group members,
allowing better calibration of reciprocity beliefs. Conversely, the perceived distribution of reciprocity levels in the
general population is more dispersed, which in turn makes the choice of an individual from the general population a
riskier prospect.
24
This was measured using a 7 point Likert-scale response to the question “To what extent do you identify with
other members of {Insert group name}?” taken from Levin and Sidanius (1999). This provides us with an additional
measure of favoritism towards the ingroup.
25
Trust is measured by a 7 point Likert scale response to the question “Generally speaking, would you say that most
people can be trusted, or that you need to be very careful in dealing with people?” taken from the 2005 version of
the World Values Survey, accessible at http://www.wvsevsdb.com/wvs/WVSAnalize.jsp. Fairness is measured by a
7 point Likert scale response to the question “Do you think that most people would try to take advantage of you if
they got a chance, or would they try to be fair?” taken from the 2005 version of the World Values Survey, accessible
at http://www.wvsevsdb.com/wvs/WVSAnalize.jsp.
26
We conduct the same analysis with the Random Choice treatment data in appendix B (table B.2).
27
Model coefficients are provided in table B.1 in appendix B for the interested reader.
13 Constant
0.733
0.723
0.773
0.803
0.815
Pseudo R-squared
Chi-squared
P-value
Log Likelihood
0.263
26.4
0.000
-37.09
0.225
22.6
0.000
-39.01
0.307
30.9
0.000
-34.86
0.416
41.9
0.000
-29.38
0.438
44.0
0.000
-28.30
78
78
78
78
78
Observations
Note: * p<0.1, **p<0.05, *** p<0.01. Probit specification, standard errors in parentheses. Table reports marginal
effects. Dependent variable takes on a value of 1 if subject selected ingroup member as partner in the trust game.
Data is from the Nepotism treatment. The “Beliefs” variable measures the difference in beliefs about reciprocity
between ingroup members and others. Thus, positive values mean subjects expect more back from their ingroup,
while negative values mean subjects expect less back from their ingroup (relative to others). The “Favoritism”
variable measures the difference in dictator giving to ingroup members and others. Thus, positive values are
subjects giving more to ingroup relative to others, while negative values are subjects giving more to others.
Models 1, 2, and 3 estimate partner choice with a dummy variable for the treatment
(equaling 1 if the ingroup respondent is less efficient, and 0 otherwise) and beliefs and/or
favoritism variables. Model 4 adds risk preferences, while model 5 adds gender and social
preference controls. We first note that the coefficient on the treatment dummy is significant and
negative, replicating what we observed in figure 2: when ingroup members are less efficient,
subjects are significantly less likely (approximately 40 percentage points less likely) to choose
ingroup members as partners (p<0.01). We also find that subjects with higher beliefs regarding
ingroup member performance (relative to others) are significantly more likely to select them
(p<0.05), and this effect does not vary by treatment (p=0.70 for the interaction term). In fact, for
every 1% increase beliefs about ingroup member performance, subjects are approximately 2.7%
more likely to select ingroup members as partners. Thus, we find that beliefs regarding partner
performance matter for nepotism.
Second, we test the relationship between favoritism towards the group and nepotism. We
find little evidence in favor of the favoritism channel. The coefficient is weakly significant in
model 2 (p<0.10), and does not remain so once we add additional controls. Similarly, group
identity is also not significant in our data. The effect of favoritism also does not vary by
treatment (p=0.53). This suggests that subjects engage in nepotism primarily due to beliefs.
Table B.2 in the appendix conducts the same analysis for the Random match treatment, where
the choice between ingroup member and other was not incentive compatible (subjects were
simply asked for their preference in the post-game survey, prior to decisions being revealed).
We find similar effects for the treatment (Low ingroup efficiency) and for beliefs regarding
14 performance. In this table, however, we also find that favoritism towards ingroup members also
predicts partner choice. This implies that favoritism may not play as much of a role when real
stakes are attached to the partner choice decision.
Finally, we note that the coefficient on risk preference is significant and negatively
related to the choice of an ingroup member as partner (p<0.05). Risk-seeking subjects are less
likely to choose ingroup members as partners in the trust game. The negative coefficient is also
significant when ingroup members are less efficient (p<0.05). This implies that risk-averse
subjects are more likely to engage in nepotism.
Overall, we find that nepotism is a strategic decision, motivated mainly by beliefs about
ingroup member performance. We do find some suggestive evidence in favor of favoritism, but
only when there are no payoff implications. We also find some evidence of risk-aversion
informing the choice to engage in nepotism. People engage in nepotism mainly because they
expect better (more profitable) outcomes for themselves.
Impact of nepotism on trust
In this section we analyze the amount sent by proposers in the trust game across all four
treatments.28 This allows us to estimate the impact of nepotism on trust. Therefore, our major
variable of interest is the Nepotism treatment dummy. Since we use the strategy method, each
subject takes a decision for ingroup members as responders, and for others as responders. We
estimate tobit models for the amount sent to the responder in the trust game, pooling all decisions
and clustering by individual. In addition to the treatment dummies, we add in controls for
whether the target responder is an ingroup member (dummy), whether the target responder is
chosen as partner (dummy), risk preferences, gender, trust and fairness perceptions, as well as
group identity. Table 3 displays the results.
28
Recall that in the Random Match treatment, subjects indicate a preference for participating in the trust game with
a member of their own group, or a randomly selected individual. This preference has no bearing on who they are
ultimately matched with. In the Nepotism treatment, counterparts are matched in accordance to the subject’s group
choice. In both treatments, subjects are asked to make both decisions (one for an ingroup responder and another for
the other responder).
15 Model 1 introduces the two treatment dummies (for the Nepotism treatment, and the
Equal efficiency treatment), while model 2 adds an interaction term. Model 3 adds control
dummies for whether the decision is for an ingroup responder, and whether the decision is for a
responder that is chosen as partner. Finally, model 4 adds the gender, risk and social preference
controls.
Table 3: Trust - Tokens sent by proposer
Dependent variable: Trust
Nepotism Treatment
(1 = Nepotism)
Equal Efficiency Treatment
(1 = Equal ingroup efficiency)
Nepotism X Equal Efficiency
Ingroup
(1 = Sent to ingroup)
Chosen as Partner
(1 = Responder chosen as partner)
Risk Preferences
(6 = Risk seeking)
Gender (D)
(1 = Female)
I
II
III
IV
1.346
(1.56)
3.267**
3.202
(1.97)
5.279**
3.204
(1.97)
5.280**
4.144**
(1.96)
5.891***
(1.56)
(2.39)
(2.39)
(2.25)
-3.802
-3.797
-4.752
(3.13)
(3.12)
(2.90)
1.484***
(0.42)
1.342***
1.474***
(0.41)
1.315***
(0.42)
(0.41)
1.153**
(0.47)
-4.701***
(1.25)
16 Group Identity
(7 = Strong ingroup identity)
Generalized Trust
(7 = More trusting)
Generalized Fairness
-0.260
(0.58)
0.823
(0.60)
0.492
(7 = Others are fair)
Constant
(0.58)
7.094***
6.128***
4.710***
-1.997
(1.26)
(1.38)
(1.41)
(4.45)
0.006
0.055
-836.80
0.007
0.062
-835.40
0.010
0.001
-833.40
0.036
0.000
-811.40
Observations
296
296
296
296
Left censored observations
38
38
38
38
Right censored observations
59
59
59
59
Pseudo R-squared
P-value
Log Likelihood
Note: * p<0.1, **p<0.05, *** p<0.01. Tobit specification, standard errors in parentheses. Dependent variable is the
number of tokens sent in the trust game. Data in models I and II is from the equal efficiency treatment, while data in
models III and IV is from the low ingroup efficiency treatment. Variables are censored at 0 (lower limit) and 20
(upper limit).
Both the Nepotism and Equal efficiency treatments have a positive coefficient in the
models. In particular, raising the efficiency of the ingroup member has a positive impact on trust
overall, significantly increasing trust, on average, by between 3 and 6 tokens ($1.50-$3.00;
p<0.05). Second, the Nepotism treatment has a positive coefficient, but it is not significant
(p=0.39). When we add the interaction term and controls, however, we find that the Nepotism
treatment in the Low ingroup efficiency treatment is significant (p<0.05), increasing trust by 4
tokens ($2.00). In the Nepotism and Equal efficiency treatment, trust is lower and marginally
significant (p=0.103). Thus, Nepotism may have some positive impact on trust, but it is not
always the case, particularly when ingroup members are as efficient as others.29 We do find,
however, that subjects trust their ingroup more overall: subjects send 1.5 additional tokens
($0.75) to their ingroup (p<0.01). In addition, subjects send 1.3 additional tokens ($0.65) to their
chosen/preferred partner (p<0.01). We also find that trust is a risky decision in this context, with
more risk-tolerant subjects trusting more (p<0.05), and that women send significantly less than
29
This result is similar to the findings of Slonim and Garbarino (2008), who find that partner choice induces higher
levels of trust. In their framework additional information regarding partner gender and age were available to
subjects, whereas in our study, the only information available is that of ingroup status and efficiency.
17 men ($2.35 less; p<0.01), which is consistent with Buchan et al. (2008), and many studies in the
survey by Croson and Gneezy (2009). 30
Overall, we find little evidence for Nepotism having a positive impact on trust. We do
find that the equal efficiency treatments have a positive and significant impact on trust overall,
even with non-ingroup members.31 Importantly, however, we find that subjects consistently send
more to their ingroup (about 1.5 tokens on average), and send more to the respondents that
choose as their partners, across all treatments.
Impact of nepotism on reciprocity
We now turn to responder trustworthiness/reciprocity in this section. We utilize the
strategy method in measuring the reciprocity levels; i.e., responders make a decision for every
possible amount sent by the proposer, making a total of 11 decisions. In addition, responders can
return the amount that they received, as well as their own initial endowment.
Figures 3 and 4 display reciprocity decisions (tokens returned) of ingroup members
(figure 3) and others (figure 4). The figures compare differences in reciprocity in the Random
Match and Nepotism for the Low ingroup efficiency treatments (figures 3a and 4a) and the Equal
efficiency treatments (figures 3b and 4b). In each graph the dotted line indicates that the
responder just returns the amount sent, and the dashed line indicates that the amount returned
equalizes the gains from the amount sent. Returns have the usual feature in that they are linear
and upward-sloping in trust. From the graphs it is plain to see that average returns in the
Nepotism treatment are nearly always higher than average returns in the Random match
treatments, both for ingroup members and for others. However, the difference in treatments in
particularly pronounced in the Low ingroup efficiency treatment, and particularly from ingroup
members. In fact, in just one case do the returns approach an equal split: among group members
in the Low treatment when nepotism is available. For the Equal efficiency treatment, subjects
appear to be unaffected by the Nepotism treatment.
30
Note that we have not explicitly controlled for expectations in the regressions, since expectations are elicited
subsequent to the trust decision and so are likely to be endogenous.
31
While we would expect that raising ingroup member efficiency has a positive impact on trust among ingroup
members, the increase in trust with others comes as a surprise. This may be due to the fact that trust decisions were
made simultaneously, and hence increased trust in the ingroup had a positive spillover effect on trust overall. Our
data does not allow us to test this motivation, however.
18 Ingroup Reciprocity -­‐ Low ingroup efficiency Ingroup Reciprocity -­‐ Equal efficiency 35 35 Random match NepoDsm Same amount returned Equal split 25 20 30 Amount Returned Amount Returned 30 15 10 5 Random match NepoDsm Same amount returned Equal split 25 20 15 10 5 0 0 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 Amount Sent 8 10 12 14 16 18 20 Amount Sent Figure 3a-b: Average returns by ingroup members by treatment Other Reciprocity -­‐ Low ingroup efficiency Other Reciprocity -­‐ Equal efficiency 35 35 30 Random match NepoDsm Same amount returned Equal split 25 20 Amount Returned Amount Returned 30 15 10 5 Random match NepoDsm Same amount returned Equal split 25 20 15 10 5 0 0 0 2 4 6 8 10 12 14 16 18 20 0 Amount Sent 2 4 6 8 10 12 14 16 18 20 Amount Sent Figure 4a-b: Average returns by others by treatment
We formally test the effect of the treatments on reciprocity separately for ingroup
members, and for others, using a linear regression with individual clusters (tobit regressions with
left censoring at zero give us similar results). The dependent variable is the amount returned by
responders. Our main independent variables are: the amount sent by the proposer, the treatment
dummies and their interaction. As above, we also control for gender, group identity (for others,
this question refers to how strongly they identify with the ingroup), generalized trust, and
fairness preferences. Models 1 and 2 are for ingroup responders, while models 3 and 4 are for
others. Models 1 and 3 use just the treatment dummies and the amount sent by the proposer,
while models 2 and 4 add controls. Table 4 provides the estimation results.
19 Table 4: Reciprocity: Tokens returned by responders to Ingroup and Others
Dependent variable: Reciprocity
Ingroup
Ingroup
Tokens Sent by Proposer
Nepotism Treatment
(1 = Nepotism)
Equal Efficiency Treatment
(1 = Equal ingroup efficiency)
II
III
IV
0.923***
0.923***
0.912***
0.912***
(0.09)
(0.09)
(0.09)
(0.09)
3.521*
(1.95)
1.439
7.058***
(2.42)
5.884**
1.707
(1.90)
0.221
4.084*
(2.24)
3.131
(1.94)
(2.55)
(1.95)
(2.16)
Gender (D)
(1 = Female)
Group Identity
(7 = Strong ingroup identity)
Generalized Trust
(7 = More trusting)
Generalized Fairness
R-squared
P-value
Observations
Others
I
Nepotism X Equal Efficiency
(7 = Others are fair)
Constant
Others
-7.671**
-4.176
(3.72)
(3.51)
-0.464
(1.80)
-0.288
(0.75)
1.218*
(0.62)
1.999***
-0.210
(1.81)
0.962*
(0.54)
1.269*
(0.64)
1.378**
(0.49)
(0.61)
-1.263
-14.10***
-0.334
-14.10***
(1.22)
(4.74)
(1.39)
(2.68)
0.313
0.000
0.417
0.000
0.235
0.000
0.365
0.000
693
693
935
935
Note: * p<0.1, **p<0.05, *** p<0.01. Linear regression with individual clusters, standard errors in parentheses.
Dependent variable is the number of tokens returned in the trust game. Data in models I and II is for ingroup
subjects while data in models III and IV is other subjects.
The regressions confirm the observations from the figures. First, we find that reciprocity
is increasing in trust: subjects return one token ($0.50) for every two tokens sent by proposers
($1.00). Furthermore, the Nepotism treatment has a positive and significant impact on
reciprocity among group members overall (p<0.10), but a smaller effect on ingroup reciprocity
when ingroup members are equally efficient as others. The Equal efficiency treatment has a
positive impact on reciprocity among group members but is not significant (p=0.46). We do
find, however, that reciprocity increases in the Random match-Equal efficiency treatment
(p<0.05), but is significantly lower in the Nepotism-Equal efficiency treatment (p<0.05). Thus,
the effect of Equal efficiency on ingroup member reciprocity varies depending on the existence
of Nepotism.
20 For others’ reciprocity, we observe a similar pattern as for the ingroup, except that the
coefficients are smaller and not as significant. We do observe a positive effect of Nepotism on
reciprocity, but only in the Low ingroup efficiency treatment (p<0.10). The independent effect of
the Equal efficiency treatment is not significant (p=0.15) and neither is the treatment interaction
(p=0.24). Finally, we also observe a significant relationship between trust and fairness with
reciprocity, meaning that more trusting subjects and those with higher perceptions of fairness are
also more reciprocal, regardless of ingroup status.
Importantly, we find that the presence of nepotism has a significant and positive impact
on reciprocity among the ingroup. This same pattern holds for others, but the effect is
considerably smaller, and not significant. Ingroup members in the nepotism treatment send back
3.5 extra tokens ($1.75) on average, while others return half that amount ($0.85). This indicates
that ingroup members reward nepotism. Knowing that they have lower productivity, ingroup
members are willing to reward generously those who choose them.
Impact of nepotism on earnings
In the analysis above, we found that nepotism is motivated by beliefs about performance
of the ingroup. We find that subjects send more to their ingroup, and that nepotism has a
positive impact on reciprocity, with a larger effect on the ingroup. We now ask whether
nepotism is profitable for ingroup members.
Since we use the strategy method, proposers make two decisions, one for each group of
responders. We can thus calculate earnings based on the mean level of reciprocity for each
amount sent (again, because of the strategy method, we collect responder reciprocity decisions
for each level of trust). Using this method, we can estimate what each proposer would earn by
being partnered with their ingroup or with the other for each treatment. We take the difference in
earnings between being paired with an ingroup member and with a non-ingroup member. Figure
7 displays the results by treatment (independent of the choice of partner), with positive amounts
indicating higher earnings from ingroup relative to others.
In all treatments except one, partnering with the ingroup is more profitable. Both
Nepotism treatments yield significantly higher earnings when pairing with the ingroup (Equal
efficiency: 1.70 tokens-$0.85; p<0.01; Low ingroup efficiency: 1.25 tokens-$0.63; p<0.01). In
the Random match treatment when the ingroup is equally efficient, ingroup members still benefit
21 from being paired with the ingroup, though the increase in earnings is modest (0.44 tokens $0.22; p<0.10). However, in the Random match treatment with lower ingroup member
efficiency, subjects are better off partnering with others, earning 1.92 tokens ($0.96) less when
being paired with ingroup members (p<0.01). Thus, we can clearly see that Nepotism has a
positive impact on proposer earnings, even when the efficiency of the ingroup members is low.
Figure 7: Relative earnings: Earnings with an ingroup less earnings with others (95% confidence
interval)
Figure 7 reports earnings for all proposers, including those that chose their ingroup
members and those that chose other responders as partners. In the next figures (8a and 8b) we
split the sample by Nepotists (those that chose or preferred an ingroup members as partner) and
Non-nepotists (those that chose or preferred a non-ingroup member as partner). We find that
earnings from pairing with group members is higher for both non-nepotists and nepotists alike.
Thus, the positive impact of nepotism on ingroup earnings is true for both sets of responders
(though this difference is not significant in the Equal efficiency treatment for non-nepotists).
Furthermore, when nepotism is available, ingroup member efficiency has no significant impact
22 on the earnings of nepotists (fig. 8a), indicating that nepotists are completely compensated for
the lowered ingroup member efficiency.32
Figure 8a-b: Relative earnings: Nepotists and non-nepotists by treatment (95% confidence
interval)
Overall, we find that nepotism is profitable for ingroup members, even when ingroup
members are less efficient than others. This is mainly due to higher levels of trust among
ingroup members, and higher levels of reciprocity when nepotism is possible. We find that both
nepotists and non-nepotists benefit from partnering with ingroup members when the choice is
available to them. Finally, we also find strong evidence of positive efficiency impacts: when
ingroup members are more efficient, ingroup members are better off partnering with ingroup
members.
CONCLUSIONS AND POLICY IMPLICATIONS
In this paper we present the results of a study designed to examine the factors that
motivate nepotistic behavior, and the impacts of such behavior on trust, reciprocity, and profits.
The study uses a variation on a well-studied experimental game: the trust game. We find that
individuals engage in nepotism due to beliefs about ingroup members. Furthermore, the
efficiency of ingroup members is a relevant factor: subjects are less likely to partner with
32
For nepotists, the premium for selecting a less efficient ingroup partner is large, and larger than for non-nepotists.
This is because nepotists have higher relative trust in ingroup members as compared with others, and that trust is
reciprocated.
23 ingroup members when they are less efficient. In addition, more risk-tolerant individuals are
more likely to partner with people outside their group.
We find that there is greater trust towards ingroup members. We also find that nepotism
has a positive impact on reciprocity: first movers are rewarded when they select fellow ingroup
members as partners.
Taken together, these results demonstrate why nepotism exists and
persists, even in the presence of costs. Becker (1971) argued that engaging in discrimination (of
any variety) would reduce profits since a more efficient worker would be available with a wider
search. However, this reasoning does not take into account that individual behavior also varies
under the two conditions. We find that nepotism compensates for decreased efficiency through
increased trust among ingroup members, and increased reciprocity by trusted ingroup
responders. We find that when nepotism is available, partnering with group members is always
profitable. Hence, we find evidence for nepotism as a social dilemma: it persists because it is
individually profitable, but still may be welfare reducing overall.
Our results are similar to those found by Brandts and Sola (2010), Fiedler et al. (2011),
and Belot and van de Ven (2011), which show that reciprocity increases with reduced social
distance. Fershtman et al. (2005) also shows that trust is higher among group members, but has
no real impact on reciprocity. We build on these studies by identifying the motives for engaging
in nepotism, and find nepotistic behavior to be rational: subjects engage in nepotism because
they believe in greater productivity by fellow group members. Nepotism is a profitable strategy.
We also find that nepotism has limited impact on trust, but greater impact on reciprocity,
yielding higher returns.
Our stylized representation of nepotism differs from the “real world” form of nepotism in
two ways. First, the analysis we present is static, i.e. the trust game is played a single time and
ends. Nepotism may have a significant long term component with impacts on inequality and
meritocracy that we do not address here. In addition, these repeated interactions may provide
further incentives for individuals to engage in inefficient behavior. Further research is needed to
estimate the long run impact of engaging in such behavior. Second, this paper is divorced from
any externalities resulting from the employment of low-quality workers, the effect of which is
(as yet) unknown.
Our results have interesting implications for policy. First, we show clear incentives for
group polarization. Partnering with ingroup members pays off, even when ingroup members are
24 relatively inefficient. Organizations that do not allow nepotism may not be availing themselves
of productivity enhancements. So, should we eliminate anti-nepotism rules? The answer lies in
the purpose of the rule itself. If the purpose is to reduce discrimination for its own sake, then
maintaining the rule is desirable. However, if the purpose is to maximize profits, then the
relationships between employers and workers need closer examination.
25 REFERENCES
Akerlof, George, and Rachel Kranton. 2000. “Economics and Identity.” Quarterly Journal of
Economics, 115(3): 715-753.
Ashraf, Nava, Iris Bohnet, and Nikita Piankov. 2006. “Decomposing Trust and Trustworthiness.”
Experimental Economics, 9(3): 193-208.
Bandiera, Oriana, Iwan Barankay, and Imran Rasul. 2009. “Social Connections and Incentives in
the Workplace: Evidence from Personnel Data.” Econometrica, 77(4): 1047-1094.
Barr, Abigail. 2003. “Trust and Trustworthiness: Experimental Evidence from Zimbabwean
Villages.” The Economic Journal, 113(489): 614-630.
Becker, Gary. 1971. The Economics of Discrimination. Chicago, IL: University of Chicago
Press.
Ben-Ner, Avner. and Louis Putterman. 2001. “Trusting and Trustworthiness.” Boston University
Rule Review, 81(2): 523-551.
Bennedsen, Morten, Kasper M. Nielsen, Francisco Perez-Gonzalez, and Daniel Wolfenzon .
2007. “Inside the Family Firm: the Role of Families in Succession Decisions and
Performance.” Quarterly Journal of Economics, 122(2): 647-691.
Belot, Michele, and Jeroen van de Ven. 2011. “Friendships and Favoritism at School: Evidence
from the Field.” Economic Journal, 121(557): 1228-1251.
Berg, Joyce, John Dickhaut, and Kevin McCabe. 1995. “Trust, Reciprocity, and Social History.”
Games and Economic Behavior, 10(1): 122-142.
Billig, Michael, and Henri Tajfel. 1973. “Social Categorization and Similarity in Intergroup
Behavior.” European Journal of Social Psychology, 3(1): 27-52.
Bohnet, Iris, and Richard Zeckhauser. "Trust, risk and betrayal." Journal of Economic Behavior
& Organization 55.4 (2004): 467-484.
Brandts, Jordi, and Carlos Solà. 2010. “Personal Relations and their Effect on Behavior in an
Organizational Setting: An Experimental Study.” Journal of Economic Behavior and
Organization, 73(2): 246-253.
Brewer, Marilynn B. 1999. “The Psychology of Prejudice: Ingroup Love or Outgroup Hate?”
Journal of Social Issues, 55(3): 429-444.
Brewer, Marilynn B. and Rupert J. Brown. 1998. “Intergroup Relations.” In Daniel T. Gilbert,
Susan T. Fiske, and Gardner Lindzey (Eds.), The Handbook of Social Psychology, 2(4),
Boston, MA: McGraw-Hill.
Brick, Ivan E., Oded Palmon, and John K. Wald. 2005. “CEO Compensation, Director
Compensation, and Firm Performance: Evidence of Cronyism.” Journal of Corporate
Finance, 12(3): 403-423.
Buchan, Nancy R., Rachel T.A. Croson, and Sara Solnick. 2008. “Trust and Gender: An
Examination of Behavior, Biases and Beliefs in the Investment Game.” Journal of
Economic Behavior and Organization, 64(3-4): 466-476.
Castillo, Marco, and Ragan Petrie. 2010. “Discrimination in the Lab: Does Information Trump
Appearance?” Games and Economic Behavior, 68(1): 50-59.
Chen, Yan, and Sherry X. Li. 2009. “Group Identity and Social Preferences.” American
Economic Review, 99(1): 431-457.
Cox, James C. 2004. “How to Identify Trust and Reciprocity.” Games and Economic Behavior,
46: 260-281.
26 Croson, Rachel T.A., and Uri Gneezy. 2009. “Gender Differences in Preferences.” Journal of
Economic Literature, 47(2): 448-474.
Davis, James H., F. David Schoorman, and Lex Donaldson. 1997. “Toward a Stewardship
Theory of Management.” The Academy of Management Review, 22(1): 20-47.
Eckel, Catherine C., and Philip J. Grossman. 2008. “Forecasting Risk Attitudes: An
Experimental Study Using Actual and Forecast Gamble Choices.” Journal of Economic
Behavior and Organization, 68 (1): 1-17.
Eckel, Catherine C., and Rick K. Wilson. "Is trust a risky decision?" Journal of Economic
Behavior & Organization 55, no. 4 (2004): 447-465.
Englebert, Pierre. 2000. “Pre-Colonial Institutions, Post-Colonial States, and Economic
Development in Tropical Africa.” Political Research Quarterly, 53(1): 7-36.
Falk, Armin, and Christian Zehnder. 2007. “Discrimination and In-Group Favoritism in a
Citywide Trust Experiment.” IZA Working Paper DP 2765.
Fearon, James D., and David D. Laitin. 1996. “Explaining Interethnic Cooperation.” American
Political Science Review, 90(4): 715-735.
Fershtman, Chaim, Uri Gneezy, and Frank Verboven. 2005. “Discrimination and Nepotism: the
Efficiency of the Anonymity Rule.” Journal of Legal Studies, 34(2): 371-394.
Fiedler, Marina, Ernan Haruvy, and Sherry X. Li. 2011. “Social Distance in a Virtual World
Experiment.” Games and Economic Behavior, 72(2): 400-426.
Fischbacher, Urs. 2007. “z-Tree: Zurich Toolbox for Ready-Made Economic Experiments.”
Experimental Economics, 10(2): 171-178.
Goette, Lorenz, David Huffman, and Stephan Meier. 2006. “The Impact of Group Membership
on Cooperation and Norm Enforcement: Evidence Using Random Assignment to Real
Social Groups.” American Economic Review, 96(2): 212-216.
Kets de Vries, Manfred F.R. 1993. “The Dynamics of Family Controlled Firms: The Good and
Bad News.” Organizational Dynamics, 21(3): 59-71.
Lentz, Bernard F. and David N. Laband. 1989. “Why So Many Children of Doctors Become
Doctors: Nepotism vs. Human Capital Transfers.” Journal of Human Resources, 24(3):
392-413.
Levin, S., & Sidanius, J. (1999). “Social dominance and social identity in the United States and
Israel: Ingroup favouritism or outgroup derogation?” Political Psychology, 20, 99 – 126.
Lewis, Amy C., and Steven J. Sherman. 2003. “Hiring You Makes Me Look Bad: Social-Identity
Based Reversals of the Ingroup Favoritism Effect.” Organizational Behavior and Human
Decision Processes, 90(2): 262-276.
McConaugby, Daniel L., Charles H. Matthews, and Anne S. Fialko. 2001. “Founding family
Controlled Firms: Performance, Risk, and Value.” Journal of Small Business
Management, 39(1): 31-49.
Messick, David M., and Diane M. Mackie. 1989. “Intergroup Relations.” Annual Review of
Psychology, 40(1): 45-81.
Perez-Gonzalez, Francisco. 2006. “Inherited Control and Firm Performance.” American
Economic Review, 96(5): 1559-1588.
Schechter, L. (2007). Traditional trust measurement and the risk confound: An experiment in
rural Paraguay. Journal of Economic Behavior & Organization, 62(2), 272-292.
Singell, Larry D., and James Thornton. 1997. “Nepotism, Discrimination, and the Persistence of
Utility-Maximizing, Owner-Operated Firms.” Southern Economic Journal, 63(4): 904919.
27 Slonim, Robert, and Ellen Garbarino. 2008. “Increases in Trust and Altruism from Partner
Selection: Experimental Evidence.” Experimental Economics, 11(2): 134-153.
Tajfel, Henri, and John Turner. 1979. “An Integrative Theory of Intergroup Conflict.” In William
G. Austin and Stephen Worchel (Eds.), The Social Psychology of Intergroup Relations,
Monterey, CA: Brooks-Cole.
Vanhanen, Tatu. 1999. “Domestic Ethnic Conflict and Ethnic Nepotism: A Comparative
Analysis.” Journal of Peace Research, 36(1): 55-73.
28 Appendix A: Game Screenshots
Figure A.1: Eckel-Grossman risk measure screenshot
29 Figure A.2: Nepotism game proposer decision screenshot
30 Figure A.3: Nepotism game responder decision screenshot
31 Appendix B: Additional Regressions
Table B.1: Probability of selecting ingroup members as partner (coefficients)
Dependent variable: Partner choice (1 = Chose ingroup member)
Treatment
(1 = Low ingroup efficiency)
Beliefs
(Ingroup less other)
Favoritism
I
II
III
IV
V
-1.063***
-1.319***
-1.209***
-1.522***
-1.521***
(0.35)
(0.34)
(0.37)
(0.45)
(0.47)
0.206
8.195**
(3.92)
0.240
9.301**
(4.22)
0.231
10.80**
(4.50)
0.252
(0.13)
(0.16)
(0.17)
(0.20)
-0.419***
-0.354**
(0.14)
(0.15)
8.080**
(3.68)
(Ingroup less other)
Risk Preferences
(6 = Risk seeking)
Gender (D)
(1 = Female)
Group Identity
(7 = Strong ingroup identity)
Generalized Trust
(7 = More trusting)
Generalized Fairness
0.613
(0.46)
0.042
(0.13)
-0.063
(0.16)
0.050
(7 = Others are fair)
Constant
(0.15)
Pseudo R-squared
Chi-squared
P-value
Log Likelihood
Observations
0.914***
1.092***
0.924***
2.701***
2.041*
(0.28)
(0.27)
(0.29)
(0.73)
(1.23)
0.263
26.4
0.000
0.225
22.6
0.000
0.307
30.9
0.000
0.416
41.9
0.000
0.438
44.0
0.000
-37.09
-39.01
-34.86
-29.38
-28.30
78
78
78
78
78
Note: * p<0.1, **p<0.05, *** p<0.01. Probit specification, standard errors in parentheses. Table reports
coefficients. Dependent variable takes on a value of 1 if subject selected ingroup member as partner in the trust
game. Data is from the Nepotism treatment. The “Beliefs” variable measures the difference in beliefs about
reciprocity between ingroup members and others. Thus, positive values mean subjects expect more back from their
ingroup, while negative values mean subjects expect less back from their ingroup (relative to others). The
“Favoritism” variable measures the difference in dictator giving to ingroup members and others. Thus, positive
values are subjects giving more to ingroup relative to others, while negative values are subjects giving more to
others.
32 Table B.2: Probability of preferring ingroup members as partner (Random match treatments)
Dependent variable: Partner choice (1 = Chose ingroup member)
Treatment
(1 = Low ingroup efficiency)
Beliefs
(Ingroup less other)
Favoritism
I
II
III
IV
V
-1.298***
-1.448***
-1.636***
-1.640***
-1.838***
(0.36)
(0.37)
(0.41)
(0.41)
(0.46)
0.242***
9.537**
(4.44)
0.213**
9.474**
(4.50)
0.214**
10.38**
(4.72)
0.234**
(0.09)
(0.09)
(0.09)
(0.10)
0.009
0.056
(0.11)
(0.13)
11.19**
(4.46)
(Ingroup less other)
Risk Preferences
(6 = Risk seeking)
Gender (D)
(1 = Female)
Group Identity
(7 = Strong ingroup identity)
Generalized Trust
(7 = More trusting)
Generalized Fairness
0.311
(0.44)
0.120
(0.16)
-0.088
(0.14)
-0.175
(7 = Others are fair)
Constant
(0.16)
Pseudo R-squared
Chi-squared
P-value
Log Likelihood
Observations
0.809***
0.884***
0.798***
0.767
0.967
(0.27)
(0.26)
(0.27)
(0.47)
(1.16)
0.271
25.0
0.000
-33.68
0.223
20.6
0.000
-35.86
0.335
30.9
0.000
-30.71
0.335
30.9
0.000
-30.71
0.373
34.4
0.000
-28.96
70
70
70
70
70
Note: * p<0.1, **p<0.05, *** p<0.01. Probit specification, standard errors in parentheses. Table reports
coefficients. Dependent variable takes on a value of 1 if subject selected ingroup member as partner in the trust
game. Data is from the Random match treatment. The “Beliefs” variable measures the difference in beliefs about
reciprocity between ingroup members and others. Thus, positive values mean subjects expect more back from their
ingroup, while negative values mean subjects expect less back from their ingroup (relative to others). The
“Favoritism” variable measures the difference in dictator giving to ingroup members and others. Thus, positive
values are subjects giving more to ingroup relative to others, while negative values are subjects giving more to
others.
33 
Fly UP