Nationalism and Social Sanctioning Across Ethnic Lines: Experimental Evidence from the Kenya-Tanzania Border * Sangick Jeon Independent Researcher [email protected]Tim Johnson Assistant Professor Atkinson Graduate School of Management Willamette University [email protected]Amanda Lea Robinson † Assistant Professor The Ohio State University 2130 Derby Hall 540 N. Oval Mall Columbus, OH 43210 P: 614-292-5210 — F: 614-292-1146 [email protected]Word Count: 4718 * This research was conducted with generous support from the Russell Sage Small Grants Program in Behav- ioral Economics, the Global Underdevelopment Action Fund at the Freeman Spogli Institute for International Studies at Stanford University, and the Stanford Center for Philanthropy and Civil Society. Valuable research assistance was provided by Chacha Maroa, Daniel Merengo, Geoffrey Ochieng, and Wango Chaula Wango, and the Shirati Health, Education, and Development group and Kenyatta University provided research affiliations for this work. The research also benefited from helpful feedback at the 2013 Midwest Group in African Political Economy meeting. † Corresponding author. Authors’ names appear in alphabetical order to reflect their equal contributions.
60
Embed
Nationalism and Social Sanctioning Across Ethnic Lines€¦ · fects sanctioning in Kenya { namely increased punishment of outgroup violations against ingroup members { is at odds
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Nationalism and Social Sanctioning Across Ethnic Lines:Experimental Evidence from the Kenya-Tanzania Border∗
∗This research was conducted with generous support from the Russell Sage Small Grants Program in Behav-ioral Economics, the Global Underdevelopment Action Fund at the Freeman Spogli Institute for InternationalStudies at Stanford University, and the Stanford Center for Philanthropy and Civil Society. Valuable researchassistance was provided by Chacha Maroa, Daniel Merengo, Geoffrey Ochieng, and Wango Chaula Wango, andthe Shirati Health, Education, and Development group and Kenyatta University provided research affiliationsfor this work. The research also benefited from helpful feedback at the 2013 Midwest Group in African PoliticalEconomy meeting.†Corresponding author. Authors’ names appear in alphabetical order to reflect their equal contributions.
Past research shows that ethnic diversity reduces the ability to sanction norm violators,ultimately undermining cooperation. We test this directly by experimentally varying theethnic composition of groups playing a dictator game with third-party punishment amongtwo ethnic groups along the Kenya-Tanzania border. We also implement a structurallyidentical game where the endowment division is randomly determined in order to isolatea punishment motivation from the motivation to rectify income inequality. While costlyincome adjustment in both games is driven primarily by norm violations and inequalityaversion, the ethnic composition of groups also influences sharing and sanctioning behav-ior in Kenya but not Tanzania, consistent with documented differences in the strengthof nationalism across the two countries. However, the way in which shared ethnicity af-fects sanctioning in Kenya – namely increased punishment of outgroup violations againstingroup members – is at odds with theories that anticipate that costly sanctioning willprimarily target coethnics.
Societies require communally-determined standards of conduct – i.e., social norms – to
function (Sober and Wilson 1998; Fehr and Fischbacher 2003; Richerson and Boyd 2005),
and individuals thus sanction violations of those norms (Boyd and Richerson 1992), even when
doing so is costly (Fehr and Gachter 2002; Fehr and Fischbacher 2004; Henrich et al. 2006).
However, scholars have argued that enforcing social norms is more difficult in ethnically-
diverse communities because individuals are less willing or able to effectively sanction across
ethnic lines (Fearon and Laitin 1996; Miguel and Gugerty 2005; Shinada et al. 2004;
Habyarimana et al. 2009).
We conduct experiments designed to test this directly, among members of two ethnic
groups in East Africa, the Luo and the Kuria. Participants in our experiments completed
dictator games with third-party punishment and random income games with third-party in-
come adjustment. The latter game is structurally identical to the dictator game except that
a randomizing device allocates income to players, thus allowing us to distinguish between
the punishment of norm violators and efforts to reduce inequality. Random assignment to
groups for each game generated variation in the ethnic make-up of experimental groups, which
allowed us to identify shared ethnicity’s role in social sanctioning and inequality aversion.
Furthermore, we conducted the behavioral games among members of the Luo and Kuria
ethnolinguistic groups living on both sides of the Kenya-Tanzania border, to assess whether
supra-ethnic nationalism can facilitate sanctioning across ethnic lines, ultimately improving
cooperation in diverse settings. Because nationalism has been much stronger in Tanzania
than Kenya (Barkan 1994; Miguel 2004), we expect that ethnic differences will be a greater
barrier to cooperation and sanctioning in Kenya than in Tanzania.
In general, we find that across all coethnicity treatments in both countries, the degree
of norm violation and income inequality are by far the strongest and most robust predictors
of costly sanctioning. We also report fairly weak but detectable differences in the degree
to which ethnicity shapes sharing and (to a lesser degree) costly sanctioning, but only in
Kenya. This suggests that Tanzania’s strong nationalism may indeed alleviate ethnic barriers
to cooperation. However, the patterns of play suggest that when punishment is conditioned on
1
ethnicity, it tends to be focused on outgroup members who fail to share with the third party
punisher’s coethnic. This is at odds with conventional political science theories of ethnic-based
sanctioning, which anticipate that costly sanctioning will be reserved for ingroup members
who violate social norms of cooperation with other ingroup members (Miguel and Gugerty
2005; Habyarimana et al. 2009).
Research Design
Our research design includes two behavioral economic games designed to separate distinct
motivations for social sanctioning. Social sanctioning that is motivated by punishment of
a norm violation is captured using a a classic dictator game with third party punishment
(DG3) (Fehr and Fischbacher 2004), while the desire to rectify inequality is isolated using
the random income game with third party income adjustment (RIG3), a modification of the
game employed in Dawes et al. (2007). Each game has three roles – A, B, and C. The two
games are depicted graphically in Figure 1.
In the first stage of the DG3, A is endowed with 10 tokens and told that she can either
keep the 10 tokens or she can divide them – in any manner – between herself and B. After A
makes her decision, C is endowed with 5 tokens, informed of A’s sharing decision, and offered
the opportunity to spend some of his 5 tokens to reduce the final income of A; for every 1
token C spent, A’s final income is reduced by 3 tokens. In this game, A’s decision about
how much to share with B indicates adherence to a sharing norm, while C’s costly decision
to punish indicates a willingness to sanction a norm violation. The RIG3 works exactly the
same way except that the initial division of the 10 tokens between A and B is determined
randomly rather than by A, and C has the opportunity to reduce the income of either A or
B. As a result, in RIG3, C’s costly decision to adjust incomes is an indicator of inequality
aversion.
Following the framework of Bernhard et al. (2006), we also manipulated the ethnic make-
up of the game partners in order to identify the impact of shared ethnicity on cooperation and
social sanctioning in DG3 (and on inequality aversion in RIG3). We implemented these games
2
Dictator Game withThird Party Punishment
(DG3)
A Bx
Stage 1
C
A
pa
B
Stage 2
PayoffsA = 10 − x− 3paB = xC = 5 − pa
Random Income Game withThird Party Punishment
(RIG3)
Nature
A
10 − x
B
xStage 1
C
A
pa
B
pbStage 2
PayoffsA = 10 − x− 3paB = x− 3pbC = 5 − pa − pb
Figure 1: Games: Each game is played in two stages. In the first stage of the DG3, playerA transferred some amount, x, to player B, keeping 10 − x for herself. PlayerC observed the amount transferred in the first stage, x, and then decided howmuch, if any, to pay to reduce the income of player A. Whatever punishment paidwas tripled and deducted from player A. RIG3 is structurally equivalent to DG3except that x was determined randomly and C was allowed to reduce the incomeof players A or B.
with the Luo and the Kuria, two ethnic groups who reside in southwest Kenya and northwest
Tanzania. In this sense our study is a lab-in-the-field experiment (Grossman 2011). The four
treatment groups, depicted in Figure 2, had the following ethnic compositions:
1. All players A, B, and C are of the same ethnic group.
2. Players A and B are of the same group while C is of another group.
3. Players A and C are of the same group while B is of another group.
3
4. Player B and C are of one group while A is of another group.
Note that each treatment group could be constituted two different ways given that we have
two different ethnic groups.
This design allows us to evaluate not just whether shared ethnicity impacts cooperation
and sanctioing, but also competing explanations about how. Miguel and Gugerty (2005) and
Habyarimana et al. (2009) argue that higher rates of cooperation among coethnics result from
costly sanctioning being ethnically-bound, with evidence coming from Kenya and Uganda,
respectively. These findings suggest that punishment, and thus cooperation, will be more
common when A and C are coethnic (conditions ABC and AC), and especially so when B is
also a coethnic (ABC> AC > AB=BC=0). Fearon and Laitin (1996) are motivated instead
by the surprisingly high rates of interethnic cooperation, and propose two different strategies
that could support cooperation in diverse contexts.1 In the in-group policing model, pun-
ishment is only targeted at coethnics, and especially when coethnics fail to cooperate with
non-coethnics, with the expectation that non-cooperation by non-coethnics will be sanctioned
by other non-coethnics (AC>ABC>AB=BC=0). In the spiral model, by contrast, punish-
ment is directed at both coethnics and non-coethnics who defect against one’s own coethnics
(ABC and BC), but not transgressions against non-coethnics (ABC=BC>AB=AC=0). Fi-
nally, Bernhard et al. (2006), directly evaluated the role of shared ethnicity in sanctioning
behavior among members of two tribes in Papau New Guinea. Contrary to all but one of
the theoretical expectations outlined above, they find that punishment is harshest for trans-
gressions against coethnics (ABC and BC), but also some leniency for coethnic transgressors
(BC>ABC>>AB=AC).
1Fearon and Laitin (1996) propose these strategies in contexts of repeating interaction.While our one-shot games do not allow such repeated play, we nevertheless derive expectationsfor behavior based on Fearon and Laitin’s strategy profiles with the expectation that subjectsoften play one-shot behavioral economic games as if they are in a context of repated socialinteractions (Hoffman et al. 1996; Habyarimana et al. 2007).
4
AB
C
C
AB
AB C
AB
AC C
AB
BC C
AB
Fig
ure
2:
Eth
nic
Con
figu
rati
on
Tre
atm
ent
Gro
up
s:S
had
edci
rcle
sre
pre
sent
pla
yer
sfr
omth
esa
me
eth
nic
grou
p.
5
We also evaluate whether the effects of ethnic affiliations are moderated by the presence
of a strong, supra-ethnic national identity. To do so, our research design exploits the natural
experiment afforded by the political border between Kenya and Tanzania, which was deter-
mined by colonial authorities in the 19th century. Miguel (2004) argues that the arbitrary
nature of this border creates laboratory-like conditions to test the effects of nation-building on
interethnic cleavages domestically, since communities share the same objective cultural differ-
ences, geography, and history on both sides of the border, but differ radically in their exposure
to nation-building policies. While ethnicity has played a central role in post-independence
Kenyan politics, concerted efforts at nation-building in Tanzania – including a common na-
tional language and public education emphasizing a common Tanzanian history and culture
– resulted in a stronger sense of a common Tanzanian identity (Barkan 1994; Miguel 2004).
Thus, implementing the lab-in-the-field experiments with the Kuria and Luo living on each
side of the border permits us to examine how different levels of popular nationalism affect sanc-
tioning patterns across an identical set of ethnic divisions.2 While the differences in national
identification between the two countries was our primary motivation in comparing behavior
on each side of the border, we recognize that there are many other differences between the
two countries that could also shape behavior (see Dunning (2012) for a discussion of “bundled
treatments” within natural experiments in general, and McCauley and Posner (2015) for a
specific discussion of the use of African borders as sources of natural experiments). We are
thus cautious in interpreting any national differences as resulting solely from differences in
nationalism.
While building on past research, our research design is novel in two important ways. First,
our study adds to the designs of Habyarimana et al. (2009) and Bernhard et al. (2006) by
incorporating the RIG3, which allows us to evaluate the effect of coethnicity on egalitarian
2In nationally representative surveys in 2011/2012, Tanzanians were more likely thanKenyans (96% vs. 91%) to say they identified with their national identity at least as much astheir ethnic identity (Afrobarometer 2012). We find similar differences in our sample usingthe same question (98% vs. 92%, t=3.5, p<0.01), as well as finding that Tanzanians weremuch more likely than Kenyans to agree that “even though there is a lot of cultural variety inTanzania [Kenya], we are more the same than we are different” (87% vs. 64%, t=6.6, p<0.01).
6
Figure 3: Map of Experiment Sites. Dotted polygon shows the ancestral homeland of theLuo, and the striped polygon shows the ancestral homeland of the Kuria (Gordon2005). Black circles indicate the two experiment sites.
motives separately from punishment of non-cooperation. Second, we build on Miguel’s (2004)
observational finding that interethnic cooperation is more robust in Tanzania than Kenya by
experimentally manipulating coethnicity directly among members of the exact same ethnic
groups on each side of the border.
Experimental Protocols and Data Collection
The behavioral economic games were implemented separately on each side of the international
border in a rural town – Karamu, Kenya and Sombanyasoko, Tanzania – near the intersection
of the Kuria and Luo homelands (see Figure 3). Participants were recruited using door-to-door
canvassing of randomly selected households in 12 largely ethnically homogenous villages near
the experiment sites. To generate ethnic diversity in each experimental session, participants
7
were recruited from 1-2 villages Luo villages and 1-2 Kuria villages for each session. Potential
participants were informed that they would receive a show-up fee approximately equal to
one day’s wages in the informal economy (300 Kenyan Shillings (KES) or 5,000 Tanzanian
Shillings (TZS) and that it would be possible to earn additional money during the study
depending on the decisions made by participants. Interested individuals were scheduled to
attend one study session and were reminded the day before via text message, when possible.
Of the 672 individuals recruited to participate, 596 (89%) showed-up, 558 were randomized
into treatment groups, and behavioral decisions from 501 were analyzed. Figure A.1 in the
appendix shows the CONSORT diagram for the study.
Research assistants explained both the DG3 and RIG3 games to the 20-30 participants
in each session as a group, and worked through examples to ensure comprehension (see the
appendix for the exact wording and examples used). To generate the random division of the
10 tokens in the first stage of the RIG3 game, we used a wheel that contained 11 values
on its face (in one unit increments from 0 to 10) and a “respin” value (see Figure A.6 in
the appendix). For each RIG3 game, Player C would spin the wheel once to determine the
incomes of A and B in the game. Participants were informed that they would not play with
actual Kenyan/Tanzanian shillings but with tokens that would be exchanged at a rate of 1
token equal to 10 Kenyan shillings or 200 Tanzanian shillings.3
Participants were then randomized into groups of three and into particular roles within
each group by drawing numbers from a hat.4 This random assignment of individuals to
groups generated the assignment to the four different coethnicity treatments in Figure 2.5
333% of our sampled reported that their household had no cash income during the pastweek and another 25% reported earning less than 500 Kenyan Shillings/7500 Tanzanian. Thus,while each token was worth a relatively low sum in absolute terms, the amount at stake ineach game relative to weekly incomes was quite meaningful.
4Only the researchers knew how the numbers drawn corresponded to groups and rolesso that all game partners remained anonymous. See Section E of the appendix for moreinformation about random assignment to groups and treatments.
5In experimental sessions with equal proportions of Kuria and Luo participants, all fourexperimental conditions (ABC, AB, AC, and BC) were equally likely. The more unbalancedthe ethnic composition of the experimental session, however, the more likely the ABC condi-tion became relative to the other three conditions. On average, our 20 sessions were 55% oneethnic group and 45% the other, but imbalance was as large as 71%-29% in one session. In
8
To minimize the possibility of spillovers between games, we designed the treatments such
that each participant made only one decision across both games (and, thus, each participant
was only called into a separate room once). This was done by assigning Player A in DG3
the role of Player B in RIG3, assigning Player B in DG3 the role of Player C in RIG3, and
assigning Player C in DG3 the role of Player A in RIG3 (see Figures A.4 and A.5). Since only
Players A and C in DG3 and Player C in RIG3 make experiment decisions (see Figure 1),
each participant only made one experiment decision for both games. And because subjects
were not informed about the game outcomes until the end of the experiment, participants’
decisions in one game could not have been conditioned on outcomes from the other game.
Subjects did not know the identity of any player during the game and, in keeping with this
design feature, made game decisions in a separate room from other participants. To avoid
revealing our interest in ethnicity, participants were not given the precise ethnic affiliation
of their partners but instead the names of the villages they were from. Because villages are
highly homogeneous in this region, home village conveyed a strong signal of partners’ ethnic
affiliations without cuing participants in to the intention of our group treatments.6 Like the
nationalism “treatment,” however, information about a partner’s village is a also bundled
treatment, conveying information about the likely ethnicity of that partner along with other
information related to village affiliation. However, we recruited from 12 different villages and
the effects of shared ethnicity among group members is averaged over lots of different village
pairings. In addition, we include village fixed effects for each decision maker in order to
this most extreme case, assignment to the ABC condition became 50% more likely than eachof the other three conditions. While such imbalance reduces observations in some treatmentgroups, and thus statistical power, it should not introduce bias in differences across treatmentssince individual characteristics are still orthogonal to treatment assignment.
694% of Kenyan participants and 93% of Tanzanian participants belonged to the ethnicgroup majority in their village. We exclude participants who report being from an ethnicgroup other than the Kuria or Luo (n = 24), and we recode treatment assignment for Kuriaand Luo ethnic minorities (n = 12). For example, a Kuria from a Luo village assigned to roleA in a group considered homogenous (ABC) based on village majorities will perceive her owntreatment as BC instead of ABC. Player C in this group, however, perceives his treatment asABC because he will infer (incorrectly in this rare case) that Player A is Luo given that she isfrom a Luo majority village. We also report the results when these participants are excluded(Tables A.5, A.6, and A.7 of the appendix).
9
capture village level differences in cooperation, sanctioning, and inequality aversion.
After making her decision, each participant completed a brief questionnaire (see Section
F of the appendix). After all decisions were made, participants were individually informed of
their game outcome and given their earnings in cash based on true game decisions. Survey
responses, as well as the experimental decisions, were collected using hand-held mobile devices
equipped with Open Data Kit (ODK).
Results
Sharing was common in the DG3 in both Kenya and Tanzania. The median allocation Player
A gave to Player B equaled almost three tokens in the overall sample. Costly sanctioning in
the DG3 was also common: roughly 56% of Players C in the DG3 spent at least one token to
punish Player A (reducing Player A’s payoff by 3 tokens), 18% spent two or more tokens, and
2% spent 3 tokens. Table 1 reports average levels of cooperative sharing and costly sanctioning
overall, as well as broken down by country and coethnicity treatment.
To determine if coethnicity influenced sharing in the DG3, we estimate a simple linear
model in which A’s transfer to B serves as the dependent variable and binary indicators
of treatment group status serve as the independent variables (with ABC as the omitted
category).7 After estimating that parsimonious model, we add covariates that condition
treatment estimates on Player A’s age, gender, income, education, and village. Estimated
coefficients from these models appear in Table 2, with Models 1-2 reporting estimates for
the Kenyan subsample and Models 3-4 reporting estimates for the Tanzanian subsample.
In Models 1 and 3, which solely contain treatment indicators, we cannot reject the null
hypothesis at the 95% confidence level that the coefficient estimates associated with our
treatment indicators equal zero. Including covariates in the models increases the variation in
sharing that we can explain, but the coefficient estimates for our treatment indicators remain
small relative to their standard errors, save for the coefficient estimate associated with the
AC indicator in Model 2, which is estimated on data from the Kenyan sample. However,
7The results are similar when estimated using an ordered probit model (Table A.2).
10
the parameterization of the model reported in Table 2 only presents comparisons between
the ABC condition and each of the other conditions. Substantively meaningful differences in
sharing might exist, however, when taking into account other comparisons across treatment
groups. As a result, we conduct a series of pair-wise comparisons presented in Table 3. We
present both unadjusted p values and p values adjusted for multiple comparisons using the
Duncan method (Duncan 1955).8
The pairwise comparisons performed in Table 3 indicate significant differences across treat-
ment groups, but only with 90% confidence. In Kenya, participants in AB and AC groups
share less than their peers in ABC groups (Table 3, Rows 1 and 2) and participants in AC
groups share less than BC groups (Table 3, Row 6). This pattern suggests that the presence of
coethnicity between players B and C, which occurs in both BC and ABC groups, yields higher
levels of sharing (almost a full token more, on average). The pattern of play in Kenya is thus
inconsistent with past political science research showing that cooperation is induced by the
threat of sanctioning from an ingroup member (e.g., Miguel and Gugerty 2005; Habyarimana
et al. 2009), and is instead in line with the findings from Bernhard et al. (2006), which show
that pro-social sharing peaks when the potential punisher comes from the same group as the
individual who receives the gains from sharing. Moreover, these results support our general
expectation that ethnicity would influence behavior in Kenya but not Tanzania; we find no
differences in levels of sharing across treatment groups in Tanzania. Thus, albeit with lower
confidence (partially due to the limited statistical power of our tests, see Appendix Section
C), we find evidence consistent with the proposition that the ethnic composition of a group
influences adherence to pro-social sharing norms.
8Our statistical power to detect treatment effects is quite limited by the small samplesizes within each of four treatment groups across the two countries (see Section C of theappendix), making the risk of a Type II error already quite high. Because multiple comparisonadjustments reduce the risk of Type I errors at the expense of increasing the chance of TypeII errors, we utilize the Duncan adjustment, which is less conservative than many alternativemethods of multiple comparisons adjustment.
11
Tab
le1:
Des
crip
tive
Sta
tist
ics
ofE
xp
erim
ent
Ou
tcom
es
Ken
yaT
anza
nia
Fu
llABCDG
,AB
DG
,ACDG
,BCDG
,ABCDG
,AB
DG
,ACDG
,BCDG
,S
am
ple
ABCRIGBCRIGAB
RIG
ACRIGABCRIGBCRIGAB
RIG
ACRIG
Dic
tato
rG
am
e:
Tok
ens
Kep
tby
A7.
326.8
37.5
07.8
66.6
17.4
17.6
17.1
37.4
5(1.9
0)
(1.6
3)(1.8
4)(1.7
1)(1.9
1)(2.1
4)(1.7
5)(2.0
0)(2.0
6)
Toke
ns
Tra
nsf
erre
dto
B2.
683.1
72.5
02.1
43.3
92.6
02.3
92.8
72.5
5(1.8
9)
(1.6
3)(1.8
4)(1.7
1)(1.9
1)(2.1
4)(1.7
5)(2.0
0)(2.0
6)
Toke
ns
Pai
dby
Cto
Red
uce
A0.7
60.6
51.0
81.0
00.7
90.4
80.7
30.5
30.7
6(0.8
0)
(0.8
0)(0.9
1)(0.8
0)(0.8
9)(0.6
5)(0.7
0)(0.7
4)(0.7
8)
Ran
dom
Incom
eG
am
e:
Toke
ns
All
oca
ted
toA
5.4
96.8
65.0
07.5
36.4
75.8
95.0
62.9
54.6
8(3.2
2)(2.8
0)(3.7
0)(2.3
2)(3.6
0)(2.7
5)(3.4
0)(2.6
9)(2.7
0)
Tok
ens
All
oca
ted
toB
4.51
3.1
45.0
02.4
73.5
34.1
14.9
47.0
55.3
2(3.2
2)(2.8
0)(3.7
0)(2.3
2)(3.6
0)(2.7
5)(3.4
0)(2.6
9)(2.7
0)
Tok
ens
Paid
by
Cto
Red
uce
A0.3
20.3
80.4
70.5
30.2
70.6
0.44
0.17
0.1
00.1
8(0.5
2)(0.5
9)(0.5
1)(0.8
0)(0.4
6)(0.5
1)(0.3
9)(0.3
1)(0.3
9)
Toke
ns
Pai
dby
Cto
Red
uce
B0.
270.3
30.1
90.
120.0
70.2
20.1
80.6
70.2
7(0.5
6)(0.7
3)(0.5
1)(0.3
3)(0.2
6)(0.5
1)(0.3
9)(0.8
0)(0.4
6)
Nu
m.
of
Gro
up
s171
2426
2118
2718
1522
Note
:M
eans
wit
hst
andard
dev
iati
ons
inpare
nth
esis
.ABC
repre
sents
trea
tmen
tgro
ups
inw
hic
hP
layer
sA
,B
,and
Care
all
from
the
sam
etr
ibe,
AB
repre
sents
trea
tmen
tgro
ups
inw
hic
honly
Pla
yer
sA
and
Bare
from
the
sam
etr
ibe,
AC
repre
sents
trea
tmen
tgro
ups
inw
hic
honly
Pla
yer
sA
and
Care
from
the
sam
etr
ibe,
andBC
repre
sents
trea
tmen
tgro
ups
inw
hic
honly
Pla
yer
sB
and
Care
from
the
sam
etr
ibe.
12
Table 2: Cooperative Sharing Across Coethnicity Treatments
Kenya Tanzania
(1) (2) (3) (4)
AB −0.667 −0.904∗ −0.204 −0.001(0.501) (0.503) (0.612) (0.589)
AC −1.024∗ −1.054∗∗ 0.274 0.190(0.529) (0.518) (0.648) (0.648)
BC 0.222 −0.071 −0.047 0.326(0.552) (0.538) (0.578) (0.589)
OLS estimates with standard errors in parentheses. ABC treatment group omitted.
The dependent variable is tokens transferred from Player A to Player B.
Village fixed effects are based on A’s village.∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01
13
Table 3: Differences in Cooperative Sharing in DG3 by Coethnicity Treatment
Kenya Tanzania
Difference Unadj. Adj. Difference Unadj. Adj.(Std. Err.) p p (Std. Err.) p p
AB vs. ABC −0.90 0.08 0.08 −0.00 1.00 1.00(0.50) (0.59)
AC vs. ABC −1.05 0.05 0.06 0.19 0.77 0.77(0.52) (0.65)
BC vs. ABC −0.07 0.90 0.91 0.33 0.58 0.58(0.54) (0.59)
AC vs. AB −0.15 0.78 0.78 0.19 0.79 0.81(0.53) (0.70)
BC vs. AB 0.83 0.12 0.14 0.33 0.61 0.61(0.53) (0.64)
BC vs. AC 0.98 0.08 0.08 0.14 0.85 0.86(0.55) (0.71)
Note: Based on Models 2 and 4 of Table 2. Differences reflect the first treatment minus the secondtreatment. Adjusted p values are adjusted using the Duncan method. ABC represents treatment groups inwhich Players A, B, and C are all from the same tribe, AB represents treatment groups in which only PlayersA and B are from the same tribe, AC represents treatment groups in which only Players A and C are fromthe same tribe, and BC represents treatment groups in which only Players B and C are from the same tribe.
14
Next, we evaluate the impact of the coethnicity treatments on Player C’s willingness to
sanction non-cooperation at a personal cost. OLS coefficient estimates with and without the
full set of controls are given in Table 4, again with the ABC treatment group as the omitted
category.9 In both countries, norm violations – as measured by the amount kept by Player A
– positively affect costly sanctioning and are, by far, the strongest predictors of sanctioning.
Moreover, the coefficient estimate associated with the amount Player A kept takes a greater
value in Kenya than Tanzania (z = 4.63, p < 0.01; Table 4, Models 1 and 3).
In terms of the effect of coethnicity, the treatment indicators included in the models of
Table 4 only focus comparisons on a treatment group’s effect relative to the ABC condition. As
a result, we use estimates from Models 2 and 4 of Table 4 to execute all pairwise comparisons
across treatment conditions and we present those results in Table 5. Again, if we lower our
confidence to the 90% level and focus primarily on unadjusted p values, we note modest effects
of coethnicity. In the Kenyan subsample, while controlling for A’s decision in the first stage,
sanctioning by C is slightly greater in BC groups than in ABC groups or AC groups, although
these effects are estimated with considerable imprecision. Thus, A players in Kenya seem
to have correctly anticipated leniency from coethnics in the AC condition and retribution
from non-coethnics in the BC condition, but incorrectly anticipated punishment in the ABC
condition, which was instead treated much more like AC than BC. These patterns are again
most consistent with Bernhard et al.’s (2006) finding that costly sanctioning is greatest in
BC, and inconsistent with the general expectation in political science that sanctioning should
be greatest in homogenous groups (ABC). In Tanzania, we see the same general pattern of
sanctioning across coethnicity treatments (BC≥AB>ABC=AC), but the differences across
treatments are not statistically significant by any conventional standard.
9The results are similar when estimated using an ordered probit model (Table A.3).
15
Table 4: Costly Sanctioning Across Coethnicity Treatments
OLS estimates with standard errors in parentheses. ABC treatment group omitted.
The dependent variable is tokens spent by C to reduce income of A.
Village fixed effects are based on A’s village.∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01
16
Table 5: Differences in Costly Sanctioning in DG3 by Coethnicity Treatment
Kenya Tanzania
Difference Unadj. Adj. Difference Unadj. Adj.(Std. Err.) p p (Std. Err.) p p
AB vs. ABC 0.12 0.45 0.48 0.23 0.33 0.38(0.16) (0.23)
AC vs. ABC −0.02 0.90 0.90 −0.09 0.68 0.68(0.17) (0.23)
BC vs. ABC 0.34 0.08 0.11 0.28 0.16 0.19(0.19) (0.20)
AC vs. AB −0.15 0.38 0.38 −0.32 0.22 0.25(0.17) (0.26)
BC vs. AB 0.21 0.28 0.28 0.06 0.81 0.81(0.20) (0.23)
BC vs. AC 0.36 0.07 0.09 0.38 0.11 0.11(0.20) (0.23)
Note: Based on Models 2 and 4 of Table 4. Differences reflect the first treatment minus the secondtreatment. Adjusted p values are adjusted using the Duncan method. ABC represents treatment groups inwhich Players A, B, and C are all from the same tribe, AB represents treatment groups in which only PlayersA and B are from the same tribe, AC represents treatment groups in which only Players A and C are fromthe same tribe, and BC represents treatment groups in which only Players B and C are from the same tribe.
17
Sanctioning behavior in the DG3 is thus consistent with past research showing that costly
punishment is more severe for non-cooperation by a noncoethnic affecting a coethnic, at least
in Kenya. However, what appears to be punishment might instead be motivated by a desire
to rectify inequality generated by norm violations, rather than punishment of norm violations
per se. Within the DG3 game alone, these two very different motivations would produce
observationally equivalent behavior. We thus compare these results to the rate of income
adjustment by Player C in the RIG3 game, where inequality between Player A and Player B
was randomly determined.
It is clear from Figure 4 that participants were willing to alter incomes, even at a cost
to themselves, in order to rectify randomly generated income. The figure shows that both
the willingness to alter incomes and the degree of alteration increased with greater inequality.
However, Figure 4 also shows that some participants (n = 16) altered incomes to increase
inequality rather than decrease it, by reducing the income of the player that received equal
to or less than her partner. Such spiteful behavior was not observed even once in the DG3,
and is quite surprising given that such income alteration was costly to player C. There was no
differences in the rate of spiteful behavior across the two countries (t = 1.04, p > 0.10), but
most spiteful income reduction (69%) was targeted at non-coethnics. Because the primary use
of the RIG3 is to determine whether the patterns of income reduction in the DG3 game are
potentially due to inequality aversion, we focus here on income adjustment patterns excluding
the 16 participants who increased inequality.10
We replicate our analysis of sanctioning behavior in the DG3 – conditional on coethnicity
treatments – for income adjustment in the RIG3. Patterns of coethnicity are again captured
by dummy variables for AB, AC, and BC conditions, with ABC as the omitted category.
Unlike in the DG3, Player C in the RIG3 was able to alter the income of either Player A or
Player B: we thus model the reduction of Player A’s randomly generated income separately
from reduction of Player B’s income. The degree of inequality randomly generated is captured
10Results with spiteful participants included in the analysis are reported in Table A.8 ofthe appendix; including those participants does not alter the substantive interpretation of theresults.
18
Figure 4: Player C’s Income Adjustment as a Function of Randomized Inequality
0.2
.4.6
.8Av
erag
e To
kens
Pai
d fo
r Inc
ome
Adju
stm
ent
0 1 2 3 4 5 6 7 8 9 10Random Allocation to Player A
Reduced A Reduced B
by a variable indicating A or B’s allocation. OLS coefficient estimates with and without the
full set of controls are reported separately for Kenya (Models 1-4) and Tanzania (Models 5-8)
in Table 6, and pairwise comparisons across treatment groups are presented in Table 7.11
The results in Table 6 show that in both Kenya and Tanzania the degree of income
inequality between Players A and B in the RIG3 is the strongest predictor of income alteration,
just as it was in the DG3. Furthermore, the ethnic composition of a participant’s group has
no consistent effect on income reduction across models. In models estimated on the Kenyan
sample, the coefficients associated with the coethnicity treatment groups appear small relative
to their standard errors, thus preventing us from rejecting the null hypothesis that those
coefficients equal zero (Table 7). There are similarly no significant effects in the adjustment
of B’s income in Tanzania. The only statistically significant patterns we observe are in the
adjustment of A’s income in Tanzania, which was significantly higher in ABC groups than
either BC or AC groups (Table 7). Given that this pattern of behavior was not apparent
in adjusting Player A’s income, and that it is not consistent with behavior in the DG3, we
11The results are similar when estimated using an ordered probit model (Table A.4).
19
cautiously interpret these findings as spurious. In short, we find very little evidence that
income reduction in the RIG3 varies with the ethnic composition of a participant’s group in
either country.
The amount by which individuals punished or reduced incomes, due to inequitable alloca-
tions, varied across the DG3 and RIG3, respectively. Upon viewing the coefficient estimates
associated with the variables relating to the amount Player A kept or the amount randomly
allocated, one can note that the amount of costly income reduction in the RIG3 per unit
of inequality appeared less than the amount of costly punishment per unit of inequality in
the DG3. To understand such differences and to assess if these two games captured differ-
ent motivations for income adjustment, we examine whether or not feelings of anger toward
human-produced inequality and a taste for retributive violence (collected during the attitu-
dinal survey after game play) correlate with income alteration. Following Fehr and Gachter
(2002), we asked participants to “imagine that in Activity 1 (the DG3), an individual gave
you 1 token and kept 9” and, then, we asked whether or not the participants would feel “not
at all angry,” “a little angry,” “angry,” “quite angry,” or “very angry.” We also asked par-
ticipants to express, on a 5-point scale, the degree to which they agreed that “in order for
justice to be served, violence should be repaid with violence.” Both anger at failing to share
(t=7.05, p<0.01) and a taste for retribution (t=6.58, p<0.01) were significantly stronger in
Kenya than in Tanzania. When comparing answers to these questions to real game behavior,
we find that decisions about how much to spend to sanction others in the DG3 is positively
correlated with both self-reported anger at non-cooperation by Player C (ρ = 0.20, p < 0.01)
and support for retribution (ρ = 0.22, p < 0.01). In contrast, there is no correlation between
income alteration in RIG3 and feelings of anger toward human-produced inequality (ρ=0.06,
p = 0.42) or taste for retribution (ρ=0.05, p = 0.54). These correlations are consistent with
the view that the costly punishment of norm violations (DG3) does not solely result from
aversion toward the material inequality resulting from norm violations.
20
Tab
le6:
Cos
tly
Inco
me
Ad
just
men
tA
cros
sC
oet
hnic
Tre
atm
ents
Ken
yaT
an
zan
ia
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Am
ou
nt
All
oca
ted
(A)
0.0
85∗∗∗
0.0
67∗∗∗
0.05
5∗∗∗
0.0
57∗∗∗
(0.0
22)
(0.0
24)
(0.0
14)
(0.0
15)
Am
ount
All
oca
ted
(B)
0.08
1∗∗∗
0.07
1∗∗∗
0.1
00∗∗∗
0.0
98∗∗∗
(0.0
16)
(0.0
18)
(0.0
19)
(0.0
20)
AB
0.127
0.15
6−
0.1
10−
0.14
4−
0.18
4−
0.19
70.
090
0.04
7(0.1
91)
(0.1
99)
(0.1
38)
(0.1
46)
(0.1
23)
(0.1
26)
(0.1
64)
(0.1
71)
AC
−0.
136
−0.1
36−
0.1
75−
0.10
9−
0.19
4∗−
0.20
5∗
−0.1
77−
0.1
69(0.1
92)
(0.2
05)
(0.1
39)
(0.1
50)
(0.1
13)
(0.1
13)
(0.1
50)
(0.1
54)
BC
0.085
−0.1
07−
0.1
010.
038
−0.
235∗∗
−0.
260∗∗
−0.1
58−
0.1
96(0.1
71)
(0.1
93)
(0.1
24)
(0.1
41)
(0.1
17)
(0.1
15)
(0.1
56)
(0.1
56)
Age
0.0
17−
0.0
300.0
75∗
−0.0
27(0.0
70)
(0.0
51)
(0.0
39)
(0.0
53)
Male
−0.0
680.2
48∗
0.10
80.0
27(0.1
72)
(0.1
25)
(0.0
93)
(0.1
26)
Ed
uca
tion
Lev
el0.
136∗
−0.1
30∗∗
−0.
048
−0.2
09∗∗
(0.0
77)
(0.0
56)
(0.0
63)
(0.0
85)
Inco
me
Lev
el0.
011
−0.0
09−
0.02
5∗−
0.0
07(0.0
25)
(0.0
18)
(0.0
13)
(0.0
17)
Con
stan
t−
0.16
6−
0.3
98−
0.0
460.
463∗
0.09
30.
062
−0.
174
0.21
0(0.1
83)
(0.3
62)
(0.1
02)
(0.2
38)
(0.1
12)
(0.2
25)
(0.1
23)
(0.3
04)
Vil
lage
Fix
edE
ffec
tsN
oY
esN
oY
esN
oY
esN
oY
es
R2
0.24
0.39
0.33
0.45
0.28
0.40
0.35
0.44
Ob
serv
atio
ns
6868
6868
7776
7776
OL
Ses
tim
ate
sw
ith
standard
erro
rsin
pare
nth
eses
.A
BC
trea
tmen
tgro
up
om
itte
d.
The
dep
enden
tva
riable
isto
ken
ssp
ent
by
Cto
reduce
inco
me
of
A(M
odel
s1-2
,5-6
)or
B(M
odel
s3-4
,7-8
)
Villa
ge
fixed
effec
tsare
base
don
C’s
villa
ge.
∗p<
0.1
0,∗∗p<
0.0
5,∗∗∗p<
0.0
1
21
Tab
le7:
Diff
eren
ces
inIn
com
eR
edu
ctio
nin
RIG
3by
Coet
hn
icit
yT
reat
men
t
Ken
yaT
an
zan
ia
Red
uct
ion
of
AR
edu
ctio
nof
BR
edu
ctio
nof
AR
edu
ctio
nof
B
Diff
eren
ceU
nad
j.A
dj.
Diff
eren
ceU
nad
j.A
dj.
Diff
eren
ceU
nad
j.A
dj.
Diff
eren
ceU
nad
j.A
dj.
(Std
.E
rr.)
pp
(Std
.E
rr.)
pp
(Std
.E
rr.)
pp
(Std
.E
rr.)
pp
AB
vs.
AB
C0.1
60.
440.4
4−
0.1
40.
330.3
3−
0.2
00.
120.
140.
050.
780.7
8(0.2
0)
(0.1
5)(0.1
3)(0.1
7)
AC
vs.
AB
C−
0.14
0.51
0.5
4−
0.1
10.
470.5
0−
0.2
10.
070.
10−
0.1
70.
280.2
8(0.2
1)
(0.1
5)(0.1
1)(0.1
5)
BC
vs.
AB
C−
0.11
0.58
0.5
80.
040.
790.7
9−
0.2
60.
030.
03−
0.2
00.
210.2
4(0.1
9)
(0.1
4)(0.1
1)(0.1
6)
AC
vs.
AB
−0.
290.
20
0.2
00.
030.
830.8
3−
0.0
10.
950.
95−
0.2
20.
240.2
6(0.2
3)
(0.1
7)(0.1
3)(0.1
8)
BC
vs.
AB
−0.
260.
24
0.2
70.
180.
270.2
9−
0.0
60.
630.
63−
0.2
40.
180.1
8(0.2
2)
(0.1
6)(0.1
3)(0.1
8)
BC
vs.
AC
0.03
0.89
0.9
00.
150.
330.3
8−
0.0
60.
670.
69−
0.0
30.
880.8
9(0.2
1)
(0.1
5)(0.1
3)(0.1
7)
Note
:B
ase
don
Model
s2,
4,
6and
8of
Table
6.
Diff
eren
ces
reflec
tth
efirs
ttr
eatm
ent
min
us
the
seco
nd
trea
tmen
t.A
dju
sted
pva
lues
are
ad-
just
edusi
ng
the
Dunca
nm
ethod.ABC
repre
sents
trea
tmen
tgro
ups
inw
hic
hP
layer
sA
,B
,and
Care
all
from
the
sam
etr
ibe,
AB
repre
sents
trea
tmen
tgro
ups
inw
hic
honly
Pla
yer
sA
and
Bare
from
the
sam
etr
ibe,
AC
repre
sents
trea
tmen
tgro
ups
inw
hic
honly
Pla
yer
sA
and
Care
from
the
sam
etr
ibe,
andBC
repre
sents
trea
tmen
tgro
ups
inw
hic
honly
Pla
yer
sB
and
Care
from
the
sam
etr
ibe.
22
Conclusion
This study examined how the experimental manipulation of a group’s ethnic composition
influenced sharing and costly income alteration in norm-laden (DG3) versus norm-free (RIG3)
contexts across Kenya and Tanzania. Across both games and in both countries, we found some
evidence that the coethnicity of group members influenced game play. We find evidence that
cooperative sharing and third-party punishment are conditioned by coethnicity, but theses
effects are only observed in Kenya. In Kenya, we find that cooperative sharing is greater when
the potential victim of non-cooperation and the third-party punisher are coethnics, regardless
of the dictator’s own ethnic affiliation. Punishment was indeed harsher for non-coethnics
who failed to share with the punisher’s coethnic, but not for coethnics who failed to share
in homogenous groups. Thus, in Kenya, we find evidence of parochialism (Bernhard et al.
2006), characterized by greater punishment of outgroup members who fail to share with in-
group members, and leniency towards selfish in-group members. However, these results derive
from estimates that exhibit considerable imprecision and we can reject the null hypotheses
with which they correspond only at confidence levels below the conventional 95% level.
Noting the uncertainty of our estimates, the patterns of ethnic-based sharing and sanc-
tioning we observe are at odds with prominent theories of coethnic cooperation in political
science. In particular, Habyarimana et al.’s (2009) coethnicity experiments in Uganda, Miguel
and Gugerty’s (2005) research on diverse communities in Kenya, and Fearon and Laitin’s
(1996) in-group policing model all anticipate that sanctioning will be concentrated on in-
group members, since its provision is costly and its benefits are shared among the whole
group. In contrast, our findings from Kenya are more consistent with punishment being used
to protect in-group members especially against out-group members. Economists and psy-
chologists have documented similar behavior in different contexts (Bernhard et al. 2006;
Baumgartner et al. 2012), which suggests that theories that assume costly punishment will
be reserved for coethnic non-cooperation should be amended to account for mounting evidence
to the contrary.
Consistent with our expectation that ethnic affiliation would shape cooperation and sanc-
23
tioning in Kenya but not Tanzania, we find no statistically significant effects of coethnicity on
behavior in Tanzania as compared to the weak evidence of ethnic effects in Kenya. We antic-
ipated this pattern due to stark differences in the degree of pan-ethnic national identification
in Tanzania compared to Kenya. However, like many studies that utilize African borders to
generate variation in an independent variable of interest (McCauley and Posner 2015), we are
unable to definitively attribute our findings to differences in nationalism alone. The location of
our study – very near the international border and with the same two ethno-linguistic groups
– holds constant geographic characteristics and local ethnic considerations that are likely to
be similar on each side of the border. Nevertheless, there are other important differences
between Kenya and Tanzania in the post-independence period in general, such as the degree
of political competitiveness and the nature of their economies, as well as differences in how
the locality on each side of the border is situated within the larger national context, includ-
ing distance to the capital and the relative sizes of the two ethnic groups. Such differences
are likely responsible for variation in play across the two countries, including higher rates of
punishment, anger at non-cooperation, and support for retribution in Kenya. In addition, the
modest difference we see in the role of ethnicity in shaping game play across these national
contexts could also result from a combination of these various factors.
We also note that the influence of coethnicity and national context appears limited in
comparison to the robust and noteworthy effects of norm violations and inequality. The most
robust predictor of costly sanctioning in our DG3 study is the amount that dictators allocated
to their game partner, which itself appears to have resulted primarily from factors beyond
the ethnic makeup of the group. Similarly, in the RIG3, randomly-generated inequality serves
as the best predictor of costly income reduction. In sum, our findings suggest that research
studying the role of ethnic affiliations and national context may prove relevant at explaining
some of the variation in the social behaviors we study, but factors shared across ethnic and
national contexts appear to be the primary drivers of both costly sanctioning and the costly
Note: Means with standard deviations in parenthesis. ABC represents treatment groups in which PlayersA, B, and C are all from the same tribe, AB represents treatment groups in which only Players A and Bare from the same tribe, AC represents treatment groups in which only Players A and C are from the sametribe, and BC represents treatment groups in which only Players B and C are from the same tribe. Age isa categorical variable representing a participants’ age (0 if between 18-29 years old, 2 if between 30-39 yearsold, 3 if between 40-49 years old, and 5 if over 50 years old). Education Level is a categorical variable thatrepresents participants’ highest level of education (0 if no formal schooling, 1 if some primary, 2 if primary, 3if some secondary, 4 if secondary, and 5 if post-secondary). Income Level is a categorical variable representingparticipants’ income in the previous week (0 if no income, 1 if KES 1-99/TZS 1-2,499, 2 if KES 100-249/TZS2,500-4,999, 3 if KES 250-499/TZS 5,000-7,499, 4 if KES 500-749/TZS 7,500-9,999, 5 if KES 750-999/TZS10,000-12,499, 6 if KES 1,000-1,249/TZS 12,500-14,999, 7 if KES 1,250-1,499/TZS 15,000-17,499, 8 if KES1,500-1,749/TZS 17,500-19,999, 9 if KES 1,750-1,999/TZS 20,000-29,999, 10 if KES 2,000-2,999/TZS 30,000+,and 11 if KES 3,000+). Male is a dummy indicating gender. Religiosity signifies the number of times aparticipant attended a religious service in the past week. Luo is a dummy indicating Luo ethnicity. EthnicMinority is a dummy indicating Luo living in Kuria-majority villages and Kuria living in Luo-majority villages.
2
Fig
ure
A.1
:C
ON
SO
RT
Dia
gram
Recruited
(n=672)
Excluded
(n=114):
a)Did
notshow
up(n=76)
b)Show
eduptoolate
(n=9)
c)Leftearly(n=3)
d)Could
notform
group(n=26)
Randomized
(n=558)
ABCD
G
ABCRIG
ABD
G
BCRIG
ACD
G
ABRIG
BCD
G
ACRIG
Assigned
(n=164)
Assigned
(n=138)
Assigned
(n=128)
Assigned
(n=128)
Excluded
(n=15):
a)Poor
comprehen
sion(n=8)
b)Ethnic
minority(n=5)
c)Spiteful(n=2)
Excluded
(n=22):
a)Poor
comprehen
sion(n=10)
b)Ethnic
minority(n=6)
c)Spiteful(n=6)
Excluded
(n=20):
a)Poor
comprehen
sion(n=9)
b)Ethnic
minority(n=7)
c)Spiteful(n=4)
Excluded
(n=16):
a)Poor
comprehen
sion(n=6)
b)Ethnic
minority(n=6)
c)Spiteful(n=4)
Analyzed(n=151):
AD
G:n=51
CD
G:n=51
CRIG:n=47
Analyzed(n=122):
AD
G:n=44
CD
G:n=40
CRIG:n=32
Analyzed(n=112):
AD
G:n=36
CD
G:n=38
CRIG:n=34
Analyzed(n=116):
AD
G:n=40
CD
G:n=39
CRIG:n=33
3
B Additional Analyses and Robustness Checks
B.1 Ordered Probit Regressions
Table A.2: Cooperative Sharing Across Coethnicity Treatments (Ordered Probit)
Kenya Tanzania
(1) (2) (3) (4)
AB −0.435 −0.729∗∗ −0.030 0.093(0.302) (0.338) (0.326) (0.333)
AC −0.611∗ −0.752∗∗ 0.128 0.099(0.319) (0.342) (0.346) (0.368)
BC 0.214 −0.014 −0.020 0.255(0.343) (0.360) (0.310) (0.341)
OLS estimates with standard errors in parentheses. ABC treatment group omitted.
The dependent variable is tokens spent by C to reduce income of A.
Village fixed effects are based on A’s village.∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01
8
Tab
leA
.7:
Cos
tly
Inco
me
Ad
just
men
tA
cros
sC
oet
hnic
Tre
atm
ents
(No
Eth
nic
Min
orit
ies)
Ken
yaT
an
zan
ia
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Am
ou
nt
All
oca
ted
(A)
0.0
92∗∗∗
0.0
75∗∗∗
0.05
5∗∗∗
0.0
56∗∗∗
(0.0
23)
(0.0
24)
(0.0
15)
(0.0
15)
Am
ount
All
oca
ted
(B)
0.09
1∗∗∗
0.07
8∗∗∗
0.0
96∗∗∗
0.0
97∗∗∗
(0.0
16)
(0.0
17)
(0.0
20)
(0.0
20)
AB
0.125
0.17
1−
0.1
35−
0.17
1−
0.18
5−
0.20
60.
084
0.02
6(0.1
92)
(0.1
99)
(0.1
36)
(0.1
42)
(0.1
30)
(0.1
35)
(0.1
71)
(0.1
81)
AC
−0.
124
−0.1
16−
0.2
18−
0.13
2−
0.20
1∗−
0.20
9∗
−0.1
61−
0.1
67(0.1
94)
(0.2
05)
(0.1
38)
(0.1
46)
(0.1
17)
(0.1
16)
(0.1
53)
(0.1
56)
BC
0.144
−0.0
42−
0.1
60−
0.00
2−
0.24
3∗−
0.25
7∗∗
−0.1
78−
0.1
98(0.1
77)
(0.1
96)
(0.1
26)
(0.1
40)
(0.1
23)
(0.1
21)
(0.1
61)
(0.1
62)
Age
−0.0
10−
0.00
10.0
81∗
−0.0
49(0.0
72)
(0.0
51)
(0.0
42)
(0.0
56)
Male
−0.0
730.2
26∗
0.10
70.0
46(0.1
72)
(0.1
23)
(0.0
97)
(0.1
30)
Ed
uca
tion
Lev
el0.
130∗
−0.1
31∗∗
−0.
056
−0.1
89∗∗
(0.0
77)
(0.0
55)
(0.0
67)
(0.0
91)
Inco
me
Lev
el0.
010
−0.0
10−
0.02
5∗−
0.0
14(0.0
25)
(0.0
17)
(0.0
14)
(0.0
19)
Con
stan
t−
0.22
9−
0.4
12−
0.0
400.
443∗
0.10
40.
073
−0.
170
0.22
8(0.1
96)
(0.3
61)
(0.1
01)
(0.2
33)
(0.1
19)
(0.2
34)
(0.1
23)
(0.3
13)
Vil
lage
Fix
edE
ffec
tsN
oY
esN
oY
esN
oY
esN
oY
es
R2
0.25
0.41
0.37
0.50
0.27
0.40
0.34
0.42
Ob
serv
atio
ns
6666
6666
7473
7473
OL
Ses
tim
ate
sw
ith
standard
erro
rsin
pare
nth
eses
.A
BC
trea
tmen
tgro
up
om
itte
d.
The
dep
enden
tva
riable
isto
ken
ssp
ent
by
Cto
reduce
inco
me
of
A(M
odel
s1-2
,5-6
)or
B(M
odel
s3-4
,7-8
)
Villa
ge
fixed
effec
tsare
base
don
C’s
villa
ge.
∗p<
0.1
0,∗∗p<
0.0
5,∗∗∗p<
0.0
1
9
B.3 Including Spiteful Income Adjustment
10
Tab
leA
.8:
Cos
tly
Inco
me
Ad
just
men
tA
cros
sC
oet
hn
icT
reat
men
ts(w
ith
Sp
itef
ul
Pu
nis
her
s)
Ken
yaTanzania
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Am
ount
Alloca
ted
(A)
0.0
68∗∗∗
0.0
52∗∗
0.0
42∗∗∗
0.0
43∗∗∗
(0.0
21)
(0.0
24)
(0.0
15)
(0.0
15)
Am
ount
Alloca
ted
(B)
0.0
69∗∗∗
0.0
66∗∗∗
0.0
83∗∗∗
0.0
83∗∗∗
(0.0
18)
(0.0
21)
(0.0
19)
(0.0
20)
AB
0.0
99
0.1
46
−0.0
93
−0.1
67
−0.2
35∗
−0.2
45∗
0.2
23
0.2
06
(0.1
88)
(0.2
01)
(0.1
57)
(0.1
70)
(0.1
23)
(0.1
27)
(0.1
58)
(0.1
66)
AC
−0.0
55
−0.0
25
−0.2
70∗
−0.2
85
−0.2
13∗
−0.2
45∗∗
−0.0
64
−0.0
81
(0.1
94)
(0.2
09)
(0.1
62)
(0.1
77)
(0.1
17)
(0.1
20)
(0.1
50)
(0.1
56)
BC
0.1
81
0.0
83
−0.2
27
−0.2
21
−0.2
40∗
−0.2
60∗∗
−0.0
73
−0.1
14
(0.1
73)
(0.1
94)
(0.1
44)
(0.1
65)
(0.1
22)
(0.1
22)
(0.1
57)
(0.1
58)
Age
0.0
06
−0.0
57
0.0
55
−0.0
16
(0.0
71)
(0.0
60)
(0.0
41)
(0.0
54)
Male
0.0
26
0.1
77
0.1
42
−0.0
63
(0.1
72)
(0.1
46)
(0.0
96)
(0.1
24)
Educa
tion
Lev
el0.0
79
−0.0
77
−0.0
01
−0.1
95∗∗
(0.0
79)
(0.0
67)
(0.0
64)
(0.0
84)
Inco
me
Lev
el−
0.0
07
0.0
22
−0.0
15
−0.0
14
(0.0
24)
(0.0
21)
(0.0
13)
(0.0
17)
Const
ant
−0.0
85
−0.1
29
0.0
79
0.3
48
0.1
99∗
0.1
07
−0.1
26
0.3
37
(0.1
85)
(0.3
70)
(0.1
18)
(0.2
91)
(0.1
16)
(0.2
36)
(0.1
28)
(0.3
04)
Villa
ge
Fix
edE
ffec
tsN
oY
esN
oY
esN
oY
esN
oY
es
R2
0.1
50.2
70.2
00.3
00.1
90.3
00.2
70.3
5O
bse
rvati
ons
74
74
74
74
87
86
87
86
OL
Ses
tim
ate
sw
ith
standard
erro
rsin
pare
nth
eses
.A
BC
trea
tmen
tgro
up
om
itte
d.
The
dep
enden
tva
riable
isto
ken
ssp
ent
by
Cto
reduce
inco
me
of
A(M
odel
s1-2
,5-6
)or
B(M
odel
s3-4
,7-8
)V
illa
ge
fixed
effec
tsare
base
don
A’s
villa
ge.
∗p<
0.1
0,∗∗p<
0.0
5,∗∗∗p<
0.0
1
11
C Statistical Power
The experiment reported in the main text involves four treatment conditions implemented across two
national contexts. By spreading the study sample across these factors, our investigation allocates a
relatively small number of observations to each treatment condition in each national context. These
allocations of experiment participants raise concerns about statistical power.
To understand the relevance of these concerns, we performed post hoc power calculations for a
range of possible effect sizes, taking as given the sample sizes used in the various pairwise compar-
isons reported in the main text. Figures A.2 and A.3 display these calculations for our Kenyan and
Tanzanian samples, respectively. Each figure consists of six panels representing each of the six pair-
wise comparisons that can be made between our treatment conditions. At the top of each panel, the
relevant pairwise comparison is listed along with the sample size for each treatment condition. In
the plotting field of each panel, we display the statistical power of a two-sample t-test with unequal
sample sizes (viz. those sample sizes displayed in the title of each panel) for a range of hypothetical,
standardized effect sizes. The hypothetical effect sizes (in standard deviation units) range from those
deemed very small (d = 0.25) to very large (d = 0.8) by past researchers (Cohen 1988). We do not
use the estimated effect sizes from our present investigation because we acknowledge that they might
themselves result from inadequate power. Instead, we draw on hypothetical effects sizes in order to
learn how large an effect would need to be in order for us to detect it, given our sample sizes.
As Figures A.2 and A.3 indicate, for small effect sizes we are grossly under-powered and we do
not reach conventional levels of power (power=0.8) even for effects that researchers would deem large.
This dearth of power appears evident across all panels, thus suggesting that we remain under-powered
for all comparisons of our focal treatment conditions.
12
Figure A.2: Power Calculations for Treatment Comparisons in Kenya
0.3 0.4 0.5 0.6 0.7 0.8
0.2
0.3
0.4
0.5
0.6
0.7
0.8
ABC (n=24) and AB (n=26)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.2
0.3
0.4
0.5
0.6
0.7
ABC (n=24) and AC (n=21)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
ABC (n=24) and BC (n=18)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.2
0.3
0.4
0.5
0.6
0.7
AB (n=26) and AC (n=21)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.2
0.3
0.4
0.5
0.6
0.7
AB (n=26) and BC (n=18)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
AC (n=21) and BC (n=18)
Hypothetical Effect Size
Pow
er
Note: The panels display the power of a two-sample t-test for varying effect sizes (in standard deviationunits), given the actual, unequal sample sizes for experiments conducted in Kenya.
13
Figure A.3: Power Calculations for Treatment Comparisons in Tanzania
0.3 0.4 0.5 0.6 0.7 0.8
0.2
0.3
0.4
0.5
0.6
0.7
ABC (n=27) and AB (n=18)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
ABC (n=27) and AC (n=15)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.2
0.3
0.4
0.5
0.6
0.7
0.8
ABC (n=27) and BC (n=22)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.1
0.2
0.3
0.4
0.5
0.6
AB (n=18) and AC (n=15)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
AB (n=18) and BC (n=22)
Hypothetical Effect Size
Pow
er
0.3 0.4 0.5 0.6 0.7 0.8
0.1
0.2
0.3
0.4
0.5
0.6
AB (n=15) and BC (n=22)
Hypothetical Effect Size
Pow
er
Note: The panels display the power of a two-sample t-test for varying effect sizes (in standard deviationunits), given the actual, unequal sample sizes for experiments conducted in Tanzania.
14
D Participant Recruitment
Before the experiment began, research teams made visits to the target villages to seek permission
from local authorities to recruit participants. Participants were recruited from 12 different villages
in southwest Kenya and northwest Tanzania – where the ancestral homelands of the Kuria and Luo
meet. This included 8 small villages near Karamu, Kenya (4 Kuria villages and 4 Luo villages) and 4
small villages near Somba Nyasoko, Tanzania (2 Kuria villages and 2 Luo villages). The RAs recruited
participants by taking a random walk from a pre-designated part of the village (typically the center)
and recruiting a random household member from every Xth home, determined by a dice roll. When
recruiting, RAs alternated between gender, and selected individuals randomly using a dice roll. These
procedures involved the following script (spoken in Swahili or participant’s mother tongue):
Hello, my name is , and we would like to recruit a member of your household
to participate in a research project involving [Kenyatta University (Kenya) / the Shirati
Health, Education, and Development Foundation (Tanzania)], and Stanford University
in the United States. If your household agrees to participate, you will be given [KES
300 / TZS 5000] cash for your time and matched with other participants from the area
to complete a series of activities. The experiment will be held next weekend here in
at 8am, and it should take no more than four or five hours total. Would
somebody here be interested in participating?
Please know that participation is completely voluntary, and you may withdraw from the
research at any time. But if you show up for the experiment you will receive [KES 300 /
TZS 5000] and a chance to win more money depending on the decisions made by you and
others in the experiment. May I schedule you for the experiment?
Great, from this household, I need to select a [male / female]. Can we gather all the
[males / females] of the house?
[SELECT BY ROLLING DICE AND RECORD RESPONDENT INFO]
You are now registered for the experiment. It will be held at . Please know
that other people from this village are being recruited to attend sessions on different days,
so you should pay attention to the day that you are invited, which is written here on the
15
experiment info sheet. Please try to arrive 15 minutes early – if you are late, we cannot
give you any money.
E Experiment Procedures
All experiment instructions given to a group were in Swahili. During one one one interactions between
the RA and a participant, either Swahili or the participant’s home language (Kuria or Luo).
Each experiment session in Kenya and Tanzania involved 30 participants at the most. When
participants arrived to the experiment site they were randomly assigned a participant ID that ranged
from 1-30 by drawing a number out of a hat. These IDs were then used to randomly assign participants
to groups of three. Figure A.4 shows the forms used to connect randomly assigned numbers to specific
groups and roles within each game for that group. We relied variation in the ethnic composition of
these randomly assigned groups to generate variation in the treatment conditions. In experimental
sessions with equal proportions of Kuria and Luo participants, all four experimental conditions (ABC,
AB, AC, and BC) were equally likely. The more unbalanced the ethnic composition of the experimental
session, however, the more likely the ABC condition became relative to the other three conditions. On
average, our 20 sessions were 55% one ethnic group and 45% the other, but imbalance was as large
as 71% one ethnic group in one session. In this most extreme case, assignment to the ABC condition
became 50% more likely than each of the other three conditions. This helps account for the greater
number of ABC groups (n = 55) than AB (n = 45), AC (n = 44), and BC (n = 42) in the DG3 game.
Participants that could not be assigned to a group were sent away with the promised show-up fee.
Each participant played one dictator game and one random-income game. To avoid the possibility
of spillovers across games we took two precautions. First, results from the experimental games were not
given to participants until after the experiment was over. Second, player assignments were designed
so that each participant made only one experiment decision (despite playing two games). This was
accomplished by assigning player B of the dictator game to the role of player C in the random-income
game, and assigning players A and C of the dictator game to the role of players A and B in the
random-income game. All experiment decisions and survey responses were recorded on mobile phones
equipped with Open Data Kit (ODK).
16
Figure A.4: Form which mapped numbers 1-30 drawn from a hat to particular groups androles for the dictator game with third party punishment.
PLA
YE
R A
SS
IGN
ME
NT
KE
Y –
Dic
tato
r (A
ctiv
ity 1
) C
ount
ry:
Ken
ya /
Tanz
ania
D
ate:
___
____
__
Village'
ID'
Interview2'
Game2'
Role2'
Grou
p2'
Allocatio
n2'
Redu
ction2
'Pa
yout2'
!1!
YES!
Dictator!
A!1!
!!
!!
2!!
Dictator!
B!1!
!!
!!
3!YES!
Dictator!
C!1!
!!
!!
4!YES!
Dictator!
A!2!
!!
!!
5!!
Dictator!
B!2!
!!
!!
6!YES!
Dictator!
C!2!
!!
!!
7!YES!
Dictator!
A!3!
!!
!!
8!!
Dictator!
B!3!
!!
!!
9!YES!
Dictator!
C!3!
!!
!!
10!
YES!
Dictator!
A!4!
!!
!!
11!
!Dictator!
B!4!
!!!!
!!
12!
YES!
Dictator!
C!4!
!!!!
!!
13!
YES!
Dictator!
A!5!
!!
!!
14!
!Dictator!
B!5!
!!!!
!!
15!
YES!
Dictator!
C!5!
!!!!
!!
16!
YES!
Dictator!
A!6!
!!
!!
17!
!Dictator!
B!6!
!!!!
!!
18!
YES!
Dictator!
C!6!
!!!!
!!
19!
YES!
Dictator!
A!7!
!!
!!
20!
!Dictator!
B!7!
!!!!
!!
21!
YES!
Dictator!
C!7!
!!!!
!!
22!
YES!
Dictator!
A!8!
!!
!!
23!
!Dictator!
B!8!
!!!!
!!
24!
YES!
Dictator!
C!8!
!!!!
!!
25!
YES!
Dictator!
A!9!
!!
!!
26!
!Dictator!
B!9!
!!!!
!!
27!
YES!
Dictator!
C!9!
!!!!
!!
28!
YES!
Dictator!
A!10!
!!
!!
29!
!Dictator!
B!10!
!!!!
!!
30!
YES!
Dictator!
C!10!
!!!!
!! R
A N
ame:
___
____
____
____
___!
17
Figure A.5: Form which mapped numbers 1-30 drawn from a hat to particular groups androles for the random income game with third party income adjustment.
PLA
YER
ASS
IGN
MEN
T K
EY –
RIG
(Act
ivity
2)
Cou
ntry
: K
enya
/ Ta
nzan
ia
D
ate:
___
____
__
Village'
ID'
Interview2'
Game2'
Role2'
Grou
p2'
Allocatio
n2'
Redu
ction2
'Pa
yout2'
!1!
!RIG!
B!1!
!!
!!
2!YES!
RIG!
C!1!
!!
!!
3!!
RIG!
A!1!
!!
!!
4!!
RIG!
B!2!
!!
!!
5!YES!
RIG!
C!2!
!!
!!
6!!
RIG!
A!2!
!!
!!
7!!
RIG!
B!3!
!!
!!
8!YES!
RIG!
C!3!
!!
!!
9!!
RIG!
A!3!
!!
!!
10!
!RIG!
B!4!
!!
!!
11!
YES!
RIG!
C!4!
!!!!
!!
12!
!RIG!
A!4!
!!!
!!
13!
!RIG!
B!5!
!!
!!
14!
YES!
RIG!
C!5!
!!!!
!!
15!
!RIG!
A!5!
!!!
!!
16!
!RIG!
B!6!
!!
!!
17!
YES!
RIG!
C!6!
!!!!
!!
18!
!RIG!
A!6!
!!!
!!
19!
!RIG!
B!7!
!!
!!
20!
YES!
RIG!
C!7!
!!!!
!!
21!
!RIG!
A!7!
!!!
!!
22!
!RIG!
B!8!
!!
!!
23!
YES!
RIG!
C!8!
!!!!
!!
24!
!RIG!
A!8!
!!!
!!
25!
!RIG!
B!9!
!!
!!
26!
YES!
RIG!
C!9!
!!!!
!!
27!
!RIG!
A!9!
!!!
!!
28!
!RIG!
B!10!
!!
!!
29!
YES!
RIG!
C!10!
!!!!
!!
30!
!RIG!
A!10!
!!!
!! R
A N
ame:
___
____
____
____
____
____
____
____
___
18
E.1 Common Group Instructions
Once all participants were present and assigned IDs, the RAs started the experiment by obtaining oral
consent and giving the experiment introduction to participants collectively:
Thank you for signing up for this experiment, and for contributing your time and effort
to this research. Today you will be matched with other participants from the area and
asked to complete a series of activities. The purpose of this experiment is to study how
people spend money. Please know that your participation here is completely voluntary
and that you are free to withdraw from this experiment at any time without any penalty.
As promised, you will receive [KES 300 (Kenya) / TZS 5000] cash for showing up today.
But if you complete the experiment, you will have a chance of winning more depending
on the decisions made by you and others in the experiment.
Do you agree to participate?
After obtaining oral consent:
Great, let’s get started. Today’s experiment may take up to 4 hours so if you think you
will not be able to stay that long let me know now. Before we begin I want to make
some general comments about what we are doing here today and explain the rules that we
must follow. We will be performing some experiments in which you can get some money.
Whatever money you will get in the experiments will be yours to keep and take home.
Maybe you won’t get any money from the experiment, but you will receive the promised
KES 300 (Kenya)/ TZS 5000 (Tanzania) for showing up today. This money is not part
of the experiment, it is yours to keep. Assistant 1 and I will be supplying the money,
but you should understand that this is not our own money. It is money given to us by a
University in the United States to use for research.
Before we proceed any further, let me stress something that is very important. You were
invited here without understanding very much about what we are planning to do today.
If at anytime you find that this is something that you do not wish to participate in for
any reason, you are free to leave. You may leave at anytime whether we have started the
experiment or not.
I will now explain the experiment to you. Afterwards each of you will come into the
adjacent room one-at-a-time with me and carry out the experiment. It is important that
19
you listen as carefully as possible, because only people who understand the experiment
will actually be able to participate. We will run through some examples here while we
are all together. You cannot ask questions or talk while here in the group. This is very
important. Please be sure that you obey this rule because it is possible for one person to
spoil the experiment for everyone. If one person talks about the experiment while sitting
in the group, we will not be able to carry out the experiment today. Do not worry if you
do not completely understand the experiment as we go through the examples here in the
group. Each of you will have a chance to ask questions in private to be sure that you
understand what you have to do.
Also, your decisions in the experiment will determine how many shillings you receive at
the end of the experiment. In the experiment we will use tokens, not shillings. Here are
100 (Kenya) / 2000 (Tanzania) shillings. (Currency is shown to participants.) Here are
ten tokens (Poker chips are shown to subjects.) Every token is worth 10 / 200 shillings.
(Subjects are shown 1 token and 10 / 200 shillings.) In the experiment we will only use
tokens, but at the end of the experiment each token will be exchanged for shillings. If you
obtain 0 tokens, you will get 0 shillings; if you obtain 1 token, you will obtain 10 / 200
shillings; if you obtain 2 tokens, you will obtain 20 / 400 shillings; if you obtain 3 tokens,
you will obtain 30 / 600 shillings; if you obtain 4 tokens, you will obtain 40 / 800 shillings;
if you obtain 5 tokens, you will obtain 50 / 1000 shillings; if you obtain 6 tokens, you will
obtain 60 / 1200 shillings; if you obtain 7 tokens, you will obtain 70 / 1400 shillings; if
you obtain 8 tokens, you will obtain 80 / 1600 shillings; if you obtain 9 tokens, you will
obtain 90 / 1800 shillings; if you obtain 10 tokens, you will obtain 100 / 2000 shillings.
After the introduction, RAs explained what the dictator and random-income games would entail,
starting with the dictator game (activity 1):
We will now explain the experiment. Please listen closely as you will need to remember
these instructions in order to complete the experiment and earn additional shillings.
You will all play two types of games: Activity 1, and Activity 2. Both activities are
similar and involve three people – Person A, Person B, and Person C. Each of you have
been assigned a random number and this determines whether you will be playing the role
of A, B, or C in these activities. The number was determined by chance, and each of you
had an equal chance of being assigned to each role. During the experiment, none of you
20
will know exactly with whom you are interacting, only which village the other person is
from. I cannot tell you the identity of the people you are matched with, so please do not
ask. I will, however, tell you that today there are people from 4 villages: W, X, Y, Z.
Here are 100 / 2000 shillings. (Currency is shown to participants.) Here are ten tokens
(Poker chips are shown to subjects.) Again, remember, every token is worth 10 / 200
shillings. (Subjects are shown 1 token and 10 / 200 shillings.) Here are 10 tokens.
In Activity 1, Person A must decide how many of these ten tokens to give to Person B
and how many to keep. Person B takes home whatever Person A gives to him, but Person
A has to wait until Person C has made a decision before finding out what he is going to
take home. Person C is given 5 tokens. Person C can do two things with his 5 tokens.
1. Person C can reduce the income of Person A. For the cost of 1 token, Person C can
reduce by 3 tokens the amount of money Person A gets to keep.
2. Person C can pay nothing, keep the 5 tokens and leave Person A with the tokens he
or she wanted to keep for him or herself untouched.
All of these decisions will be made anonymously in the adjacent room. We will call each
of you in order of your participant ID. When it’s your turn to do the experiment, you
can come inside the adjacent room. I will tell you whether you are Person A, Person
B, or Person C, explain the experiment again, and ask you to work through a couple of
examples to be sure that you understand. After you have completed the experiment, you
can come back out to this room and wait for others to complete the experiment. Please
know that it is perfectly acceptable to keep all the money given to you in the activity.
Some people must take home the money because they have a sick child or must pay for
their children’s school fees. It is also ok to use the money to pay for food and other bills.
The random-income game (activity 2) was then described:
Now there is also a second activity. In Activity 2, there are still three players, A, B, C.
Rules are similar to Activity 1, in that Persons A and B receive some amount of money,
and Person C gets a chance to modify the income of Person A or B. However, in this
activity, the amount of money that Person A and B receive is determined randomly by
this wheel (see Figure A.6). Each wheel is divided into 12 sections. Each section has a
value in it from 0 to 10 and one re-spin. Each wheel will be spun once for each group and
21
this will determine the incomes of A and B in the activity. The likelihood of the marker
landing on any given number is the same for all numbers. The number that the marker
lands on indicates the number of tokens given to Person A. For example, if the marker
lands on 6, then Person A is given 6 tokens. Person B gets the remaining tokens; that is,
Person B gets 10 minus the number of tokens shown on the wheel.
Figure A.6: Wheel used to generate random division of 10 tokens in the RIG3.
Before Person A and Person B are given these tokens, however, Person C must make a
decision. Person C is given 5 tokens. Person C can do three things with his 5 tokens.
1. Person C can reduce the income of Person A. For the cost of 1 token, Person C can
reduce by 3 tokens the amount of money obtained by Person A.
2. Person C can reduce the income of Person B. For the cost of 1 token, Person C can
reduce by 3 tokens the amount of money obtained by Person B.
3. Person C can pay nothing, keep the 5 tokens and leave the money of Person A and
Person B untouched.
Again, this will be done anonymously in the adjacent room. We will call each of you in
22
order of your participant ID. When it’s your turn to do the experiment, you can come
inside the adjacent room. I will tell you whether you are Person A, Person B, or Person
C, explain the experiment again, and ask you to work through a couple of examples to be
sure that you understand. After you have completed the experiment, you can come back
out to this room and wait for everybody else to complete the experiment.
When you have finished there, you have to wait until everybody has performed the exper-
iment. Remember that you are not allowed to come and talk to the people still waiting
to carry out the experiment. When everyone has finished the experiment, I will again call
you in one-at-a-time and pay you your experiment winnings. Again, please know that it
is perfectly acceptable to keep all the money given to you in the experiment.
Are there any questions?
After the common group instructions were given, RAs took participants to a private room one by
one to conduct the games.
E.2 Dictator Game Interview for Player A
Player A of the dictator game received the following instructions:
Hello. I will now interview you to play Activity 1. You’ve been selected to play the role
of A. As I have told you, there are three persons in this experiment – Person A, Person B,
and Person C. None of you will know exactly with whom you are interacting, only which
village they are from. Only I know who will be interacting with whom; I will never tell
anyone else. Now you yourself are Person A, and you are playing with a B from village
and a C from village .
[SHOW PROPS HERE DISPLAYING A, B, C, THEIR VILLAGES, AND ARROWS
INDICATING THE DIRECTION OF TRANSFER]
Here are 10 tokens. You must decide how much of this money you want to give to Person
B from village and how much you want to keep for yourself. Person B
takes home whatever you give him, but you will have to wait until Person C from village
has decided what he wants to do before finding out how much money you
23
can take home. Person C from village will be given 5 tokens. Person C can
do two things with his 5 tokens.
1. Person C can reduce the income of Person A. For the cost of 1 token, Person C can
reduce by 3 tokens the amount of money Person A wants to keep.
2. Person C can pay nothing, keep the 5 tokens and leave the money Person A wanted
to keep for himself untouched.
Now I would like to ask you some questions to make sure you understand the activity:
1. You are A, and imagine that you allocate 1 token to person B. How many tokens do
you have at that point? [Answer: 9 tokens]
2. Ok, so you have 9 tokens. Say that C reduces your income by 6. How many tokens
do you have at that point? [Answer: 3 token]
3. How many tokens does B have at that point? [Answer: 1 token]
4. How many tokens did C have to pay to reduce your income by 6? [Answer: 2 tokens]
[FURTHER EXAMPLES AND TEST QUESTIONS, IF NEEDED]
Now it is your turn to play. Here are 10 tokens. You can now decide how much of this
money you want to give Person B and how much money you want to keep for yourself.
Please divide this money into two piles and put the amount that you wish to give to
Person B from village in the B cup and the amount that you wish to keep
for yourself in your A cup, and remember it is ok to keep all the tokens for yourself.
Okay, we will split the money as you decided between yourself and Player B. To see how
much you can take home of the money you have kept for yourself, we first have to find
out what Person C decides to do. We will tell you how much you can take home at the
end of the experiment, after everybody has played. In the meantime, please do not talk
to anybody else about the activities you played.
E.3 Dictator Game Interview for Player C
Player C of the dictator activity received the following instructions:
24
Hello. I will now interview you to play Activity 1. You’ve been selected to play the role of
C. As I have told you, there are three persons in this experiment ? Person A, Person B,
and Person C. None of you will know exactly with whom you are interacting, only their
village. Only I know who is to interact with whom and I will never tell anyone else. Now
you yourself are Person C, and you are playing with an A from village and
a B from village . [Show props here displaying A, B, C, their villages, and
arrows indicating direction of transfers?]
Person A was given 10 tokens, worth 100 / 2000 shillings. Person A told me how much
of these 10 tokens he wants to give to Person B from village , and how much
he wants to keep for himself to take home. Now I will give you 5 tokens. With these 5
tokens you can do one of two things:
1. You can reduce the income of Person A. For the cost of 1 token, you can reduce by
3 tokens the amount of money Person A wants to keep. The most you can reduce
Person A’s money to is zero tokens.
2. You can do nothing, keep the 5 tokens and leave the money Person A wanted to
keep for himself untouched.
Now I’d like to ask you some questions to make sure you understand the activity:
1. Imagine that person A allocates 1 token to person B. How many tokens does A have
at that point? [Answer: 9 tokens]
2. Ok, so A has 9 tokens, and you are C. Say that you reduce A’s income by 6. How
many tokens does A have at that point? [Answer: 3 token]
3. How many tokens does B have at that point? [Answer: 1 token]
4. How many tokens did you have to pay to reduce A’s income by 6? [Answer: 2 tokens]
[FURTHER EXAMPLES AND TEST QUESTIONS, IF NEEDED]
Now its your turn to play. The allocation that Person A from village has
made to Person B from village is written on the paper in front of me. It
says that A kept and gave to B. Would like to reduce Person
25
A’s income? If yes, please tell me how many tokens you’d like to reduce A’s income by.
Remember you must pay 1 token for every 3 tokens by which you reduce A’s income, and
its ok to keep all 5 tokens for yourself.
[DIVIDE The 5 TOKENS VISUALLY TO CONFIRM RESPONSE]
Okay, we will reduce A’s income by the amount you wish. We will tell you how much you
can take home at the end of the experiment, after everybody has played the two activities.
In the meantime, please do not talk to anybody else about the activities you played.?
E.4 Random Income Activity Interview for Player C
The following instructions were given to player C of the random-income game:
Hello. I will now interview you to play Activity 2. You’ve been selected to play the role of
C. As I have told you, there are three persons in this experiment ? Person A, Person B,
and Person C. None of you will know exactly with whom you are interacting, only their
village. Only I know who is to interact with whom and I will never tell anyone else. Now
you yourself are Person C, and you are playing with an A from village and
a B from village .
[SHOW PROPS HERE DISPLAYING A, B, C, THEIR VILLAGES, AND ARROWS
INDICATING THE DIRECTION OF TRANSFER]
In a moment I will spin a wheel and it will determine how much Person A and B receive
in this activity. Person A from will receive the amount of tokens equal to the value on
which the marker landed. Person B will receive 10 minus that number of tokens. But
first, I will give you 5 tokens. With these 5 tokens you can do three things:
1. You can reduce the income of Person A. For the cost of 1 token, you can reduce by
3 tokens the amount of money obtained by Person A. You can reduce the amount
of money Person A wants to at most zero tokens.
26
2. You can reduce the income of Person B. For the cost of 1 token, you can reduce by
3 tokens the amount of money obtained by Person B. You can reduce the amount of
money Person B wants to at most zero tokens.
3. You can pay nothing, keep the 5 tokens and leave the money of Person A and Person
B untouched.
Now I’d like to ask you some questions to make sure you understand the activity:
Imagine that I spin this wheel and person A gets 9 tokens and person B gets 1 token. Say
that you reduce A’s income by 3 tokens.
1. How many tokens does A have at that point? [Answer: 6 tokens]
2. How many tokens does B have at that point? [Answer: 1 token]
3. How many tokens did you have to pay to reduce A’s income by 3? [Answer: 1 token]
4. If you keep all 5 tokens, how many tokens do A and B have at that point? [Answer:
9 tokens for A and 1 token for B]
[FURTHER EXAMPLES AND TEST QUESTIONS, IF NEEDED]
Now I will spin this wheel to determine Person A and B’s incomes. [Spin Wheel] The
wheel marker landed on . This means that Person A currently will receive
tokens and Person B will receive tokens. Now, do you want
to pay to reduce one of the players’ incomes? If yes, whose income, and by how many
tokens? Remember you must pay 1 token for every 3 tokens by which you reduce A or
B’s income, and it is ok to keep all 5 tokens for yourself.
[DIVIDE The 5 TOKENS VISUALLY TO CONFIRM RESPONSE]
Okay, we will reduce A’s and B’s incomes by the amount you wish. We will tell you how
much you can take home at the end of the experiment, after everybody has completed the
experiment. In the meantime, please do not talk to anybody else about the activities you
played.
27
F Exit Survey
F.1 Instructions
After completing the experimental games, the RAs thanked participants for their time and informed
them that they would be called back into the adjacent room one at a time to take the exit survey and
receive their winnings.
The exit survey was administered using the following script:
The experiment is now complete. Thank you again for your participation. Your group
responses have been recorded, and in just a moment I will pay you your winnings from
both activities.
Before I pay you, however, I’d like to ask you a few questions about your background, your
interests, and what you thought about the experiment. Your responses to these questions
will not affect your winnings so please answer the questions honestly. Your responses will
remain completely confidential, and they will be used only by researchers to study how
people make decisions involving money.
After finishing the exit survey, participants were given their experiment winnings and asked not
to speak with other participants until everybody has been interviewed and paid out.
F.2 Exit Survey Questionnaire
1. Participant ID.
2. Experiment Session ID / Date / Country.
3. Participant Gender.
4. What is your age?
(a) 18-29
(b) 30-39
(c) 40-49
(d) 50-59
(e) 60+
5. Have you ever lived outside your current district for more than 6 months? Yes or No
28
6. What is your highest level of education?
(a) None
(b) Some primary
(c) Finished primary
(d) Some secondary
(e) Finished secondary
(f) Post-secondary school (university)
7. Which languages can you speak well enough to have a conversation?
(do not read list, check all those named)
(a) English
(b) Kuria
(c) Luo
(d) Swahili
(e) Other
8. What is your tribe?
(a) Kuria
(b) Luo
(c) Other
9. In the past week, how many times did you attend a religious service, if at all?
10. Did you earn any cash income in the past week? Yes or No
If yes, how much?
11. In the past week, how many days did you read the newspaper, if at all?
12. Imagine that in Activity 1 of this experiment, an individual gave you 1 token and kept 9. Please
indicate your feelings toward this person.
(a) Not at all angry
(b) A little angry
29
(c) Angry
(d) Quite angry
(e) Very angry
13. Think about all Tanzanians [Kenyans]. Which of the following statements is closest to your
view?
(a) I see myself as quite similar to most Tanzanias [Kenyans].
(b) I see myself as quite different from most Tanzanians [Kenyans].
14. Think about all Tanzanians [Kenyans]. Which of the following statements is closest to your
view?
(a) Because there is a lot of cultural variety in Tanzania, there is very little that makes us the
same.
(b) Even though there is a lot of cultural variety in Tanzania, we are more the same than we
are different.
15. Think about all Luos [Kurias]. Which of the following statements is closest to your view?
(a) I see myself as quite similar to most Luos [Kurias].
(b) I see myself as quite different from most Luos [Kurias].
16. Do you personally know anybody that has married a member of a different tribe?
(a) Yes
(b) No
17. Let us suppose that you had to choose between being a [national group] and being a [ethnic
group. Which of the following statements best expresses your feelings?
(a) You feel only [national].
(b) You feel more [national] than [ethnic].
(c) You feel equally [national] and [ethnic].
(d) You feel more [ethnic] than [national].
(e) You feel only [ethnic].
30
18. How much do you agree with the following statement: In order for justice to be served, violence