Content uploaded by Abhijit Ramalingam
Author content
All content in this area was uploaded by Abhijit Ramalingam on Dec 03, 2017
Content may be subject to copyright.
Accepted Manuscript
Title: ‘Doggedness’ or ‘disengagement’? An experiment on
the effect of inequality in endowment on behavior in team
competitions
Author: Shaun P. Hargreaves Heap Abhijit Ramalingam
Siddharth Ramalingam Brock V. Stoddard
PII: S0167-2681(15)00269-3
DOI: http://dx.doi.org/doi:10.1016/j.jebo.2015.10.002
Reference: JEBO 3683
To appear in: Journal of Economic Behavior & Organization
Received date: 8-2-2014
Revised date: 9-9-2015
Accepted date: 1-10-2015
Please cite this article as: Heap, S.P.H., Ramalingam, A., Ramalingam, S., Stoddard,
B.V.,‘Doggedness’ or ‘disengagement’? An experiment on the effect of inequality in
endowment on behavior in team competitions, Journal of Economic Behavior and
Organization (2015), http://dx.doi.org/10.1016/j.jebo.2015.10.002
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Page 1 of 30
Accepted Manuscript
‘Doggedness’ or ‘disengagement’? An experiment on the effect of
inequality on effort in team competitions
Highlights
Lab experiment examining individual public good contributions in competing teams
First investigation of effect of inequality in endowments between competing teams
Competition boosts effort up to moderate inequality but not at high levels of
inequality
The effect is strongest for the poor - they respond doggedly at moderate inequality
The 'rich' disengage from competition at high inequality levels
Page 2 of 30
Accepted Manuscript
2
‘Doggedness’ or ‘disengagement’? An experiment on the effect of
inequality in endowment on behavior in team competitions
Shaun P. Hargreaves Heap a, Abhijit Ramalingam b*,
Siddharth Ramalingam c, Brock V. Stoddard d
a Department of Political Economy, King’s College London, The Strand, London WC2, UK
s.hargreavesheap@kcl.ac.uk, Ph: +44-20-7848-1689.
b School of Economics and Centre for Behavioural and Experimental Social Science,
University of East Anglia, Norwich NR4 7TJ, UK, a.ramalingam@uea.ac.uk,
Ph: +44-1603-597382.
c IDFC Foundation, Naman Chambers, G Block, BKC, Mumbai 400051, India
siddharth.ramalingam@idfc.com, Ph: +91-22-42222000.
d Department of Economics, University of South Dakota, 414 E Clark St, Beacom Hall 326,
Vermillion, SD 57069, USA, brock.stoddard@usd.edu, Ph: +1-605-677-6643.
September 9, 2015
Abstract
Teams often suffer from a free rider problem with respect to individual contributions. That
putting teams into competition with each other can mitigate this problem is an important
recent insight. However, we know little about how inequality in endowment between teams
might influence this beneficial effect from competition. We address this question with an
experiment where teams contribute to a public good that then determines their chances of
winning a Tullock contest with another team. The boost to efforts from competition
disappears when inequality is high. This is mainly because the ‘rich’ ‘disengage’: they make
no more contribution to a public good than they would when there is no competition. There
is evidence that the ‘poor’ respond to moderate inequality ‘doggedly’, by expending more
effort compared to competition with equality, but this ‘doggedness’ disappears too when
inequality is high.
JEL Codes: C91, C92, D63, H41
Keywords: public goods, experiment, team competition, inequality, doggedness,
disengagement
* Corresponding author: abhi.ramalingam@gmail.com
Page 3 of 30
Accepted Manuscript
3
1. Introduction
Few doubt some degree of inequality is unavoidable. For instance, if everyone received the
same pay-off, independent of their efforts, there would be no incentive to make costly efforts.
But can there be too much inequality? Does inequality ever discourage effort? If so, what
level of inequality is optimal for the encouragement of effort? We address a specific instance
of this question with an experiment. We focus on outcomes that are determined through a
competition between teams, where individual contributions to team effort affect the
likelihood of a team’s success.
One example of this type of interaction is a competition between political parties or
lobbying groups that depend on the voluntary contributions of their members to fund a
campaign budget. Team sporting contests are another. Companies are similarly formed by
groups of people whose combined efforts influence the likelihood of their success; and,
within organizations, the tournament system of remuneration often creates such contests
between sales, design, or production teams (Bandiera et al, 2013). These examples suggest
such contests between teams are an important class of economic and social interactions.
Given monitoring difficulties, there is likely to be some individual contribution to a
collective enterprise like a campaign group, a company, a sports or workplace team that is
subject to the free rider problem. This is the aspect of individual behavior that concerns us
because it has been argued competition between teams (or its creation) can help each team to
overcome this free rider problem (see Bornstein et al, 1990, Bowles and Gintis, 2011, Tan
and Bolle, 2007, Ishida, 2006, Guntthorsdottir and Rapoport, 2006, and Marino and Zábojník,
2004). This is an important and relatively recent insight concerning the benefits of
competition. However, teams are rarely equally endowed and little is known about how this
type of inequality between teams might affect this benefit. This is why our specific question
is potentially important for policy. We are concerned with whether inequality of endowment
between teams affects the benefit from competition for the free rider problem over individual
contribution to a collective enterprise.
The question is particularly suited to experimental investigation. In part, this is
because the influence of competition in this respect is often much larger than the standard
rational choice models predict, suggesting competition has some additional, non-standard
motivational power that can be fruitfully examined in the laboratory (see, for example,
Page 4 of 30
Accepted Manuscript
4
Sheremeta, 2010 for evidence on behavior in contests). In addition, there are two common
but conflicting intuitions about how this specific motivational power of competition might be
affected by inequality; and experiments might help settle this dispute. One intuition comes
from when ‘weaker’ sports teams seem to perform better than expected against ‘stronger’
ones (as in ‘Miracle’, the movie version of the US ice hockey victory over the USSR in the
Lake Placid Winter Olympics). The underdog is spurred by adversity, so to speak, into a
special show of ‘doggedness’. Alternatively, members of both teams may feel, given the level
of inequality, the result is already a foregone conclusion and so all make lower contributions.
This is the intuition that inequality can cause ‘disengagement’.
There is some evidence of ‘disengagement’ in individual sporting tournaments when
there is inequality (e.g. Brown, 2011 and Franke, 2012). There is also experimental evidence
that investment in conflict falls in individual contests when there are differences in ability
(see Fonseca 2009, Anderson and Freeborn, 2010, Deck and Sheremeta, 2012, and
Kimbrough et al, 2014). The experimental evidence on free riding in team competitions is
more mixed and focuses primarily on inequality in the number of team members. The rational
choice theoretical expectation is that small teams will contribute more because the prize is
worth more to each member of a small team than a big one. Kugler et al (2010) find
contributions are boosted beyond these expectations and there is a small difference between
the small and the big teams but in the reverse direction to that predicted by rational choice
theory (Abbink, et al, 2010, have a similar result). Against this Zhang (2012), in a voter
participation experiment, finds members of small groups are more likely to participate (i.e.
the equivalent of contribute more) but the reverse is the case once communication within
teams is allowed. Levine and Palfrey (2007), in a voting participation experiment, find
evidence that as group size increases, voting falls. They also find participation increases
when the election is evenly balanced, a ‘competition’ effect. Further, they also find an
underdog effect whereby members of small teams are more likely to vote. A difference in the
number of team members is one aspect in which teams may be unequal. In our experiment, in
contrast, we introduce inequality through differences in the individual endowment across
teams.1
We examine this type of inequality for three reasons. First it maps team inequality on
to the more familiar form of individual endowment inequality. Second, in some settings,
1 Bornstein et al (2005) introduce a different kind of inequality by making one team always the winner in the
event of a tie but this has less obvious counterparts outside of committee procedures.
Page 5 of 30
Accepted Manuscript
5
inequality is not or cannot be expressed through differences in numbers in each team, but it
can be, and often is, manifest through differences in team endowments that are relevant in
determining the outcomes of competitions. This is the case in campaigning groups where
total campaign budget matters; and in sporting contests where the number of players is fixed.
For instance, the cost of a player typically depends positively on that player’s potential skills
and where this relation is linear, the average potential skill of a team player will depend on
the team’s overall budget. Each player then faces a choice over how much of their potential
skill, indexed in these conditions by the amount spent on his or her services, to contribute to
the competition. Third, inequality in the endowment of each team does not affect the rational
choice prediction of team contributions to the public good. This is potentially important for
the interpretation of any differences in behavior we observe. It is well known from the
individual contest literature that there tends to be over investment in conflict relative to the
equilibrium prediction. In this context, if inequality affected both the rational choice
equilibrium prediction and actual behavior, it would become difficult to interpret whether this
arises from the change in the equilibrium or a change in the determinants of the out-of-
equilibrium play. We avoid this difficulty: in our experiment, if inequality affects behavior,
then it is because it influences behavior for non-rational choice reasons.2
The design of the experiment captures the free rider aspect of the individual effort
decision via a public goods contribution problem. We examine the influence of competition
between teams on these public good decisions by making a team’s contribution to its public
good affect its chances of success in a lottery (Tullock, 1980) contest.3 In this we follow
Guntthorsdottir and Rapoport (2006); our innovation is to introduce team inequality through
the members of the ‘poor’ team having a smaller individual endowment than the members of
the ‘rich’ team.
We find, like others (see Guntthorsdottir & Rapoport 2006), the existence of the Tullock
contest does boost contributions to the public good as compared with the level of
contributions when there is no contest; and it does so by more than standard game theory
2 This difficulty arises in some of the ‘number’ inequality team contests and in those individual contests where
inequality is captured by particular kinds of ‘ability’ differences. It is also worth noting that there is another
aspect of inequality: that between members of an individual team. This aspect of inequality within a group
appears to influence contributions to a group public good (e.g. see Buckley and Croson, 2006). Since, we are
concerned with the effect of inequality between team resources, we deliberately avoided introducing this source
of difference in behavior into our experiment by giving each individual in a team the same endowment.
3 Savikhin and Sheremeta (2012) also study both public goods games and contests. However, they study
individual decisions when the two separate games are played simultaneously, while the two games are linked in
our experiment.
Page 6 of 30
Accepted Manuscript
6
predicts. In this sense, competition between teams does mitigate the free rider problem within
a team and so promotes efficiency in our experiment. We also find this boost is sensitive to
the level of inequality between the teams. The boost occurs with equality in later periods of
the game, but not when inequality is ‘high’. The magic of competition in this respect, so to
speak, disappears with ‘high’ inequality. This is largely because the ‘rich’ become
‘disengaged’ when inequality is ‘high’: that is, they contribute no more than they would in a
simple public goods game where there is no competition with another group. There is some
evidence the ‘poor’ respond ‘doggedly’ when there is moderate inequality by contributing
more than they do when competition occurs between equally endowed teams. But this
’doggedness’ also disappears under ‘high’ inequality.
Section 2 presents a model of competition, Section 3 explains the design of the
experiment in detail and Section 4 gives the results. We discuss the results in Section 5 and
conclude the paper in Section 6. The online Electronic Supplementary Material contains
additional analysis (Appendix A) and the experimental instructions for the Moderate
inequality competition treatment (Appendix B).
2. Theoretical Model
We closely follow the model of team competition presented in Guntthorsdottir and Rapoport
(2006). They have a standard public goods decision that individuals in each team make. Each
player i in group k composed of m players receives an endowment of ek > 0 and must decide
how much to invest 0 ≤ xik ≤ ekin a public good. The remainder, (ek- xik), is automatically
invested in a private good. The return from the private good is 1 while the return from the
public good depends on the total contribution to the public good in group k, denoted by Xk =
∑i xik. Each player in group k receives a fraction, g (0 < g < 1 and mg > 1), of Xk, i.e., the
marginal per-capita return (MPCR) from the public good is g. Thus the payoff to player i in
group k is given by
(1)
The Nash equilibrium in this game is for each player to contribute nothing to the
public good (while the social optimum is full contribution by all players). However, this
public goods decision is connected to a team competition. Two groups compete for a prize in
a standard Tullock contest where the probability of success depends on relative contributions
to their public goods; and this changes the equilibrium. In particular, with two groups, k and
Page 7 of 30
Accepted Manuscript
7
l, competing for a prize S that is split equally amongst its m members and where each group
wins the prize with a probability determined by a Tullock Contest Success Function given by
(2), the pay-off to each individual is now given by (3).
(2)
(3)
The first-order condition for player i in group k and player j in group l are respectively
(4)
(5)
The two first-order conditions imply that, in equilibrium the aggregate contribution to
the public good must be the same: Xk = Xl. As long as both groups have sufficient
endowments, in an interior equilibrium we have4
(6)
Without further assumptions, such as symmetry, we cannot solve for individual
contribution levels. In equilibrium, they should sum to the value given by (6) but since this
can be generated by a variety of individual contributions, there are multiple equilibria.
Nevertheless, it is clear that the competition for a prize can raise the equilibrium
contributions of groups above zero, the equilibrium in the standard public goods game (while
the social optimum remains full contribution by all individuals).
Furthermore, as long as both groups have sufficient funds (as is the case in our
experiment), the equilibrium group allocation is independent of the endowments of both
groups. Thus, the equilibrium analysis of the interaction suggests inequality will have no
effect on the public goods contributions. This enables a clean test of whether inequality has
some further effect on behavior not captured by this model.
3. Experimental Design
4 Under certain conditions, there can also exist an equilibrium where both groups contribute zero. See
Chowdhury and Sheremeta (2011).
Page 8 of 30
Accepted Manuscript
8
We examine whether contributions to the public good change as the equality/inequality of
team endowments change across competition treatments.5 This comparison is important but it
does not necessarily reveal whether ‘doggedness’ or ‘disengagement’ as a distinct reaction to
inequality has been activated. This is because, when team inequality changes, individual
endowments change and there may be an effect from the altered endowment on behavior in a
public goods game that explains why behavior changes with inequality. As a result, we have
two kinds of treatments. There are, as just sketched, the team competition (COMP)
treatments that are distinguished by the degree of equality/inequality in team endowments. In
addition, there are simple public goods games (VCMs) treatments that are distinguished by
differences in individual endowments (where these differences reflect those found in the team
competition treatments). It is through the comparison of the team-competition contributions
with the contributions under the baseline VCMs with same endowments that we examine the
effect of competition after controlling for endowments. This is the pure boost from
competition in each case. In particular, we perform this comparison for equal endowments
and for High (see below) inequality of endowments to see whether the boost from
competition is also found when inequality is High. If the boost is not found when inequality
is high, we say subjects become ‘disengaged’ from the competition because their behavior in
the competition is no different from a simple VCM with the same endowment.
3.1 Linear VCM Treatments
The simple public goods decision uses the Voluntary Contributions Mechanism (VCM) in
groups of three subjects. Each subject in a group received the same endowment of tokens that
could be allocated to either a private account or a group account. Subjects earned one token
for each token allocated to the private account. For each token allocated to the group account,
each member of the three-person group earned 0.5 tokens, i.e., MPCR = g = 0.5. Once all
subjects made their contribution decisions, each subject was informed of his/her group’s total
contribution to the public good in that period and his/her individual earning from his/her
private account and the group account.
This decision setting was repeated for 20 periods. Earnings from a period could not be
carried forward to future periods. In each period, each subject received a fresh endowment.
5We also depart from the Guntthorsdottir and Rapoport (2006) set up by repeating the interaction 20 times
(rather than 80) and by having partner rather than stranger matching. The former brings our experiment closer to
the common practice in repeated public goods games and the latter is to enhance the statistical power.
Page 9 of 30
Accepted Manuscript
9
Further, all members of a group received the same endowment each period. At the end of a
session, tokens were converted to cash at the rate of 150 tokens to £1. To control for possible
wealth effects from different levels of subject endowments, we ran 3 treatments of the VCM
environment, each with different individual endowments.6 We ran three sessions where each
subject received a per-period endowment of 50 tokens and two sessions each where each
subject received a per-period endowment of 20 tokens or 80 tokens.7
3.2 Competition with Equally-Endowed Teams (COMP(50-50))
In the competition treatment with equality, each member in a team of three members makes
the same linear public goods decision as in the VCM treatments. Each member is individually
endowed with 50 tokens. In addition, paired teams compete in a Tullock contest where
contribution decisions in the first stage determine the probability of winning a prize of S =
120 tokens. The competitors in each Tullock contest were chosen randomly at the beginning
and then kept fixed throughout a session.
In this second stage, each group was informed of their total allocation to the group
account, the allocation to the group account of the competing group, the winning probabilities
for each group and the winning group (one's own group or the other group) in the period.
Winning probabilities were given by a Tullock contest success function given by (2).
3.3 Competition with Unequally-Endowed Teams
The three competition with inequality treatments varied the per person endowment of each
team. In our High inequality treatment, members of one group in a competing pair received a
per-period endowment of 20 tokens each while members of the other group received a per-
period endowment of 80 tokens each. In our Moderate inequality treatment, members of one
group in a pair received a per-period endowment of 40 tokens each while members of the
other group received a per-period endowment of 60 tokens each. In our Low inequality
treatment, members of one group in a pair received a per-period endowment of 45 tokens
6 While previous work has looked at the issue of differing endowment sizes, they do not provide such a direct
test of a “wealth effect”. For instance, Isaac and Walker (1988) and Isaac et al (1994) also examine VCM
treatments with different individual endowments. However, they also simultaneously change group size and the
MPCR from the group account across treatments.
7We initially planned to run further VCMs with intermediate endowment levels of 40, 45, 55 and 60 (matching
the full set of endowments in the COMP treatments) to examine the wealth effects in VCMs. However, for
reasons that are explained later, this did not prove necessary.
Page 10 of 30
Accepted Manuscript
10
each while members of the other group received a per-period endowment of 55 tokens each.
In all cases, all participants knew the endowments.
We say more about the choice of endowment and prize values below but we note now
the total combined resources of both competing groups were the same in all competition
treatments, i.e., 300 tokens.8 Thus wealth is kept the same across competition treatments,
only its distribution changes. For the same reason, to keep earnings comparable under the
VCM control where there is no prize, final token earnings were converted to cash at the rate
of 200 tokens to £1 in the four competition treatments.
We ran five sessions each of the four competition treatments, for a total of 22 groups
and 11 competing pairs in each. Due to the potential effects of competition on the decisions
of both groups in a pair, each comparison pair forms an independent unit of observation. We
thus have data on 11 independent pairs in each competition treatment.
In the no-competition VCM treatments, zero contribution is the unique subgame
perfect equilibrium in each period. Figure 1 shows the reaction functions of the two groups in
the competition treatments.
Figure 1. Reaction Functions of the two competing groups
0 20 40 60 80
Contribution of Group 1
0 20 40 60 80
Contribution of Group 2
8 Further, given our parameters, the equilibrium remains unchanged in all competition treatments. See below.
Page 11 of 30
Accepted Manuscript
11
Two things to note from this are that a) the equilibrium group contribution is 20
tokens in all competition treatments and b) there is no incentive for one group to ‘overbid’
because the other group has contributed more than 20: i.e., neither the ‘poor’ nor the ‘rich’
should, under any set of beliefs about the other, want contributions to exceed 20. This means
rational contribution in this model is always feasible for the ‘poor’ team, as its smallest total
endowment when there is High inequality is 60. Finally, we chose S = 120 so that we would
have at least one case each where the individual prize (40 tokens per group member) would
be higher than, equal to and lower than the individual endowment.
3.4 Procedures
There were 27 experimental sessions conducted at the University of East Anglia (UEA). For
each session, 12 to 18 subjects were recruited from the student body at UEA. Subjects were
anonymously and randomly assigned to three-person groups (m = 3). The composition of
these groups remained the same throughout each session, i.e., all sessions used partner
matching.
At the beginning of each session, instructions were read out by an experimenter.
Subjects also had a copy they could refer to at any time during the experiment. Subjects then
took a quiz to ensure understanding. They could not proceed until all questions were
answered correctly. Once the experiment began, subjects made decisions privately at their
computer terminals. The experiment was programmed in z-Tree (Fischbacher 2007). A total
of 366 subjects participated in our experiment. Table 1 summarizes our treatments and lists
the number of observations in each treatment. A session lasted 45 minutes on average and
each subject earned between £10 and £11 on average including a £2 show-up fee.
Table 1. Summary of Treatments
Treatment Endowments # groups # pairs
No Competition
VCM(20) 20 11 -
VCM(50) 50 12 -
VCM(80) 80 11 -
Competition Treatments
No Inequality 50-50 22 11
Low Inequality 45-55 22 11
Medium Inequality 40-60 22 11
High Inequality 20-80 22 11
Total - 122 44
Page 12 of 30
Accepted Manuscript
12
4. Results
We begin by considering the ‘pure’ effect of competition. This effect is isolated through a
comparison of the COMP treatment with the VCM with same endowments. In particular,
does the boost from competition on contributions found when teams are equal survive/change
when endowments are highly unequal? We then shift to the question of the effect of
inequality on behavior, given competition. We address this by comparing the COMP
treatments.
4.1 Effects of competition between equals
Figure 2 presents the average group contributions over time by independent groups (12
groups) in VCM(50) and by competing pairs of groups (11 pairs) in COMP(50-50). We
observe the usual pattern of contributions in the VCM (see, for instance, Fehr and Gächter
2000). They start high (here, about two-thirds of the endowment) and collapse by period 20
to about one-fifth of the group’s endowment (about 30 tokens). This collapse is particularly
evident in the last 10 periods of the game.
Figure 2. Average Group Contributions in VCM(50) and Pairs in Competition
Treatments over time
Page 13 of 30
Accepted Manuscript
13
Contributions in COMP(50-50) start at similar levels to VCM(50) and evolve in the
same way until about period 10 and become notably different about period 15. Thereafter, it
seems average contributions show less deterioration with equality competition than in the
VCM; and groups contribute more than the equilibrium prediction of 20 tokens.
At this point, we are interested in establishing the boost from competition and not an
investigation of the effects of inequality. However, for purposes of visual comparison, Figure
2 also presents the time trends of the average group contributions of the other COMP
treatments. We note that patterns of contribution are similar in all COMP treatments. In
particular, competition stems the decline in contributions observed in VCM(50). We leave a
full analysis of the inequality treatments until later.
Table 2. Average Group Contributions by Competing Pairs
Obs. Periods 1-5 Periods 6-15 Periods 16-20 All Periods
VCM(50) 12 97.73 77.03 47.33 74.78
(44.08) (50.89) (48.13) (45.99)
COMP(50 -50) 11 94.28 88.00 76.83** 86.77
(17.84) (20.08) (17.08) (17.72)
Notes: Figures in parentheses are standard deviations. The unit of observation in VCM(50) is an independent
group instead of a pair. A ** superscript indicates that the values are pairwise significantly different from each
other at the 5% level (Wilcoxon z = -2.031, p = 0.0423).
Table 2 presents summary statistics on average group contributions by groups in the
VCM and competing pairs in the equality competition treatment. We disaggregate into the
initial five periods, the middle 10 periods, the last 5 periods as well as presenting data for all
periods combined. While the contribution level is higher in COMP(50-50) in almost all
periods, the difference is only statistically significant in the last 5 periods (p = 0.0423).
Page 14 of 30
Accepted Manuscript
14
Table 3. Proportion of Zero Contributors
Periods
1-5 Periods
6-15 Periods
16-20 All
Periods
VCM(50) 0.20 0.31 0.53 0.34
COMP(50-50) 0.14 0.16 0.20** 0.16
The unit of observation is the proportion of zero contributors in a COMP(50-50) pair or VCM(50) group. A **
superscript indicates that the value is pairwise significantly different from the corresponding VCM value in the
same sub-period at the 5% level (n=23 in each Wilcoxon test).
Table 3 reports the proportion of subjects contributing zero in VCM(50) and
COMP(50-50). Although zero contribution is more difficult to interpret in this context than
the standard linear public goods game, it is nevertheless illustrative of changes in behavior
that occur under competition. There is a lower level of zero contributions in all sub-periods
and across all periods with COMP(50-50). The difference becomes particularly pronounced
in the last 5 periods when zero contributing increases greatly in VCM(50) and only slightly in
COMP(50-50). This suggests there is a general influence on behavior from competition. It
does not yield a statistically significant difference in average contributions in the first 15
periods, but as zero contributions jump and contributions generally drop during the last 5
periods of the VCM, the difference with respect to contributions when there is competition
becomes significant. We summarize this pure boost from competition in Result 1.
Result 1: Competition under equality reduces the incidence of zero contributions
compared with the VCM, this is most marked in the last 5 periods with the result that
average contributions to public goods are also higher under competition in the last 5
periods.
4.2 Effects of competition between unequals: the case of High inequality
Figure 3A presents average contributions of the ‘poor’ and the ‘rich’ in COMP(20-80) and
the relevant comparison without competition, respectively, of average contributions in
VCM(20) and VCM(80). Figure 3B compares the average aggregate contribution to the
public good in COMP(20-80) with the average aggregate contributions in VCM(20)
combined with VCM(80).
Page 15 of 30
Accepted Manuscript
15
Figure 3A: Average Contributions of `Poor’ and `Rich’ in Comp(20-80) and the
Relevant Comparisons without Competition, VCM(20) and VCM(80).
Figure 3B: Average Aggregate Contribution in COMP(20-80) with Combined
VCM(20)-VCM(80)
Figure 3A shows the contributions of the ‘poor’ in COMP(20-80) stay steady over
time, unlike in VCM(20) where contributions decline steadily over time. However, the
patterns of contributions of the ‘rich’ in COMP(20-80) are indistinguishable from
contribution patterns in VCM(80). Statistical tests confirm these impressions. There is weak
statistical evidence of a boost in contributions from competition among the ‘poor’ in the last
5 periods relative to VCM(20) [28.73 vs. 14.93, n = 11, p=0.094]. There is no evidence the
Page 16 of 30
Accepted Manuscript
16
behavior of the ‘rich’ is affected by competition [last 5 periods 107.47 vs. 85.65, n = 11,
p=0.694].9
The latter is important because the ‘rich’ make larger contributions than the ‘poor’
and so exercise a disproportionate influence on total contributions. Figure 3B suggests that
aggregate contributions in COMP(20-80) are not very different from contributions in a
typical VCM(20)-VCM(80) combination. Tests show that there is no significant difference in
total contributions in COMP(20-80) as compared with a VCM(20) combined with a
VCM(80).10 In other words, the boost from competition found under equality in COMP(50-
50) compared with VCM(50) has disappeared.
Table 4 contrasts the proportion of zero contributors among the ‘poor’ in COMP(20-
80) with VCM(20); and the proportion among the ‘rich’ in COMP(20-80) with that in
VCM(80). The only significant difference is, once again, among the ‘poor’ in the last 5
periods. This reinforces what has been said before: the behavior of the ‘rich’ under
competition with High inequality is indistinguishable from that of the behavior of people with
the same endowment in a simple VCM.
Table 4. Proportion of Zero Contributions: Comparing Poor and Rich Comp 20-80 with
VCM 20 and VCM 80
Treatment Periods 1-5 Periods 6-15 Periods 16-20 All Periods
VCM 20 0.19 0.33 0.52 0.34
‘Poor’ Comp 20-80 0.24 0.30 0.30** 0.28
VCM 80 0.13 0.21 0.29 0.21
‘Rich’ Comp 20-80 0.07 0.13 0.21 0.13
A ** superscript indicates that the value is pairwise significantly different from the corresponding VCM value
in the same sub-period at the 5% level (n=22 in each Wilcoxon test).
Result 2: The boost from competition may still be present in the behavior of the
‘poor’ under High inequality. It is evident in a lower proportion of zero contributors
and weakly so in terms of total contribution in the last 5 periods. However, there is no
effect on the behavior of the ‘rich’ from competition when inequality is High. As a
9 There are no significant differences (at the 10% level) in any of the other sub-periods for the ‘rich’ and the
‘poor’.
10 For the combined VCM, groups from VCM(20) and VCM(80) were matched using three different random
matching protocols. Standard deviations and statistical tests were qualitatively similar across the three matching
protocols.
Page 17 of 30
Accepted Manuscript
17
consequence of the latter, the boost to aggregate contributions from competition
disappears when inequality is High.
This result suggests the ‘rich’ become ‘disengaged’ from the competition in COMP(20-80):
their contributions are no different from those they would make in a VCM(80). The evidence
on the behavior of the ‘poor’ is more difficult to interpret. In part, this is because the positive
effect of competition on contributions is statistically weak. In addition, even if the effect of
competition is granted, it does not necessarily represent ‘doggedness’ because there is no
evidence from this comparison that the effect of being the ‘poor’ team in a competition is
different to the effect of competition when the teams are equal. To explore this issue, we shift
the comparison to that between the different competition treatments. To facilitate this, we
begin by examining the effects of varying endowment levels revealed in the VCMs.
4.3 VCM contributions with different endowment levels
The left and right panels of Figure 4 plot, respectively, the absolute and % (of endowment)
contributions in VCM(20), VCM(50) and VCM(80). They show effects from differences in
endowment. However, there are no apparent effects on behavior from endowment differences
when behavior is represented by contributions as a % of endowment. This is supported by
statistical tests for significance in differences reported in Tables A1 and A2 in Appendix A.
Figure 4. Average Group Contributions (Absolute & Percentage) in VCM Treatments
over time
Result 3: While absolute contributions to the VCM(20), VCM(50) and VCM(80) are
different, the % contributions are not statistically significantly different across these
endowment levels.
Page 18 of 30
Accepted Manuscript
18
From this Result, we draw the more general inference that % contribution does not change
with endowments over this range. As a result, it is reasonable to infer that any difference
observed in the % contribution across COMP treatments reflects something other than a
behavioral response to differences in endowment. This is so for the endowment levels studied
above: 20, 50 and 80. In addition, since the 40, 45, 55 and 60 endowment levels fall within
the (20, 80) interval and are close to 50, it is plausible to assume that the same applies for
these endowment levels as well.
4.4 The effects of inequality when there is competition: the behavior of ‘rich’ and ‘poor’
Following Result 3, we focus on the % contributions of the ‘rich’ and the ‘poor’ and consider
first the average % contributions across COMP treatments. Table 5 gives these %s
disaggregated across sub-periods for the ‘rich’ and ‘poor’ respectively in each inequality
treatment.
Table 5. Average Percentage Contributions in the Competition Treatments
‘Rich’ groups
Treatment Obs. Periods
1-5 Periods
6-15 Periods
16-20 All
Periods
COMP(50-50) 11 0.63
(0.12) 0.59
(0.13) 0.51
(0.11) 0.58
(0.12)
COMP(45-55) 11 0.63
(0.18) 0.57
(0.22) 0.50
(0.29) 0.57
(0.20)
COMP(40-60) 11 0.67
(0.17) 0.60
(0.20) 0.52
(0.23) 0.60
(0.19)
COMP(20-80) 11 0.60
(0.16) 0.52
(0.18) 0.45
(0.20) 0.52
(0.16)
Page 19 of 30
Accepted Manuscript
19
‘Poor’ Groups
Treatment Obs. Periods
1-5 Periods
6-15 Periods
16-20 All
Periods
COMP(50-50) 11 0.63
(0.12) 0.59
(0.13) 0.51
(0.11) 0.58
(0.12)
COMP(45-55) 11 0.67
(0.22) 0.60
(0.26) 0.54
(0.25) 0.60
(0.23)
COMP(40-60) 11 0.73
(0.21) 0.76
(0.27) 0.67
(0.28) 0.73
(0.23)
COMP(20-80) 11 0.58
(0.25) 0.52
(0.30) 0.48
(0.29) 0.53
(0.25)
Notes: Figures in parentheses are standard deviations.
Although the ‘rich’ contribute on average less as a % of their endowment in each sub-period
when there is High inequality, this difference is not statistically significant. In fact, using
Mann-Whitney tests, none of the pairwise treatment comparisons is statistically significant in
each sub-period. The ‘poor’ contribute on average the highest % of their endowment in
COMP(40-60) in all sub-periods and the lowest % in COMP(20-80). The ‘poor’s’ %
contribution is weakly statistically significantly higher in COMP(40-60) than COMP(50-50)
in the first 5 periods (p = 0.071) and their contribution is significantly lower in COMP(20-80)
than COMP(40-60) in the middle 10 periods (p = 0.053) and overall (p = 0.053).11 This
suggests Moderate inequality may produce some ‘doggedness’ (that is, a response greater
than when there is equality in competition) but this disappears under High inequality. We
check whether these differences in the behavior of the ‘poor’ yield any statistically significant
difference between their average behavior and that of the ‘rich’ in these (or any sub periods)
of the various COMP treatments. They do not (see Table A3 in Appendix A). Given the
earlier noted fall in average % contributions of the ‘rich’ in COMP(20-80), this is not so
surprising.
There is also support for differences in behavior of the ‘poor’ in the analysis of
individual % contributions. Table 6 reports panel random effects regressions on % individual
contributions.
11 None of the other pairwise treatment comparisons for ‘poor’ groups is statistically significant at the 10% level
in any of the sub-periods.
Page 20 of 30
Accepted Manuscript
20
Table 6. Individual Percentage Contributions of the ‘Poor’ and the ‘Rich’: All Periods
Dependent Variable: Individual percentage contribution in each period
Independent Variables ‘Poor’ ‘Rich’
Lagged Own Percentage
Contribution 0.78***
(0.04) 0.75***
(0.05)
Lagged Deviation from Average
Pct Contribution of Others -0.32***
(0.03) -0.21***
(0.04)
Lagged Percentage Contribution of
Competing Group -2.47
(2.76) -1.05
(2.81)
Lagged Win -0.22
(1.54) -0.80
(1.52)
COMP(45-55) -1.82
(1.70) -1.01
(2.04)
COMP(40-60) 4.30**
(1.86) 0.44
(2.03)
COMP(20-80) -3.16
(2.17) -2.34
(1.85)
Period -0.46***
(0.08) -0.40***
(0.08)
Female -2.42*
(1.36) -0.98
(1.82)
International Student -3.20**
(1.51) -4.11**
(1.89)
Experienced -3.21**
(1.52) -0.60
(1.83)
Age 0.15
(0.16) 0.04
(0.15)
Constant 20.73***
(4.91) 22.43***
(5.93)
Observations 3135 3135
*p < 0.10, ** p < 0.05, *** p < 0.01. Random effects panel regression model with std. errors clustered on
independent (44) pairs in parentheses. Experienced subjects are those who have participated in more than 3
economics experiments prior to this one.
The individual regressions in this table control for the standard variables in the public
goods (for instance, Sefton et al 2007) and contest literature (for instance, Sheremeta, 2011
and Dechenaux et al, 2014) found to influence contributions: lagged contributions, deviations
Page 21 of 30
Accepted Manuscript
21
from average contributions and lagged wins. They also control for the time trends evident in
the contributions. The omitted treatment is COMP(50-50). The coefficient on the COMP(40-
60) dummy is positive and significant in the equation for the ‘poor’: that is, there is a boost
from Moderate inequality as compared with Equality. The coefficient on COMP(20-80) is
negative for the ‘poor’ and while not significant (so cannot be distinguished from Equality), it
is statistically significantly different from the coefficient on COMP(40-60) [post-regression
Wald test, p=0.004]. 12 So, the ‘doggedness’ among the ‘poor’ induced by Moderate
inequality disappears under High inequality. We also note there are no treatment effects for
the ‘rich’. The coefficient on the COMP(20-80) dummy is negative for the ‘rich’, but not
statistically significant.13
We next investigate if the boost in percentage contributions of the ‘poor’ under
moderate inequality is driven by a reduction in the proportion of zero contributors. Table 7
presents the proportion of zero contributors among the ‘poor’ in the inequality treatments in
all sub-periods. For purposes of comparison, it also presents the corresponding proportion in
Comp(50-50).
Table 7. Proportion of Zero Contributors among the ‘poor’
Periods
1-5
Periods
6-15
Periods
16-20
All
Periods
COMP(50-50) 0.14 0.16 0.20 0.16
COMP(45-55) 0.10 0.13 0.19 0.11
COMP(40-60) 0.06 0.09 0.19 0.14
COMP(20-80) 0.24 0.30 0.30 0.28
As before, we find the increase in contributions by the ‘poor’ in COMP(40-60)
relative to groups in COMP(50-50) is driven by a reduction in the number of zero
contributors. The proportion of zero contributors is significantly lower in COMP(40-60) than
12 The coefficient of Comp(40-60) is also significantly higher than that of Comp(45-55) [post-regression Wald
test, p = 0.005].
13 To further investigate treatment effects on individual contribution decisions, we performed two robustness
checks. First, we examined the models in Table 6 with the 1-period percentage change in contributions as the
dependent variable. The results from this robustness check are qualitatively the same as those reported in Table
6. Second, we examined the cumulative distribution of average absolute individual contribution by treatment in
three comparisons: VCM(50) vs. COMP(50-50), VCM(20) vs. ‘Poor’ COMP(20-80), and VCM(80) vs. ‘Rich’
COMP(20-80). We also examined the cumulative distribution of average percentage individual contribution by
treatment of all COMP treatments. Visually and statistically there are no significant differences in cumulative
distributions across these comparisons.
Page 22 of 30
Accepted Manuscript
22
in COMP(50-50) in periods 1-5 (p = 0.031) and in periods 6-15 (p = 0.074). Further, the
proportion is lower in COMP(40-60) than in COMP(20-80) in periods 1-5 (p = 0.002), in
periods 6-15 (p = 0.022) and over all periods (p = 0.023).14
Result 4: Moderate inequality when there is competition significantly reduces the
proportion of zero contributions by the ‘poor’ and boosts their overall % contribution
relative to competition under Equality. This reduction and boost relative to Equality
disappears under High inequality.
4.5 Optimal inequality for absolute contributions
Result 4 suggests, from a social welfare perspective, COMP(40-60) might be the best
arrangement because, with the ‘doggedness’ of the ‘poor’ in this treatment, total
contributions might be higher under the Moderate inequality of COMP(40-60) than any other.
Table 8 gives the absolute level of contributions by treatment, as this is what matters for
welfare and not % contributions.
Table 8. Average Absolute Group Contributions by Competing Pairs
Treatment Obs. Periods
1-5 Periods
6-15 Periods
16-20 All
Periods
COMP(50-50) 11 94.28
(17.84) 88.00
(20.08) 76.83
(17.08) 86.77
(17.72)
COMP(45-55) 11 97.02
(21.56) 87.52
(30.33) 77.09
(34.76) 87.29
(27.43)
COMP(40-60) 11 104.60
(15.53) 99.36
(23.42) 87.13
(30.37) 97.61
(21.44)
COMP(20-80) 11 89.57
(15.46) 77.54
(19.88) 68.1
(25.00) 78.19
(17.74)
Notes: Figures in parentheses are standard deviations. The unit of observation in is an independent group.
The absolute level of contributions does rise as inequality increases and is highest in
COMP(40-60), thereafter descending to a level in COMP(20-80) that is indistinguishable
from what happens in standalone VCMs with endowments of 20 and 80. However, the
difference between COMP(50-50) and COMP(40-60) is not statistically significant.15
14 We conducted the corresponding across-treatment comparisons for ‘rich’ groups as well. We do not find any
treatment differences, thus supporting the regression results for the ‘rich’. For the sake of brevity, we do not
report these tests.
15 Aggregate contributions in COMP(40-60) are higher than in COMP(20-80) in periods 1-5 (p = 0.053), periods
6-15 (p = 0.061) and over all 20 periods (p = 0.045). We also compared the proportions of zero contributors in
competing pairs across treatments. Unsurprisingly, the proportion is lower in COMP(40-40) than in COMP(20-
80) in the same sub-periods. Differences across the other treatments are not significant.
Page 23 of 30
Accepted Manuscript
23
Result 5: Contributions in the aggregate are higher with Moderate inequality than
with Equality but the difference is not statistically significant.
5. Discussion
Results 1 and 2, taken at face value, are important. It is known that competition
between teams can help teams overcome the free-rider problem with respect to individual
contributions within each team (e.g. Guntthorsdottir & Rapoport, 2006, and our Result 1).
But competition between teams is rarely characterized by equal team resources and it is an
open question whether the boost from competition generalizes to unequal endowment
competitions. Result 2 suggests that it does not. The boost to public goods contributions from
competition disappears when there is a High (20-80) level of inequality of endowment
between teams.
There are two conflicting intuitions in the literature over how people might respond to
inequality in a competition. People, particularly the ‘underdogs’, could become more
‘doggedly’ determined in the competition; alternatively people might become ‘disengaged’
from the competition because the result can appear to be a foregone conclusion. There is
evidence for both in the experimental literature when inequality occurs along the dimension
of numbers of team members in team competitions and for a fall in contributions when there
are differences in ability in individual contests (e.g. Kugler et al 2010, Levine & Palfrey 2007
and Deck and Sheremeta, 2012). We contribute to this debate. Our experiment considers a
different dimension of inequality: team resources. This is both common and has the
advantage that it yields an equilibrium prediction of no change in behavior in response to a
change in inequality. In particular, Result 2 suggests the disappearance of the competition
boost when inequality of endowment is High occurs because the ‘rich’ become ‘disengaged’
in a precise sense. They behave in the same way as they would when playing a simple VCM
with the same endowment.
Result 4, trading on the insight of Result 3 with respect to the invariance of %
contribution to changes in endowment, is also important. It suggests the ‘poor’ respond
‘doggedly’ to Moderate (40-60) inequality but this ‘doggedness’ disappears under High
inequality. In other words, the relationship between inequality and team effort is non-
monotonic for the ‘poor’ in our experiment. This may explain why earlier experiments have
Page 24 of 30
Accepted Manuscript
24
found evidence for both types of effects from inequality: they could be sampling different
portions of an underlying non-monotonic relationship between inequality and team effort.16
Result 5 suggests Moderate inequality (40-60) may be optimal for aggregate
contributions to the public good in our experiment. There is weak evidence that total
contributions are higher under competition with Moderate inequality than with equality.
More clearly, if inequality grows to the high level (20-80), then the magical effect of
competition disappears altogether. Of course, although it is natural for a firm concerned with
profit or the designer of sporting contests to be concerned with aggregate contribution to the
team public goods, there are other welfare standards. One might, for instance, assess the
outcomes using a Rawlsian criterion by focusing on the absolute contributions of the ‘poor’.
From this perspective, it is difficult to distinguish between COMP(50-50) and COMP(40-60),
but both are clearly better for the ‘poor’ than COMP(20-80) (see Table A4 in Appendix A).
Thus, we conclude that as far as welfare is concerned, a state of competition and moderate
inequality has much to commend. In comparison, competition and High inequality, as
compared with competition with Equality or Moderate inequality, has no welfare advantage
either in terms of efficiency or for a Rawlsian.
These are the insights when the Results are taken at face value. We turn now to
whether the results should be taken in this way. We begin with an important qualification
regarding our use of the terms ‘doggedness’ and ‘disengagement’ so as to focus on what is
important about these results and then turn to how our results cohere with those in the
literature. The results refer to changes in behavior in relation to team competition and its
interaction with inequality of endowments. It is these changes that are important, not the
interpretative glosses that come from attaching the terms ‘doggedness’ and ‘disengagement’
16 It may be useful to consider the degree of inequality, in this endowment sense, that can characterise real team
competitions. This is relatively easy to do for sporting contests. The expenditure of Manchester United and
Manchester City both topped £330m in the 2011/12 UK Premiership soccer season and there were 7 of the
remaining 18 teams in the premiership whose expenditure in that season was less than ¼ of these values (i.e.
they exceeded the 20-80 ratio of our High inequality treatment, see
http://www.theguardian.com/football/2013/apr/18/premier-league-finances-club-by-club). The salary caps in the
NFL yield much less inequality. In the 2012/13 season, the Detroit Lions had the top total salary bill of
$128.7m, while the lowest salary bill was at the Oakland Raiders with $83.3m (see
http://www.theguardian.com/sport/interactive/2013/jan/30/nfl-salaries-team-position#baltimore-ravens,san-
francisco-49ers).
Page 25 of 30
Accepted Manuscript
25
to them. These terms reflect one set of intuitions about how such changes might be explained,
we find them plausible but they are not the only ones.17
One respect in which our results cohere with others is with respect to the coefficients
in the individual contribution regressions on variables like lagged contribution and deviation
from average contribution; they take on similar values as those found in other public goods
experiments (see, for instance, Sefton et al 2007).18 Likewise, our key baseline result (Result
1) that competition with equality of endowment boosts contributions to team public goods,
reproduces a common result in the literature on competition between teams. Furthermore,
although this is less remarked in the literature, we also appear to reproduce a feature of those
results: that the boost occurs in the last part of the repeated interaction. For example,
Nalbantian and Schotter (1997) report, in a related experiment, on the differences between
remuneration schemes for the last 5 periods, but not on differences in earlier periods.
Likewise, Erev et al (1993) have two observations on what is the effective contribution to the
public good in their field experiment and the difference that emerges under competition is
only present in the last of these observations. Similarly, although not part of their formal test,
it would seem visually that the main difference under competition occurs in the later periods
in Guntthorsdottir and Rapoport (2006).
This last commonality between our result and that of others is important in another
respect: the interpretation of why the competition boost occurs. Another interpretation of why
competition boosts contributions might be that it comes from the change in the Nash
equilibrium. The Nash equilibrium on this account could be acting like a reference point that
other-regarding motives build on. Since the VCM Nash equilibrium is for 0 contribution by
each group and in the competition treatments the Nash equilibrium shifts to 20, one might
expect to get what other-regarding motives supply in the VCM plus 20 when there is
competition. The average boost across all 20 periods in our COMP(50-50) treatment
17 For example, it has been argued in the individual contest literature that endowments have an effect on
behavior through the element of ‘free endowment’ and through the influence of mistakes (see Sheremeta, 2011,
Price and Sheremeta 2011, 2014). It is possible this or some elaboration of these ideas could explain our results.
However, it is perhaps worth noting these ‘endowment’ effects appear not to be associated with the % of
endowment invested in the contest in these works; and our results on doggedness refer to differences in the %
contribution that occurs when inequality changes.
18 The coefficients on the individual controls in these equations are also broadly consistent with what has been
found in public goods studies. The only possible unusual one is the negative coefficient on the ‘experienced’
dummy in the ‘poor’ equation. It is also negative for the ‘rich’, but is not significant. We interpret a general
negative effect as reflecting the way subjects who may have played public goods games are less likely to
cooperate. Alternatively, since contributions typically fall in contests games as well when the contest is
repeated, it may reflect experience of that game too.
Page 26 of 30
Accepted Manuscript
26
compared to our VCM(50) treatment is just over 13, so the figure is somewhat close to the 20
difference in Nash equilibria. One piece of evidence that counts against this interpretation,
however, is one might expect to see this throughout the 20 periods. Instead, the boost only
occurs in the last 5 periods and it is at a higher value in these last periods: over 30 more than
the VCM. The dramatic boost in the last 5 periods is difficult to reconcile with an influence
from social preferences.
Furthermore, there is no evidence in the analysis of outcomes that a specific social
preference for inequality aversion (see Fehr and Schmidt, 1999), which might be triggered by
inequality of endowments in our experiment, has been in play. There is, for example, no
significant difference in the response of the ‘rich’ to High inequality as compared with
Equality or Moderate inequality (as one would expect if inequality aversion was
underpinning behavior). In addition, in the High inequality treatment (as in all the treatments)
the distribution of resources at the end of the experiment almost exactly matches the initial
inequality in initial endowments (see Table A5 in Appendix A). Again, if inequality aversion
had been driving our results, there should have been some reduction in team inequality over
the course of the experiment.
Our specific result on ‘disengagement’ by the ‘rich’ and the disappearance of
‘doggedness’ among the poor when inequality of endowments is High is also consistent with
a field experiment studying inequality of abilities. Erev et al (1993) find differences in team
abilities are negatively associated with the productivity boost from competition for both the
relatively ‘high’ and ‘low’ ability teams in their experiment. There is a difficulty with
attributing these changes in productivity to the influence of inequality in abilities because
these differences were not known to the subjects as part of the experimental design. It is
possible, since it was a field experiment and there does not seem to have been any effort to
prevent communication between subjects, the subjects might nevertheless have come to know
something about these differences. In addition, they have heterogeneity within each team and
this can affect behavior (see Rapoport and Suleiman, 1993, and Buckley and Croson, 2006,
and Tan, 2008; Bartling and Siemens, 2011 do not find an effect on behavior). These factors
make interpretation of their results difficult. Nevertheless, it is reassuring that in so far as
inequality is playing a role in their experiment, it is broadly consistent with what we find
when inequality is High.
Page 27 of 30
Accepted Manuscript
27
6. Conclusion
At least since Adam Smith, economists have recognized the beneficial effects of competition
in markets. The possible positive influence of competition on the free rider problem within
teams is a more recent insight (e.g. see Bornstein et al, 1990, and Bowles and Gintis, 2011).
This is an important development because the free rider problem exists to some degree in
most teams and because many outcomes in economic and social life depend on competition
between teams. However, teams are rarely endowed equally and, while there are conflicting
intuitions about how this might affect individual contributions to the team effort, little is
actually known about the influence of this type of inequality. This is the question we
addressed with an experiment.
We find the boost from competition disappears when there is High inequality. This is
largely because the ‘rich’ become ‘disengaged’: they behave no differently than they would
in a VCM with the same endowment. We also find evidence of a non-monotonic relationship
between inequality and the boost in contributions to the team public good made by the ‘poor’.
A moderate level of inequality raises the contributions of the ‘poor’ as a % of their
endowment relative to equality. However, this ‘doggedness’ disappears when inequality
becomes High.
These results suggest that while competition with Moderate inequality has much to
commend it, there is nothing to be gained through competition when inequality is High.
Acknowledgements
The authors thank Jordi Brandts, Gary Charness, Enrique Fatas, Ragan Petrie, Daniela
Puzzello, Sudipta Sarangi, Bob Sugden, Steve Tucker, Ted Turocy, Jimmy Walker, Arlington
Williams, the Associate Editor, two anonymous referees and seminar participants at the
University of East Anglia, Indiana University and the Southern Economic Association
Annual Meetings 2012 for helpful advice, comments and suggestions. Funding from the
School of Economics, University of East Anglia and the Department of Political Economy,
King’s College London is gratefully acknowledged. Hargreaves Heap’s work was supported
by the Economic and Social Research Council through the Network for Integrated
Behavioural Science (Grant reference ES/K002201/1).
Page 28 of 30
Accepted Manuscript
28
References
Abbink, K., J. Brandts, B. Herrmann, and H. Orzen (2010) “Intergroup Conflict and Intra-
Group Punishment in an Experimental Contest Game”, American Economic Review,
100(1), 420-447.
Anderson, L.A. and B.A, Freeborn (2010). “Varying the intensity of competition in a
multiple prize rent seeking experiment”, Public Choice, 143, 237-254.
Bandiera, O., I. Barankay, and I. Rasul (2013) “Team incentives: Evidence from a firm level
experiment”, Journal of the European Economic Association, 11(5), 1079-1114.
Bartling, B., and F.A. von Siemens (2011) “Wage inequality and team production: An
experimental analysis”, Journal of Economic Psychology, 32, 1-16.
Bornstein, G., I. Erev, and Rosen, O. (1990) “Intergroup competition as a structural solution
to social dilemmas”, Social Behavior, 5, 247-260.
Bornstein, G., T. Kugler, and S. Zamir (2005) “One Team Must Win, the Other Need Only
Not Lose: An Experimental Study of an Asymmetric Production Game”, Journal of
Behavioral Decision Making, 18(2), 111-123.
Bowles, S., and H. Gintis (2011) A Cooperative Species: Human Reciprocity and its
Evolution. Princeton: Princeton University Press.
Brown, J. (2011) “Quitters Never Win: The (Adverse) Incentive Effects of Competing with
Superstars”, Journal of Political Economy, 119(5), 982-1013.
Buckley, E., and R. Croson (2006) “Income and wealth heterogeneity in the voluntary
provision of linear public goods”, Journal of Public Economics, 90(4-5), 935-955.
Chowdhury, S.M. and R.M. Sheremeta (2011) “Multiple Equilibria in Tullock Contests”,
Economics Letters, 112(2), 216-219.
Dechenaux, E., D. Kovenock, and R.M. Sheremeta (2012) “A Survey of Experimental
Research on Contests, All-Pay Auctions and Tournaments”, Working Paper.
Deck, C., and R.M. Sheremeta (2012) “Fight or flight? Defending against sequential attacks
in the game of siege”, Journal of Conflict Resolution, 56, 1069-1088.
Erev, I., G. Bornstein, and R. Galili (1993) “Constructive Intergroup Competition as a
Solution to the Free Rider Problem: A Field Experiment”, Journal of Experimental
Social Psychology, 29(6), 463-478.
Fehr, E., and S. Gächter (2000) “Cooperation and Punishment in Public Goods Experiments”,
American Economic Review, 90(4), 980-994.
Fehr, E., and K. M. Schmidt (1999) “A Theory of Fairness, Competition and Cooperation”,
Quarterly Journal of Economics, 114(3), 817-868.
Fischbacher, U. (2007) “z-Tree: Zurich Toolbox for Ready-made Economic
Experiments”, Experimental Economics, 10(2), 171-178.
Page 29 of 30
Accepted Manuscript
29
Fonseca, M.A. (2009) “An experimental investigation of asymmetric contests”, International
Journal of Industrial Organization, 27, 582-591.
Franke, J. (2012) “The incentive effects of levelling the playing field – an empirical analysis
of amateur golf tournaments”, Applied Economics, 44(9), 1193-1200.
Gunnthorsdottir, A., and A. Rapoport (2006) “Embedding social dilemmas in intergroup
competition reduces free-riding”, Organizational Behavior and Human Decision
Processes, 101(2), 184-199.
Isaac, R. M., and J.M. Walker (1988) “Group size effects in public goods provision: The
voluntary contributions mechanism”, The Quarterly Journal of Economics, 103(1),
179-199.
Isaac, R.M., J.M. Walker, and A.W. Williams (1994) “Group size and the voluntary
provision of public goods”, Journal of Public Economics, 54, 1-36.
Ishida, J. (2006) “Team Incentives under Relative Performance Evaluation”, Journal of
Economics and Management Strategy, 15(1), 187-206.
Kimbrough, E.O., R.M. Sheremeta, and T. Shields (2014) “When parity promotes peace:
Resolving conflict between asymmetric agents”, Journal of Economic Behavior and
Organization, 99, 96-108.
Kugler, T., A. Rapoport, and A. Pazy (2010) “Public good provision in inter-team conflicts:
Effects of asymmetry and profit-sharing rule”, Journal of Behavioral Decision Making,
23(4), 421-438.
Levine, D.K., and T.R. Palfrey (2007) “The Paradox of Voter Participation? A Laboratory
Study”, American Political Science Review, 101(1), 143-158.
Marino, A., and J. Zábojník (2004) “Internal competition for corporate resources and
incentives in teams” The Rand Journal of Economics, 35(4), 710-727.
Nalbantian, H., and A. Schotter (1997) “Productivity under Group Incentives: An
Experimental Study”, American Economic Review, 87(3), 314-341.
Price, C.R., & Sheremeta, R.M. (2011) “Endowment effects in contests”, Economics Letters,
111, 217-219.
Price, C.R., & Sheremeta, R.M. (2014) “Endowment origin, demographic effects and
individual preferences in contests”, Journal of Economics and Management Strategy,
forthcoming.
Rapoport, A., and R. Suleiman (1993) “Incremental contribution in step-level public goods
games with asymmetric players”, Organizational Behavior and Human Decision
Processes, 55, 171-194.
Savikhin, A.C., and R.M. Sheremeta (2012) “Simultaneous Decision-Making In Competitive
And Cooperative Environments”, Economic Inquiry, 51(2), 1311-1323.
Page 30 of 30
Accepted Manuscript
30
Sefton, M., R. Shupp and J.M. Walker (2007) “The Effects of Rewards and Sanctions in
Provision of Public Goods”, Economic Inquiry, 45(4), 671-690.
Sheremeta, R.M. (2010) “Experimental comparison of multi-stage and one-stage contests”,
Games and Economic Behavior, 68(2), 731-747.
Sheremeta, R.M. (2011) “Contest design: An experimental investigation”, Economic Inquiry,
49, 573-590.
Tan, F. (2008) “Punishment in a Linear Public Good Game with Productivity Heterogeneity”,
De Economist, 156(3), 269-293.
Tan, J.H.W. and F. Bolle (2007) “Team competition and the public goods game”, Economics
Letters, 96(1), 133-139.
Tullock, G. (1980) “Efficient Rent Seeking”, in J.M. Buchanan, R.D. Tollison and G. Tullock
(eds.) Toward a theory of the rent-seeking society, College Station, TX: Texas A & M
University Press, p. 97-112.
Zhang, J. (2012) “Communication in Asymmetric Group Competition Over Public Goods”,
University of Zurich Working Paper No. 69.