Unearned Inequality and Biased Distributional Preferences: Evidence from a Lab Experiment in Ghana

Abstract

I use a dictator game, with a preceding production phase, to explore how different sources of unearned inequality influence distribution decisions. The experiment introduces three inequalities: (1) inequality due to investment success, (2) inequality of initial endowments, and (3) inequality in a real-effort task requirement for receiving the initial endowment. I find that participants in rural Ghana respect each inequality, adjusting their allocation decisions in response to investment success, the distribution of initial endowments, and the effort requirement. Their distributional preferences suggest an interaction between the sources where those with high initial endowments are more likely to respect project success outcomes. Moreover, participants demonstrate a self-serving bias in fairness preferences, where those with high initial endowments and successful projects are more likely to respect inequality. Finally, I find a significant increase in self-allocations when participants must exert effort to receive their endowment.

Keywords:

• inequality
• fairness views
• experimental games
• distributional preferences

JEL CLASSIFICATION CODES:

• D31
• D63
• H31

Acknowledgements

This research was funded by the Feed the Future Innovation Lab for Markets, Risk & Resilience and made possible by the generous support of the American people through the United States Agency for International Development (USAID); Award Number RUCA# A14-0028-004. The contents of this publication are the responsibility of the author and do not necessarily reflect the views of USAID or the United States Government. I would like to thank John P. Dougherty, Jon Einar Flatnes, Eric Acheampong, William Alale, Richard Alebina, Samuel Asitanga, Thomas Debura, and Anthony Abanga for their feedback and assistance with field work. The experiment was pre-registered with the AEA: AEARCTR-0005290. The analysis was approved by the Institutional Review Board at The Ohio State University: 2019B0521. Declarations of interest: none. Supplementary Materials, including data and code for replication and additional analysis, are available via the online version of this journal.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 An egalitarian view would imply that all inequality is unfair (Cappelen et al., 2007; Jakiela, 2015).

2 The Ex Ante Inequality treatment introduces a change in the ex ante expected value of one’s project, which is known to influences distributional decisions (Brock et al., 2013).

3 The real-effort task was easily completed by all participants. The task does not influence income (it does not change the size of the endowment or investment success in any way), therefore, it does not proxy for differences in productivity. Rather, it may influence behavior by altering participants’ perceptions of ownership of the initial endowment via an endowment effect. I consider the effort inequality to be ‘unearned’ because it does not influence productivity and participants do not make a choice to engage in the effort task. However, the effort task introduces one characteristic of ‘earned’ inequality: a greater sense of ownership over income.

4 A fairness view refers to a distribution of resources that an agent deems as fair. I propose two fairness views that relate to ex ante inequality: the equality view holds that resources should be divided equally, regardless of the distribution of initial endowments, while the proportionate view holds that resources should be divided in a way that is proportionate to the distribution of initial endowments.

5 The experiment was conducted alongside two other experimental games that link inequality and risk management in a developing country context. The other two games are presented elsewhere (Gallenstein, 2024).

6 The investment decision was framed in terms of investments in farming to improve comprehension. See Supplementary Materials SM.7 for the experiment script for the framing. Moreover, the Supplementary Materials includes game labels that are used in the experiment. The supplementary materials are available online in file: Unearned_Inequality_SupplementaryMaterials. The materials in the local language are available upon request.

7 To be clear, let the income of participant i be $Y_i=3\omega$ if their project succeeds and Y _i = 0 if it fails. The group total between a participant i and their partner j is then $X=Y_i+Y_j.$

8 Allocation options were restricted to this set of fixed proportions to improve comprehension among participants, who had relatively low levels of educational attainment. In fact, when pretesting a continuous allocation choice, participants often requested help from enumerators to divide the group total income into one of these proportions.

9 If both projects fail, there are no resources to divide and no allocation decision is observed for that dictator game.

10 Ex Ante refers to before the realization of the investment outcome.

11 The participant was informed about the size of their endowment before needing to complete the task, therefore, the participant would not have assumed that the size of their endowment resulted from their performance on the task.

12 To summarize, in the five rounds, the Ex Ante Equality treatment was played for one round, the Ex Ante Inequality-No Effort treatment was played for two rounds, and the Ex Ante Inequality-Effort treatment was played for two rounds.

13 It is important to note that effort is only present in the Ex Ante Inequality treatment so the impact of effort is estimated always in the presence of Ex Ante Inequality. Therefore, in the analysis below, estimates of the impact of effort are estimated as interaction effects with ex ante inequality.

14 Ex ante inequality results from a randomization process implemented by the research team, while the investment outcome is determined by one’s own draw from a bag. Therefore, there may be differences in views of process fairness (Trautmann, 2009).

15 It is a limitation of my work here that I cannot rule out the possibility that participants made portfolio formation decisions across the five dictator decisions.

16 The incentive structure was a random lottery incentive mechanism (RLIM), which is commonly used as an incentive mechanism in experiments that have multiple lotteries (Cubitt, Starmer, & Sugden, 1998; Cubitt et al., 1998). The incentive payouts were made based on a randomly chosen round from one of the three experiments played in each session, as eluded to in footnote 5. For context on the size of the incentives, local agricultural daily wage was around GHS 15. $1 =GHS 5.64 at the time of the experiment.

17 The comprehension questions were as follows: (1) If you start with GHS 6, how much will you get if you draw a white ball?, (2) Consider a case in which you and your partner start with GHS 6 and you both draw white balls. What is the total you will have together?, (3) If you chose insurance in a round, is it possible for you to earn GHS 0 in that round?, and (4) If you started a round with GHS 4, how much did your partner start with? White balls refer to the color of the success ball draw. The reference to insurance in the third question stemmed from another activity played during the session that is discussed elsewhere (Gallenstein, 2024).

18 The registered pre-analysis plan (PAP) does not include the reduced form analysis presented but does include the structural estimation for the equality and proportionate fairness views presented in Table 4 presented below.

19 Due to low literacy, participants made their decisions in front of an enumerator who recorded their responses. I cannot rule out that participants’ overt self-interest was mitigated by this feature of the experimental design.

20 Model 1 uses session fixed effects rather than individual fixed effects because the equality treatment was played only once per person. Therefore, Model 1 is across person identification while the remaining models are within person.

21 In Supplementary Materials Table SM.1 I report the impact of the SF state relative to the FS state to get a pure comparison of successful and unsuccessful participants. I find that being successful increases self-allocations under the equality treatment and for HWIs in the inequality treatment.

22 In Supplementary Materials Table SM.2 I report the impact of ex ante inequality without dis-aggregating by HWIs and LWIs under the various investment outcome states. I find that ex ante inequality increases self-allocations in the SS state but not the SF or FS states.

23 To be specific, the SF state increases self-allocations, relative to the SS state, by roughly 21% when there is no effort and by roughly 14% when the dictator exerts effort.

24 In Supplementary Materials SM.4, Table SM.3, I report estimates of the structural parameters for the ex ante inequality treatments separating out the sample by effort. I find that those who must exert effort to receive their endowment have a stronger bias toward the self-serving fairness view than found in Table 4. That is, when HWIs (LWIs) exert effort, they become more likely to hold the proportionate (equality) fairness view than when they do not earn their endowment.

25 In Supplementary Materials SM.5, I also present a comparison between average contributions in this experiment to those in similar experiments conducted in developed countries. I find that the participants in this experiment give more on average than participants in the other experiments.