Theft in South Africa: An Experiment to Examine the Influence of Racial Identity and Inequality

Abstract

This article uses a dictator game to determine the impact of racial identity and inequality on theft in South Africa. A dictator can take money from a receiver's endowment. The money vulnerable to theft is constant across groups, and there is no threat of punishment. Any differential in the amount taken is due to considerations of the racial identity of the potential victim or to variation in receiver endowments. The results provide weak evidence that theft is impacted by inequality. However, racial identity significantly influences theft decisions. Contrary to expectations, black participants take 15 per cent more from other black participants.

Acknowledgements

This work was supported by the Office of International Programs at the University of Minnesota. The Center for International Food and Agricultural Policy at the University of Minnesota and the University of the Free State also provided financial support at the project development stage.

We thank Survey Services in the UMN College of Liberal Arts for assistance with the design of the survey instrument. We also thank faculty, staff and students at University of Minnesota, the University of the Free State and Carleton College for their interest and feedback on this work. We are responsible for any remaining errors.

Notes

1. See Ahmed (2010a, 2010b), Bahry and Wilson (2004), Bertrand and Mullainathan (2004), Bossuroy and Selway (2011), Fershtman and Gneezy (2001), Fong and Luttmer (2009, 2011), Habyarimana, Humphreys, Posner, and Weinstein (2004), Hanson and Hawley (2011) and Whitt and Wilson (2007) for examples.

2. According to the World Bank, South Africa's 2000 Gini Index was 0.58. A 2005–2006 estimate by Statistics South Africa estimates a Gini coefficient of 0.73. According to the South African Regional Poverty Network, Gini coefficients are rising over time within racial groups.

3. Kahneman, Knetsch, and Thaler (1986) were the first to publish a hypothetical dictator game experiment.

4. For additional work on social capital and trust in South Africa see Carter and Castillo (2003) and Ashraf, Bohnet, and Piankov (2003).

5. See Demombynes and Özler (2005) and the references therein for reviews of economic and sociological theories on crime as well as the international literature on the links between inequality and crime.

6. Bosman and van Winden (2002), List (2007), Bardsley (2008), Jakiela (2009) and Halliday (2011) use variations of a taking game. Falk and Fischbacher (2002) and Kimbrough, Smith, and Wilson (2010) also explore theft through experiments.

7. Two distinct campuses were chosen to ensure a substantial Sotho speaking population for a later experiment. The Qwaqwa campus is located in a region that was designated as the Sotho homeland during the apartheid era.

8. Oxoby and Spraggon (2008) varied whether the dictator or receiver earned the resource to be allocated. In cases where the receiver earned wealth, dictators allocated far more resources to the receiver highlighting the importance of property rights. See Hoffman, McCabe, Shachat, and Smith (1994), Cherry, Frykblom, and Shogren (2002), Jakiela (2009) and Bosman and van Winden (2002) for similar work.

9. Zitek, Jordan, Monin, and Leach (2010) used three experiments to show that feeling wronged leads to a sense of entitlement and thus more selfish behaviour. This finding supports the hypothesis that racial identity and income inequality exacerbate theft due to South Africa's apartheid past.

10. See Van Der Merwe and Burns (2008) for a discussion of using surnames to signal racial identity. Their examples of black and white sounding South African names are included in Appendix A.

11. Qwaqwa participants played several variations of a dictator-like game. Only the first of these variations is discussed here. Only one randomly selected decision would ultimately impact payments.

12. The Mann-Whitney U test is commonly used to test for differences between independent samples. It does not assume normality and is often used as an alternative to a two-sample t-test.

13. The Kruskal-Wallis test extends the Mann-Whitney U test to three or more independent groups. It does not assume normality and is often used as an alternative to one-way ANOVA.

14. See Appendix B for a summary of relevant covariates.

15. Potential reasons the inequality variable is insignificant are presented in the discussion.

16. The data to understand perceptions of ‘worthiness’ and how this influences taking decisions were not collected. A series of questions to determine how individuals perceived their own ‘worthiness’ or that of the other player were included in the pilot phase of the survey. Due to negative feedback in the pilot survey, these questions were removed to avoid aggravating local sensibilities.

17. See Appendix E for the results of an ordered probit.

18. An anonymous African born colleague attended a conference in South Africa after the end of apartheid. After asking questions of white presenters, he was approached during a break and asked to divulge his nationality. He was told that it was clear he was not South African because black South Africans would not ask questions of a white presenter.

19. Although decisions were anonymous and participants were assured that subjects would not be identified, ingrained fear of retribution from white receivers could also form part of this socialisation hypothesis.

20. Full model (4) results are located in Appendix C. The p-value on the interaction dummy is high (0.78). Although approximately one-third of respondents indicated that one of their three closest friends was a member of a different racial group, demographics in the region and the lack of visible social interactions between white students and black students suggest that interracial friendships are not primarily between white students and black students. If true, this likely reduces the size and precision of the estimate and contributes to a high p-value.

21. This evidence is also consistent with the commonly held belief that increased integration mitigates discrimination.

22. See Appendix C for full model (5) results.

23. Although the results are not presented here, Appendix D contains additional models that examine the ethnicity of the receiver, the ethnicity of the receiver and dictator and the role of takers and receivers sharing an ethnic identity. Of relevance to the results in Appendix D, some surnames may be more obviously connected to a particular ethnic group. Unfortunately, we do not possess the data to incorporate this into our analysis. In a pilot version of the study, participants were asked to select the ethnic group to which their partners likely identified and their level of confidence in their selection. Due to negative feedback in the pilot survey, these questions were removed to avoid aggravating local sensibilities.

24. The importance of inequality considerations and its impact on decision-making is debated. For a sampling of this literature and competing theories see: Fehr and Schmidt (1999); Bolton and Ockenfels (2000); Charness and Rabin (2002); Falk and Fischbacher (2006) and Falk, Fehr, and Fischbacher (2008).