Theft, Gift-Giving, and Reciprocity: A South African Experiment

Abstract

This paper uses a taking game to examine how South African subjects alter the amount they choose to ‘steal’ in response to a resource transfer from the potential victim. Any positive resource transfer significantly reduces the amount taken. ‘Small’ transfers reduce a victim’s total losses, including the transfer and the subsequent ‘theft’. Larger transfers increase a victim’s total losses. This study failed to find that differences in the frame of a transfer (i.e. gift, as a bribe, or as a payment) influenced a taker’s response to the transfer.

Acknowledgements

The authors would like to thank the Center for International Food and Agricultural Policy in the Department of Applied Economics and the Office of International Programs at the University of Minnesota for financial support at the project development stage. The Office of International Programs provided additional financial support for project implementation. We would also like to thank Survey Services in the UMN College of Liberal Arts for assistance with the design and implementation of the online instrument.

Notes

1. In the 2007/2008 fiscal year, livestock theft in South Africa included 60,000 cattle, 80,000 sheep, and 35,000 goats, valued at ZAR358 million, or approximately USD48 million in stolen property (Brand, 2010).

2. A homeland refers to lands allocated to black ethnic groups during the apartheid era.

3. The word ‘transfer’ is used to describe varied phrasing in the actual study. Approximately 67 per cent of the Bloemfontein participants were presented with the opportunity to ‘give’ up to ZAR15 to the other player. Half of this first group of Bloemfontein participants also learned that their partner would be able to take an additional ZAR15 from their endowment after receiving the gift. The remaining 33 per cent of the Bloemfontein participants were presented with the opportunity to ‘pay’ the other player up to ZAR15. In the ‘pay’ treatment, Bloemfontein participants were also aware that the other player would be able to take an additional ZAR15 after receiving the payment. The purpose of this variation is to determine if framing the transfer either as a gift, bribe, or payment influences the behaviour of either the ‘giver’ or the ‘taker’.

4. See the Online Appendix for a list of black- and white-sounding surnames from Van Der Merwe and Burns (2008).

5. Some players may be able to infer a player’s ethnicity from a surname. As the experiment design did not specifically control the ethnic makeup of dyads, this is a potential source of variation between the two treatments.

6. In accordance with concerns at the design stage, over 40 per cent of Bloemfontein participants offered the maximum transfer of ZAR15. The average transfer was ZAR9.6 with 20, 10, and 70 per cent of the actual transfers classified as small, medium, and large, respectively. Though this is an unusually generous allocation for a dictator game, the behaviour of the Bloemfontein participants is not the subject of this paper and not discussed further.

7. See Pecenka and Kundhlande (2013) for detail on how the results of the baseline treatment were influenced by receiver race and inequality between participants.

8. Mann-Whitney p-value = 0.09.

9. A number of individuals did not indicate the number of economics courses they had taken or the race of their closest friends. These 32 individuals were excluded from the likelihood ratio test. This exclusion permitted the likelihood ratio test to be completed with a sample common to both specifications. Omitting these individuals did not appreciably change regression results.