Cooperation and Punishment in an Economically Diverse Community in Highland Tanzania

P cross-cultural variation in economic experiments was unexplained by individuallevel variables such as wealth, income, gender, and education, even though variation across societies was explained by market integration and potential returns to cooperation (Henrich et al. 2001; Henrich et al. 2004). However, the previous data contained relatively poor measures of individual income, wealth, and market integration. Additionally, most of the societies sampled contained relatively small amounts of variation in these measures. Thus, we do not know if the absence of within-group effects is due to causal unimportance or to insufficient variation. In this chapter, I analyze the game behavior of an economically and ethnically diverse community in southern Tanzania that exhibits substantial variation in these measures. Nevertheless, I find very limited evidence that individual economic variables explain variation in game behavior, echoing the weak and inconsistent results found in other societies in this volume. These results cast doubt on the conjecture that the failure of individual-level variables to explain game behavior results simply from inadequate variation within each site. However, error in measurement might be responsible for the lack of explanatory power at the individual level, while the same variables are explanatory across groups. A tremendous amount of data now exist supporting the conclusions that human behavior is poorly predicted by standard economic preferences (Camerer 2003; Fehr and Gächter 2002; Hoffman et al. 1998; Roth 1995). In addition, as more non-Western and nonstudent populations are sampled, it appears that no single revision to Homo economicus will be sufficient (Henrich et al. 2004), or at least that the variables that may eventually explain the tremendous amounts of variation in game behavior will fall outside existing theories. Adequately addressing the role of individual-level variables such as income and education, however, requires good measures of these variables, and this is complicated by the broad crosscultural nature of the data. It may not be reasonable to suggest that education or income has the same role in all groups. Nevertheless, it is reasonable to ask if such variables have powerful effects on game behavior in any significant proportion of the groups sampled. If so, then it will be more reasonable to infer that across-group variation may be a function of the same variables. Previous attempts at correlating cross-cultural game behavior with individual-level variables has met with failure, however (Henrich et al. 2001; Henrich et al. 2004). Two problems arise in these analyses. First, because many of the groups sampled had inadequately measured variables comparable to those of other groups, it was unclear whether education or income or wealth in different samples measured the same econometrically important variable. Second, most of Chapter 15