Attitudes toward Risk: Theoretical Implications of an Experiment in Rural India

In economics, empirical work with models of behaviour under risk whether security based or utility based usually involves comparing the models' predictions with the real world decisions of a sample of individuals or firms (for utility-based examples in agriculture, see Anderson et al., I977, or Halter and Dean, I97I; for a safety-based example, see Roumassett, I973). The advantage of this approach is that the analysis focuses on decisions that people actually must make in the course of their economic activities. The disadvantage is that it is difficult to determine the relative influence of risk and other factors on these decisions. This difficulty has led experimental psychologists and, to a much lesser extent, economists to design specific experiments to test the implications of utility theories in the laboratory. (For reviews of this experimental work, most of which has been confined to utility-based theories, see Luce and Suppes, I965, and Grether, I978.) This paper, following the lead of the experimental psychology literature, discusses a relatively simple but fairly large-scale experiment in ruralIndia that the writer and colleagues at ICRISAT designed to measure attitudes toward risk. We chose an experimental approach when it became clear that we could not obtain reliable estimates of risk aversion by the usual interview techniques of eliciting certainty equivalents. (For details of the difficulties encountered in this study and a description of the entire experiment see Binswanger, I980.) We have validated our experimental measures of risk aversion by showing that a portion of the observed variation among individual farmers' agricultural decisions can be related in a systematic manner to variations in the same farmer's experimentally measured degrees of risk aversion, the more risk averse choosing more conservative options (Binswanger et al. i980).1 This observed relationship between experimental behaviour and actual farm decisions suggests the importance of examining the significance of our findings for a number of models of behaviour under risk, focusing on the consistency of our findings