Is Risk Aversion Really Correlated with Wealth? How estimated probabilities introduce spurious correlation

Economists attribute many common behaviors to risk aversion and frequently focus on how wealth moderates risk preferences. This paper highlights a problem associated with empirical tests of the relationship between wealth and risk aversion that can arise when the probabilities individuals face are unobservable to researchers. The common remedy for unobservable probabilities involves the estimation of probabilities in a profit or production that includes farmer, farm and agro-climatic variables. Unfortunately, these variables are often correlated with wealth such that estimated probabilities are likely to leave statistical fingerprints on subsequently-estimated risk aversion coefficients and may thereby introduce spurious correlations between wealth and risk preferences. In this paper, we use data from an experiment conducted among 290 Indian farmers to detect these spurious correlations. We estimate coefficients of risk aversion with known probabilities and with estimated probabilities and compare subsequent correlations with wealth and other farmer traits. We estimate 'unobservable' probabilities in conjunction with risk preferences following a standard field data approach. We explore the statistical implications of estimating probabilities by comparing correlations between wealth and these two sets of estimated risk preferences. These comparisons show how estimated probabilities can change risk aversion coefficients substantially and introduce spurious correlations between risk aversion and wealth.