Understanding inconsistencies in risk attitude elicitation games: Evidence from smallholder farmers in five African countries

Recent empirical studies eliciting farmers' risk attitudes through lab-in-the-field experiments have reported high levels of inconsistency in responses. We investigate inconsistencies in risk attitudes elicitation games using data from incentivized lotteries involving 2,319 smallholder farmers from Eastern Africa (Kenya, Uganda, Tanzania) and Northern Africa (Tunisia, Morocco). Our sample demonstrates high levels of inconsistent behavior, with 48 % of the farmers exhibiting some type of inconsistency. Depending on the country, inconsistencies are explained by poverty, gender, and/or the interaction of gender and level of education. We find no significant impact (negative or positive) of education alone in all but one country model. Furthermore, we find session fixed effects to significantly explain inconsistencies in many cases, suggesting that session-specific circumstances, including inconsistencies across enumerators, play a crucial role in the successful implementation of these experiments. Our findings suggest that using risk attitude parameters without accounting for the presence and the potential causes of inconsistency may lead to unreliable results. This study may guide practitioners in identifying farmer typologies more prone to inconsistent decisions and inform policymakers about factors influencing operators' choices.