Risk attitudes: The central tendency bias

Abstract

Unincentivized measurement instruments of risk attitudes suffer from several weaknesses. One is that respondents do not consistently assign themselves to their respective risk preference categories. In particular, they are subject to a central tendency bias and classify themselves as risk-neutral when they are in fact not. We test the robustness of the central tendency bias in lottery-type questions for risk evaluations and offer an explanation of why respondents behave in a way that contradicts plausible utility models. We explore a wide range of alternative influencing factors, including careless responding, stake levels, deviations in expected value, the cognitive abilities of the respondents, self-assessment of risk attitudes, and monetary incentives. We find that careless responding and higher stakes increase the central tendency bias in risk assessment, while cognitive capabilities and extreme risk self-assessments (both positive and negative) decrease the bias. Deviations in expected value and incentives do not affect the bias. Our study further points to the fact that such problems have to be taken care of explicitly when eliciting risk attitudes.