Temporal Stability, Cross-Validity, and External Validity of Risk Preferences Measures: Experimental Evidence from a UK Representative Sample

Abstract

We conduct an “artefactual” field experiment to incorporate three different risk preferences measures within the Innovation Panel (IP) of the UK Household Longitudinal Survey (UKHLS). We randomly allocate to an experimental module a nationally representative sample of 661 adult respondents to the IP Wave 6 (IP6). These subjects respond to the incentive-compatible tasks by Holt and Laury (2002) (HL), and by Binswanger (1980, 1981) and Eckel and Grossman (2008) (B-EG), and to the SOEP survey questions by Dohmen et al. (2011) for self-reported willingness to take risks in general (SOEP-G), in finance (SOEP-F), and in health (SOEP-H). One year later (IP7) the same measures are repeated for 413 of these respondents. This design allows us to systematically test, for a UK representative sample, the validity of the three measures along three dimensions. First, we look at cross-validity by testing how responses at one point in time correlate across the three tasks, assuming a Constant Relative Risk Aversion (CRRA) utility function. Second, we look at temporal stability by comparing the responses across IP6 and IP7. Third, we look at external validity by considering a range of risky health and financial behaviors in the UKHLS. We have three main findings. First, concerning cross-validity, we find evidence that the different measures generally correlate and map into each other, although their associations are not perfect. Second, concerning temporal stability, there are significant and positive correlations of the B-EG, HL, and SOEP measures across IP6 and IP7. Finally, we find mixed evidence concerning external validity.