The Multiple Price List (MPL) procedure of Holt and Laury (2002) is one of the most widely used methods to measure risk attitudes. It enables to size, in the Lab, the intensity of risk aversion of subjects, both in the gain and loss domains (Chakravarty and Roy (2009)). In particular, owing to the reflection effect of Kahneman and Tversky (1979), such a measure, when implemented in the loss domain, makes it possible to segment the population into two groups: risk lovers[1] and risk averters. As shown in Corcos et al. (2017), this segmentation of risk attitudes contributes to the understanding of the insurance demand behavior. However, several papers, as in Hershey and Schoemaker (1980), documented the fact that individuals were more risk averse when the same decisions were framed as insurance choices rather than decontextualized risky choices.

To go further, our paper aims at comparing two methods of classification of risk attitudes. The first results from the standard procedure of Holt and Laury (2002) implemented in the loss domain, while the second is based on the contextualized hedging choices observed in the lab for insurance and self-insurance. Our basic intuition has been to categorize our subjects, as risk-averters (RA) or risk-lovers (RL), according to their decisions to buy or not a positive hedging for insurance (I) and/or self-insurance (SI). A preliminary statistical analysis shows the robustness of the Holt and Laury's measure.