Testing the « Lemons » Model: an Experimental Market for Insurance

Akerlof’s (1970) classical paper demonstrated that adverse selection can eliminate markets. These findings by Akerlof gave rise to a proliferation of theoretical models on insurance with adverse selection. However, there exists no empirical test for the model of Akerlof in the context of insurance. In fact, it is not possible to observe a market a la Akerlof in a real context, since that market would have in fact disappeared. The main objective of this work is to test the predictions made by Akerlof, in the context of an experimental insurance market. This experiment is, to our knowledge, the first done in this area. The major interest of this experiment lies in that it allows us to bypass the usual obstacles for testing the model of Akerlof. To fit in perfectly within the framework of analysis drawn by Akerlof, we integrate the assumption that individuals differ in their level of risk aversion (heterogeneity). Also, we allow for the possibility that the insurers adopt strategies of pricing contracts that differ according to the reasoning adopted. We conduct eleven independent experimental sessions in which two insurers in Bertrand competition interact over 60 rounds with height potential insurance contributors. The latter differ in their risk level and can not be differentiated by insurers. The data shows that market functioning depends on insurers’ reasoning (sophisticated versus naïve) and individuals’ risk-aversion levels. Yet, in line with the theoretical prediction, low-risk individuals are driven out of the insurance market.