Econometric Methods for Testing for Asymmetric Information: A Comparison of Parametric and Nonparametric Methods with an Application to Hospital Daily Benefits*

Since Chiappori and Salanié, testing for asymmetric information has become tantamount to testing for a positive relation between risk and coverage which is predicted by (equilibrium) models for adverse selection and moral hazard. In this paper we review parametric tests and a recently developed nonparametric test (Su and Spindler) and discuss the underlying assumptions. As an illustrative example, we apply the tests to hospital daily benefits which is part of the accident/disability insurance. We compare the results and show that the pattern of asymmetric information is mixed. The main finding is that we detect asymmetric information for low insured sums, but not for high insured sums which is a surprising result. We can also show that in certain ranges both types of test deliver similar results, while under certain circumstances, approaches using aggregation to binary variables might run into difficulties.