Session: Assessing Health Status
Room: Hollister McManus Lounge
Time: Tue 15:00-16:30

Presenter: Amelie Wuppermann (University of Munich. Economics)

Discussant: Jennifer KohnDrew University

Abstract

Motivation
While economic theory stresses the importance of adverse selection in competitive insurance markets (Rothschild and Stiglitz (1976), Wilson (1977), Chiappori et al. (2006)), the empirical evidence on adverse selection is mixed (Cutler et al. (2008)). Multiple dimensions of asymmetric information, e.g. private information on risk preferences and risk type, have been stated as a reason for the mixed evidence (Finkelstein and McGarry (2006)). In insurance markets with a thorough underwriting process, lack of private information and thus missing scope for adverse selection could be a different explanation for the scant evidence. This study analyzes empirically whether applicants for health-related insurances have private information about their risk despite careful medical underwriting.

Data
The analysis is conducted using data from the English Longitudinal Study of Ageing (ELSA). ELSA is one of the few longitudinal data sets that provides health data which are objectively measured by a nurse. In addition to age, gender, smoking status, occupational status, pre-existing medical conditions and family health history, results of a blood sample analysis, blood pressure measurement, and body mass index can be included in the analysis. This captures variables typically made available to insurers through underwriting.

Estimation
A measure of self-rated health (SRH) on a 5-point scale serves as proxy for private information. SRH and all other controls from the first wave are used to predict health events in subsequent ELSA waves. Significant information in SRH when controlling for the information that is available to the insurer is interpreted as evidence for private information.
Like in other longitudinal surveys, item non-response and attrition are present in ELSA. Both mechanisms are health-related and simply ignoring them might bias the results. This paper uses a new strategy to correct for the two mechanisms. Inverse probability weighting is employed that allows for a correlation between the unobservable influences on the two mechanisms by jointly estimating the selection probabilities using a bivariate probit.

Results and Conclusion
SRH contains significant additional information for the future occurrence of functional limitation. Death and the onset of specific diseases are not significantly predicted by SRH when medical information is controlled for. Correcting for attrition and item non-response does not change the main results. Overall, the mixed evidence on adverse selection in markets for life and health insurance can potentially be explained by the differential reduction of private information through different underwriting processes undertaken by insurers.

Key Terms

ELSA, adverse selection, objective health measures, correlated selection mechanisms

Authors:

Amelie Wuppermann (University of Munich (LMU). Economics)