Examining the Size of the Justification Bias in the Effect of Health on Retirement Behavior

This paper seeks to determine the size of the so-called justification bias arising from using self-reported health measures in assessing the effect of health on the early retirement decisions. For this purpose, we use detailed objective diagnosis codes from the Danish administrative register data covering all medical and hospital records from 1997 to 2012, combined with the Danish Longitudinal Survey of Ageing on retirement and health for a sub-sample of the Danish population. This specific combination of objective and subjective health measures will be a valuable contribution to the existing literature seeking to assess the effects of health on retirement, since the existing literature primarily have been using survey-based health measures in their studies. Thus, in terms of understanding the importance of the justification bias, our paper will be a valuable contribution to researchers trying to examine the effects of health on retirement. We adopt a tractable, relatively flexible, and robust hazard model specification of the duration until retirement, using the actual timing of labor supply events and changes in health, financial, demographic, and socioeconomic variables over a period of 15 years. In particular, we use the competing risk framework in order to account for the different exit routes into retirement. In order to allow for interpretation of the estimates from the competing risk model, we elaborate on the result found by Thomas (1996). Our results suggest that there are sizable differences in using objective compared to subjective health measures in assessing the effect of health on retirement, supporting the justifications bias hypothesis. Thus, one should be very careful when using subjective health measures as a proxy for true, individual health conditions.

Reference: Thomas, J. M. (1996) "On the interpretation of covariate estimates in independent competing risk models", Bulletin of Economic Research Vol. 48, pp. 27-39.