Do the sick retire early? Chronic illness, asset accumulation and early retirement

Abstract

Our objective is to determine how chronic illness affects asset accumulation and retirement. Previous studies have found that poor health leads to early retirement, but those studies failed to look at the indirect impact of chronic illness on retirement. Using data from the Health and Retirement Study, we define an illness as chronic if the individual reports having asthma, cancer, heart disease, stroke or diabetes for four or more years. We first estimate how a chronic illness influences asset accumulation. We then estimate how asset accumulation and current poor health influence retirement. We observe that the vast majority of the chronically ill population do not report their general health to be poor nor do they report functional limitations in activities of daily living. Nevertheless, our results indicate that chronic illness leads these people to accumulate fewer assets during their working years and consequently retire later. Neither researchers nor policy-makers discussing the many critical issues surrounding illness and retirement have addressed this issue.

Acknowledgements

The authors thank Patricia Reagan, members of the Department of Economics at Northern Illinois University, and an anonymous referee for helpful comments.

Notes

¹ Other studies of retirement include Reno 1971, Burkhauser 1979, Burtless and Moffit 1984, 1985, 1986, Moffitt (1984, 1987), Burtless (1986, 1987), Butler et al. 1989 and Gustman and Juster 1995.

² Wilson 2001, p. 1139.

³ See discussion in Ettner (1996) and Attanasio and Emmerson 2003.

⁴ See Ettner (1996) for a discussion of this literature.

⁵ Conforming to the HRS data, this peri-retirement period begins at age 51.

⁶ If an improvement in health or additional time spent in health production decreases both the marginal disutility of work and the marginal utility of consumption, then ∂S ₂/∂H ₂ > 0.

⁷ This assumption is similar to the effect of poor health on labour supply in period 2 that was posited in hypothesis H1.

⁸ We are not able to assess the possibility of sample selection in our multivariate analysis because our regression and probit analyses require information on the present value of social security and pension income that is unavailable for persons excluded from the study sample.

⁹ To evaluate what length of time most accurately represents a ‘long-term’ illness, we examined two additional specifications of CHRONIC in which the number of years for which the person has had the chronic condition was increased to ‘six or more’ and ‘nine or more’. The results were qualitatively similar to those reported in this article, although due to the reduction in the number of persons with chronic conditions, the statistical significance of the variable is reduced. In addition to using the number of years since onset to define which illnesses are chronic, we were advised by a physician about diseases that are inherently chronic.

¹⁰ Using current earnings for retirees would introduce simultaneity between earnings and retirement status. However, the availability of pre-retirement earnings data in the HRS allows us to avoid this problem.

¹¹ Again, we also examined two specifications of CHRONIC in which the number of years for which the person has had the chronic condition was increased to ‘six or more’ and ‘nine or more’. Once again, the results were qualitatively similar to those reported in this article.

¹² Six percent of the respondents in the sample have had diabetes for four or more years, while 3% of the respondents have had cancer for at least 4 years. Another 2% of the respondents in the sample had a heart attack at least 4 years ago, or coronary heart disease or other heart problems for at least 4 years and 1% of respondents had a stroke at least 4 years ago. Five percent of respondents had asthma at the time of the survey.

¹³ The Smith–Blundell Chi-square value for the fullest specification for men is 0.013. With one degree of freedom, this yields a p-value of 0.91, leading us to reject endogeneity. Similarly, the Chi-square value for women is 0.08. With one degree of freedom, this yields a p-value of 0.78, leading us to again reject endogeneity.

¹⁴ Adding current poor health to the explanatory variables does not change the effects of CHRONIC or LIMIT on asset accumulation.

¹⁵ Note that the signs of education and marital status are somewhat sensitive to whether ASSETS are treated endogenously.

¹⁶ There may be an omitted variable bias in the estimated effects of some of the independent variables, but these biases do not appear to differ between workers with and without chronic health problems.