151
What Explains The Difference in The Effect of Retirement on Health?: Evidence From Global Aging Data
This paper reviews the studies of the effect of retirement on health and considers why these studies report different results of the effect of retirement on health. A number of studies have investigated the relation between retirement and health over the last two decades. However, there is no unifying view about the impact of retirement on health. Some studies conclude that retirement has a positive impact on health defined as mental health or physical health. On the other hand, other studies conclude that retirement has no or negative effect. We want to consider why these results are different. Some studies (Eibichi, 2015 and Motegi et al., 2015) consider that the changes in health behavior after retirement were different in each study. These differences in the change of health behavior after retirement (lifestyle habits including drinking, smoking, exercise and sleeping) explain the differences in the results. In this study, we focus on the effect of the more basic elements that could explain the differences in the estimated results reported by the literatures. We validate the effect of the differences in surveyed countries, analysis methods and the definition of variables on the estimated results of the effect of retirement on health. We analyze the influence of each factor on the estimated results of the effect of retirement on health by changing only one factor (e.g. only analysis method) while we keep the other factors unchanged.
METHODS:
We examine the effect of three factors one by one by changing only one factor while we keep the other factors unchanged. The three factors are the definition of retirement (e.g. not working for pay or retire), analysis method (e.g. OLS, FE, IV) and surveyed country (e.g. the U.S., Korea, Japan). We review main analysis methods and main health outcomes that the related literatures have used. Based on this information, we decide what we include in each factor. For example, we include “not working for pay” or “retire” in the definition of retirement. We use the sister datasets of the Health Retirement Survey (HRS). We use HRS, ELSA, SHARE, KLOSA, CHARLS and JSTAR. We report the sensitivity of the three factors on cognition ability indicators, subjective health, subjective well-being, BMI, CESD, IADL, ADL and grip strength.
RESULTS:
We sum up the influences of three factors on the estimated results of the retirement on health.
(1) The difference in surveyed country weakly influences the estimated results except cognitive ability indicators. The estimated results do not largely change as far as we see the estimated IV results except cognitive ability indicators.
(2) The difference in analysis methods strongly influences the estimated result. When we change analysis method, the estimated result can lead to opposite conclusion in many health indexes.
(3) The difference in the definition of retirement weakly influences the estimated result. The estimated results change slightly in all health outcomes even when we change the definition of retirement.