Several studies have investigated the relationship between retirement and elderly’s health in developed countries, however, the effects on health are inconsistent among these studies. This study examines this relationship from the point of view of working hours and social activity before they retire. Japanese workers are known as workaholic, they usually do not have good balance between work and life. Those who overwork or do few social activities before retirement are predicted to worsen their physical and mental health after retirement.

The data used in this study is Longitudinal Survey of Middle-age and Elderly Persons (LSMEP), nationally representative longitudinal survey in Japan, conducted by Ministry of Health, Labour and Welfare (MHLW). From the LSMEP, we use wave one (2005) and wave 11 (2015). The activities of Daily Living (ADL), as a physical health indicator, and the Japanese version of Kessler Psychological Distress Scale (K6), as a mental health indicator, are used as dependent variables.

Respondents’ retirement decision is endogenous, being affected by respondent’s attributes, for example, age, education and marital status. This endogeneity causes the bias in the retirement effect. To solve the endogeneity problem, the Inverse Probability Weighting Regression Adjustment (IPWRA) method, known as doubly robust estimator, is applied. Estimation is conducted using some subsamples divided by working hours, working more than 40 hours in a week or no more than, and by the number of social activities involved, having more than three social activities or no more than.

The results obtained are summarized as follows: (1) In general, retirement deteriorates ADL and K6 for both male and female. (2) Retirement negative effect on ADL is larger for the respondents who overworked or did few social activities before retirement than who did not for both male and female. (3) Retirement negative effect on K6 contradicts the hypothesis for female.