181
Retirement And Cognitive Decline: Evidence From Global Aging Data

Tuesday, June 14, 2016
Lobby (Annenberg Center)

Author(s): Yoshinori Nishimura

Discussant:

OBJECTIVES:

This paper examines the hypotheses that Rohwedder and Wills (2010) discuss. Rohwedder and Wills (2010) suggest that cognitive ability through mental exercise can avoid cognitive decline. They set up two hypotheses why retirement cause cognitive decline. One hypothesis is that workers do more mental exercise than retiree because work environments are more cognitively challenging and stimulating for workers. Another is a hypothesis that retirement causes the disincentive to invest human capital. The difference in incentive to invest human capital explains the degree to reduce mental exercise. In any case, these hypotheses indicate the negative relationship between retirement and cognitive ability. To empirically prove this relationship, they estimate a simple IV regression describing the relationship between retirement and cognitive ability indicator. They report the negative significant result of retirement on cognitive ability. However, we doubt whether there is a negative relationship between retirement and cognitive function based on the hypotheses. We analyze the relationship between retirement and cognitive ability based on their suggestions. We check robustness of the simple regression analysis by Rohwedder and Wills (2010). We also check other important suggestions that Rohwedder and Wills (2010) do not analyze.

METHODS:

First, we replicate a similar simple regression analysis by Rohwedder and Wills (2010). To check robustness of this analysis, we add some control variables (e.g. income, education) and use some similar instrumental variables. Second, we analyze the effect of retirement on cognitive function by using longer panel data. We analyze the effect on cognitive function by omitting the individual fixed effect and time effect. When we do this analysis, we analyze the effect to vary across individuals (the effect of education, environment and other factors) which Rohwedder and Wills (2010) suggests. Finally, we analyze the effect of duration after retirement on cognitive function which Rohwedder and Wills (2010) do not analyze.

RESULTS:

We sum up the results.

(1) The simple regression result by Rohwedder and Wills (2010) is not robust. When we change the pattern of control variables and the pattern of instrumental variables, we sometimes get positive effect of retirement on cognitive function.

(2) We suggest that there is no effect of retirement duration on cognitive decline velocity when we omit the individual fixed effect. Even if we separate the sample depending on education level, gender difference and cohort groups, we cannot confirm that there is a negative effect of retirement duration on cognitive decline velocity.

We suggest that the hypothesis proposed by Rohwedder and Wills (2010) is doubtful.