Pekka Ilmakunnas1, Seija Ilmakunnas2. 1. 1 Aalto University School of Business, Helsinki, Finland. 2. 2 Labour Institute for Economic Research, Helsinki, Finland.
Abstract
AIM: We examine the relationship between the subjective assessment of health status and retirement by using information on expected and actual retirement ages. METHODS: Subjective data from cross-sectional surveys, conducted in Finland in 2003 and 2008, are linked to information on actual retirement age from register data from 2003 to 2013. Regression models are estimated for actual and expected retirement ages. RESULTS: While the health status is positively correlated with both actual and anticipated full-time retirement age, the actual age of retirement is less sensitive to health. On average, individuals tend to retire later than they had anticipated. We examine potential biases in the health-retirement relationship. Measurement error in regard to health status biases the results downwards. Using data on observed retirement ages, omitting those who do not retire during the data period, leads to a selection problem. Ignoring the selection also leads to a downwards bias in the health-retirement age connection. As a more exogenous health variable we use health shocks, which are measured by average annual days of absence due to sickness in the follow-up period. These shocks are negatively related to retirement age in a subsample of initially healthy individuals. CONCLUSIONS: When subjective assessment of health is used for explaining retirement behavior, the effects of health can often be underestimated rather than overestimated. To lengthen working careers, attention should be given to both the ability (health) and willingness (perceptions of proper retirement age) of people to continue longer at work.
AIM: We examine the relationship between the subjective assessment of health status and retirement by using information on expected and actual retirement ages. METHODS: Subjective data from cross-sectional surveys, conducted in Finland in 2003 and 2008, are linked to information on actual retirement age from register data from 2003 to 2013. Regression models are estimated for actual and expected retirement ages. RESULTS: While the health status is positively correlated with both actual and anticipated full-time retirement age, the actual age of retirement is less sensitive to health. On average, individuals tend to retire later than they had anticipated. We examine potential biases in the health-retirement relationship. Measurement error in regard to health status biases the results downwards. Using data on observed retirement ages, omitting those who do not retire during the data period, leads to a selection problem. Ignoring the selection also leads to a downwards bias in the health-retirement age connection. As a more exogenous health variable we use health shocks, which are measured by average annual days of absence due to sickness in the follow-up period. These shocks are negatively related to retirement age in a subsample of initially healthy individuals. CONCLUSIONS: When subjective assessment of health is used for explaining retirement behavior, the effects of health can often be underestimated rather than overestimated. To lengthen working careers, attention should be given to both the ability (health) and willingness (perceptions of proper retirement age) of people to continue longer at work.