The life expectancy at retirement age has been steadily increasing over the past fifty years for both men and women. In 1960, women of 65 living in OECD countries had a remaining life expectancy of about 15 years.
Today, women of the same age can expect to live another 20 years, and a typical person approaching retirement can expect to spend over 20% of his or her life in retirement. However, enjoying these golden years might impose a hidden cost on seniors.
“Use it or lose it”
Psychologists have long argued that the brain needs to be constantly exercised in order to maintain its efficiency. As retirement is meant to lessen the daily toll imposed on the mind and body of retirees, under this “use it or lose it” hypothesis their cognitive functions might deteriorate at a faster rate, accelerating the onset of senility and damaging their ability to manage their wealth and assets.
While the correlation between cognitive functions and retirement is strong both in aggregated and individual data, estimating the causal effect of retirement on cognitive functions (or “mental retirement”) is a daunting task. More lucid and sprightly individuals are likely to retire later, and a similar, if not stronger, relationship exists between cognitive functions and schooling.
Retirement eligibility as an instrumental variable
Researchers usually perform causal inference by applying instrumental variables estimation. In the mental retirement case, a popular strategy is to exploit differences in retirement eligibility across observable groups. For example, a 61-year-old man in Denmark is eligible for early retirement, while a 61 year-old man in the US is not. Similar cross-country differences exist between men and women, and for early and old-age retirement eligibilities.
As long as these differences are purely institutional and do not contribute to the formation of cognitive skills except through retirement, it is possible to exploit them to quantify the causal relationship between retirement and cognitive skills, and filter out spurious correlations. However, this assumption does not always hold.
Using a set of international surveys from Europe and the US (for a total of 13 countries, including Denmark), we show that this identification strategy yields inflated estimates of the effect of retirement on cognitive skills. We show that differences in retirement eligibility across countries, gender and cohorts are systematically related to differences in average years of schooling, leading to a negative bias in instrumental variable estimates. Accounting for these correlations, we can explain about half the previously estimated mental retirement effect.
These correlations are, at least partially, due to institutional differences. More specifically, differences in retirement eligibilities are correlated to differences in the schooling systems in place half a century ago. While a 59-year-old Italian received five years of compulsory schooling, a 59-year-old Dane received seven years of compulsory schooling instead.
An active life is important
The contributions from our article are both methodological and substantive. While we show that researchers ought to tread lightly when using cross-country differences in retirement laws as a source of variation, we conclude that transition into pension is unlikely to affect cognitive functions in itself.
Our results do not contradict the “use it or lose it” hypothesis. Quite the contrary, we find that a healthy and engaging lifestyle, including participating in different forms of associations, voluntary work or simply taking care of family members, is associated with sharper cognitive skills. Moreover, according to our estimates, this is precisely what people tend to do once their working life is over. Enjoying the golden years is unlikely to affect mental abilities; apathy very well might.
Paul Bingley and Alessandro Martinello: Mental retirement and schooling. European Economic Review 63 (2013) 292-298.