Long-run health trends in Europe

Abstract

Long-run health trends are worsening for middle-aged baby-boomers in the US. In contrast to this alarming development, Abeliansky and Strulik (2019, A&S), in one of the few studies of long-term health developments in Europe, find improving health trends in Europe across a broad range of cohorts. We conducted a cohort analysis of health deficits similar to A&S but focused on middle-aged individuals. We find that the positive health trend has all but stalled for the more recent birth cohorts in Europe. Our results are robust to a large range of different definitions of the health index. We argue that the difference between A&S and our results is due to longitudinal sample selectivity bias. Our findings have implications for future public and private costs of health care, and they are relevant for the ongoing discussion on how to achieve longer healthy working lives.

Introduction

European citizens enjoy a higher life expectancy (OECD, 2019), a higher healthy life expectancy (WHO, 2019) and better health on average (Avendano et al., 2009, Avendano and Kawachi, 2014) than US citizens. Moreover, not only the increase in US life expectancy has stalled as recent data show¹, but also US baby boomers, despite their longer life expectancy over previous generations, have higher rates of chronic disease, more disability, and lower self-rated health than members of the previous generation at the same age (King et al., 2013).

The paper by Abeliansky and Strulik (2019, A&S), published recently in this journal, is one of the few studies of long-term health developments in Europe. A&S use panel data from the Survey of Health, Aging and Retirement in Europe (SHARE) and compute the health deficits of individuals aged 50 to 85 in 14 European countries by birth cohort. They find that European health trends stand in stark contrast to the developments in the United States: health deficits still keep declining for older individuals as well as for more recently born cohorts, both for men and women. A&S argue that this continuous trend approximates the rate of medical progress.

While this is a pleasant result for Europeans, there are several reasons to be surprised by the findings of A&S. Medical progress is shared globally without much delay. If anything, medical technology – procedures, equipment and drugs – is more advanced in the US than in Europe. One may argue that in the US top medical technology is only accessible for the well-to-do while it is more equitably distributed in Europe. However, also Europe has faced a trend in rising inequality and an increase in the socio-economic gradient of health and life expectancy (e.g., Mackenbach et al. (2016)). Since the research by A&S is an important contribution to understanding cross-national health differences and their causes, it is essential to check the robustness of the A&S results.

We therefore investigated the sensitivity of their cohort analysis to different definitions of the applied health deficit index. We are able to replicate the results by A&S but found that they are sensitive to the selection of variables included in the health deficit index, in particular to including variables that are not observed in the later waves of the SHARE panel. We therefore employ a health index that only includes the variables that are available in all years of the panel data. We find similar results for older cohorts but reach strikingly different conclusions for middle-aged individuals. This difference is driven by only 4 out of the 38 health measures used by A&S. These variables are available in the early but not the most recent waves of SHARE. In econometric terms, this creates a sample selectivity bias since younger cohorts are more exposed to these missing variables. We argue that the disappearance of the variables from the panel artificially lowers the value of the index for the younger cohorts, thus creating the illusion of improving health conditions across Europe while, in fact, they are stagnating.

Our result of a stagnating health trend is robust to a large range of different definitions of the health index as long as the health measures are available throughout all waves. E.g., in order to avoid taking any stance on which variables are more important than others, we randomly choose sets of 35 and 40 variables out of 52 variables available in all waves and always reach similar results.

Our findings are relevant for the ongoing discussion on how to achieve longer healthy working lives and suggest that not only the US but also Europe may see a stalling of the long-term upwards health trend among individuals in their second half of life.