We analyse YoGL for Europe based on SHARE, a multidisciplinary cross-country survey providing micro-level information on the health, well-being and socio-economic characteristics of around 140,000 individuals, based on data collected in around 380,000 interviews (Börsch-Supan et al. 2013). These survey data are especially well-suited for analysing YoGL in European countries: first, because the data are ex-ante harmonised; and, second, because the data include information on all four dimensions needed to compute the indicator. The survey’s target population consists of all non-institutionalised individuals aged 50 and older. This analysis is based on the most recent Wave 7 from 2017, which includes European countries as well as Israel (Bergmann et al. 2019;Börsch-Supan 2019b). We keep all European individuals, which results in 70,191 observations from 26 countries: namely, Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Italy, Latvia, Lithuania, Luxembourg, Malta, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden and Switzerland. In addition, information from Wave 5 (2013) is utilised for out-of-sample predictions of physical health (Börsch-Supan 2019a;Malter and Börsch-Supan 2015). Wave 5 includes only 14 European countries: namely, Austria, Belgium, Czechia, Denmark, Estonia, France, Germany, Italy, Luxembourg, Netherlands, Slovenia, Spain, Sweden and Switzerland. In total, they provide 59,713 observations.

Before describing in detail the operationalisation of the different YoGL dimensions, we should stress that the concept of the indicator is based on one specific interpretation of human well-being that focuses on longevity combined with being above minimum thresholds in further dimensions of well-being. This focus on basic material needs, basic physical and mental functioning, and having at least a medium level of life satisfaction is highly sensitive to improvements at the tail end of the population distribution, but is insensitive to improvements in the upper parts. Our decision to focus on the tail end rather than the mean of the distribution is in line with the “leave no-one behind” principle, which is also promoted in the Sustainable Development Goals (UN System Chief Executives Board for Coordination 2017).

The calculation of YoGL is based on demographic life table methods (Sullivan method) in which age-specific person-years lived at each age are multiplied by age-specific proportions considered to be above the critical threshold in all four YoGL dimensions (Sullivan 1971). Consequently, a year is only counted as a good year for those individuals who are above the critical thresholds in all four dimensions; and no trade-offs or compensatory mechanisms are assumed between the dimensions. Summing up the age-specific person-years of good life for all of the remaining age groups above the age at which remaining life expectancy will be assessed results in the expected total years of good life. Formally, the calculation can be summarised in the following mathematical notation:

where lx denotes the number of survivors at age xi (beginning of the interval i); Li indicates the number of person-years lived in the age group i; and πi denotes the prevalence of the state of interest, i.e., country-specific proportions of the population above a critical level in all four dimensions by 10-year age groups and gender.

As with total life expectancy, YoGL can be assessed at birth, as well as at any other age considered appropriate. As the focus of this paper is, however, on the well-being of older adults, we calculate YoGL at age 50.

By decomposing YoGL into four different dimensions, we are able to better understand why a population is losing good years of life. The methodology we apply here resembles the calculation of YoGL using the Sullivan method; however, in this case, the prevalence of the state of interest is decomposed into four different proportions: proportion of the population who are out of poverty (πip), proportion of the population with basic cognitive ability (πic), proportion of the population without severe activity limitations (πih) and proportion of the population with positive life satisfaction (πis).

This approach enables us to calculate four sub-indicators, which we call “years of life out of poverty”, “years of life without severe activity limitations”, “years of life with basic cognitive ability” and “years of life with positive life satisfaction”. Given that the YoGL indicator is based on no implicit trade-off assumptions or weighting structures, it must hold that the number of years of good life is less than or equal to each of the four sub-indicators for every (sub-)population.

The YoGL calculations are based on data from Wave 7, because it is the most recent wave, and includes more countries than any previous SHARE survey wave. Wave 7 does not, however, include information on the performance-tested ability to stand up from a chair. Thus, the data are extrapolated from Wave 5, which is the most recent SHARE wave in which the chair stand test was conducted. For the out-of-sample predictions, the information on the ability to stand up from a chair from Wave 5 (2013) is regressed on crucial determinants on both the individual and the country level. These determinants include demographic and socio-economic characteristics, information on physical and cognitive health (most importantly, the self-reported ability to stand up from a chair), country-level health expenditure, region dummies, as well as the other three well-being dimensions (being out of poverty, having positive life satisfaction and basic cognitive ability). Using the same determinants from Wave 7 (2017) enables us to predict levels of physical health for 2017. A detailed list of variables along with summary statistics is provided in Table A.2 in the Appendix.

Since the outcome variable is binary, a logistic regression model is utilised for the regression. Standard errors are clustered at the household level. The output table is provided in the Appendix (Table A.3). Overall, the non-linear extrapolation model yields a relatively high Pseudo R2 of 0.27. The model fits the underlying data well, as shown in Figure A.1 in the Appendix. In all of the countries that participated in Wave 5, the share of individuals with physical limitations based on the predicted data is virtually identical to the share based on the observed data. Exceptions include Czechia, Luxembourg, Austria and Spain; but even in those countries, the difference never exceeds four percentage points. Nevertheless, we conduct robustness analyses in which we compare YoGL based on predicted physical limitations with YoGL based on observed physical limitations for Wave 5 (Section 5.3).

Figure 2 displays YoGL at age 50 by country and gender. It also shows life expectancy at age 50, and ranks countries accordingly, which allows us to analyse the relationship between remaining years of life and remaining years of good life. On average, the YoGL at age 50 are 19.7 years for women and 18.6 years for men. The results also show, however, that the numbers of YoGL vary considerably between countries. In 2017, 50-year-old Swiss women can expect to live another 36.6 years, of which 28.7 are considered good years. Hence, these women can expect to spend 78.4% of their remaining lifetime as good years. By contrast, 50-year-old Bulgarian women have a life expectancy of 30.4 additional years in 2017, of which just 12.1 years are considered good years, resulting in a proportion of only 39.8%. The findings for men are very similar. The projected numbers of YoGL for men are highest in Switzerland (27.2 years) and are lowest in Lithuania (10.6 years) and Bulgaria (12.5 years).

Overall, life expectancy at age 50 is highest in Southern Europe and Switzerland, followed by Northern and other Western European countries. For example, 50-year-old women in Spain can expect to live another 37.1 years, and 50-year-old men in Switzerland can expect to live another 33.1 years. The high life expectancy in Southern Europe does not, however, translate into more YoGL. Indeed, despite its relatively high life expectancy, especially for women, Portugal is among the countries with the lowest numbers of YoGL. Instead, the Northern and Western European countries are the leaders in YoGL for both men and women. The highest numbers of YoGL are observed in Switzerland and in Northern European countries such as Sweden and Denmark, followed by in other Western European countries. Consequently, these are also the countries with the highest proportions of remaining years of good life compared to the total remaining years of life. However, Central and Eastern European countries have both low life expectancy and lower numbers of YoGL. These differences between overall life expectancy and YoGL confirm the importance of differentiating between mere survival and well-being.

Figure 3 shows the gender differences in YoGL and life expectancy at age 50 by country. In general, women have higher life expectancy than men. While this pattern holds for all countries, the gender gap in life expectancy is largest in Latvia (7.6 years) and Lithuania (7.5 years), and is smallest in Sweden (2.7 years). On average over all of the countries in our sample, women’s life expectancy exceeds men’s life expectancy by 4.9 years. Interestingly, however, the differences in YoGL are very small. On average, the gap in YoGL between women and men is only 1.1 years. This suggests that women’s advantage in life expectancy does not translate into an equal advantage in good years. The only exceptions to this general pattern are Denmark and Sweden, where the gender gap in YoGL is only slightly smaller than the gender gap in life expectancy. In seven countries (Hungary, Greece, Italy, Portugal, Cyprus, Bulgaria and Romania), the numbers of YoGL at age 50 are even lower for women than for men.

As a result of the gender disparities noted above, women have, on average, a much lower proportion of remaining years of good life than of remaining years of life. While men at age 50 can expect, on average, to spend 62.7% of their remaining time as good years, the corresponding share for women is just 57.1%. The proportions of remaining years of good life are particularly low for women in Portugal (38.6%), Romania (39.1%) and Bulgaria (39.8%). These results are related to the “male-female health-mortality paradox” (di Lego et al. 2020), a term that is used to refer to the phenomenon that, on average, women live longer than men, but spend a larger proportion of their life in poorer health (Luy and Minagawa 2014;Pongiglione et al. 2015).

Decomposing YoGL into four dimensions allows us to better understand why countries are losing good years of life. The results presented in Figures 4 and 5 clearly show that, overall, most good years are lost due to limitations in physical health, followed by limitations in cognitive health. By contrast, poverty has only small effects on YoGL, since most Europeans are considered to be out of poverty according to the definitions described above. The countries with the lowest numbers of years of life out of poverty are all Central and Eastern European countries; but even in those countries, more YoGL are lost due to deductions in the other dimensions. Related literature has, however, suggested that poverty plays an important role in the lower levels of well-being observed in low-income countries outside of Europe, and in trends in YoGL over time (Reiter and Lutz 2020;United Nations Development Programme 2019).

Analysing the decomposition separately by country and gender reveals that there are big differences in the composition of YoGL. While the reduction in good years due to low life satisfaction plays a small role in most countries, it is highly relevant for Central and Eastern European countries. In Bulgaria, Hungary, Lithuania and Latvia, it is even the main reason for the reduction in good years; thus, in these countries, life satisfaction has a bigger impact on YoGL than physical and cognitive health. These results are very much in line with previous findings showing that in Central and Eastern European countries, life satisfaction is persistently low (Dingemans and Henkens 2018;Guriev and Zhuravskaya 2009). This pattern has often been attributed to the impact of the post-communist transition and the related increases in inequality and uncertainty (Djankov et al. 2016;Sanfey and Teksoz 2007).

Women lose most of their YoGL due to their physical limitations. Among men, however, the negative effects of physical and cognitive limitations are almost identical. These results have to be interpreted against the backdrop of differences in overall life expectancy. As we mentioned above, women spend a larger proportion of their overall lifetime in poor health. Similar results are found for life satisfaction. While women spend more years with positive life satisfaction than men, they spend a smaller proportion of their lifetime with positive life satisfaction than men.

Sensitivity analyses displayed in Table A.4 and Figures A.2, A.3 and A.4 in the Appendix show that the results discussed above are robust. As we discussed in Section 3.1, we consider different thresholds for physical limitations, cognitive limitations and subjective well-being. Furthermore, we compare YoGL based on predicted and observed physical limitations for countries that participated in Wave 5.

The results of the main analysis show that 63.6% of the population are simultaneously out of poverty, report positive life satisfaction, have basic cognitive ability and suffer from no severe activity limitations. This share hardly changes when the robustness analyses are applied. Decreasing the threshold for cognitive limitations to two words instead of three words increases the share of individuals who are above the threshold in all YoGL dimensions to 68.6%; while decreasing the threshold for positive life satisfaction from five to four increases the share to 68.7%.

As can be seen in Figure A.2, changing the threshold for cognitive limitations leaves the results virtually unchanged for most countries. Although YoGL increase slightly when the more lenient threshold is applied, the country ranking and related results are hardly affected. One exception is Portugal, where YoGL increases substantially based on the new threshold, especially for men. The reason for this increase is that cognitive ability in Portugal has a different distribution than it does in the other countries. First, as mentioned above, Portugal is the only country in which no participant remembered more than eight words – in all other countries, the maximum is 10 words. Second, while the number of words recalled is normally distributed in most countries, the Portuguese distribution has an additional concentration around three words. Hence, whether the threshold is set at two or three words makes a greater difference for Portugal than it does for the other countries. When we look at the distribution of the number of words in previous waves in Portugal, this curious distribution does not appear. Thus, we conclude that the results for Portugal need to be treated with caution.

Decreasing the threshold for positive life satisfaction also increases YoGL slightly for all countries simultaneously; however, since the country ranking hardly changes (Figure A.3), the conclusions drawn from the results are virtually unaffected.

Figure A.4 shows the difference in the YoGL when using predicted values of physical limitations rather than observed values of physical limitations for Wave 5, which covered a much smaller group of countries than Wave 7. The results show that when the predicted rather than the observed values are used, the YoGL hardly change, and the country ranking remains exactly the same for both men and women. We thus conclude that our predictions are robust.