Article Text

Download PDFPDF

Trends of healthy and unhealthy working life expectancy in Germany between 2001 and 2020 at ages 50 and 60: a question of educational level?
  1. Stefanie Sperlich,
  2. Johannes Beller,
  3. Jelena Epping,
  4. Siegfried Geyer,
  5. Juliane Tetzlaff
  1. Medical Sociology, Hannover Medical School, Hannover, Germany
  1. Correspondence to Dr Stefanie Sperlich, Medical Sociology, Hannover Medical School, Hannover 30625, Germany; sperlich.stefanie{at}


Background Extending the number of active working years is an important goal both for maintaining individual quality of life and safeguarding social security systems. Against this background, we examined the development of healthy and unhealthy working life expectancy (HWLE/UHWLE) in the general population and for different educational groups.

Methods The study is based on data from the German Socio-Economic Panel study, including 88 966 women and 85 585 men aged 50–64 years and covering four time periods (2001–05, 2006–2010, 2011–2015 and 2016–2020). Estimates of HWLE and UHWLE in terms of self-rated health (SRH) were calculated using the Sullivan’s method. We adjusted for hours worked and stratified by gender and educational level.

Results Working-hours adjusted HWLE at age 50 increased in women and men from 4.52 years (95% CI 4.42 to 4.62) in 2001–2005 to 6.88 years (95% CI 6.78 to 6.98) in 2016–2020 and from 7.54 years (95% CI 7.43 to 7.65) to 9.36 years (95% CI 9.25 to 9.46), respectively. Moreover, UHWLE also rose with the proportion of working life spent in good SRH (health ratio) remaining largely stable. At age 50, educational differences in HWLE between the lowest and highest educational groups increased over time in women and in men from 3.72 to 4.99 years and from 4.06 to 4.40 years, respectively.

Conclusions We found evidence for an overall increase but also for substantial educational differences in working-hours adjusted HWLE, which widened between the lowest and highest educational group over time. Our findings suggest that policies and health prevention measures at workplace should be more focused on workers with low levels of education in order to extend their HWLE.


Data availability statement

Data are available on reasonable request. The raw data were drawn from the German Socio-Economic Panel Study. The datasets used are available from the corresponding author on reasonable request. German data privacy laws necessitate that all users sign a data user contract with DIW Berlin.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • To date, there have been few studies on trends in healthy working life expectancy (HWLE), particularly those on socioeconomic differences are lacking.

  • Previous studies suggest a positive association between levels of education and HWLE.

  • HWLE and unhealthy working-life expectancy (UHWLE) increased over the last decades.


  • We found working-hours adjusted HWLE and UHWLE at ages 50 and 60 to be increasing between 2001 and 2020 for both genders.

  • Gender inequalities in working-hours adjusted HWLE largely persisted to the disadvantage of women.

  • The rise in HWLE was mainly driven by the increase of employment rates while no significant improvements in the proportions of good self-rated health were found.

  • Educational inequalities in HWLE to the detriment of the lowest educational group have widened over time.


  • Our findings suggest that the greatest potential for increasing HWLE is among individuals with the lowest level of education.

  • Hence, priority should be given to tailored workplace health promotion measures for low-educated employees.

  • In addition, increased efforts should be directed towards reducing gender inequalities in employment.


The participation in the labour market is a central part of the productive involvement of older people in social life and is, therefore, considered a critical criterion for successful ageing.1 2 The extent to which gainful employment is conducive to the quality of life for older employees depends on the subjective importance of employment as well as on factors related to the working conditions.3 4 Workers with lower levels of education are more likely to experience unfavourable working conditions, such as high physical demands and low job control and job rewards.5

Germany, such as other Western societies, is facing demographic ageing that has far-reaching implications for the further development of the federal pension system, long-term care and statutory health insurance. Particular challenges arise from the fact that the share of the working population is declining while that of benefit recipients is growing. In Germany, the number of people aged 67 and older is projected to increase by 22% between 2020 and 2035, from 16 million to at least 20 million.6 In order to meet the challenge of demographic ageing, many countries have introduced measures aimed at increasing the length of working life. In Germany, for example, early retirement has become more restrictive through a series of reforms.7 In addition, the age limit for the statutory retirement pension in Germany will be gradually raised to 67 by 2029.8

One central concern associated with increasing pension ages is whether employees are sufficiently healthy to be able to work longer. In the case of a negative scenario, employees will be forced to continue working despite poor health or drop out of the labour market before reaching retirement age. Extending the working life would, therefore, primarily be possible for healthy employees and those with less physically demanding work, with the consequence that existing social inequalities could be exacerbated.

Different population indicators have been proposed in order to systematically examine changes in the length of working life, such as ‘working life expectancy’ (WLE) and ‘healthy/unhealthy WLE’ (HWL/UHWLE).1 9 The systematic review by Parker et al from 2020 identified only four studies that estimated HWLE in later working life.10 Consistent with these findings, recent studies from Germany have determined that WLE and HWLE have increased over time, particularly among women.11 12 To the best of our knowledge, there have been no studies from Germany on the development of social inequalities in the HWLE over time. Addressing this gap, the aim of this study is to investigate trends in WLE, HWLE and UHWLE for men and women in Germany according to different educational groups. The focus of this study is on whether HWLE and UHWLE have increased in the last two decades and what role educational attainment has played in these trends.


Data source

The analyses are based on data from the German Socio-Economic Panel (GSOEP V.31), conducted by the German Institute for Economic Research. The GSOEP is a representative annual survey of individuals aged 18 and older in private households in Germany that started in 1984.13 Data are collected by face-to-face interviews using different questionnaires for individuals, households or specific subgroups. The GSOEP population is updated regularly with new survey samples to reflect compositional changes in the German population and in order to compensate for drop-outs occurring over time. Further information on GSOEP can be obtained from Goebel et al.13

We included participants between 50 and 64 years of age as our focus lies on the older working age. Our analyses are based on a pooled dataset including the waves from 2001 to 2020, which were divided into four time periods to balance out annual fluctuations (2001–2005, 2006–2010, 2011–2015 and 2016–2020). We merged the data into one dataset in which each year of participation corresponds to one row. This means that within one time period the same person could be included several times in the data. However, since we calculated prevalence rates for each single age between 50 and 64 years, each subject occurs only once at one given age within the 5-year period, avoiding the problem of nested data when calculating age-specific prevalence rates. We used cross-sectional weights that are assumed to produce a nationally representative sample. The proportion of missing values for the variables used to calculate HWLE was low: 0% for age, sex and employment status; 0.1% for self-rated health (SRH) and education level according to the Comparative Analysis of Social Mobility in Industrial Nations (CASMIN classification); and 1.9% for weekly working hours. Respondents with missing information on these variables were excluded, except for the variable ‘weekly working hours’. For this variable, we replaced the missing values with the respective mean by gender and education level to ensure that all employed persons were included in the analyses. We used the guidelines 'Strengthening the Reporting of Observational studies in Epidemiology' (STROBE).14


Self-rated health

We used SRH as a health measure to calculate (U)HWLE, which has proven to be a reliable and valid health indicator predicting healthcare utilisation, current and future health problems, and mortality.15–17 SRH is one of the indicators that has been frequently used to study healthy life expectancy.18 It reflects different domains of health such as mental health,19 20 chronic diseases,21 chronic pain22 and functional limitations.23 It has been shown that reporting a poor SRH was related higher to early retirement than diagnosed diseases.24 Poor SRH was also more strongly associated with whether older workers believed they would not be able to work beyond age 65 than when they stated that a diagnosed disease is a barrier to their daily work.24 In line with this, Pietiläinen et al25 found that SRH is a strong predictor of disability retirement. Hence, poor SRH provides a useful marker for increased risk of work disability and subsequent disability retirement.

SRH was measured by the question ‘In general, how would you rate your current health status?’ on a 5-point-scale (very good, good, satisfactory, less good and bad). Individuals who respond ‘less good’ and ‘bad’ were considered to have an ‘unhealthy’ status in terms of UHWLE. We opted for a broader definition of good health that also includes the intermediate category of ‘satisfactory’. The reason for this was that we focused on older workers, for whom (very) good health can be less expected compared with younger individuals. In addition, satisfactory health was considered a sufficient health status for employment. Accordingly, individuals who respond ‘very good’, ‘good’ and ‘satisfactory’ were assigned to have a ‘healthy’ status in terms of HWLE.

Employment status

In line with the International Labour Organization (ILO), we defined a person as employed if he or she declared to have worked in the last week for at least 1 hour for pay or profit. This also includes persons who are temporarily not at work but have an attachment to their job (eg, due to holidays, sick leave or job-related training).26


Educational attainment was measured using the CASMIN educational classification that is based on two primary classification criteria: first, the differentiation of an educational hierarchy (elementary, intermediate, maturity and tertiary), and second, the differentiation between ‘general’ and ‘vocationally oriented’ education.27 The original nine CASMIN-levels were merged into five educational groups: lowest (levels 1a and 1b), low (level 1c), intermediate (levels 2a und 2b), high (levels 2c_gen, 2c-voc, 3a) and highest (level 3b) (online supplemental table 1).

Supplemental material

Statistical analysis

First, we analysed the temporal development of the relevant study variables, that is, employment status, hours worked weekly and good/poor SRH stratified by gender and educational level, using linear and logistic regression techniques. We calculated predicted probabilities and used cluster-robust SEs to adjust for the panel structure of the data.

We used Sullivan’s method28 to calculate WLE and HWLE/UHWLE, which has been used in former studies.18 29 In line with a previous study,29 we did not account for mortality because no information was available in the data. This means that the estimated expectancies should be interpreted as being condition on survival between ages 50 and 64 years. Dudel et al30 showed that ignoring mortality rates up to 64 years of age is acceptable since mortality is generally rather low up to that age.

In a first step, we calculated WLE for each time period as the sum of the age-specific proportions of men and women working between the ages 50 years and 64 years, using 1-year age groups. We adjusted the WLE for working hours to take into account gender differences in the number of working hours. The adjustment for working hours was achieved by weighting employment rates with the ratio of actual work hours of the individual to 40 hours, which corresponds to a full-time work schedule. This means that if, for example, the average number of working hours is 30, the employment rate is multiplied by 0.75 (30/40).30 For this purpose, working hours of over 40 hours were limited to the maximum of 40 hours. Then, we calculated HWLE/UHWLE as the sum of the age-specific proportions of men and women working in good or poor SRH, respectively. HWLE measures the total remaining lifetime spent working while in good health, whereas UHWLE captures the remaining lifetime spent working in poor health. With a retirement age of 65 years, these indicators can theoretically reach a maximum of 15 and 5 years at ages 50 and 60, respectively. We calculated 95% CIs based on the sum of the variance associated with each age-specific proportion and the unweighted number of observations. Moreover, we determined the working ‘health ratio’ as the proportion of remaining working years expected to be spent in good health of total WLE. The SEs for calculating the CIs for the health ratio are based on the standard errors of HWLE multiplied with 100 and divided by WLE (formula: 100%×SE of HWLE/WLE). For each time period, all estimates were calculated separately for women and men and each educational class according to the manual by Jagger et al.31


On average, both female (n=88 966) and male participants (85 585) were 56 years old (SD=4.3) while 50.5% were female. Employment rates, occupational position, adjusted household net income and the share of the non-German population increased while sickness absence and early retirement decreased with higher educational attainment (table 1).

Table 1

Weighted sample characteristics in % by gender and educational level (CASMIN)

Temporal trends in employment, working hours and SRH

Employment rates increased in both genders but more for women, while the number of hours worked per week has hardly changed for both (figure 1A,B). Among employed individuals, the proportion of those in good SRH has remained largely constant (figure 1C). In contrast, the proportion of good SRH among those not employed has declined significantly, especially among women (figure 1D).

Figure 1

Predicted probabilities (predicted means) and cluster-robust SEs of employment (A), working hours (B), good self-rated health (SRH) in employed (C) and non-employed persons (D), GSOEP from 2001–2005 to 2016–2020, aged 50–64 years, stratified by gender. GSOEP, German Socio-Economic Panel study.

The share of persons with lower education decreased over time, while that of intermediate and higher education increased (figure 2A). Women in particular showed a clear trend towards higher education. Moreover, employment rates increased for all educational groups but less pronounced for those with the lowest educational attainment (figure 2B). The number of hours worked per week has not changed over time among those with intermediate and higher education, while it slightly decreased among the less educated, leading to widening educational inequalities in the working hours (figure 2C).

Figure 2

Predicted probabilities (predicted means) and cluster-robust SEs of educational groups according to CASMIN (Comparative Analysis of Social Mobility in Industrial Nations) (A), employment (worked in the last week for at least 1 hour for pay or profit) by CASMIN educational groups (B) and working hours by CASMIN educational classification (C), GSOEP from 2001–2005 to 2016–2020, aged 50–64 years, stratified by gender. GSOEP, German Socio-Economic Panel study.

Despite fluctuations between time points, the proportions of persons with good SRH among employed individuals remained largely unchanged for all education groups (online supplemental figure 1A). In contrast, the proportion of persons reporting good SRH among those not employed has declined for both genders, particularly among women with the lowest level of education. Correspondingly, educational inequalities among non-employed women have increased (online supplemental figure 1B).

Supplemental material

Trends in working-hours adjusted WLE

WLE measured in full-time equivalents (working-hours adjusted WLE) increased over time for both genders while gender differences remained largely stable (table 2). Among women and men aged 50 years, WLE rose from 5.47 years (95% CI 5.29 to 5.66) in 2001–2005 to 8.25 years (95% CI 8.12 to 8.39) in 2016–2020 and from 8.96 years (95% CI 8.86 to 9.07) to 11.17 years (95% CI 11.04 to 11.31), respectively. At age 60, the remaining WLE increased in women from 0.76 years (95% CI 0.64 to 0.88) to 2.16 years (95% CI 2.06 to 2.25) and in men from 1.65 years (95% CI 1.59 to 1.71) to 3.17 years (95% CI 3.11 to 3.24). In contrast, the working-hours unadjusted estimates suggested a convergence of the gender gap in WLE (online supplemental table 2). In general, the adjusted WLE are smaller compared with the unadjusted estimates. This is because marginally employed individuals are given the same weight as full-time workers in the unadjusted estimates, while they are given a lower weight in the adjusted estimates according to the number of hours worked per week.

Table 2

Working hours-adjusted working life expectancy (WLE) in years at ages 50 and 60, GSOEP from 2001–2005 to 2016–2020 by gender and educational level (CASMIN classification)

Stratified by educational level, it revealed that WLE increases steadily with increasing schooling (table 2). Over time, the difference in WLE at age 50 between the lowest and highest educational group increased from 3.76 to 5.10 years for women and from 3.83 to 4.53 years for men. In 2016–2020, women and men at age 50 with the lowest educational attainment could expect to work further 4.61 years (95% CI 4.29 to 4.93) and 7.76 years (95% CI 7.41 to 8.10) in full-time equivalents while the corresponding figures for their highest educated counterparts were 9.71 years (95% CI 9.46 to 9.96) and 12.29 years (95% CI 12.11 to 12.48), respectively. Moreover, the gap between the lowest and the low educational group widened, while they narrowed between the two higher educational groups.

Trends in working-hours adjusted HWLE/UHWLE

Working-hours adjusted HWLE at age 50 years increased in women and men from 4.52 years (95% CI 4.42 to 4.62) in 2001–2005 to 6.88 years (95% CI 6.78 to 6.98) in 2016–2020 and from 7.54 years (95% CI 7.43 to 7.65) to 9.36 years (95% CI 9.25 to 9.46), respectively (table 3). At age 60, the corresponding increases were from 0.62 years (95% CI 0.58 to 0.67) to 1.77 years (95% CI 1.71 to 1.83) and from 1.40 years (95% CI 1.35 to 1.46) to 2.35 years (95% CI 2.28 to 2.42), respectively. Clear educational differences for both genders were also found for HWLE. Over time, the difference in HWLE at age 50 between the lowest and highest educational group increased for women from 3.72 to 4.99 years and for men from 4.06 to 4.40 years. In 2016–2020, women and men at age 50 with the lowest educational attainment could expect to work healthy further 3.48 years (95% CI 3.19 to 3.76) and 6.51 years (95% CI 6.18 to 6.84) in full-time equivalents. The corresponding figures for the highest educated women and men were 8.47 years (95% CI 8.21 to 8.73) and 10.91 years (95% CI 10.70 to 11.13), respectively. Again, the gap between the two lower educational groups widened, while it narrowed between the two higher educational groups. As with the WLE values, the working-hours unadjusted HWLE values were higher, since no weighting was applied here according to the extent of employment (online supplemental table 3).

Table 3

Working hours-adjusted healthy working life expectancy (HWLE) in years at ages 50 and 60, GSOEP from 2001–2005 to 2016–2020 by gender and educational level (CASMIN classification)

With increasing WLE also UHWLE increased over time in women and men. At age 50, the increases were, for example, from 0.95 years (95% CI 0.85 to 1.05) in 2001–2005 to 1.37 years (95% CI 1.27 to 1.48) in 2016–2020 and from 1.42 years (95% CI 1.32 to 1.53) to 1.81 years (95% CI 1.71 to 1.92), respectively (online supplemental table 4). With few exceptions, increasing UHWLE was found for all educational groups. The working-hours unadjusted UHWLE estimates show the same pattern, again with overall higher values (online supplemental table 5).

Trends in the health ratio

At age 50, the proportion of working life expected to be spent in good SRH of total WLE (health ratio) remained largely stable for both genders (online supplemental table 6). For example, in 2001–2005, women aged 50 years could expect to spend 82.6% (95% CI 80.8 to 84.5) of their remaining working life of 5.47 years (95% CI 5.29 to 5.66) (full-time equivalents) in good SRH (table 2). This corresponds to 4.52 healthy working years (95% CI 4.42 to 4.62) (table 3) while 0.95 years (95% CI 0.85 to 1.05) are expected to be spend in poor health (online supplemental table 4). In 2016–2020, the health ratio was 83.4% (95% CI 82.1% to 84.6%) of 8.25 remaining working years (95% CI 8.12 to 8.39) that can be divided into 6.88 healthy (95% CI 6.78 to 6.98) (table 3) and 1.37 unhealthy working years (95% CI 1.27 to 1.48) (online supplemental table 4). For men, the corresponding figures for the first and last period were 84.1% (95% CI 82.9% to 85.3%) of 8.96 remaining working years (95% CI 8.86 to 9.07) and 83.8% (95% CI 82.8% to 84.7%) of 11.17 remaining working years (95% CI 11.04 to 11.31) that are expected to be spent in good SRH.

At age 60, the least qualified women experienced a significant decline in the health ratio over time, from 86.4% (95% CI 72.0% to 100.8%) to 73.2% (95% CI 58.6% to 87.8%), suggesting that the relative proportion of women working in poor SRH has increased in this group. Among men of that age, a marked decrease in the health ratio was observed for all educational groups, except for those with the highest tertiary education, for whom the proportion of working life expected to be spent in good SRH slightly increased.


Over the past decades, Germany has introduced pension and labour market reforms that aimed at increasing the length of working life among the older working population.7 32 Indicating the success of these efforts, we found that working hour-adjusted WLE at ages 50 and 60 steadily increased over time for both genders. These findings are consistent with other studies from Germany,11 12 30 Finland33 and the USA34 that also found that the share of remaining life spent in work has increased across successive cohorts. However, comparing our results with other studies is limited due to the high heterogeneity between the studies, for example, regarding the statistical approach, the choice of health outcomes and the time period under consideration. Moreover, we adjusted the WLE for working hours as proposed by Dudel et al,30 which is usually not the case. Women as well as educationally disadvantaged groups are not only less likely to be employed, but, if they are employed, also work significantly fewer hours a week. This fact is also likely to have a significant impact on income levels and subsequent pension entitlements, and thus on the extent of social inequality between genders and educational groups. In order to adequately represent and not underestimate these social differences, we decided to focus on the working-hour adjusted estimates. Following this approach, we found that the gender gap in WLE to the disadvantage of women remained largely constant over time, while the results of the working hour-unadjusted WLE indicated that it had narrowed significantly. This illustrates that results on the temporal development of gender inequality in WLE also depend on whether the number of working hours is taken into account or not.

While it has been shown that the extent of WLE increases with socioeconomic status,35–37 little is yet known about how these social inequalities have evolved over time. The few available studies suggest that the trends in WLE were similar across social classes and educational groups.30 33 Differing from this, we found a significantly smaller increase in WLE in the lowest educational group. Thus, we determined both, an overall increase of educational attainment as well as widening educational inequalities in WLE between those with the highest and lowest level of education.

Temporal development of HWLE and UHWLE

Along with increasing WLE, we found that also HWLE rose over time. However, no concomitant improvement in the proportion of employed persons with good SRH was observed. This might indicate that the increase in HWLE was mainly driven by the increase in employment rates. Parker et al10 summarised that the length of HWLE at age 50 was predominantly less than 10 years. Consistent with this, at this age, our estimates showed a maximum of further 10.9 years expected to be spent in good health at work. Hence, the possible maximum number of 15 years until official retirement was far from being reached. However, we calculated HWLE as full-time equivalents, which naturally are smaller compared with the unadjusted estimates.

Moreover, similar to Heller et al,12 we found that not only healthy, but also unhealthy WLE (UHWLE) increased over time. This finding also corresponds to a previous study that demonstrated that increases in the total number of working years were accompanied by increases in the number of years spent working with a chronic disease.29 As both HWLE/UHWLE increased, the health ratio remained largely stable over time. However, among the non-working population we found that the proportion of persons with good SRH declined over time. This could be a consequence of the labour market reforms, which have led to early retirement becoming increasingly unattractive. As a result, the share of those who leave the labour force not voluntarily but for health reasons might have increased. The diminishing reservoir of healthy individuals in the non-working population suggests that it will become increasingly difficult in the future to further increase the proportion of HWLE.

Temporal development of educational inequalities in HWLE

Consistent with the study by Parker et al,38 we found that HWLE rose with levels of education. This finding also fits with studies that have shown that occupational disabilities as well as unemployment and early retirement significantly decreased with increasing levels of education.39 40 The expansion of higher educational attainment that has taken place in recent decades (figure 2A) can, therefore, be seen as an important driver of the increase in HWLE in the general population. Our study found that the differences in HWLE and WLE between CASMIN education levels 4 and 5 decreased over time. This suggests that it has become increasingly irrelevant for the extent of (H)WLE whether one has obtained a university degree (CASMIN level 5) or a (vocational) maturity certificate or lower tertiary education (both CASMIN level 4). At the same time, we found that education-related differences in HLWE persisted and even widened between the highest and lowest educational groups. Moreover, at age 60 we found that the proportion of working life spent in good SRH (health ratio) in men only increased in the highest educational group. The most negative trend in the HWLE is exhibited by individuals with no or only basic school-leaving qualifications and no further vocational training. The proportion of these people has decreased continuously over the past decades, which can clearly be seen as a positive development from a public health perspective. At the same time, however, our results indicate that the disadvantages in terms of labour force participation and SRH has increased among this subgroup. To prolong healthy working lives in educationally disadvantaged groups, a combined strategy from different policy areas is needed, including education-related measures aiming to promote the further qualification of employees as well as workplace health promotion measures to reduce the risk of work-related early retirement. Moreover, labour market policy measures aimed at reducing the high unemployment rates among low-educated workers are also relevant. Since particularly men of non-German origin are over-represented in the lowest educational group (table 1), extending HWLE is also an issue of migrants' vocational education. In principle, the question arises as to whether an extension of the working life can be realised for all employees, especially as full-time employment. For physically demanding work, the appropriate amount of working hours and the timing of retirement might be adjusted to reduce the risk of early retirement for health reasons, potentially increasing social and health inequalities among older employees.

Strengths and limitations

This study provides, to the best of our knowledge, the first estimates of time trends in HWLE for Germany stratified by educational groups. The data were drawn from the German Socio-Economic Panel (GSOEP), the largest and most comprehensive household survey in Germany that claims to be representative of the German population. As in other survey samples, the proportion of individuals in good SRH is likely to be overestimated since institutionalised individuals and those who could not participate for health reasons did not participate in the survey. Therefore, the proportion of individuals with good SRH in the general population may be lower than determined in this study, which may have led to an overestimation of the HWLE and an underestimation of UHWLE. Similarly, other population groups with poorer SRH, such as the non-employed, could also be under-reported.12 In addition, it should be considered that the meaning of SRH may have changed over time. For example, Galenkamp et al41 found that between 1992 and 2009 poor SRH was determined less by chronic diseases and more by severe disability. The authors suggest that in recent decades there has been a shift in the factors considered important for SRH, with functional health gaining in importance. Finally, we used the Sullivan’s method, which can be recommended for its simplicity, relative accuracy and ease of interpretation.31 The Sullivan’s method has the advantage that it has much lower data requirements and is less susceptible to low case numbers. As has been shown by Mathers and Robine,42 results obtained with the Sullivan’s and the multistate method, respectively, are very similar as long as changes in prevalence rates are gradual and not abrupt. Since we generally did not find such sudden changes in our analyses, our results could be considered valid. However, in some cases, we found fluctuations between time points as well as decreasing proportions of individuals reporting good SRH, especially among women with the lowest educational level. It should, therefore, be noted that this partial violation of this assumption might have affected the validity of results. Finally, it should be noted that the Sullivan’s method, as a period measure, only allow valid trends for the future under the assumption that the underlying influencing factors do not change over time.


For both genders, we found evidence for increasing HWLE but also UHWLE over time. Moreover, we determined substantial educational inequalities in the length of HWLE that even widened over time. Measures aiming to increase HWLE should therefore focus on educationally disadvantaged groups.

Data availability statement

Data are available on reasonable request. The raw data were drawn from the German Socio-Economic Panel Study. The datasets used are available from the corresponding author on reasonable request. German data privacy laws necessitate that all users sign a data user contract with DIW Berlin.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants but ethics board approval was not required, because we only conducted analyses of completely anonymised GSOEP-datasets. Participants gave informed consent to participate in the study before taking part.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors SS and JT developed the idea and research questions of the study. SS analysed the data and wrote the first draft of the manuscript. JB, JE, SG, and JT contributed to the conception and discussion of the study and reviewed the work critically. All authors read and approved the final version of the manuscript. SS takes full responsibility for the work and the decision to publish it.

  • Funding German Research Foundation, grant number TE 1395/1–1.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.