HEALTH CAPACITY TO WORK AT OLDER AGES IN FRANCE

Transcription

1 HEALTH CAPACITY TO WORK AT OLDER AGES IN FRANCE OECD, April 2016 Didier Blanchet Eve Caroli Corinne Prost Muriel Roger

2 General context From a low point at the end of the 1990s, French LFP and ER for older workers have started re-increasing significantly Partly due to structural effects (women s LFP) But also due to the pension reforms that took place since 1993 And also to other labor market changes Progressive closure of the main early-retirement schemes Stricter regulation of unemployment benefits (e.g. active job search now required for older unemployed) Or maybe a more global evolution of attitudes toward older workers

3 General context Employment rates

4 General context This process is expected to go on and contribute significantly to the stabilization of pension expenditures Last projections by the Pensions Advisory Committee (COR) anticipate the average RA to increase up to 64 years by 2040 Combined with expected drops in global replacement rates, this should even lead to slightly declining pensions/gdp ratios under median or optimistic macroeconomic scenarios

5 General context In this context, much lower need to convince that increasing RAs is a possible answer to population ageing People anticipate that they will have to work longer: until 68 according to the Drees barometer! The age beyond which they find it unacceptable to be forced to work longer remains much lower (64) but not incompatible with expected projections.

6 General context However, knowing whether health or other employment problems are a barrier to increasing RAs remains an interesting issue If the answer turns out to be positive, this may suggest that RA policies could be re-oriented in a less restrictive way In case of a negative answer, it suggests that there exists an additional margin of manoeuver at the RA level, allowing for instance for less severe policies in terms of pension levels Additional Work Capacity: measure of the distance between current retirement ages and what we call the health barrier, i.e. the age at which health prevents people from working longer.

7 Aim of the paper The aim of the paper is to discuss some of the tools proposed in the literature to evaluate this potential health barrier I. Employment rates and mortality: the Milligan and Wise approach (2012) II. Employment rates and self-assessed health implementing the MW approach III. Simulating employment rates at older ages: the Cutler, Meara and Richards-Shubik approach (2013) IV. Conclusion

8 I The mortality approach The Milligan-Wise methodology consists in using mortality as a proxy of health and then compute a measure of additional work capacity for older workers (aged 55-69), based on the gap in employment rates across time for given mortality rates. According to the MW definition of additional work capacity, strong increase from the mid 70s to the mid 2000s. The trend has started to reverse since then.

9 Employment versus mortality - Men Mortality data and French Labor Force Survey, INSEE

10 Mortality-based additional work capacity Men 55 to 69 Mortality data and French Labor Force Survey, INSEE

11 Additional work capacity for men in 2012, using 1977 to 2011 as reference years Mortality data and French Labor Force Survey, INSEE

12 Heterogeneity in mortality-based work capacity: managers and professionals vs blue-collars life tables disaggregated according to socio-economic status are released by Insee for aggregate time periods data: civil registration data (exact age at death) are matched with successive censuses for a 1% representative longitudinal sample of the French population (panel EDP échantillon démographique permanent) disaggregation according to occupations. Unfortunately no disaggregation according to education levels yet.

13 Heterogeneity in mortality-based work capacity: managers and professionals vs blue collars In , life expectancies at age 55 were 23.3 for managers/profesionals and 19.2 for blue collars, i.e. a 4.1 year gap. In , they were respectively 28.4 and 23.5, i.e. a 4.9 year gap. Incorporating this into the MW framework? We do not make use of socio-economic mortality differentials for remote time periods : average mortalityemployment relationships of past periods are used as benchmarks and applied to mortality rates currently observed in the two social groups we consider.

14 Employment versus mortality - managers and professionals / blue collars

15 Additional work capacities by social group in 2004, using 1977 to 2003 as reference years. Men 55-69

16 Additional work capacities by social group in 2004, using 1977 to 2003 as reference years. Men 55+

17 MW : The limits MW gives, at most, an extreme upper bound for the «health barrier» Even like that, some problems remain Objection 1. Conventional character of the reference period The «health barrier» can be raised arbitrarily by choosing a sufficiently remote reference period (times where there was no retirement system at all?). difficult to accept : we need to consider changes in norms regarding the compatibility between work and bad health/exposure to mortality risk.

18 MW : The limits Objection 2. Mortality is not necessarily a good proxy for health Objection 3. Mortality is not a good proxy for employability Objection 4. Mortality/health may react, in turn, to changes in RA policies Next section focuses on point 2, by looking at time trends in SAH and trying to replicate MW with such indicators Point 1 will be implicitly addressed by the CM approach, and 4 discussed further in this context

19 II SAH Enquête Santé et Protection sociale (ESPS) Periodicity and sample size Yearly from 1992 to 1997 with respondents/year Every 2 years from 1998 to 2010 with respondents/year Self-assessed health = a note ranging from 0 to 10 We use note<8 as the criterion for «poor/fair health» Age gradient. Education gradient but not very large. No improvement of SAH over time.

20 Poor SAH and mortality - Men

21 Poor SAH and mortality - Women

22 Poor SAH by age group across time - Men

23 Poor SAH by age group across time - Women

24 Poor SAH by education quartile - Men

25 Poor SAH by education quartile - Women

26 Employment versus poor SAH - Men

27 Additional work capacity for men in 2010, using 1992 to 2010 as reference years

28 SAH with ESPS data: discussion Issue of potential declaration biases As health systems become more efficient, do individuals raise their health expectations? Sieurin et al. (2011): in France, male life expectancy at 50 increased by about 22% between 1990 and 2008, whereas life expectancy without functional limitations increased by only 0.6% over the same period. Life expectancy without sensory and cognitive limitations actually decreased (by 3 to 8%).

29 III The Cutler-Meara method We estimate an employment model at age based on health and demographic variables, using waves 1, 2, 4 and 5 of Share. We estimate separate models by gender We simulate the employment rates for older cohorts. The simulations are based on the coefficients of the regression model for the age group and the health and demographic characteristics of the older age groups. Simulations are made by gender but also by gender and education levels.

30 This method allows us to provide some information on the work capacity of older workers under the assumptions that: Health would be the same at older ages if people were working (i.e. being retired does not induce better health at older ages) A given illness has the same impact on work capacity at each age the ability to work at a given age does not depend on anything else but health as in the previous sections. In particular, no interaction between age, health and human capital

31 Results Large additional work capacities on average after 60 yearold. It roughly increases with the level of education. Not so large for age 55-59, especially for the least educated group. This suggests that in the year-old group, there is not much room for maneuver to increase employment rates, except in the most educated group, once taken into account their health conditions.

32 PVW synthetic health index, by age and gender. Averages over 2004, 2007, 2011 and Men Women

33 Employment rates and PVW health index Men and women aged Dependent variable: Employment rate Men Women Constant 0.567*** 0.553*** (0.052) (0.042) PVW synthetic health index 0.004*** 0.004*** (0.001) (0.000) No diploma or primary school *** *** (0.041) (0.039) Lower secondary education *** (0.060) (0.042) Higher secondary education (0.032) (0.029) High school diploma and above Ref Ref - - Observations R-squared

34 Work capacity by education Men and women 55 to 64 Age % of individuals in employment Men Women Actual Predicted Estimated WC Actual Predicted Estimated WC No diploma or primary 49.43% 58.93% 9.50% 65.99% 69.07% 3.08% school Lower secondary education 65.06% 65.67% 0.61% 82.61% 82.41% -0.20% Higher secondary 65.54% 84.71% 19.17% 70.80% 82.92% 12.12% education High school diploma and above 79.50% 92.53% 13.04% 75.70% 85.93% 10.23% Age No diploma or primary school 9.23% 58.61% 49.39% 18.47% 67.86% 49.39% Lower secondary education 18.18% 65.09% 46.91% 17.71% 82.87% 65.16% Higher secondary education 13.43% 83.37% 69.94% 19.46% 81.96% 62.51% High school diploma and above 30.10% 91.19% 61.09% 26.30% 84.75% 58.45%

35 Work capacity by education Men and women 65 to +70 Men Age Women Estimate d WC Actual Predicted Estimated WC Actual Predicted No diploma or primary school 2.42% 57.47% 55.06% 2.05% 67.67% 65.62% Lower secondary education 4.26% 65.33% 61.07% 2.00% 81.08% 79.08% Higher secondary education 1.39% 82.30% 80.92% 0.87% 80.60% 79.73% High school diploma and above 6.33% 88.11% 81.78% 2.69% 82.24% 79.55% Age +70 No diploma or primary school 0.08% 51.62% 51.53% 0.11% 61.32% 61.21% Lower secondary education 0.74% 54.56% 53.83% 0.72% 72.50% 71.78% Higher secondary education 0.71% 77.09% 76.38% 1.27% 74.77% 73.50% High school diploma and above 0.58% 83.07% 82.50% 1.07% 77.66% 76.59%

36 Conclusion and Discussion Reminder of the context: RA are currently increasing and are expected to continue doing so, in a way that, at this stage, is predicted to be large enough to ensure the equilibrium of public pension schemes. In this context, it is not obvious that increasing RA still further than expected is an absolute necessity. However, understanding whether health or other employment problems are a barrier to increasing RAs remains an interesting issue.

37 Conclusion and Discussion The results from the mortality approach predict a high level of unused work capacity. However, an increase in life expectancy does not always entail an increase in healthadjusted life expectancy. Using subjective health data, we don t find the same predictive results as for the mortality approach. This remains true after accounting for heterogeneity, at least in education. Results obtained using this education gradient also show the importance of looking beyond averages. The health barrier is not necessarily the same for everybody.

38 Conclusion and Discussion The micro approach confirms the importance of accounting for heterogeneity when studying the health barrier. It permits to uncover the health factors that most affect working capacity. It underlines that reverse effects of work on health at older age have to be taken into account when estimating the unused work capacity.

39 Conclusion and Discussion To conclude, we have estimated an upper bound on unused work capacity due to health problem but we have showed that global policies that aim at uniform increases in retirement ages ignoring health differentials raise problems of fairness. Moreover, even if assuming that we can perfectly measure the health gradient, the question remains of how far retirement age should be delayed, considering that, even a population in perfect health may choose to allocate part of additional years of life to retirement rather than to work.

40 Appendix

41 0.8 MEN Any IADL Any ADL Difficulty walking Poor SAH

42 0.8 WOMEN Any IADL Any ADL Difficulty walking Poor SAH

43 0.7 MEN Heart problems High blood pressure Stroke Diabete Lung disease Arthritis Cancer

44 0.8 WOMEN Heart problems High blood pressure Stroke Diabete Lung disease Arthritis Cancer

45 Employment rates and Poor SAH, ADL, IADL and difficulty walking Men and Women aged Dependent variable: Employment rate Men Women Constant 0.908*** 0.851*** (0.035) (0.031) Poor SAH *** *** (0.034) (0.031) Any IADL *** * (0.079) (0.057) Any ADL *** *** (0.048) (0.045) Difficulty walking ** *** (0.081) (0.070) Observations R-squared

46 Age % of individuals in employment Men Women Actual Predicted Estimated WC Actual Predicted Estimated WC No diploma or primary school 49.43% 60.20% 10.77% 65.99% 71.90% 5.90% Lower secondary education 65.06% 70.32% 5.26% 82.61% 87.21% 4.61% Higher secondary education 65.54% 86.39% 20.85% 70.80% 83.64% 12.84% High school diploma and above 79.50% 93.79% 14.29% 75.70% 88.63% 12.93% Age Men Women Actual Predicted Estimated WC Actual Predicted Estimated WC No diploma or primary school 9.23% 61.64% 52.41% 18.47% 72.11% 53.64% Lower secondary education 18.18% 70.29% 52.10% 17.71% 86.05% 68.34% Higher secondary education 13.43% 85.41% 71.97% 19.46% 83.29% 63.84% High school diploma and above 30.10% 91.77% 61.67% 26.30% 87.28% 60.98% Age Men Women Actual Predicted Estimated WC Actual Predicted Estimated WC No diploma or primary school 2.42% 61.31% 58.89% 2.05% 71.49% 69.44% Lower secondary education 4.26% 69.04% 64.79% 2.00% 80.83% 78.83% Higher secondary education 1.39% 85.03% 83.64% 0.87% 82.85% 81.98% High school diploma and above 6.33% 90.98% 84.66% 2.69% 84.63% 81.94% Age +70 Men Women Actual Predicted Estimated WC Actual Predicted Estimated WC No diploma or primary school 0.08% 48.82% 48.74% 0.11% 57.50% 57.38% Lower secondary education 0.74% 53.44% 52.70% 0.72% 71.19% 70.48% Higher secondary education 0.71% 76.98% 76.27% 1.27% 73.10% 71.82% High school diploma and above 0.58% 84.45% 83.87% 1.07% 78.76% 77.69%

47 Employment rates and most serious health conditions Men and Women aged Dependent variable: Employment rate Men Women Constant 0.899*** 0.849*** (0.037) (0.032) Heart problem (0.065) (0.083) Blood pressure (0.038) (0.034) Stroke ** *** (0.128) (0.100) Diabete *** (0.057) (0.069) Lung problem *** (0.088) (0.073) Arthritis ** ** (0.039) (0.030) Cancer *** (0.098) (0.059) Observations R-squared

48 Age % of individuals in employment Men Women Actual Predicted Estimated WC Actual Predicted Estimated WC No diploma or primary school 49.43% 59.11% 9.68% 65.99% 67.29% 1.30% Lower secondary education 65.06% 65.62% 0.56% 82.61% 81.74% -0.87% Higher secondary education 65.54% 84.57% 19.03% 70.80% 83.93% 13.13% High school diploma and above 79.50% 94.35% 14.86% 75.70% 87.12% 11.42% Age Men Women Estimated WC Actual Predicted Estimated WC Actual Predicted No diploma or primary school 9.23% 58.57% 49.34% 18.47% 65.31% 46.84% Lower secondary education 18.18% 64.58% 46.39% 17.71% 79.39% 61.68% Higher secondary education 13.43% 82.38% 68.94% 19.46% 83.05% 63.59% High school diploma and above 30.10% 91.13% 61.03% 26.30% 85.80% 59.50% Age Men Women Estimated WC Actual Predicted Estimated WC Actual Predicted No diploma or primary school 2.42% 56.12% 53.71% 2.05% 64.01% 61.97% Lower secondary education 4.26% 63.10% 58.84% 2.00% 80.10% 78.10% Higher secondary education 1.39% 81.18% 79.80% 0.87% 82.12% 81.25% High school diploma and above 6.33% 89.16% 82.83% 2.69% 84.17% 81.48% Age +70 Men Women Estimated WC Actual Predicted Estimated WC Actual Predicted No diploma or primary school 0.08% 53.20% 53.12% 0.11% 60.98% 60.87% Lower secondary education 0.74% 55.45% 54.72% 0.72% 74.84% 74.13% Higher secondary education 0.71% 78.12% 77.41% 1.27% 77.31% 76.04% High school diploma and above 0.58% 85.72% 85.15% 1.07% 81.14% 80.07%