Long-term care reform and the labor supply of household members Evidence from a quasi-experiment

Transcription

1 Long-term care reform and the labor supply of household members Evidence from a quasi-experiment Johannes Geyer (DIW) Thorben Korfhage (RWI) 9 th European Workshop on Labour, Health and Education under Demographic Change, Vienna 17Sep2015

2 Outline 1 Motivation 2 Data & Model 3 Results 4 Discussion Long-term care reform 2 / 33

3 Motivation I Demographic ageing The risk of long term care dependency increases with age Prevalence rates The share of the oldest old (80+) will increase from 3.8 Mio. in 2008 to 11.6 Mio. in Prevalence rates will double (Schulz, 2008). At the same time: working age population will decrease Ratio of contributors and beneciaries will decline The number of potential caregivers will decline Financial pressure on public health and long-term care insurance Long-term care reform 3 / 33

4 Motivation II Long-term care insurance (LTCI) In 1995 Germany introduced the LTCI providing incentives for family carers to organize care at home Family care is the main pillar of the long term care system. About 70% of all LTCI beneciaries receive care at home Shares ) Budgetary perspective: family care is a cost-saving alternative to formal care at home or to stationary care Long-term care reform 4 / 33

5 Research question Labor supply and long-term care provision Support for family carers is supposed to strengthen (not to replace) family care; to enable families to organize care at home (Lilly et al., 2010) The trade o between supporting family carers and saving money on formal care depends on opportunity costs (reduced labor supply of family carers) RQ: What is the impact of the introduction of the LTCI in 1995 on the labor supply of co-residential carers? Long-term care reform 5 / 33

6 Previous literature Existing literature primarily focuses on general relation between care and labor supply: Usually no or negative eects of care responsibilities on labor supply (e.g., Ettner, 1996; Johnson and Lo Sasso, 2000; Carmichael and Charles, 2003; Heitmueller, 2007). The impact of the institutional setting is less well analyzed Europe: Heger (2014); Geyer and Korfhage (2015); Schaer (2015) US: Skira (2015) Japan: Sugawara and Nakamura (2014) Long-term care reform 6 / 33

7 Long-term care reform I Before 1995 No uniform long-term care system; system relied more than today on family care Support for dependent people by the health insurance and (means tested) social assistance Private long-term care insurance since mid-80s, very low coverage rates Long-term care reform 7 / 33

8 Long-term care reform II Introduction of LTCI in 1995 Compulsory long-term care insurance for the whole population Eligibility depends on care needs and is not means tested. Focus on permanent (at least six months) and substantial physical impairments: At least two activities of daily living (ADL), e.g. mobility, dressing, eating, continence, toileting and washing and bathing. At least one instrumental activity of daily living (IADL), e.g. obtaining food, cooking, laundering, housecleaning, managing one's medications, phone use. Long-term care reform 8 / 33

9 Long-term care reform III LTCI Depending on the level of impairments, three care levels are distinguished Benets: benets in cash, benets in kind, combination of both: Care level in cash in kind mininum per day (share '96) (rel. to wages) I (40%) 205 (10%) min II (43%) 410 (21%) min III (17%) 665 (33%) min Cash benets neither earmarked nor is spending monitored LTCI does not provide full coverage Long-term care reform 9 / 33

10 Estimation strategy & Identication Idea: identify households aected by the reform and comparable households that were not directly aected We use multi-person households with a person in need of care and estimate their labor supply decision DiD estimation (employment decision / working hours): y it = α + β(tr i Post t ) + λtr i + δpost t + X it γ + e it (1) Important assumptions: Treatment did not aect the control group Treatment did not lead to self-selection into the treatment group Common trend Long-term care reform 10 / 33

11 Data I LTC(I) & SOEP SOEP contained a (HH) question on care need already before 1995: Does someone in your household need help on a constant basis due to age, sickness or medical treatment? The question is not related to eligibility for benets from the LTCI No incentive to self-select into the treatment group after 1995 Question changed between 1990 (permanent need of care) and 1991 (permanent need of help) Long-term care reform 11 / 33

12 Data II Sample Exclude 1995, benets in kind not available until July 1995 Exclude sub-sample D (rst surveyed in 1995) Exclude East Germany due to the transition period Exclude people who do not participate on the labor market (e.g. permanently sick or disabled) We keep pensioners who retired in t-1, others are dropped Focus on people aged 45 to 65 in multi-person households Small sample: Pool data two years before and after treatment Large sample: Pool data from 1991 until 2007 Long-term care reform 12 / 33

13 Data III Treatment & Control Group Treatment group: all individuals in the sample who belong to a HH with a person in need of care. 89 males (88 females) before 1995 and 71 (92) after 1995 Control group: all individuals in the sample who do not belong to a HH with a person in need of care 2,455 males (2,246 females) before 1995 and 2,301 (2,217) after 1995 Long-term care reform 13 / 33

14 Employment Rates (Female) Employment Rate (%) Sample Reform Period Treatment Group Control Group Year Descriptives Women

15 Employment Rates (Male) Employment Rate (%) Sample Reform Period Treatment Group Control Group Year Descriptives Men

16 Table : Regression on Female Employment (1) (2) (7) Post (0.013) (0.012) (0.011) Tr (0.060) (0.057) (0.096) Post95 Tr (0.071) (0.065) (0.062) Age (0.020) (0.036) Controls No No Yes Obs Note: Values denote estimated coecients. Standard errors are clustered on household level and reported in parentheses. Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEP, own calculation. Long-term care reform 16 / 33

17 Table : Regression on Male Employment (1) (2) (7) Post (0.013) (0.011) (0.010) Tr (0.060) (0.055) (0.104) Post95 Tr (0.082) (0.074) (0.068) Age (0.022) (0.034) Controls No No Yes Obs Note: Values denote estimated coecients. Standard errors are clustered on household level and reported in parentheses. Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEP, own calculation. Long-term care reform 17 / 33

18 Robustness checks Subsample analysis: drop all pensioners; no retirees Placebo regressions: with & without covariates covering the period between 1986 and 2007 (2008 was the rst large reform of the LTCI since its introduction). placebo Time span: Using the large sample extended time span Selection: Control for constant HH composition constant hh-composition Matching matching Long-term care reform 18 / 33

19 Discussion I Labor market eects of the introduction of the LTCI dicult to anticipate ex ante Benets in cash increase household's non-working income while benets in kind provide a substitute for informal care The insurance does not fully cover the risk of LTC; a certain amount of informal care has always to be provided We know from survey data that care recipients prefer family members as caregivers. In our sample: only multi-person households (co-residential carers), most likely cash benet recipients Long-term care reform 19 / 33

20 Discussion II Results suggest that men and women reacted dierently to the introduction of the LTCI Large point estimates but also large standard errors Normally we nd that female labor supply reacts more elastically to nancial incentives than male's Female employment rates and working hours were low (mid 90s), in particular if they were family carers; LTCI provided little incentives to further reduce employment. This was dierent for men. A second reason for the strong eect for men is that the reform aected incentives to retire Long-term care reform 20 / 33

21 References I Carmichael, F. and Charles, S., Benet payments, informal care and female labour supply. Applied Economics Letters, 10(7): Ettner, S. L., The opportunity costs of elder care. The Journal of Human Resources, 31(1): Geyer, J. and Korfhage, T., Long-term care insurance and carers' labor supplya structural model. Health Economics. Heger, D., Employment and well-being of Informal caregivers. Three essays on informal care, health, and education. Ph.D. thesis, Queen's University. Heitmueller, A., The chicken or the egg?: Endogeneity in labour market participation of informal carers in england. Journal of Health Economics, 26(3): Long-term care reform 21 / 33

22 References II Johnson, R. W. and Lo Sasso, A. T., The trade-o between hours of paid employment and time assistance to elderly parents at midlife. The Urban Institute, pages 140. Lilly, M. B., Laporte, A., and Coyte, P. C., Do they care too much to work? the inuence of caregiving intensity on the labour force participation of unpaid caregivers in canada. Journal of Health Economics, 29(6): Schaer, S. K., The eect of free personal care for the elderly on informal caregiving. Health economics, 24(S1): Schulz, E., Zahl der Pegefälle wird deutlich steigen. DIW Wochenbericht, 75(47): Skira, M. M., Dynamic wage and employment eects of elder parent care: Dynamic eects of caregiving. International Economic Review, 56(1):6393. Sugawara, S. and Nakamura, J., Can formal elderly care stimulate female labor supply? The Japanese experience. Journal of the Japanese and International Economies, 34(C): Long-term care reform 22 / 33

23 Benet recipients of the LTCI: Prevalence rates by gender and age (2011) Return Men Women Prevalence rate (in %) to to to to 90 Average 0 15 to to to to 90 Average Graphs by gender Age group

24 Benet recipients of the LTCI: Shares of care settings (2011) Return 47.3% 23% 29.7% Informal care at home Stationary care Informal & formal care at home

25 Working Hours (Female) Hours Worked Period with differnt care question 2 years before and after reform Treatment Group Control Group Year Long-term care reform 25 / 33

26 Working Hours (Male) Hours Worked Period with differnt care question 2 years before and after reform Treatment Group Control Group Year Long-term care reform 26 / 33

27 Return Table 2: Descriptive Statistics for Treatment and Control Group (Female) tr before 95 tr after 95 control before 95 control after 95 Employed Working hours Retired Age Age Migration background Working experience in years Years of education Health status: good very good satisfying poor very poor Married Other household income / Household size Community size: <20, , , >100, HH-member needs help with: no help or not known getting around outside the house household chores, preparing meals washing, dressing, etc getting into and out-of-bed, etc Observations Note: We performed t-tests to check whether means a significantly different before and after treatment (within

28 Return Table 1: Descriptive Statistics for Treatment and Control Group (Male) tr before 95 tr after 95 control before 95 control after 95 Employed Working hours Retired Age Age Migration background Working experience in years Years of education Health status: good very good satisfying poor very poor Married Other household income / Household size Community size: <20, , , >100, HH-member needs help with: no help or not known getting around outside the house household chores, preparing meals washing, dressing, etc getting into and out-of-bed, etc Observations Note: We performed t-tests to check whether means a significantly different before and after treatment (within treatment or control group). indicates statistical significance on the 5% level. All means are calculated

29 Return Table : Placebo Regressions on Male Employment (without covariates) Post Tr Post Tr Controls No No No No No No No No No No No No No No No Obs Note: Each model uses observations of the two years before and after the year reported in the top row. The year itself is omitted. Values denote estiamted coecients. Standard errors are clustered on household level. Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEP, own calculation. Table : Placebo Regressions on Male Employment (with covariates) Post Tr Post Tr Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Obs Note: Each model uses observations of the two years before and after the year reported in the top row. The year itself is omitted. Values denote estiamted coecients. Standard errors are clustered on household level. Due to dierent availability control-variable dier before From they include: a dummy for age 60, age 2, age, migration background, working experice in years, years of education, selfreported health status, marital status, householdsize, community size and the amount of help needed by household-memeber. Before 1993 selfreported health status and the amount of help needed by household-memeber is not included. Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEP, own calculation.

30 Return Table : Regressions on male employment with dierent time-spans before and after treatment (without covariates) Post Tr Post95 Tr Controls No No No No No No No No No No No No Obs Note: The year 1995 is omitted in all models. Values denote estiamted coecients. Standard errors are clustered on household level. Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEP, own calculation. Table : Regressions on male employment with dierent time-spans before and after treatment (with covariates) Post Tr Post95 Tr Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Obs Note: The year 1995 is omitted in all models. Values denote estiamted coecients. Standard errors are clustered on household level. Controls include a dummy for age 60, age 2, age, migration background, working experice in years, years of education, selfreported health status, marital status, householdsize, community size and the amount of help needed by household-memeber. Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEP, own calculation.

31 Return Table : Male sample without pensioners Employment Employment Hours worked Hours worked without cov. with cov. without cov. with cov. Post (0.010) (0.010) (0.499) (0.486) Tr (0.051) (0.102) (2.805) (4.908) Post95 Tr (0.071) (0.068) (3.047) (2.889) Controls No Yes No Yes Obs Obs. in Tr Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEPv29, own calculation.

32 Return Table : Male sample with constant household composition Employment Employment Hours worked Hours worked without cov. with cov. without cov. with cov. Post (0.014) (0.012) (0.653) (0.535) Tr (0.071) (0.110) (3.606) (5.097) Post95 Tr (0.091) (0.073) (3.780) (2.991) Controls No Yes No Yes Obs Obs. in Tr Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEPv29, own calculation.

33 Return Table : Regressions on matched sample (male) Employment Employment Hours worked Hours worked without cov. with cov. without cov. with cov. Post (0.027) (0.025) (1.191) (1.161) Tr (0.063) (0.109) (3.159) (5.098) Post95 Tr (0.093) (0.080) (3.947) (3.418) Controls No Yes No Yes Obs Obs. in Tr Signicance levels: p <0.10, p <0.05, p <0.01 Source: SOEPv29, own calculation.