The impact of a longer working life on health: exploiting the increase in the UK state pension age for women

Transcription

1 The impact of a longer working life on health: exploiting the increase in the UK state pension age for women David Sturrock (IFS) joint with James Banks, Jonathan Cribb and Carl Emmerson June 2017; Preliminary, do not quote without permission Funded by the Economic and Social Research Council through the More Years, Better Lives Joint Programming Initiative

2 Motivation In the UK, the female state pension age for women has been increasing and will reach 65 by This reform has increased labour supply, but what is impact of longer work on health? Health impact is clearly of direct concern for policy-makers Heterogeneous impacts of particular interest Could there be knock-on public service and public finance impacts?

3 Overview of method, data and results We exploit changes in the UK female state pension age Since 2010, female SPA has been gradually increasing from 60 to 65 (still ongoing); Has been shown to increase labour force participation of women Use an instrumental variables approach Use state pension age increase as source of exogenous variation in labour supply Data from English Longitudinal Study of Ageing Provisional findings: No overall impact of work on mental well-being, work better for those in more physical jobs Some evidence that work maintains cognitive ability Work prevents onset of moderate mobility problems, with weaker effect for those in more physical jobs

4 Related Literature Effect of work/retirement on mental health unclear Positive impact of retirement: Grotting and Lillebo (2017). Coe and Zamarro (2011) ; Gorry, Gorry and Slavov (2016) find little/no effect Some papers find that work is good for cognitive function Use it or lose it hypothesis (Rohwedder and Willis 2010). Coe and Zamarro (2011); Bingley and Martinello (2013) contest these findings Mechanisms are identified in a number of studies Eibich (2015), Zhu (2016): those in work do less exercise, smoke more Heterogeneity: suggestion that physical work is worse for health Eibich (2015); Grotting and Lillebo (2017)

5 Female State Pension Age in the UK Over the period , female State Pension Age (SPA) in UK is being raised from 60 to 65 Achieved by raising eligibility birth date by 1 month every 2 months (more slowly than prior to reform) Cribb, Emmerson and Tetlow (2016) demonstrated reforms caused increased labour supply: Employment of women at ages 60 and 61 increased by 6.3ppts

6 6th January th April th July th October th January th April th July th October th January th April th July th October th January th April th July th October th January th April th July th October th January th April 1955 Female state pension age Female State Pension Age in the UK Date of Birth Pre 1995 Pensions Act Post 1995 Pensions Act Post 2011 Pensions Act

7 Empirical Methodology We want to identify the causal impact of work on health, but likely to be endogeneity problems: Retirement decision may depend on health status: simultaneity bias We use an IV method where we use being under the state pension age as an instrument for being in work Valid so long as no direct health effects of being eligible for state pension and no cohort and time specific health changes which confound our instrument Due to the reform, individuals of the same age (in years) can be either over or under their SPA

8 Empirical Methodology We run a two-stage least squares model where our first stage is: 66 Our second stage is: 2015Q2 (1) W SPA ( age a) 1( t ) X it 1 it a it 2 it it a Q Q2 (2) H Wˆ ( age a) 1( t ) X it 1 it a it 2 it it a Q2 W it H it = in paid work = health outcome SPA it X it = under state pension age = vector of controls In a second specification, we control linearly for birth cohort to control for any (smooth) unobservable changes in health across cohorts In both cases, we are identifying a Local Average Treatment Effect (LATE) increase in the state pension age for women in the UK

9 Data Use data from English Longitudinal Study of Aging (ELSA) Panel in its 7 th wave. We use data from waves 3 to 7 (2006/ /15) Representative sample of individuals aged 55 and above We keep all women born in the financial years 1948/49 to 1954/55. We observe each woman s date of birth so know their SPA date Data on many types of health outcomes. We focus on 3 broad types: mental health and well-being cognitive function mobility and disability (difficulties with Activities of Daily Living)

10 Health outcomes Mental health and well-being: 1. CES depression score (CES-D) : number of depressive symptoms 2. Binary variable for whether scored 4 or more on CES-D 3. Mental well-being score: principal component of CES-D score, CASP-19 score and sum of number of questions indicating loneliness Cognitive Function: 4. Verbal fluency score assigned 0 to 100 based on percentiles of scores 5. Cognitive index sums score from questions on orientation in time, scores from 2 verbal learning tests, and verbal fluency percentile divided by 10 Mobility: 6. 3 binary variables for having any moderate mobility/adl/iadl problems, any severe mobility/adl/iadl problems, and any problems at all (i.e. moderate or severe) Note: More details on outcome measures available in appendix slides

11 Summary of outcome measures (59 year-old women pre-reform) Measure Scale Mean Std deviation CES-D score 0 to CES-D 4 or more 0 or PC of mental well-being N/A Cognitive index 0 to Moderate mobility problems 0 or Severe mobility problems 0 or Any mobility problems 0 or Note: This table uses pre-reform data from ELSA Wave 4 ( )

12 OLS regressions of health on work status Simple OLS shows all outcomes are better for those in work Outcome Impact of being in work on health No cohort control Linear cohort control Coefficient Std error Coefficient Std error CES-D score *** (0.074) *** (0.074) CES-D 4 or more *** (0.013) *** (0.013) PC of mental well-being *** (0.044) *** (0.044) Cognitive index *** (0.198) *** (0.198) Moderate mobility problems *** (0.018) *** (0.018) Severe mobility problems *** (0.018) *** (0.017) Any mobility problems *** (0.018) *** (0.018) Note: both specifications include controls for year of age (in dummies), quarter of interview (in dummies), marital status, education level, region, quadratic in partner s age, partner s education, and dummies for partner being over SPA and aged *** denotes that the effect is significantly different from zero at the 1% level, ** at the 5% level and * at the 10% level.

13 Employment rate SPA reform has increased labour supply among affected year-olds 80% 70% Female employment rate by age and wave of ELSA 60% 50% 40% 30% 20% 10% 0% We exploit the exogenous variation in work status caused by the reform

14 First Stage Regression Increasing the State Pension Age caused an increase in the employment rate of year-olds of around 12 percentage points Effect of being under State Pension Age on being in work No cohort controls Linear cohort controls Coefficient 0.128*** 0.120*** Standard error (0.028) (0.028) Number of observations 7,302 7,302 R-squared Note: both specifications include controls for year of age (in dummies), quarter of interview (in dummies), marital status, education level, region, quadratic in partner s age, partner s education, and dummies for partner being over SPA and aged *** denotes that the effect is significantly different from zero at the 1% level, ** at the 5% level and * at the 10% level.

15 Results 1: mental well-being We find no overall effect of work on mental health and well-being Effect of being in work on health outcome Outcome CES-D score (n=7014) CES-D 4 or more (n=7014) PC of CASP, CES score, loneliness index (n=6092) No cohort controls Linear cohort controls (0.831) (0.875) (0.158) (0.168) (0.509) (0.543) Note: Estimated using two-stage least squares and include a vector of controls as described on slide 14.

16 Results 2: cognition Some evidence that cognitive ability is improved by being in work Effect of being in work on health outcome Outcome Cognitive Index (n=5651) Verbal fluency score (n=5653) No cohort controls Linear cohort controls 9.642* 10.78* (5.403) (6.328) 12.79* (7.377) (8.348) Note: Estimated using two-stage least squares and include a vector of controls as described on slide 14. Results are driven by improved performance in the verbal fluency test We find a large coefficient which is imprecisely estimated

17 Results 3: mobility Staying in work causes fewer mobility problems Individuals who stay in work are less likely to have at least one moderate mobility problem But no significant impact on severe mobility problems Results driven by individuals being less likely to report stooping, kneeling or crouching problems. Effect of being in work on health Outcome No cohort controls Linear cohort controls Any mobility problems (n=7302) Moderate mobility problems (n=7302) Severe mobility problems (n=7302) ** ** (0.230) (0.249) ** ** (0.239) (0.261) (0.203) (0.217) Note: Estimated using two-stage least squares and include a vector of controls as described on slide 14.

18 Heterogeneity by physicality of job: mental well-being Individuals are classified by whether their most recent job was mainly sedentary, mainly involved standing or was a physical or heavy manual job Those in more physical jobs have a more positive impact of work on mental health and well-being Effect of being in work on health Outcome Sedentary Additional impact for Standing Additional impact for Physical/heavy manual CES-D score (n=5620) CES-D 4 or more (n=5620) PC of CASP, CES score, loneliness index (n=4963) *** * (0.939) (0.345) (0.468) ** * (0.175) (0.059) (0.085) ** (0.563) (0.201) (0.256) Note: Coefficients are reported for the interaction of work status and physicality of last job status. 3 instruments are created by interacting whether under SPA with physicality of last job status.

19 Heterogeneity by physicality of job: cognition No significant differences in the impact of work on cognitive ability according to the physicality of last job Effect of being in work on health Outcome Sedentary Additional impact for Standing Additional impact for Physical/heavy manual Cognitive Index (n=4526) Verbal fluency score (n=4528) (10.46) (2.414) (3.275) (12.59) (2.907) (3.933) Note: Coefficients are reported for the interaction of work status and physicality of last job status. 3 instruments are created by interacting whether under SPA with physicality of last job status.

20 Heterogeneity by physicality of job: mobility Those in standing jobs have less of a positive impact of work on mobility Effect of being in work on health Outcome Any mobility problems (n=5774) Moderate mobility problems (n=5774) Severe mobility problems (n=5774) Sedentary Additional impact for Standing Additional impact for Physical/heavy manual ** (0.310) (0.123) (0.118) *** 0.229* (0.327) (0.130) (0.154) (0.245) (0.097) (0.118) Note: Coefficients are reported for the interaction of work status and physicality of last job status. 3 instruments are created by interacting whether under SPA with physicality of last job status.

21 Potential mechanism: exercise No significant impact of likelihood of being engaged in exercise at different levels of intensity and frequency Effect of being in work on exercise Outcome None or low exercise (n=7302) Medium exercise (n=7302) High exercise (n=7302) No cohort controls Linear cohort controls (0.200) (0.217) (0.243) (0.263) (0.208) (0.221) Note: Estimated using two-stage least squares and include a vector of controls as described on slide 14.

22 Provisional conclusions We find that longer work improves some aspects of health but not others and that effects differ depending on the type of work Not found evidence of an impact on mental well-being though some evidence that longer work has a more positive effect for those in more physical jobs. Evidence that work maintains cognitive ability Results indicate that work reduces likelihood that have mobility problems, though smaller impact for physical jobs These results highlight potential importance of looking at how the effect of longer working differs by job type and the way in which reforms may have consequences for particular aspects of health

23 Next steps Explore further health outcomes, e.g. walking speed Investigate other mechanisms through which work could affect health: Quantity/quality of sleep Social activities undertaken/ how often see friends Examine further dimensions of heterogeneity: Marital status Attitude towards job Occupational classification of job Consider alternative measures of work e.g. part-time/full-time, number of hours worked

24 Appendix slides

25 Full details of CES-D Now think about the past week and the feelings you have experienced. Please tell me if each of the following was true for you much of the time during the past week: [Yes/No for each] 1. You felt depressed 2. You felt that everything you did was an effort 3. Your sleep restless 4. You were happy 5. You felt lonely 6. You enjoyed your life 7. You felt sad 8. You could not get going

26 Full details of CASP-19 Individual is presented with following 19 statements and asked to choose between often, sometimes, not often or never for each: 1. My age prevents me from doing the things I would like to 2. I feel that what happens to me is out of my control 3. I feel free to plan for the future 4. I feel left out of things 5. I can do the things that I want to do 6. Family responsibilities prevent me from doing what I want to do 7. I feel that I can please myself what I do I can do the things that I want to do 8. My health stops me from doing things I want to do 9. Shortage of money stops me from doing things I want to do 10. I look forward to each day 11. I feel that my life has meaning 12. I enjoy the things that I do 13. I enjoy being in the company of others 14. On balance, I look back on my life with a sense of happiness 15. I feel full of energy these days 16. I choose to do things that I have never done before 17. I feel satisfied with the way my life has turned out 18. I feel that life is full of opportunities 19. I feel that the future looks good for me

27 Full details of loneliness questions Individual is presented with following 5 questions and asked to select one of hardly ever or never, some of the time, often : 1. How often do you feel you lack companionship? 2. How often do you feel left out? 3. How often do you feel isolated from others? 4. How often do you feel in tune with the people around you? 5. How often do you feel lonely?

28 Full set of mobility Qs Individual is asked which, if any, of the following 10 activities they have problems with: 1. Climbing several flights of stairs without resting 2. Stooping, kneeling or crouching 3. Pushing and pulling large objects 4. Walking 100yds 5. Sit for 2 hours 6. Get up from a chair 7. Climb a flight of stairs without resting 8. Reach above shoulder level 9. Lift more than 10 lb 10. Pick up a 5p coin from a table Problems with any of 1 to 3 are deemed moderate and 4 to 7 are deemed severe (classification as per Banks et al., 2016)

29 Characteristics of women aged 59 Characteristic Percentage of women aged 59 In work 55% Married 72% Have a working partner 63% Own home outright 56% Paying off mortgage on home 31% No educational qualifications 36% O-Level or equivalent qualifications 29% A-level or higher qualification 35% Characteristic Percentage of women aged 59 and in employment Sedentary occupation 41% Standing occupation 37% Physical or heavy manual occupation 22%

30 Characteristics of women in work (pre-reform) Characteristic Proportion of 58 and 59- year-old women in work Proportion of 60 and 61- year-old women in work Married 71% 71% Married and partner in work 59% 54% Full time work 37% 26% Median net annual full time earnings 15,484 16,243 Works in a sedentary job 41% 41% Works in a standing job 36% 39% Works in a physical job 22% 19% Works in public sector 46% 53% No educational qualifications 33% 34% Has O-level or equivalent 30% 29% Has A-level or equivalent 37% 36% Home owner 91% 91%