Health consequences of higher State Pension Age in the UK

Transcription

1 Health consequences of higher State Pension Age in the UK Ludovico Carrino, Karen Glaser and Mauricio Avendano WPEG Annual Conference Sheffield University July 27-28, 2017 Ludovico Carrino 1

2 Increase in UK State Pension Age (SPA) Pension Act 1995, 2011 and 2014 Affected women born on or after April 1950 SPA increases since May 2010 Single woman: 1 year delay loss between 14,008 (if fully eligible 1 st and 2 nd tier) and 5,587 (if 1 st tier only) (Cribb et al., 2016) 2

3 RESEARCH QUESTIONS AND DATA 1. Does a policy that postpone retirement age have an impact on the health of women? 2. Is the length of the extension associated with stronger effects? Data from Understanding Society, six waves Socioeconomic, employment and health information We cannot observe women turning SPA before the reform (would require data for 2008 and earlier) Health measures: SF-12 (mental, physical), between 0 (bad) and 100 (good) Eight dimensions: physical/social function, role physical/mental, pain, mental health, energy, health perception GHQ-12 psychological status, between 0 (good) and 36 (bad) Inability to carry out normal functions, appearance of new and distressing psychological phenomena 3

4 Retirement, emotions and health Retirement can be both good and bad for health Conflicting empirical evidence Positive effect: Coe & Zamarro (2011), Eibich (2015) and Kolodziej & García-Gómez (2017) Negative effect: Bonsang et al. (2012), Behncke (2012), Mazzonna and Peracchi (2012, 2016, but job heterogeneity) Emotional reactions to reforms may matter more than the shift itself (Lazarus 1991) Shock in the (anticipated) disengagement process: lack of agency, goal obstruction, anger, worry. (De Grip et al. 2011) Heterogeneity by job/ses (Van Solinge &Henkens 17) Our focus: impact of the policy as such (reduced-form model) 4

5 8/4/2017 King s Ca Foscari College University London of Venice 5

6 Sample selection and statistics Between , we can observed women aged being either affected or unaffected by SPA-reform (due to birth-date) Younger than 59 are all affected, older than 64 are all unaffected The number of unaffected women aged 59 is small (5%) Main selection: women aged between 2009 and 2015, having worked during their life (no proxy interv.) 8407 observations. 6

7 60yo 61yo 62yo 63yo 64yo Tot Above SPA 488 1,129 1,584 1,695 1,657 6,553 Below SPA 1, ,854 Tot 1,641 1,726 1,688 1,695 1,657 8,407 Born before Apr ,255 1,489 4,774 Born on-after Apr ,295 1, ,633 Tot 1,641 1,726 1,688 1,695 1,657 8,407 8/4/2017 King s Ca Foscari College University London of Venice 7

8 Variable Whole sample By whether had SPA postponed (born after By SPA status March 1950) No Yes Above Below Mean Mean Mean Ndiff Mean Mean Ndiff Age *** Married/couple 70.2% 71% 69.2% Widowed/Divorced 25.3% 25.1% 25.5% * 25.5% 24.2% Single 4.5% 3.9% 5.3% 3.9% 6.3% Number of children GCSE or higher ed. 78.9% 76.1% 84.2% 77.8% 85.9% Employed 41% 35.6% 48.2% ** 36.2% 58.2% *** Unemployed 1% 0.3% 1.8% 0.3% 3% *** Retired 52.1% 59.6% 41.9% ** 59.4% 25.6% *** Sick or caregivers 5.9% 4.4% 7.8% 3.9% 12.6% *** Routine 38.9% 41.6% 35.4% 40.2% 34.6% Intermediate 28.3% 28.4% 28% 28.6% 26.9% * Higher occupation % 36.6% 31.2% 38.6% SF-12 physical * /4/2017 King s Ca Foscari College University London of Venice 8

9 Meryl Streep Born in 1949 When turning 60 (in July 2009) she is ABOVE SPA and can retire Isabelle Huppert Born in 1953 When turning 60 (2013) she is BELOW SPA and cannot retire. She can retire when turning 63 (2016) Is Isabelle s mental health at 60 different from Meryl s (at 60) because of her SPA status accounting for their characteristics and for the fact that they turn 60yo in different years? 8/4/2017 King s Ca Foscari College University London of Venice 9

10 First stage : effect on retirement rates We look at age groups in different (quarter) years Being under SPA significantly decreases retirement rates (empirical evidence in Cribbs et al., 2016) Also: working rates increase but do not fully compensate higher rates of sick / home-caring BELOW SPA 10

11 Descriptives: effect of working rates 11

12 Descriptives: sick/homecaring rates 12

13 Descriptives: effect on SF-12 mental Also: Detrimental effect on GHQ index No effect for SF- 12 physical health index BELOW SPA 13

14 Descriptives: GHQ depression index 14

15 Descriptives: SF-12 physical component 15

16 Estimation strategy/1 Women of same age may have different SPA status, depending on date of birth and time of the interview above/below SPA at time of interview SPA extension varies from zero months to 5 years We build two individual-level variables: Being above/below SPA when interviewed. Individual & time specific. Extent of SPA extension Increases with month-of-birth only for those born after 03/

17 Estimation strategy/2 We exploit non-linearities of SPA-status, and compare the health status of women of similar age but interviewed at different times. Reduced form to evaluate policy effect. OLS Linear regression (Staubli et al.,.2013; Cribbs et al., 2016) y ict = α + βi age < SPA ict + γ i + δ c + θ t + X ict ξ + ε icta Health outcomes Age-in-quarters fixed effects Variable for month-of-birth Year-in-quarters fixed effects X controls for education, marital status, number of children(last) job-category, country Second model has shift in SPA (in months) as main indep variable Interacting main independent variables with job-category Standard errors clustered at month-of-birth or at individual level 17

18 SF-12 MENTAL score Detrimental effect of being under SPA Worse status for SF-12 (only mental) and GHQ effect of being under SPA on SF-12 mental index p-value=0.027 Being under SPA GHQ SF-12 Mental SF-12 physical 0.519** *** (0.242) (0.392) (0.591) Obs (fixed effects for age ad years, in quarters, plus controlling for month of birth and other covariates; s.e. clustered by month-of-birth) ABOVE SPA SPA status when interviewed BELOW SPA 18

19 Being under-spa interacted with job type GHQ SF-12 Mental SF-12 physical Being under SPA (for routine workers) 1.127*** ** *** Routine job (ref.) Intermediate manager *** 1.302*** 2.367*** Intermediate * being under SPA *** 1.231* 2.659** Manager * being under SPA ** ** Obs (fixed effects for age ad years, in quarters, plus controlling for month of birth and other covariates; s.e. clustered by month-of-birth) 19

20 SF-12 mental score Detrimental effect is heavier for routine workers being under SPA, SF-12 mental score & job type lower values mean worse health above SPA SPA status when interviewed below SPA routine management intermediate 8/4/2017 King s Ca Foscari College University London of Venice 20

21 Extent of SPA-shift interacted with job SPA shift in months (for routine workers) GHQ SF-12 Mental SF-12 physical 0.033*** ** *** Routine job (ref.) Intermediate manager *** 1.247*** 2.151*** Intermediate * shift in months ** 0.034* 0.067** Manager * shift in months ** *** Obs

22 SF-12 mental score In a graph (using categorical variable for monthsof-postponement in SPA) extent of SPA postponement, SF-12 mental score & job type higher values mean better health SPA postponement in months routine management intermediate 8/4/2017 King s Ca Foscari College University London of Venice 22

23 Clinical relevance GHQ caseness: 3+ (out of 12), or 13+ (out of 36), Goldberg et al., 1997; Kelly et al., SF-12 mental caseness: 45.8 (Vilagut et al., 2013) Being under SPA (for routine workers) GHQ 12 caseness GHQ 36 caseness SF-12 MCS caseness 0.055** 0.069** 0.079*** (0.028) (0.029) (0.025) Routine job (ref.) Intermediate (0.141) (0.019) (0.021) manager ** *** *** Intermediate * being under SPA Manager * being under SPA (0.121) (0.016) (0.017) ** ** (0.238) (0.032) (0.036) * (0.211) (0.030) (0.031) Obs (fixed effects for age ad years, in quarters, plus controlling for month of birth and other covariates; s.e. clustered by month-of-birth) 23

24 Robustness tests Sample selection: Narrower: 60-62yo (5055 obs.), so that for every age-group there is a control and a treated group (63-64yo are always observed above-spa) Larger: including also 59-57yo Clustering standard errors at individual level Alternative controls for age/year/cohort second-order polynomial for age in quarters and year in quarters, together with a linear control for year-and-month of birth second-order polynomial for age in quarters, fixed-effects for year in quarters, fixed effects for birth-year Dropping individuals who have not worked since 2000 Alternative measure of GHQ (12-valued scale) Instrumental variable (working = under-spa), no interactions with job-type Instrument strength F-test=21.5 Results are confirmed for GHQ (p-value 5%) and SF-12 MCS (p-value 3.8%) 24

25 Discussion The UK pension reform led to an increase in working careers but a negative effect on mental health longer postponement (born 53 or later) worsening outcomes Routine workers: mental and physical health decline Will the effect last or is it only short-lived? Is it the news of the postponement? Were women given advance notice? Cridland report: people need 10-year notice Is it because of the postponement itself? Income and poverty effect (Cribbs & Emmerson, 2017) Support may be needed for those who find it difficult to continue to work Implications for care-giving 25

26 8/4/2017 King s Ca Foscari College University London of Venice 26

27 SF-12 mental score Detrimental effect of longer SPA shift extent of SPA postponement and SF-12 mental index GHQ SF-12 Mental SF-12 physical SPA postponement in months 0 months (ref.) months (0.297) (0.572) (0.923) 6-24 months 0.589* (0.334) (0.588) (0.779) months 0.857* * (0.46) (0.778) (1.119) 36+ months 1.241** ** (0.508) (0.813) (1.377) Obs (fixed effects for age ad years, in quarters, plus controlling for month of birth and other covariates; s.e. clustered by month-of-birth) 27