retirement patterns in europe: the effect of health agar brugiavini enrica croda franco peracchi
TRANSCRIPT
Retirement Patterns in Europe: the Effect of Health
Agar BrugiaviniEnrica Croda
Franco Peracchi
2
Motivation
Europe is aging faster than other parts of the world: – the highest proportion of population aged 65 or
over (about 15%) – The old-age dependency ratio is expected to
increase from about 27% in 2000 to 39% in 2025, and to 53% in 2050
Reforms throughout Europe
Need to know more about retirement behavior
3
This paper
This paper uses SHARE (Survey of Health Ageing and Retirement in Europe) to investigate the retirement decisions of older Europeans
In particular: – retirement (and labor force participation)
decisions may depend on a number of factors, including health
– individual factors matter, but we also hypothesize that institutions matter as well
4
Theoretical Framework Following Grossman (1972a, 1972b, 1999) and
Currie and Madrian (1999), assume individuals derive utility from consumption, leisure and also, directly from health, i.e. have the following the period utility function:
Health is valued by individuals both for its own sake and because being sick is assumed to take time away from market and non-market activities
),;,,( tttttt XLCHUU
5
Theoretical Framework Individuals maximize intertemporal Utility function:
subject to the following constraints:
t
tT
1t
Uδ1
1max
tttt
ttttt
tttttt
tttttt
STHTWL
YrATWwI
YAAMpC
VTHMHHH
)(
),;,,(
1
1
6
Theoretical Framework
The model yields a conditional labor supply function depending on (endogenous) health, and more generally a demand function for health
Empirically, health must be treated as an endogenous choice
Note that:– both equations are intrinsically dynamic– causality between health outcomes and labor
force participation can go both ways (alternative use of time)
*AT
SE
DKDE
CH
FR
SP IT
GR
BENL PL
Israel
USA (HRS)
First wave 2004/05: 28.000 individuals
Korea
Japan
China
UK (ELSA)
*PL
IR
*CZ
Data: SHARE
8
Distribution of Economically Active IndividualsMen
0.2
.4.6
.81
SE DK DE NL BE FR CH AT ES IT GR
Men working
50-54 55-59 60-64 65+
9
Economic Activity and Physical Health
0.2
.4.6
SE DK DE NL BE FR CH AT ES IT GR
Economic activity of 'functioning' respondents
worker retired but work retired allother
10
Generosity of Pension Systems
Measure of Social Security and Pension Wealth (SSW): – SSW defined as present discounted value of
expected future benefits from social security and pensions, discounted by both a given interest rate and the conditional survival probability
– In the absence of longitudinal data, one cannot measure dynamic incentives of the welfare system
11
Distribution of Social Security Wealth Median - by age and gender
120
140
160
180
200
med
ian
SS
W (
'00
0s)
50 52 54 56 58 60 62 64 66 68 70age of respondent
men women
by age and genderSocial Security Wealth
12
Social Security Wealth and Household IncomeMedian
50,0
0010
0000
1500
0020
0000
SE DK DE BE FR CH AT ES IT GR
by countrySocial Security Wealth and Household Income
median SSW median hhld income
13
Social Security WealthMedian - by activity status
050
,000
1000
0015
0000
2000
00
SE DK DE BE FR CH AT ES IT GR
by country and activity status Median Social Security Wealth
pensioners workers
14
Econometric Evidence
Probability [being Retired] Dependent variable: – Retired indicator = 1 if (self-reported as) currently
retired
0 otherwise– Sample: Workers and Pensioners - aged 50-70
about 13,000 observations
IV-probit
15
Econometric Evidence
Basic regressors common to both models– indicators for gender, marital status, indicators for age 60
and age 65, age, age squared, years of schooling– With country dummies (Germany dummy omitted)
Measure of Social Security and Pension Wealth: SSWREL: SSW/total household income
IADL-Index for health: cumulative number of failures in
instrumental activities and activities of daily living
16
Econometric Evidence
Instrumental variables
– Social Security Wealth (SSW)
Instrumented with occupational indicators
– IADL – index
Retrospective questions (“ever smoked”, “ever been depressed”, age of parents at death or if parent survived a target age), material inputs for health (vigorous physical activity) plus subjective survival probability
17
Table 8. Probit Estimates - Marginal Effects (1) (2)
Respondent is male -0.138 -0.150(0.014) (0.016)
Respondent is married 0.172 0.171(0.021) (0.021)
Years of schooling -0.002 -0.002(0.001) (0.001)
(Age/10) -3.038(0.414)
(Age/10) squared 0.336(0.035)
Respondent is 60 0.019 0.475(0.024) (0.022)
Respondent is 65 0.151 0.551(0.037) (0.012)
SSWrel 0.062 0.070(0.002) (0.003)
IADL-index 0.016 0.018(0.002) (0.002)
age dummies yesinteraction dummies
(country*age) yes
18
Table 8. Probit Estimates - Marginal Effects , cont (1) (2)
(0.027) (0.069)DK 0.078 0.231
(0.028) (0.053)NL -0.172 -0.162
(0.030) (0.071)BE 0.086 0.143
(0.034) (0.073)FR 0.102 0.104
(0.024) (0.056)CH -0.169 -0.175
(0.041) (0.103)AT 0.292 0.366
(0.016) (0.036)ES -0.080 0.017
(0.041) (0.089)IT 0.217 0.246
(0.020) (0.051)GR 0.192 0.355
(0.023) (0.040)age dummies yesinteraction dummies
(country*age) yes
19
Table 9. Instrumental Variable Estimates - IV Probit
Coeff. M.E.
Respondent is male -0.256 -0.097(0.087)
Respondent is married 0.761 0.291(0.200)
Years of schooling 0.002 0.001 '(.003)
(Age/10) -11.049 -4.178( 2.137)
(Age/10) squared 1.074 0.406(0.189)
Respondent is 60 0.120 0.045(0.141)
Respondent is 65 0.074 0.028(0.186)
SSWrel 0.254 0.096(0.061)
IADL-index 0.087 0.033(0.026)
20
Table 9. Instrumental Variable Estimates - IV Probit cont
Coeff. M.E.SE -0.268 -0.104
(0.129)DK 0.443 0.154
(0.164)NL -0.362 -0.141
(0.149)BE -0.071 -0.027
(0.188)FR 0.221 0.081
(0.136)CH -0.054 -0.021
(0.188)AT 0.573 0.192
(0.222)ES -0.262 -0.102
(0.195)IT 0.253 0.092
(0.171)GR 0.276 0.099
(0.231)
21
Conclusions
Social Security Wealth and health still relevant after including a whole battery of dummies; still of the expected sign with IV estimate
Institutions matter, even after controlling for other factors
Future plans: model labor supply (hours of work) and labor demand