Is retirement bliss? What international comparisons can teach us about individual and population ageingAxel Börsch‐SupanDirector, Munich Center for the Economics of Aging (MEA) of theMax‐Planck‐Institute for Social Law and Social PolicyCoordinator of SHARECEPRA Lecture, Lugano, 01 Octover 2013

Background

Population ageing is one of the challenges of the 21st century affecting: Pensions, health and long‐term careEconomic growth and living standardsSocial (esp. intergenerational) cohesion

Financial, debt and economic crisis has made matters worse

Genuine EU challenge, not only member states 

Requires monitoring and benchmarking

International comparisons are eye‐opener….

…and age 50+ shows accumulation of welfare state interventions over the life‐course: health, wealth, and social networks: a magnifying glass 2

3

0

5

10

15

20

25

30

2007 2010 2015 2020 2025 2030 2035 2040 2045 2050

Austria

Belgium

Canada

CzechRepublic

Denmark

Estonia

Finland

France

Germany

Greece

Hungary

Ireland

Italy

Luxembourg

Netherlands

Norw ay

Poland

Portugal

SlovakRepublic

Slovenia

Spain

Sw eden

Sw itzerland

UnitedKingdom

UnitedStates

EU27

MAX-PLANCK-INSTITUT FÜR SOZIALRECHT UND SOZIALPOLITIK

4

2. Pension costs and public debt

MAX-PLANCK-INSTITUT FÜR SOZIALRECHT UND SOZIALPOLITIK

4

Explicit and implicit debt[%GDP, 2013Q1]

1608886

12810049130

477238816586 125 18

Benchmarking economics: financial distress

5

Cavasso, Weber

Health: poor self‐rated health

6

Lindholm‐Eriksen, Vestergaard, Andersen‐Ranberg

Benchmarking health: poor self‐rated health

7

Stoeckel, Litwin

Benchmarking social environment: social network composition

for the 21st century‘s ageing societies

• Will we resolve the health care dilemma?• Will we be able to adapt the active part of our life courses to

the new time frame?• Will we be able to adapt the retirement part to the new time 

frame?• Will we able to maintain intergenerational cohesion when

ressource conflicts between generations become clearer?

Can evidence from comparing „younger“ and „older“ countries help in designing good long‐run health care, pension, economicand social policies?

(Hendrik Jürges with SHARE data)

R2 = 0,1007

10

20

30

40

50

60

70

0 5 10 15

SpainNetherlands

Unemployment rate

Shar

e of

60-

64 y

ear o

ldm

ales

alre

ady

retir

ed

Japan

USA Sweden

FranceBelgium

Italy

CanadaGermany

UK

Share of early retireesamong males 60-64 (in %)

Unemployment rate (in %)

„Place madefree bythe old“

„Lack of place forfor the young“

„Lump of labor fallacy“(Börsch-Supan with OECD data)

(Rohwedder and Willis 2010 with SHARE data)

Figure 11: Relative Generosity to the Elderly vs. the Young (Expenditure per capita devoted to the elderly versus per capita spending devoted to the young, Euro PPP)

EU90

DK90

DE90

GR90

ES90

FR90

IE90

IT90

NL90AT90

PT90

UK90

EU95

DK95

DE95

GR95

ES95

FR95

IE95

IT95

NL95AT95

PT95

FI95SW95

UK95

EU00

BE00

DK00

DE00

GR00ES00

FR00

IE00

IT00

NL00

AT00

PT00

FI00SW00

UK00

EU03

BE03

DK03

DE03

GR03ES03

FR03IE03

IT03

NL03

AT03

PT03

FI03SW03

UK03

050

010

0015

0020

00yo

ung

1000 2000 3000 4000 5000old

IT

DK

FR

EUDE

Per capita spending on old

Per c

apita

spen

ding

on y

oung

(Börsch-Supan/Reil-Held with ESS/SHARE)

• Different languages and interpretations• Different histories and institutions• Different methods

Methodological challenges

1. Distinguish cultural and methodological differencesfrom genuine policy effects:

2. Causal attribution in macro and micro data:• Usually many other influential variables• Often simultaneously determined, even across domains• Gold standard: Randomly assigned controlled trials• Usually not an option for population‐based policies• Exploit „historical experiments“• Need longitudinal data („before“ and „after“) • Regression discontinuity design, time as instrument

14

► Needto understand the interactions across domains: betweenhealth,labour force participation,institutional conditions,and socialenvironment

► Need cross-national variation in policies,histories,cultures toidentify causes and effects of welfare state interventions„instruments“

► Need longitudinal data for „policy experiments“ and to observeageing as aprocess,notastate

SocialLiving arrangements, partnership, family, social networks, social support 

HealthPhysical and mental,

health care, disability, morbidity, mortality  

EconomicIncome security, personal wealth, education

dynamic

longitudinal

Context

SocialLiving arrangements, partnership, family, social networks, social support 

HealthPhysical and mental,

health care, disability, morbidity, mortality  

EconomicIncome security, personal wealth, education

dynamic

longitudinal

Context

PLCZ

IEEE

PTSI

HU

SE

DK

DE

CHFR

SP IT

GR

BENL

LUATWave 4 (2010):

plus EE, LU, HU, PT

Wave 1 (2004):

11 countries: NL, DE, AT, DK, BE, FR, CH, SP, IT, GR, SE (+UK)

Waves 2 and 3 (2006 and 08):

plus CZ, PL, IE, IL: 15 countries

IL

UK

KoreaJapanChina

India

150,000 interviews in 19 countries

Mexico, Brazil, Argentina

International laboratory

Wave 5 (2013): plus LUX & CAT

Different languages

Different institutions

Different interpretations

Different methods

Main design challenge

Distinguish methodological effectsfrom genuine policy effects: 

Ex ante/ex postharmonizationEx ante/ex postharmonization

17

Objective measures of health help distinguishing actual differences in health from different response styles to extract genuine policy effects

Source: Jürges, 2006

Culture, language and interpretations

www.share-project.org

Theme 1: Effectiveness ofhealth care spending

(Hendrik Jürges with SHARE data)

Theme 2: Old-age employmentPension policy changes and reactions (Börsch-Supan/Schnabel)

Fig. IV-1: Average retirement ages - West German men

19601961

1962

1963

19641965

1966196719681969197019711972

1973

1975

1976

1977

1978

19791980

198119821983

198419851986198719881989

1990199119921995

Reform 1972

58

59

60

61

62

63

1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Year

56

57

58

59

60

61

62

63

1965 1970 1975 1980 1985 1990 1995-1

1

3

5

7

9

11

13

Retirement AgeUnemployment RateUnempl.R.(50-55)

Exploit specific „historical experiments“ to show the power of economic incentives

(Börsch-Supan / Schnabel)

Labor Force Participation of youth, young and elderly males

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

year

parti

cipa

tion

rate

lfpm25-54

lfpm55-59

lfpm20-24

lfpm60-64

lfpm15-19

Source: German Mikrozensus

Shocks to the system: 1972, 1984 and 1997

Theme 2: Old-age employmentPension policy changes and reactions (Börsch-Supan/Schnabel)

Exploit specific „historical experiments“ :

(Börsch-Supan / Schnabel)

Theme 3: Is retirement really bliss? Mental retirement : early retirement and cognition(Rohwedder and Willis 2010)

(Rohwedder and Willis 2010 with SHARE data)

Life satisfaction after early retirement

-4 -3 -2 -1 0 +1 +2 +3 +4

(Börsch-Supan und Jürges, 2007, GSOEP)

Controls for selectivity effects

Relative tonormal retirement

Is retirement bliss?

24

Börsch-Supan and Jürges 2006: Life satisfaction after early retirement

Adam, Bonsang, Perelman et al. 2007: Depreciation of cognitive reserve after early retirement

Coe, Lindeboom (et al.) 2008+: Does early retirement kill?

Zweimüller et al 2010: Plant closures and mortality

Rohwedder and Willis 2010: Mental retirement

Bonsang, Perelman et al. 2010: Cognitive functioning

Coe, Gaudecker, Lindeboom & Maurer 2012: Retirement and cognition

Mazzano 2013: Early retirement and cognition

Early retirement: bliss or detriment?

Which causal mechanisms? Controversialsincecausalitycouldruninbothdirections:

earlyretirement health,cognition&well‐being health,cognition&well‐being earlyretirement

Possible causal mechanims:

income support without working more leisure time might improve well‐being and health may losepurpose inlife ‐ might decrease subjective well‐being and

mentalhealth less stimulation through work place even if hated loss of social contacts anchoring function of work which inturn

stimulate cognition

Whichroledosocialnetworksplayinthisrelationship? Socialnetworksasan“cognitive‐enrichment”environment Focusonnon‐family members inthe social network excl.helpers

25

Cognition by ageand retirement pathway

26

5

6

7

8

9

10

11

60 65 70 75 80 85

Cog

niti

on s

core

Age

Early

Normal

Poly. (Early)

Poly. (Normal)

(Börsch.Supan and Schuth, 2013)

27

0.4

0.5

0.6

0.7

0.8

0.9

1

60 65 70 75 80 85

Num

ber

of f

rien

ds a

nd c

olle

ague

s

Age

Early

Normal

Poly. (Early)

Poly. (Normal)

Social inclusion by ageand retirement pathway

(Börsch.Supan and Schuth, 2013)

© SHARE-Consortium

Social networks

Name generator

Now I am going to ask some questions about your relationships with other people. Most people discuss with others the good or bad things that happen to them, problems they are having, or important concerns they may have.

Looking back over the last 12 months, who are the people with whom you most often discussed important things? These people may include your family members, friends, neighbors, or other acquaintances.

Please refer to these people by their first names. [0 – 7]

© SHARE-Consortium

Mode = 1, Median = 2, Mean = 2.44, SD = 1.60, Range = 0-7, Skewness = 0.82

Results: Network size

© SHARE-Consortium

Who is named? (conditional)

0

10

20

30

40

50

60

70

80

90

100

Spouse

Children

Siblings

Parents

Grand

child

renForm

al help

ers

Friends

Others

In network Not in network

1. Cognition ‐> Retirement:Those with low cognition tend to retire earlier

2. Retirement ‐> Cognition:Those who retire earlier loose their cognition faster

3. Cognition ‐> Social networks :Smart and succesful individuals have more friends

4. Social networks ‐>  Cognition:Those who retire earlier loose their friends faster

31

Identification strategy 1: RET

Instrumentsforyearssinceearlyretirement: Timesinceeligibleforearlyretirement Timesinceeligiblefornormalretirement

32

Rohwedder/Willis 2010JEconPersp

First stage: instruments for early RET

33

Table 5: First stage: Time elapsed since early retirement ---------------------------------------------------------------------------- (1) (2) (3) (4) ERdist ERdist ERdist ERdist ---------------------------------------------------------------------------- LERdist 0.740*** 0.608*** 0.581*** 0.367*** (0.037) (0.044) (0.045) (0.045) LNRdist -0.506*** -0.424*** -0.391*** 0.004 (0.038) (0.046) (0.048) (0.047) Demographics No yes yes yes Health No no yes yes Country/age effects No no no yes ---------------------------------------------------------------------------- N 19944 19944 18531 18531 F 300.787 255.826 199.315 231.266 Fp 0.000 0.000 0.000 0.000 ---------------------------------------------------------------------------- Standard errors in parentheses, * p<0.10, ** p<0.05, *** p<0.01 F test of excluded instruments: F(4,19924)=310.77 F(4,19919)=51.73 F(4,18501)=46.21 F(4,18488)=41.49 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Angrist-Pischke multivariate F test of excluded instruments F(2,19924)=303.82 F(2,19919)=18.05 F(2,18501)=20.17 F(2,18488)=21.21 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Demographics: age, age squared, living together with spouse/partner, gender and years of education. Health variables: functional abilities (basic and instrumental activities of daily living, ADL and IADL), global activity limitation indicator (GALI), presence of one or more chronic illnesses, grip strength measured in kilogram Fixed effects: country and country-age interaction

First stage: instruments for normal RET

34

Table 6: First stage: Time elapsed since normal retirement ---------------------------------------------------------------------------- (1) (2) (3) (4) NRdist NRdist NRdist NRdist ---------------------------------------------------------------------------- LERdist -0.769*** -0.928*** -0.876*** -0.684*** (0.082) (0.125) (0.127) (0.142) LNRdist 1.421*** 0.964*** 0.938*** 0.504*** (0.080) (0.064) (0.065) (0.068) Demographics No yes yes yes Health No no yes yes Country/age effects No no no yes ---------------------------------------------------------------------------- N 19944 19944 18531 18531 F 762.791 743.866 588.702 596.536 Fp 0.000 0.000 0.000 0.000 ---------------------------------------------------------------------------- Standard errors in parentheses, * p<0.10, ** p<0.05, *** p<0.01 F test of excluded instruments: F(4,19924)=2192.68 F(4,19919)=79.10 F(4,18501)=70.35 F(4,18488)=20.08 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Angrist-Pischke multivariate F test of excluded instruments: F(2,19924)=586.17 F(2,19919)=10.35 F(2,18501)=11.89 F(2,18488)=26.85 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000

Second stage: RET ‐> COG

35

Table 9: Second stage IV-estimation: The effect of (early) retirement on cognition

---------------------------------------------------------------------------- (1) (2) (3) (4) cogn cogn cogn cogn ---------------------------------------------------------------------------- ERdist -0.255*** -0.218* -0.214* -0.259*** (0.023) (0.119) (0.119) (0.084) NRdist -0.166*** -0.173*** -0.180*** -0.172*** (0.009) (0.065) (0.062) (0.052) Demographics No yes yes yes Health No no yes yes Country/age effects No no no yes ---------------------------------------------------------------------------- N 20348 20348 18906 18906 F 252.401 318.501 265.404 169.087 Fp 0.000 0.000 0.000 0.000 ---------------------------------------------------------------------------- Standard errors in parentheses, * p<0.10, ** p<0.05, *** p<0.01

Identfication Strategy 2: SN

Instrumentsfornumberofnon‐familymembersinthesocialnetwork:“socialcapital”attheregionallevel NUTS‐1

1. Trustinotherpeopleamongthepopulation50 separatedbycountryandgender Higherlevelof aggregated trust‐ moresocialcontacts

2. Populationdensityatnationallevel

Higherpopulationdensity‐ morenon‐familysocialcontacts?

Higherpopulationdensitiescreatepublicdistrust andincreasesneedforprivacy Brueckner &Largey 2006,Collier1998

36

37

First stage: instruments for SN

Table 7: First stage: Size of social network (friends and (ex-)colleagues ---------------------------------------------------------------------------- (1) (2) (3) (4) sn_fc sn_fc sn_fc sn_fc ---------------------------------------------------------------------------- agg_trust_~2 0.167*** 0.166*** 0.144*** 0.138*** (0.052) (0.051) (0.054) (0.054) lpden 0.082*** 0.056*** 0.061*** 0.062*** (0.019) (0.018) (0.019) (0.019) Demographics No yes yes yes Health No no yes yes Country/age effects No no no yes ---------------------------------------------------------------------------- N 19944 19944 18531 18531 F 60.674 89.676 71.463 50.639 Fp 0.000 0.000 0.000 0.000 ---------------------------------------------------------------------------- Standard errors in parentheses, * p<0.10, ** p<0.05, *** p<0.01 F test of excluded instruments: F(4,19924)=54.06 F(4,19919)=10.44 F(4,18501)=9.19 F(4,18488)=10.00 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Angrist-Pischke multivariate F test of excluded instruments: F(2,19924)=19.35 F(2,19919)=12.58 F(2,18501)=11.32 F(2,18488)=11.11 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000 Prob>F=0.0000

Second stage: RET & SN ‐> COG

38

Table 10: Second stage IV-estimation: The effect of (early) retirement and social networks on cognition

---------------------------------------------------------------------------- (1) (2) (3) (4) cogn cogn cogn cogn ---------------------------------------------------------------------------- ERdist -0.218*** -0.149 -0.180* -0.185** (0.027) (0.099) (0.104) (0.088) NRdist -0.138*** -0.106 -0.136** -0.120* (0.012) (0.065) (0.064) (0.063) sn_fc 1.919*** 1.177** 1.067** 1.037** (0.473) (0.507) (0.512) (0.516) Demographics No yes yes yes Health No no yes yes Country/age effects No no no yes ---------------------------------------------------------------------------- N 19944 19944 18531 18531 F 185.946 272.813 228.672 155.855 Fp 0.000 0.000 0.000 0.000 ---------------------------------------------------------------------------- Standard errors in parentheses, * p<0.10, ** p<0.05, *** p<0.01

39(Börsch-Supan and Schuth, 2013)

Sources of variation in cognitive aging:

Second stage: RET & SN ‐> COG

40

Table 10: Second stage IV-estimation: The effect of (early) retirement and social networks on cognition

---------------------------------------------------------------------------- (1) (2) (3) (4) cogn cogn cogn cogn ---------------------------------------------------------------------------- female 0.664*** 1.420*** 1.411*** (0.194) (0.192) (0.193) age 0.167 0.130 0.162 (0.118) (0.132) (0.130) age_q -0.001** -0.001 -0.001 (0.001) (0.001) (0.001) couple 0.692*** 0.627*** 0.624*** (0.208) (0.219) (0.213) edu_years 0.161*** 0.147*** 0.148*** (0.023) (0.024) (0.023) maxgrip 0.043*** 0.043*** (0.004) (0.004) longill -0.212*** -0.210*** (0.073) (0.073) adl -0.028 -0.029 (0.044) (0.043) iadl -0.310*** -0.313*** (0.044) (0.045) gali -0.117* -0.116* (0.062) (0.062) ---------------------------------------------------------------------------- Standard errors in parentheses, * p<0.10, ** p<0.05, *** p<0.01 Column (4) also includes country dummies and age-interactions with country dummies

Second stage: RET & SN ‐> COG

41

Table 11-14: Second stage IV-estimations: The effect of (early) retirement

and several alternative measures of size and intensity of the social network on cognition

---------------------------------------------------------------------------- (1) (2) (3) (4) cogn cogn cogn cogn ---------------------------------------------------------------------------- ERdist -0.215*** -0.134 -0.167 -0.180* (0.030) (0.103) (0.107) (0.092) NRdist -0.139*** -0.100 -0.131** -0.120* (0.013) (0.067) (0.066) (0.065) sn_contactint 0.633*** 0.371** 0.334** 0.328* (0.165) (0.166) (0.170) (0.172) ---------------------------------------------------------------------------- ERdist -0.219*** -0.150 -0.178* -0.177* (0.028) (0.099) (0.104) (0.090) NRdist -0.135*** -0.107 -0.132** -0.121* (0.012) (0.065) (0.065) (0.063) sn_distance 0.604*** 0.400** 0.389** 0.380** (0.154) (0.175) (0.181) (0.182) ---------------------------------------------------------------------------- ERdist -0.216*** -0.150 -0.182 -0.199** (0.033) (0.108) (0.111) (0.094) NRdist -0.142*** -0.117* -0.146** -0.139** (0.013) (0.067) (0.066) (0.064) sn_fc_x_cont 0.157*** 0.090** 0.078* 0.077* (0.045) (0.043) (0.044) (0.044) ---------------------------------------------------------------------------- ERdist -0.219*** -0.183* -0.208* -0.216** (0.031) (0.105) (0.110) (0.089) NRdist -0.141*** -0.136** -0.161** -0.150** (0.012) (0.065) (0.064) (0.061) sn_fc_x_dist 0.146*** 0.091** 0.086* 0.083* (0.040) (0.042) (0.044) (0.044) ----------------------------------------------------------------------------

Evidence from international comparisons can help in designinggood long‐run health care, pension, economic and social policies, and to answer key questions such as:

• Will we resolve the health care dilemma? Health as investment

• Will we be able to adapt the active part of our life courses tothe new time frame? No crowding out

• Will we be able to adapt the retirement part to the new time frame? Retirement not bliss in all dimensions

• Will we able to maintain intergenerational cohesion whenressource conflicts between generations become clearer?