1/29

THE INCREASING LONGEVITY GAPAND THE PENSION SYSTEM

Peter Haan Daniel Kemptner Holger Lüthen(FU Berlin and DIW Berlin)

6.6.2018

2/29

Table of Contents g

Introduction

Institutional background and data

Methodology

Life expectancy at age 65

Distributional implications

Conclusion

3/29

Motivation g

• Life expectancy is increasing with economic status: wage, differentmeasures of annual and lifetime income, or accumulated wealth

• Kitagawa and Hauser (1973), Lleras-Muney (2005), Chetty et al. (2016)

• Earnings-related heterogeneity in life expectancy can make apension system less progressive or even regressive

• Gustman and Steinmeier (2001), Feldstein and Liebman (2002), Breyer andHupfeld (2010), Auerbach et al. (2017)

• Importance likely increases: Increasing lifetime earnings inequality• Kopczuk, Saez, and Song (2010) for the US, Boenke, Corneo, and Lüthen

(2015) for Germany

4/29

Contribution g

• Evolution of heterogenous mortality by lifetime earnings over along period:• we focus on West German men born between 1926-1949

• Exploit universe of retirees and use pension entitlements as proxyfor lifetime earnings

• Focus on the individual and the household context: life expectancyof widows

• Distributional implications of increasing longevity gap through thepension system (Old age pensions and disability insurance)

• How do the distributional implications of the German pensionsystem change when accounting for heterogeneity in lifeexpectancy?

5/29

Table of Contents g

Introduction

Institutional background and data

Methodology

Life expectancy at age 65

Distributional implications

Conclusion

6/29

Pension insurance in Germany

• Mandatory for employees; Defined benefit system: pensionsstrongly linked to prior contributions

• Pension level depends on retirement age (early retirees getdeductions) and accumulated earnings points (EP)

• 1 EP is awarded for average contributions in a year; translates inpension of about 30e in 2016

• Accumulated EP represent ranks of lifetime earnings foremployees

7/29

Redistribution through the pension system

Two counteracting effects in the German pension system:

• Progressive component: insurance against disability⇒ disabilitypension or early retirement

• Regressive component: insurance against longevity⇒heterogeneous mortality: high earners have prolonged benefitperiod

8/29

Two different datasets g

Dataset 1: mortality (SK90, waves 1992-2015):

• stock of pensions of West German men:∼66.5 million obs with ∼3.4 million cases of death

• stock of survivor pensions of West German widows:∼29.5 million obs with ∼2 million cases of death

Dataset 2: distributional effects (VSKT, waves 2002, 2004-2015):

• biography data from the pension insurance (∼13,500 WestGerman men), monthly contributions from ages 14 to 66 andpension prospects• In the analysis we focus only on individuals with at least 30 EP ( 25

EP give similar results)

9/29

Descriptives: Observed survival rates0

.2.4

.6.8

1

65 70 75 80 85 90Age

1st 2nd 3rd 4th 5th6th 7th 8th 9th 10th

(a) By decile

0.2

.4.6

.81

65 70 75 80 85 90Age

26-28 29-31 32-34 35-3738-40 41-43 44-46 47-49

(b) By cohort group

Rat

e

10/29

Table of Contents g

Introduction

Institutional background and data

Methodology

Life expectancy at age 65

Distributional implications

Conclusion

11/29

Estimation g

• Logit:

logPr(deathitcd |survival until age t)

1− Pr(deathitcd |survival until age t)=

β0 +4∑

p=1

βptp +4∑

p=1

βpd tp + µd + ηc + νcd

• Cohorts grouped into 3-year cohorts; lifetime earnings into decilesat age 65; age: 4th order polynomial

• Mortality rates predicted for a grid of age×cohort group×decile

• Age range: 65-99

12/29

Table of Contents g

Introduction

Institutional background and data

Methodology

Life expectancy at age 65

Distributional implications

Conclusion

13/29

Life expectancy of West German men at 6514

1618

2022

26-28 29-31 32-34 35-37 38-41 41-43 44-46 47-49 Cohort group

1st 2nd 3rd 4th 5th6th 7th 8th 9th 10th

14/29

Life expectancy of widows

• So far the literature has only concentrated on the relationshipbetween earnings inequality and individual life expectancy.

• Implications for the household context are also important.

• This is particularly true for female spouses, who often have lowerlifetime earnings and pension entitlements and rely on theirhusband’s entitlements.

• We use the same method and estimate the effect of husband’spension entitlements on life expectancy of widows

15/29

Life expectancy of widows if husband dies at 6520

2224

26

26-31 32-37 38-43 44-49 Cohort group

1st 2nd 3rd 4th 5th

16/29

Table of Contents g

Introduction

Institutional background and data

Methodology

Life expectancy at age 65

Distributional implications

Conclusion

17/29

Distributional implications g

• Pension wealth:• Pension wealth is calculated as 2015 real present value of expected

pensions at age 65.

• Internal rates of return (IRR):• The IRR is the average individual interest rate that needs to be paid

on pension contributions to yield the expected stream of pensions.• The distribution of IRR provides direct information about the

distributional effects of the pension system.

• Compare pension wealth and IRR assuming homogeneous andheterogeneous mortality by deciles and cohorts

18/29

Pension wealth with heterogeneous mortality0

2000

0040

0000

6000

00

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49E

uro,

201

5 re

al v

alue

s

19/29

Difference: PW with and without differential mort.0

5000

010

0000

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49E

uro,

201

5 re

al v

alue

s

20/29

Ginis of pension wealth with and withoutdifferential mortality

Cohort 35-37 38-40 41-43 44-46 47-49

Heterogenous mortality 0.162 0.166 0.180 0.188 0.193

Homogenous mortality 0.117 0.116 0.122 0.126 0.131

Difference 0.045 0.05 0.058 0.062 0.062

21/29

IRR: Homogeneous mortality0

.01

.02

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49IR

R

22/29

IRR: Heterogeneous mortality0

.01

.02

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49IR

R

23/29

Extensions g

1. Survivor benefits• Additional returns⇒ insures living standard of survivors

• We know likelihood and level from the data

2. Mortality before 65• Extrapolation of mortality at age 65 to ages prior to 65

• Rates calibrated to meet average of official mortality statistics bycohort and sex

24/29

Pension wealth including survivor pensions0

2000

0040

0000

6000

00

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49E

uro,

201

5 re

al v

alue

s

25/29

IRR including survivor pensions0

.01

.02

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49IR

R

26/29

PW accounting for mortality before age 650

2000

0040

0000

6000

00

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49E

uro,

201

5 re

al v

alue

s

27/29

IRR accounting for mortality before age 650

.01

.02

2 4 6 8 10

Cohort 35-37

2 4 6 8 10

Cohort 38-40

2 4 6 8 10

Cohort 41-43

2 4 6 8 10

Cohort 44-46

2 4 6 8 10

Cohort 47-49IR

R

28/29

Table of Contents g

Introduction

Institutional background and data

Methodology

Life expectancy at age 65

Distributional implications

Conclusion

29/29

Conclusion g

• Longevity gap by lifetime earnings is growing

• Drivers: large increases in life expectancy for high earners versussmall increases for low earners

• Heterogeneous mortality turns otherwise progressive systemregressive⇒ regressive longevity dominates progressive disability

• Survivor pensions mitigate regressive effect but do not suppress it

• Mortality before age 65 likely to amplify regressive effect