how does delayed retirement affect mortality? - zulkarnain.pdf · and studies to date, even on...
TRANSCRIPT
Alice Zulkarnain and Matthew S. Rutledge
Center for Retirement Research at Boston College
20th Annual Meeting of the Retirement Research Consortium
Washington, DC
August 3, 2018
How Does Delayed Retirement
Affect Mortality?
Older Americans are retiring later for a
number of reasons.
• More education;
• Less physically demanding jobs;
• Improved health and longevity;
• Shift from DB to DC pensions; and
• Changing Social Security incentives.
1
This project: how does delayed retirement
affect health and mortality?
• Previous studies have explored the effects of early retirement,
and focused on specific sectors or on older birth cohorts.
• Yet, it is unclear that the relationship is symmetric: the
marginal worker deciding to retire early is likely different than
the worker deciding to delay retirement.
• This study exploits a policy from the Netherlands – the
“Doorwerkbonus” – that incentivized later retirement to study
the causal effect of delayed retirement on mortality.
2
After all, the relationship between retirement
and mortality is unclear.
• Removal of physical strain or stress, and more time for health
investments, could reduce mortality.
• Loss of network, identity, and purpose, less physical exercise,
and increased smoking and drinking could increase mortality.
• The relationship is difficult to establish since work and health
are jointly determined – e.g., healthy people with lower
mortality tend to work longer.
3
And studies to date, even on early
retirement, have provided mixed evidence.
• One recent study, focused on public sector workers, found that
early retirement decreased the five-year mortality rate by
2.6 percentage points (Bloemen, Hochguertel, and Zweerink
2017).
• While another, focused on blue-collar workers, found that
early retirement increased the risk of dying by age 67 by
2.4 percentage points (Kuhn, Wuellrich and Zweimueller
2010).
4
This study uses a Dutch policy to offer new
evidence focused on delayed retirement.
• Dutch policymakers have focused on encouraging later
retirement since the 1990s.
• For example, in the mid-1990s the option of early claiming in
DB pension plans was phased out, and since 2002, regulations
have tightened exit-pathways through DI and UI.
• The focus of this paper is yet another policy targeting later
retirement that was introduced in 2009, the “Doorwerkbonus.”
5
The “Doorwerkbonus” (DWB):
• Provides a bonus for each year worked after age 62 that
depends on age, birth cohort, and income earned.
• Different from the Delayed Retirement Credit because it is
only received while working and does not affect lifetime
benefits.
• The policy was amended in 2012 and then replaced by a
different policy in 2013.
6
This table below shows how the
“Doorwerkbonus” (DWB) is calculated.
Source: Tax and Customs Administration. 2009-2011. The Netherlands.
DWB Bonus Percentages
Year
2009 2010 2011
Birth cohort Age Bonus Age Bonus Age Bonus
1939 70 1 % 71 1 % 72 1 %
1940 69 1 70 1 71 1
1941 68 1 69 1 70 1
1942 67 1 68 1 69 1
1943 66 2 67 1 68 1
1944 65 2 66 2 67 1
1945 64 10 65 2 66 2
1946 63 7 64 10 65 2
1947 62 5 63 7 64 10
1948 - - 62 5 63 7
1949 - - - - 62 5
1950 - - - - - -
Income cap € 54,776 € 55,831 € 56,280
Income floor € 8,860 € 9,041 € 9,209
7
Men and women eligible for the DWB were
more likely to work between ages 62 and 65.
Share of Men Working
by Age and DWB Eligibility, 2008-2011
Share of Women Working
by Age and DWB Eligibility, 2008-2011
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 2008-2011.
0%
20%
40%
60%
62 63 64 65
2008 (No DWB)
2009-2011 (DWB)
0%
20%
40%
60%
62 63 64 65
2008 (No DWB)
2009-2011 (DWB)
8
Five-year Mortality Rate for Women
Ages 62-65, 1999-2011
0%
2%
4%
6%
8%
10%
1999 2001 2003 2005 2007 2009 2011
Not working
Working
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 1999-2016.
Those working have lower mortality before
and after 2009, with parallel trends before.
0%
2%
4%
6%
8%
10%
1999 2001 2003 2005 2007 2009 2011
Not working
Working
Five-year Mortality Rate for Men
Ages 62-65, 1999-2011
9
The study relies on administrative data
collected by Statistics Netherlands.
• The data contain the entire Dutch population (16.5 million in
2009) and were collected from various administrative sources
e.g., the population registry and tax office.
• The data contain information on labor, income, date of death,
demographics, and prescription drugs.
• These data are linkable by a personal identifier and cover the
period 1999-2016.
10
• 𝐷𝑖,𝑡+5 indicator for whether the person is deceased at t+5.
• 𝑊𝑜𝑟𝑘𝑖𝑛𝑔𝑖,𝑡 indicator for whether person i is working in year t.
• 𝑋𝑖,𝑡 contains income in 1999, demographic characteristics, and
in some models, pension, welfare, DI, and UI benefits.
• 𝐴𝑘 a vector of age-fixed effects, k ranges from ages 51-65.
• 𝑆𝑛 a vector of industry fixed effects.
The project first estimates an OLS regression.
𝐷𝑖,𝑡+5 = 𝛼 + 𝑋𝑖,𝑡′ 𝛽 + 𝛾𝑊𝑜𝑟𝑘𝑖𝑛𝑔𝑖,𝑡 + 𝛿𝑦𝑒𝑎𝑟 + 𝐴𝑘 + 𝑆𝑛 + 휀𝑖,𝑡
11
• First Stage:
𝑊𝑜𝑟𝑘𝑖𝑛𝑔𝑖,𝑡 = 𝜋 + 𝑋𝑖𝑡′ 𝜌 + 𝜆𝐷𝑊𝐵𝑖,𝑡 + 𝜃𝑦𝑒𝑎𝑟 + 𝐴𝑘 + 𝑆𝑛 + 휀𝑖,𝑡
o 𝐷𝑊𝐵𝑖,𝑡 a binary indicator of eligibility for the DWB.
• Second Stage:
𝐷𝑖,𝑡+5 = 𝛼 + 𝑋𝑖,𝑡′ 𝜃 + 𝛿𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝑖,𝑡 + 𝜋𝑦𝑒𝑎𝑟 + 𝐴𝑘 + 𝑆𝑛 + 𝜂𝑖,𝑡
o 𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝑖,𝑡 predicted working status from the first stage.
Since work and health are jointly determined,
the project also estimates a 2SLS regression.
12
13
The OLS results suggest working longer
reduces mortality for both men and women.
OLS Estimates of the Relationship Between Working and the Five-year Mortality Risk
Notes: Sample contains men and women born between 1943 and 1950 observed from 1999 through 2011. For cohorts born between 1943
and 1945, observations after 2009 are not included. All models include controls for marital status, income in 1999, a linear year control, and
age fixed effects. Benefit controls include indicators for whether a person receives a pension, welfare, UI, DI, or other social benefits.
Robust standard errors clustered at the birth cohort – year level in parentheses. *** p<0.01.
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 1999-2016.
Men Women
(1) (2) (3) (4) (5) (6)
Working (0/1) -0.021 *** -0.021 *** -0.019 *** -0.012 *** -0.011 *** -0.008 ***
(0.001) (0.001) (0.000) (0.001) (0.000) (0.000)
Age-fixed effects Yes Yes Yes Yes Yes Yes
Industry-fixed effects No Yes Yes No Yes Yes
Benefit controls No No Yes No No Yes
5-year mortality non-working 2008 7.4 % 7.4 % 7.4 % 4.4 % 4.4 % 4.4 %
5-year mortality working 2008 4.3 4.3 4.3 2.4 2.4 2.4
R2 0.015 0.016 0.019 0.006 0.007 0.009
Observations 10,437,307 10,437,307 10,437,307 10,143,975 10,143,975 10,143,975
13
14
But the 2SLS regression only finds a
significant reduction for men. 2SLS Estimates of the Effect of Working on the Five-year Mortality Risk
Notes: Sample contains men and women born between 1943 and 1950 observed from 1999 through 2011. For cohorts born between 1943
and 1945, observations after 2009 are not included. DWB instrument is 1 after 2009 for cohorts born between 1945-1949. All models
include controls for marital status, income in 1999, a linear year control, and age fixed effects. Benefit controls include indicators for
whether a person receives a pension, welfare, UI, DI, or other social benefits. The table reports the Kleibergen-Paap F statistic. Robust
standard errors clustered at the birth cohort – year level in parentheses. * p<0.10, ** p<0.05, *** p<0.01.
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 1999-2016.
Men Women
(1) (2) (3) (4) (5) (6)
-0.015 *** -0.016 *** -0.021 *** 0.090 -0.001 0.004
(0.005) (0.004) (0.005) (0.239) (0.021) (0.016)
Age-fixed effects Yes Yes Yes Yes Yes Yes
Industry-fixed effects No Yes Yes No Yes Yes
Benefit controls No No Yes No No Yes
First stage
F statistic 22 23 23 4 4 6
DWB coefficient 0.056 *** 0.057 *** 0.046 *** 0.011 * 0.011 0.014 **
(0.012) (0.012) (0.010) (0.006) (0.006) (0.006)
R2 0.015 0.016 0.019 0.090 0.006 0.009
Observations 10,437,307 10,437,307 10,437,307 10,143,975 10,143,975 10,143,975
14
15
It is useful to put the 2SLS regression results
into the perspective of extra months of life.
• For men, a 1.6-percentage-point reduction is a 22-percent
reduction relative to the five-year mortality risk of non-
workers in 2008.
• If this 22-percent reduction occurs only between ages 60-64,
then it increases age 60 life expectancy by 2 months.
• Of course, if the effect instead lasted beyond these ages, the
increase in life expectancy would be larger.
15
Conclusion
• The results from this study suggest that working longer may
lead to longer lives, especially for men.
• However, unless the gains are long-lasting, the actual
additional life is predicted to be relatively small.
• Still, policymakers may want to consider the interaction
between mortality and work as people continue to extend their
careers.
16
Next steps
• The project recently gained access to prescription drug data,
which will allow us to investigate the impact of delayed
retirement on health.
• Specifically, the project will study how delayed retirement
affects:
o Diabetes and depression, two health outcomes that clearly
connect to specific prescription drug classes – e.g., anti-
depressants or insulin.
o Overall health as measured by a comorbidity index based
on prescription drug use.
17
Appendix A: Institutional Setting
18
Retirees in the Netherlands can receive
income from three sources.
• Pillar I: Public Pension
o Flat rate, not means tested.
o € 1,000 for singles, € 1,400 for couples.
o Eligible at age 65 (cohorts before 1948).
• Pillar II: Collective (Employer Provided) Pensions.
o 90 percent of employees participate, majority DB plans,
fully funded.
• Pillar III: Private Individual Pension Products.
o Self-employed, and employees in sectors without
collective plan.
o Anyone can purchase a product in the third pillar.
19
Great Recession is less of a concern in the
Netherlands between 2009-2011 than in U.S.
Source: Statistics Netherlands, 2002-2014.
0%
2%
4%
6%
8%
2003 2005 2007 2009 2011 2013
Hu
nd
red
s
Men
Women
Unemployment Rate Ages 45-70, by Gender
20
Appendix B: Descriptive Statistics
21
22
Men who work at ages 62-65 have a lower
mortality risk than those who do not work.
Averages of Characteristics of Men Ages 62-65, by DWB Eligibility
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 1999-2016.
Eligible for DWB Not eligible for DWB
5 year mortality risk 5.6 % 6.1 %
Working 48.1 39.1
Working Not working Working Not working
5 year mortality risk 3.9 % 7.3 % 4.3 % 7.3 %
Married 80.7 74.0 82.5 77.6
Widowed 2.9 3.9 2.9 4.0
Divorced 10.6 12.5 9.7 10.7
Pension recipient 54.3 75.5 56.3 73.5
UI 4.0 5.1 4.3 7.7
DI 0.5 4.9 0.5 4.6
Welfare 4.2 10.7 4.7 11.2
Other social benefits 11.0 30.8 13.7 32.1
22
23
Women do not show the same pattern.
Eligible for DWB Not eligible for DWB
5 year mortality risk 3.8 % 3.9 %
Working 27.3 20.2
Working Not working Working Not working
5 year mortality risk 2.4 % 4.4 % 2.4 % 4.3 %
Married 66.3 72.9 66.7 72.8
Widowed 8.4 10.5 9.8 11.5
Divorced 18.8 11.6 17.9 11.1
Pension recipient 48.7 48.2 50.5 43.3
UI 2.9 1.8 2.9 2.7
DI 1.0 6.2 1.1 6.2
Welfare 1.8 2.9 2.0 3.2
Other social benefits 5.9 17.0 6.7 17.0
Averages of Characteristics of Women Ages 62-65, by DWB Eligibility
Source: Statistics Netherlands. 1999-2016. Microdata.
23
Overview of cohorts in the sample
Year 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
1943 56 57 58 59 60 61 62 63 64 65
1944 55 56 57 58 59 60 61 62 63 64
1945 54 55 56 57 58 59 60 61 62 63
1946 53 54 55 56 57 58 59 60 61 62 63 64 65
1947 52 53 54 55 56 57 58 59 60 61 62 63 64
1948 51 52 53 54 55 56 57 58 59 60 61 62 63
1949 50 51 52 53 54 55 56 57 58 59 60 61 62
1950 49 50 51 52 53 54 55 56 57 58 59 60 61
Source: Authors’ calculations.
24
Appendix C: Other Robustness Checks
25
26
Eliminating those who were not working at
age 55 does not change the results. 2SLS Estimates of the Effect of Working on the Five-year Mortality Risk for Those Working at Age 56
Notes: Sample contains men and women born between 1943 and 1950 observed from 1999 through 2011 who worked at age 56. For
cohorts born between 1943 and 1945, observations after 2009 are not included. DWB instrument is 1 after 2009 for cohorts born between
1945-1949. All models include controls for marital status, income in 1999, a linear year control, and age fixed effects. Benefit controls
include indicators for whether a person receives a pension, welfare, UI, DI, or other social benefits. The table reports the Kleibergen-Paap F
statistic. Robust standard errors clustered at the birth cohort – year level in parentheses. *** p<0.01.
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 1999-2016.
Working at age 56
-0.017 ***
(0.005)
Age-fixed effects Yes
Industry-fixed effects Yes
Benefit controls No
First stage
F statistic 18.9
DWB coefficient 0.062 ***
(0.014)
Mean mortality risk non-working 6.0 %
Mean mortality risk working 3.3
R2 0.01
Observations 5,077,113
26
27
As would be hoped, no effect is found for a
“placebo” policy. 2SLS Estimates of the Effect of Working on the Five-year Mortality Risk
Notes: Sample in column 1 includes cohorts born between 1940-1946, observed between 1999-2008, but excludes observations after 2006
for the cohorts born between 1940-1942. The placebo DWB indicator is 1 for cohorts born between 1943-1946 after 2006, and 0 for
everyone else. The model include controls for marital status, a linear year control, and age fixed effects. The table reports the Kleibergen-
Paap F statistic. Robust standard errors clustered at the birth cohort – year level in parentheses.
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 1999-2016.
Placebo
0.015
(0.060)
Age-fixed effects Yes
Industry-fixed effects No
Benefit controls No
First stage
F statistic 1.1
DWB coefficient -0.017
(0.045)
Mean mortality risk non-working 7.3 %
Mean mortality risk working 3.5
R2 0.003
Observations 5,749,281
27
28
Leaving out the control for income modestly
increases the size of the effect. 2SLS Estimates of the Effect of Working on Mortality Risk with and without Baseline Income Control
Notes: Sample contains men and women born between 1943 and 1950 observed from 1999 through 2011. For cohorts born between 1943
and 1945, observations after 2009 are not included. DWB instrument is 1 after 2009 for cohorts born between 1945-1949. All models
include controls for marital status, a linear year control, and age fixed effects. Benefit controls include indicators for whether a person
receives a pension, welfare, UI, DI, or other social benefits. The table reports the Kleibergen-Paap F statistic. Robust standard errors
clustered at the birth cohort – year level in parentheses. *** p<0.01.
Source: Authors’ calculations using non-public microdata from Statistics Netherlands, 1999-2016.
With income control Without income control
-0.016 *** -0.020 ***
(0.001) (0.004)
Age-fixed effects Yes Yes
Industry-fixed effects Yes Yes
Benefit controls No No
First stage
F statistic 23 23
DWB coefficient 0.057 *** 0.063 ***
(0.012) (0.011)
Mean mortality risk non-working 7.0 % 7.0 %
Mean mortality risk working 2.8 2.8
R2 0.015 0.015
Observations 10,437,307 10,437,307
28