The Impact of Early Retirement Incentives on
Labor Market Participation: Evidence from a
Parametric Change in the Czech Republic
David Kocourek∗
Filip Pertold†
September 2010
Abstract
We investigate the impact of a change in the Czech early retirement
scheme on the labor force participation of older male workers. Using
the difference-in-differences method we find that a reduction in early
retirement benefits by 2–3 % leads to approximately the same decrease
in the probability of being inactive. Our finding implies high elasticity
of older male workers participation rate. The public policy implication
is that a reduction in early retirement benefits can serve as a very
effective tool to increase the participation of older men in the Czech
labor market.
JEL classification: J21, J26
Keywords: early retirement, labor market participation, Czech Republic
∗
Czech National Bank, david.kocourek@cnb.cz
†
CERGE-EI, filip.pertold@cerge-ei.cz
We have benefited from valuable comments by R´obert Ambriˇsko, Libor Duˇsek, Kamil
Galuˇsˇc´ak, ˇStˇep´an Jurajda, Jan Kmenta, Daniel M¨unich, Sarah Smith, Zbynˇek ˇStork and
participants at a ˇCNB seminar, the EALE conference and the ˇCSE conference. Filip
Pertold worked on this paper as a research fellow at the Czech National Bank. All errors
and omissions are ours. The views expressed in this paper are those of the authors and
do not necessarily represent those of the Czech National Bank.
1
Nontechnical Summary
The aging society is a crucial issue for the Czech Republic, and the Czech
government has introduced various policy measures. In July 2001, penalization
for early retirement with permanently cut benefits was increased and
hence early retirement benefits were lowered after the change.
Changes in early retirement schemes have been introduced in various
countries. There is no full consensus in the recent literature on the final
impact on the labor market behavior of older workers. Gruber and Wise
(2002) provide a cross-country comparison of social security incentives and
suggest that they play an important role in retirement and labor market
decisions. Brinch, Hernaes, and Strom (2001) show that the introduction of
the early retirement option has decreased the labor supply in Norway. On
the other hand, Baker and Benjamin (1999) provide evidence from Canada
and the USA where the reaction to changes in early retirement benefits was
modest or even non-existent. Based on this literature we test the hypothesis
that the change in early retirement benefits in the Czech Republic increased
the labor market participation of older males.
First, we quantify the real change in benefits and monetary incentives
using various simulations. We use various indicators for representative individuals.
We find a decrease in early retirement benefits by 2–3 %; in terms
of the net wage it is approximately 1–2 %. We also computed social security
wealth, accrual rate, peak value and option value as dynamic indicators of
incentives to retire. Generally, we found that social security wealth from
early retirement substantially decreased, but the optimal retirement age did
not change substantially.
After the reform, the social security statistics show a substantial decrease
in the number of newly granted early retirement benefits. This suggests that
the reform strongly affected labor market participation. To test this we use
Czech Labor Force survey data and we employ the difference-in-difference
method (Baker and Benjamin 1999) to evaluate the effect of reform on males
labor market behavior. Our treatment group contains individuals who are
eligible for early retirement benefits; younger individuals are in the control
group.
We find that a reduction in early retirement benefits by 2–3 % leads to
approximately the same decrease in the probability of being inactive. This
finding was confirmed by various robustness checks. The results are not
dependent on length and number of periods before and after the reform.
2
1 Introduction
As policy makers face the commonly known problem of an aging society, the
labor supply of older workers becomes more important. The labor market
decisions of older workers influence government expenditure on various social
programs. For example, the way incentives to retire are formed is a crucial
issue in keeping the pension system sustainable while the population is aging.
Governments thus attempt to change the design of social security systems in
order to respect demographic changes.
The Czech Republic is an example of an aging society1
. The Czech government
has reacted to this development and has decreased the incentives to
retire early created by the social security system. Policy makers expect this
step to reduce the number of people who receive retirement benefits and at
the same time increase the number of contributors to the pension system.
These unambiguous advantages make this policy step popular also among
many other governments facing the issue of aging.2
The policy relevance of this topic is reflected in the current empirical literature.
But it doesnt exist clear answer about the causal impact of retirement
incentives on the labor supply of older workers.
Cross-country comparisons show a strong negative relationship between
early retirement incentives and labor force participation (Gruber and Wise
2002; Borsch-Supan 2000). Papers examining changes in national policies
suggest that the introduction of early retirement benefits as a specific form of
retirement incentive decreases labor force participation (e.g. Brinch, Hernaes,
and Strom (2001)).
By contrast, other studies do not find clear evidence about the sensitivity
of the labor supply of older workers to changes in the early retirement scheme.
For example, Baker and Benjamin (1999) provide evidence from the USA and
Canada which shows a relatively modest or non-existent reaction of the labor
supply to changes in the early retirement scheme. Similarly, Moffitt (1987)
finds relatively small effects of social security law on the labor supply of older
workers in the USA.
There are only a few papers about the labor supply of Czech workers.
Direct evidence concerning the labor supply of older workers is provided in
1
According to the projection of the Czech Statistical Office, the share of people aged
60 years and over will double in the next 30 years. Babecky and Dybczak (2009) try to
model this aging scenario using an OLG model.
2
It needs to be emphasized that the overall fiscal balance is improved unless retirees are
proportionally compensated for longer service and unless employees leave the labor market
and become unemployed or accept disability social assistance and/or become recipients of
support from other social programs.
3
Galuscak (2001) and Bicakova, Slacalek, and Slavik (2006). Galuscak (2001)
shows that the introduction of an earnings test, which imposed a benefit
eligibility constraint on working pensioners, led to a significant and substantial
decrease in the participation rate of workers who had reached statutory
retirement age, whereas Bicakova, Slacalek, and Slavik (2006) estimated the
effect of tax changes on the labor supply of average Czech workers as being
relatively modest. There is no direct evidence about the causal impact of
early retirement incentives and the participation of older workers.
Retirement incentives can take various forms: explicit and implicit taxation
and/or legal rules that restrict full-time work at a certain age. In our
case we investigate the effect of reducing early retirement benefits, which are
offered as non-labor income for individuals three years before the statutory
retirement age. The policy change became effective in July 2001 and cut
early retirement benefits by approximately 3 % for new claimants. To illustrate
this we also compare several incentive measures before and after the
reform.
The social security statistics show that one year after the policy change,
the number of new early retirees had decreased by half. This suggests that the
direct impact of this policy step was strong. However, as we describe in the
next section, older workers face several options regarding how to become nonemployed
(retire early3
, become unemployed, or enter disability retirement4
).
The positive causal effect of the policy change on the labor supply of older
workers is under question.
In order to find the causal impact of the policy step, we use the differencein-differences
estimation method. The treatment group includes workers who
are eligible for early retirement benefits (at most three years before the statutory
retirement age). The control group contains workers who are just about
to enter the eligibility age for early retirement, six to three years before the
statutory retirement age to be more specific. The eligibility age for entering
early retirement starts three years before the statutory retirement age. In
particular, a marginal probit model is used for testing whether the policy
change affects the participation rate of individuals who are eligible for early
retirement, controlling for other characteristics of the individuals.
Our analysis shows that this policy increased the probability of a male
participating in the labor market by 2–3 % for those eligible for early retirement.
This paper is organized as follows. The next section provides a
detailed insight into the social security system in the Czech Republic. The
3
The exact preconditions for early retirement are described in Act No. 155/1995 Coll.
4
To enter disability retirement certain health criteria have to be met. Hence, it is not
a free choice of the individual.
4
official statistics and simulations of the policy change on individuals are described
in section 3. Section 4 covers the data description of the treatment
and control group. A graphical overview is presented in section 5, the econometric
methodology is explained in section 6, and the results are described
in section 7. Section 8 concludes.
2 Institutional Setting
The Czech retirement scheme is a standard pay-as-you-go (PAYG) system
with mandatory participation for all employees and the self-employed as well.
The basic features of the Czech pension system were inherited from the system
run under the communist regime. A few legislative changes were implemented
in the years after the fall of communism, but the basic features
remained unchanged. The statutory retirement age is different for male and
female workers; the retirement age of the latter depends on the number of
children raised. Beside this differentiation the retirement age has been prolonged
by two months for males and four months for females per year after
1996 to the year the male or female was supposed to retire under the former
conditions. The retirement age for males in 1996 was set at 60 years.5
The
retirement age for females without children was 57 and each child raised reduces
the retirement age by one year. At the time of the policy change the
average retirement age was approximately 61.
Pension benefits are computed based on a formula that has an individual
specific part (a percentage-based assessment) and a part which is the same
for everybody (the basic amount). The basic amount is the amount of money
laid down by law that is received by everybody who is an old-age pension
recipient. It can be understood as the minimum pension. The individual part
reflects individual-specific characteristics, such as the earning history since
1986 and number of years in service. The wage history is discounted to the
current value and then modified by reduction limits and reduction percentages
to a calculation base (CB). The calculation CB represents the crucial
step in the Czech pension formula and causes a high degree of redistribution
in the system. Amount that is lower than the first reduction limit is fully
included. But 30 % of the amount between the first and second reduction
limit is included and only 10 % of the reminder which is above the second
5
After that there is no single retirement age for the male population in a given year.
The exact formulation is that the retirement age is prolonged by two months for each
initiated age-year after December 31, 1995 before the individual reaches the age of 60. In
practice this means that if a worker is 60 in February 2000, then his retirement age is 60
plus ten months. Therefore, the men from this example will retire in January 2001.
5
reduction limit. The number of years in service proportionately increases
(1.5 percentage points per year) the size of the adjustment percentage (AP)
and therefore the size of the percentage of the CB which will be counted as
the percentage-based assessment (PA) in the pension formula. The longer
an individual is in service, the higher the PA and therefore the higher the
pension benefit will be. The exact formula can be found in Appendix A.
This formula is applied to every kind of retirement benefits, including
early retirement benefits.6
The early retirement benefits are lower than the
standard ones, because they are reduced by an adjustment coefficient (rPYI),
which was subject to the policy change. In particular, the penalty for early
retirement before the reform was 0.6 percentage points and 0.3 percentage
points7
per each 90 days remaining to the standard retirement age before the
policy was introduced. The policy step changed the degree of penalization for
early retirement. In fact, both rates that adjust early retirement benefits (0.6
percentage points and 0.3 percentage points) were increased to 0.9 percentage
points . For example, considering an individual who retires one year before
her retirement age (a 0.6 percentage points reduction applied before the
reform), the adjustment percentage of her benefit decreased by 3.6 percentage
points after the reform instead of by 2.4 percentage points which would apply
before the policy change—lower by 1.2 percentage points.
This decrease in the adjustment percentage proportionally decreases the
pension benefit and hence has an influence on the motivation of workers to
stay active on the Czech labor market until the statutory retirement age.
Table 1 shows the drop in officially newly granted early retirement benefits.
The fall was approximately 10 % of regular pension benefits. This
observed change is most likely caused by two effects. The first one is driven
by the change in early retirement benefits. The second one is driven by
6
The Czech social security scheme recognizes two types of early retirement. One is
with permanently cut benefits, which allows individuals to retire at most three years
before the eligibility age and the individual is not allowed to work after retiring. The
decreased pension benefits are collected for the rest of the individuals life. The second
is early retirement with temporarily cut benefits, which allows the individual to retire at
most two years before the eligibility age and is tied to unemployment status for half of the
year at least. The decreased pension benefits are recalculated when the eligibility age is
reached and increased to the level as if one had retired at the eligibility age. Apart from
that, two more ways of escaping employment status are available: becoming unemployed
and becoming disabled. However, social support for disabled people is strictly tied to the
health situation of the individual and hence cannot be regarded as a fully free choice of the
individual, though it can be influenced by the individual exerting pressure on the doctor
who makes the decision about the disability pension.
7
This applies for the case when an individual who applies for early retirement benefits
and is aged 60 or more. For all other cases the permanent penalty is then just 0.6 pecentage
points per each 90 days before the standard retirement age.
6
a change in the characteristics of workers who applied for early retirement
before and after the policy step.
Table 1: Newly granted pensions (in CZK)
1999 2000 2001 2002 2003 2004 2005
(1) all pensions 5,991 6,106 6,399 7,055 7,224 7,760 8,391
(2) at retirement
age
6,222 6,485 6,823 7,226 7,512 7,968 8,693
(3) after retirement
age
7,272 7,485 7,916 8,621 9,157 9,410 10,306
(4) early ret. temp.
cut
5,370 5,513 5,838 5,917 6,224 6,404 6,836
(5) early ret. perm.
cut
5,593 5,659 5,844 5,667 5,996 6,261 6,984
(5)/(2) (in %) 90 87 86 78 80 79 80
Source: MLSA (2006), own computation of averages
The comparison of newly granted early retirement benefits before and
after the reform does not provide a clear picture about the effect of the policy
on benefits. It is probable that workers who applied for early retirement after
the reform had stronger preferences toward leisure than workers who applied
before the reform, and they might also have had different working histories8
, which determine their benefits. Therefore, we attempt to isolate the pure
policy change effect from the sorting effect. For that purpose we create several
typical individuals with different wage histories, which serve—together with
length of service—as a major input for the computation of benefits.
We also compute the early retirement benefits before and after the change
for individuals with virtually the same characteristics. The only parameter
that changes is the degree of penalization, which was subject to the policy
change. Our computations show that the net decrease in early retirement
benefits was approximately 2–3 % (CZK 120–250 per month in absolute
terms). The cut corresponds approximately to 1–2.5 % of the average net
wage for male workers in the economy.
8
Different wage histories and number of years in service, etc.
7
Table 2: Changes in early retirement benefits due to the policy change
Years before
eligible age T
Absolute
decrease before/after
(in
CZK/month)
Relative
decrease in
early retirement
benefit
before/after
(in %)
Change in
terms of net
wage (in
percentage
points)
70% of
avg. wage
T-3 191 -3 -2.4
T-2 133 -2 -1.6
T-1 131 -2 -1.1
Avg. wage
T-3 218 -3 -1.9
T-2 149 -2 -1.3
T-1 152 -2 -1.3
150% of
avg. wage
T-3 237 -3 -1.3
T-2 162 -2 -0.9
T-1 166 -2 -0.9
Source: Own computation based on the official formula published in MLSA (2002).
Note: Benefits are computed for 46 years of service. The net wage is CZK 11,324 in
2001. Three income groups were chosen arbitrarily. 70 % of the average wage reflects
approximately the group of workers with the median wage and 150 % of the average
wage represents managers and high-paid workers in the Czech economy.
The ratio of the net wage to early retirement benefits (the net replacement
rate) decreased by 0.9–2.4 percentage points. Generally, the highest decrease
applied to those who wanted to enter early retirement three years before the
eligibility age. Lower-income workers were penalized relatively more than
upper-income groups. This is a result of the pension formula: benefits are
relatively higher for low-income than for high-income workers. This implies
that the policy change affected more strongly individuals who face a relatively
disadvantaged position on the labor market.
Another way to assess the effect of this policy change is suggested in
Borsch-Supan (2000). The author stresses the importance of the time dimension
how much it is worth to give up one year of retirement in terms
of net benefit or social security wealth (SSW) computed as the difference
between the expected discounted stream of all future benefits and social security
taxes paid, which are computed as a percentage of gross earnings. The
SSW formula, which states how to compute the social security wealth for an
individual at age S planning to retire at age R, is
8
SSWS(R) =
E∑
t=R
π(t|S) · δt−S
· Bt(R) −
R−1∑
t=S
π(t|S) · δt−S
· c · Wt,
where:
SSW - social security wealth,
S - planning age,
R - planned retirement age,
E - expected age of death at age S,
π(t|S) - probability of being alive at age t conditional on being alive at age S,
Bt(R) - pension at age t for retirement at age R,
Wt - wage at age t,
δ - discount factor,
c - social security contribution rate.
SSW is very sensitive to many assumptions.9
We employ the values for
the discount factor and wage growth10
from Coile and Gruber (2007) to keep
the analysis consistent with the analysis of peak value (Coile and Gruber
2007) and option value (Stock and Wise 1990). In our computation of SSW
we do not assume any indexation. The process of indexation in the Czech
Republic depends very much on government discretion, as described in Dusek
(2007) and Dusek and Kopecsni (2008).
Table 3 shows the basic computations of retirement incentives employing
the lifetime budget constraint for an average earner.
Each row corresponds to the age at which a worker enters retirement. In
this exercise we assume for the sake of simplicity that the statutory retirement
age is 61. This means that everybody who enters retirement before the age of
61 is in early retirement regime and the worker is eligible for early retirement
benefits at 58.
Comparing SSW before and after the reform, one can see a decrease in
SSW for those who enter early retirement. SSW before and after the reform
are highest at 58. The higher pension after longer time contributing to the
social system cannot compensate for the social security contribution and
hence SSW steadily decreases and therefore it is the best decision to retire
as soon as possible since it maximizes the SSW.
9
Assumptions regarding the individual discount rate, the future indexation of benefits
under PAYG, the interest rate path, wage growth, etc.
10
For simplicity we assume the same wage growth for all income groups.
9
Table 3: Monetary incentives before and after the reform (average earner)
Last age
of work
Replace.
rate–
before
Replace.
rate–
after
SSW–
before
SSW–
after
Accrual
rate–
before
Accrual
rate–
after
58 0.837 0.828 699,347 690,703 -0.007 -0.007
59 0.870 0.864 650,158 644,474 -0.076 -0.072
60 0.906 0.903 598,921 595,727 -0.086 -0.082
61 0.936 0.936 545,586 544,716 -0.098 -0.094
62 0.964 0.964 489,416 488,365 -0.115 -0.115
63 1.012 1.012 445,006 443,768 -0.100 -0.100
64 1.037 1.037 389,143 387,718 -0.145 -0.145
65 1.105 1.105 352,27 350,657 -0.105 -0.106
Note: SSW – social security wealth – is defined as the sum of all discounted pension
benefits and social security contributions. The accrual rate is defined as the relative
year-to-year change in SSW.
A forward-looking approach to assessing the incentives created by the
pension system can be studied using peak value and option value. Peak value
(Coile and Gruber 2007) is defined as all discounted benefits from entering
retirement. In fact, it is maximized when SSW reaches its maximum. We
performed this analysis and it obviously supports the preceding analysis that
the reform has increased the incentives for the average earner to stay on the
labor market. The second approach to assessing financial incentives is the
option value model (Stock and Wise 1990). The option value attempts to
evaluate the optimal retirement age in utility terms and involves calculating
the forgone earnings that could have been earned on the labor market. It is
defined as the change in utility that results from working to the optimal age,
which is determined by maximizing the lifetime utility over consumption and
leisure. The problem of this approach is that one needs to employ certain
assumptions about wage profile in the final career stage.
We employ the standard assumption of a linear wage profile, which is not
necessarily a realistic assumption. Our results are summarized in Appendix
B and suggest that both according to the peak value and option value the
optimal retirement age was not changed by the reform and is at the age of
58 in the case of option value and at 56 in the case of peak value. However,
there is one small exception of high earner whose option value reacts to the
policy change and the optimal retirement age is moved by one year from 59
to 60.
One of the questions that this reform raised is what margin of the labor
supply is affected, and in particular whether the reform affected the extensive
10
or intensive margin of the labor supply of older workers. The extensive
margin is affected only since the labor code restricts early retirement benefits:
people who retire earlier (claim early retirement benefits) are not allowed to
work at all.
3 Data Description and Treatment and Control
Group
For the purposes of our research we use Czech Labor Force Survey data from
19982005 containing detailed information about the labor market status of
a representative sample of 60,000 individuals and their households. On a
rotating panel base, individuals and their households are surveyed during
five consecutive quarters. Therefore, one fifth of the sample is replaced every
quarter. We choose the subsample of males who are in the age window of six
to zero years until the statutory standard retirement age. Hence, our sample
includes 50,152 observations for 11,843 individuals. Summary statistics for
the treatment and control groups can be found in Appendix D.
We divide this sample into four time periods one period before the reform
and three periods after the reform. Participation in the survey is restricted to
up to five quarters. Within this period, we do not observe a sufficient number
of changes in labor market status, thus we treat our sample as repeated crosssectional
data. The reason we choose only one period before the policy change
is the low stability of the social security system: the legal system was stable
for only two years before the policy change and approximately four years after
the policy change. Our time span also reflects the comparability of the data.
We define four consecutive periods, each 1.5 years long. The first is before
the policy change (1Q2000—2Q2001), the second is immediately after the
policy change (3Q2001—4Q2002), the third is from 1Q2003 to 2Q2004, and
the fourth covers 3Q2004—4Q2005. We also try alternative time spans, but
this does not change our results significantly (see Appendix F). This division
of the total time span into four periods covers the most institutionally stable
period before and after the reform. On top of that, the results for several
time periods after the reform confirm that the impact of the policy change
is the same over time.
The important problem is the actual eligibility age, since the statutory
retirement age has been lengthening by two months per year and gives additional
noise to our data. To diminish this problem we calculate the individual
statutory retirement age as defined by law. For that purpose we
have to approximate the actual age of the respondents in the Labor Force
11
Survey, because the survey per se does not provide information about the
exact actual age (the accuracy is yearly frequency). Thus, we use only those
individuals for which we observe a change in age during the period they were
surveyed (Galuscak 2001). Using these individuals we approximate the exact
individual age at an accuracy of one quarter and calculate the actual
individual statutory retirement age and simultaneously the eligibility age for
the early-retirement. Based on this approximation we can also calculate the
number of years to retirement. This makes our analysis more accurate and
allows us to disentangle the effect of the early retirement change from the
prolonging of the retirement age.
Using the number of years to the statutory retirement age we define the
treatment and control groups. The treatment group contains people who
are eligible for early retirement: up to three years before their standard retirement
age. The younger individuals (more than three years before the
eligibility age) are in the control group, because they were not directly affected
by the policy. The relatively broad definition of the treatment group
allows us to capture all individuals who were eligible for early retirement
and could make the decision during the entire period of three years before
reaching the statutory retirement age. The disadvantage is that in the period
after the policy change the treatment group consists of two types of retirees:
men who entered early retirement in the old system and those who entered
in new system. This is reflected in our analyses and we interpret the results
with respect to this fact.
The LFS data contain information about individual characteristics that
are important for our analysis. For the purposes of our analysis we used
the following characteristics: education, family status, number of persons
in the household, and geographical location. The data do not include any
information about wages or retirement benefits.
4 Graphical Overview
As we described above, the change in the early retirement scheme increases
the incentive to stay in the labor market. As a preview of our results we
present the official statistics of newly granted pensions (Fig. 1) . The share
of newly granted pensions for this particular pension scheme dropped significantly
(the solid line).
This suggests that this reform could have a strong impact on the labor
market decision. However, the total impact on the participation rate can be
questioned, because the share of the other options for early exit could be
used, as can be seen in Figure 1.
12
Figure 1: Newly Granted Pensions (men - in % of total)
policy change
0510152025303540
2000 2001 2002 2003 2004 2005
at the retirement age after the retirement age
early ret. − temp. cut early ret. − perm. cut
disability retirement
Source: Czech Social Security Administration, own calculation
Note: The short time span before the actual policy change is given by the limitation of official statistics.
The remainder to 100% are e.g. widower’s and orphan’s pensions.
Further, we present the behavior of individuals using the Labor Force
Survey data described above. Figure 2 depicts the participation rate of the
control and treatment groups during 1998—2005. The participation rate
of the treatment group increased by around ten percentage points between
2001 and 2004. The participation rate also increased in comparison with
the control group. This suggests that our treatment group was subject to a
specific shock that did not affect the control group. One can observe that this
increase continued at a lower rate even during the period from the second half
of 2003 to almost the end of 2004. It also contains the effect of the policy
change, because in the first period after the policy change, the treatment
group still contains older cohorts that entered early retirement before the
policy change and remain in the treatment group. Due to data limitations
and the institutional set-up, we cannot define the treatment group more
precisely than 0—3 years before retirement.
In Figure 3 we can see how the participation rate changes over time in
different years to/after retirement age. This quasi-cohort approach shows
that the participation rate during the early retirement window (between -3
and 0) is the lowest in the period before the reform was introduced. Moreover,
the trend that we observe in Figure 3 is clearly increasing. The difference
between the pre-reform period and the last period studied at one year before
the statutory retirement age is 12 percentage points.
13
Figure 2: Participation rate of control and treatment group in 1998-2005
policy change
1525354555657585
(in%)
1998 1999 2000 2001 2002 2003 2004 2005
control group treatment group difference
Source: Labor Force Survey, own calculation
Figure 3: Participation rate in different distances to/after retirement age
1030507090
(in%)
−7 −6 −5 −4 −3 −2 −1 0 1 2 3 4
years before(−)/after(+) retirement age
eligibility age
1Q2000 − 2Q2001 (period before change)
3Q2001 − 4Q2002 (1. period after change)
1Q2003 − 2Q2004 (2. period after change)
3Q2004 − 4Q2005 (3. period after change)
Source: Labor Force Survey, own calculation
We also present an alternative indicator the hazard rate representing
the probability of labor force withdrawal due to retirement. Figure 4 depicts
14
the hazard rates for two periods: before and 3—4.5 years after the policy
change. In the cross-sectional setting, the definition of the hazard rate is one
minus the retention rate, which is the participation rate of workers at age t
divided by the participation rate of workers aged t–1 in the given year (Hurd
1996).
Figure 4: Hazard rates in different distances to/after retirement age
−.10.1.2.3.4
−7 −6 −5 −4 −3 −2 −1 0 1 2 3
years before(−)/after(+) retirement age
eligibility age
1Q2000 − 2Q2001 (period before change)
3Q2004 − 4Q2005 (3. period after change)
Source: Labor Force Survey, own calculation
The line in Figure 4 representing the period before the policy change has
two peaks: the first one (around -2, two to three years before the statutory
retirement age) reflects entering early retirement before the policy change,
while the second (around 0) represents entering standard retirement. The
line for the period three years after the policy change shows a substantial
change in the behavior of retirees. One can see the hazard rate smoothed
over the number of years before/after retirement. Although early retirement
frequently occurs, one cannot observe any particular peak before the statutory
retirement age in the period starting with the third quarter of 2004.
This is most probably an effect of the treatment we study. One can also see
that it is also more common to retire after the statutory retirement age. This
is in line with the hypothesis that workers generally stay longer in their jobs.
We also consider the problem of unemployment, which can potentially
change over time and therefore raise questions about our results. Figure 5
shows the development of the unemployment rate over time. Unemployment
rate is defined for each group separately so that we can control for the changes
15
in labor force in particular group. The trend in unemployment is not clear,
despite an upward movement of unemployment in the treatment group right
after the policy change. However, one needs to be aware that the number
of unemployed individuals in our sample is relatively small and this change
is most probably not statistically significant. Moreover, the dynamics of the
increase is slower when we calculate the unemployment rate using the labor
force across the groups.
Figure 5: Unemployment rate
policy change
0123456
(in%)
1998 1999 2000 2001 2002 2003 2004 2005
treatment group control group
Source: Labor Force Survey, own calculation
This graphical overview suggests that our treatment group was hit by
an external shock around the year 2001 which influenced its participation
in the labor market. We believe that this shock was with high probability
the change in the early retirement setting. This is, of course, not a rigorous
analysis, because we cannot say whether the shift in participation in the labor
market is statistically significant. The next sections thus provide a formal
econometric analysis and computation of the increase in the probability of
staying in the labor force.
5 Methodology of Econometric Analysis
As an identification strategy we use difference-in-differences (Baker and Benjamin
1999). The treatment group includes workers who are eligible for early
16
retirement benefits (at most three years before the actual statutory retirement
age). The control group contains workers between 6–3 years to the
statutory retirement age. The time periods chosen for the estimation are the
following: 1.5 years before the policy change and 4.5 years after the policy
change, divided into three periods of equal length. The increase in the total
number of early retirement benefits was dramatic in the late 1990s. We do
not want to mix the previous changes in the social security system into our
analysis, so we use only one period before the policy as a benchmark for our
analysis. The basic specification is the following:
yit = αi + β1OLDit + β2AFTER1it + β3AFTER2it + β4AFTER3it +
β5(OLDit · AFTER1it) + β6(OLDit · AFTER2it) +
β7(OLDit · AFTER3it) + β8Xit + ϵit,
where yit is one if an individual i is inactive (out of the labor force) at
time t and zero when an individual is active in the same period. OLDit is a
dummy for the treatment group. AFTER1it, AFTER2it and AFTER3it are
dummy variables for the three consecutive periods (1.5 years long) after the
policy change. The period before the policy change is defined as 1.5 years
before the policy change became effective. Xit is the vector of observable individual
characteristics (basic demographic characteristics: education, number
of people in the household, marital status, geographical location) and ϵit is
the error term. This model is estimated by a probit model with the standard
maximum likelihood estimation technique.
The estimated coefficient β1 captures all differences between the treatment
and control groups that are unrelated to the policy change. β2, β3 and
β4 capture all the period-specific changes that influence the probability of
being employed for the control and treatment groups. β5, β6 and β7 are the
coefficients of interest. They reflect the impact of the policy change on the
inactivity of the treatment group relative to the control group. The vector of
coefficients β8 captures the influence of major demographic characteristics.
6 Results
Our final sample contains 50,152 observations, 26,735 from the treatment
group and 23,417 from the control group. The estimated coefficients indicate
that the treatment significantly increased the labor supply of the treatment
group. The coefficients have the expected sign; however, the first period after
the change does not have a significant impact on the labor supply. The
17
reason is that our treatment group also contains people who entered early
retirement under the previous system. Therefore, the pass-through to the
participation rate of the treatment group is lagged and becomes visible only
in periods AFTER2 and AFTER3. β5 is not significant in our specification,
and β6 together with β7 are negative and significant. After controlling for
other observable characteristics, the results change mainly in the significance
of the coefficients. The other controls are significant with the expected signs:
higher education decreases the probability of being inactive. The number
of household members has the same effect. We do not include the labor
market status of spouses, because the labor market activity of spouses can
also potentially be affected by the reform and thus it is an endogenous variable.
To reveal the magnitude of the estimated effects—the impact on the
probability—the marginal effects are presented in Table 4.
Table 4: Estimated coefficients from the probit model in three different spec-
ifications
Model (1) (2) (3)
OLD*AFTER1 -0.0159 -0.0108 -0.0096
(0.0180) (0.0182) (0.0182)
OLD*AFTER2 -0.0509*** -0.0340* -0.0318*
(0.0179) (0.0184) (0.0184)
OLD*AFTER3 -0.0457** -0.0354* -0.0317
(0.0187) (0.0189) (0.0191)
Personal characteristics X X
District dummies X
N 50,152 50,152 50,152
Pseudo R-squared 0.07 0.10 0.14
Note: Coefficients are recalculated into the probability measure (min 0, max 1). The
excluded variables are dummies for: control group, one period before policy change,
interaction of control group and all periods. Full results are presented in Appendix E.
Standard errors are in parentheses. We also performed linear probability estimation
with OLS and it does not change the significance of the results.
*** p<0.01, ** p<0.05, * p<0.1
We estimated three different specifications. The most extended version
contains individual characteristics and 76 dummies for districts. In all models
this effect remains negative. The marginal effect of the reform on the
probability of being inactive is close to -0.03, which can be interpreted as a
3% drop in the probability of being inactive for workers who are at most three
18
years before the statutory retirement age. These results show that inactivity
significantly decreased in the treatment group during 2003–2005 relative to
the control group and the period before. Our results also show that there
is no significant effect of the policy change in the period immediately after
the policy change. This is probably due to the fact that the left-hand-side
variable is a stock (the probability of being inactive) and thus the treatment
group in the first period after the policy change contains a lot of individuals
who entered early retirement before the policy change.
We are also aware of the problem with expectations, which might have
influenced the behavior of people right before the reform became effective.
In our case it would mean that people entered early retirement earlier just
because the policy change occurred. This fact would bias our results. We
cannot fully account for this phenomenon owing to data limitations. Thus,
we did a robustness check and skipped the first half of 2001, since the law
introducing the reform was passed in the Czech parliament at the beginning
of 2001 and became effective in July 2001. We thus shorten the baseline
period to one year. The results are summarized in Table 5 and suggest that
even in this setting the reform decreased the inactivity rate among older
workers. The size of this effect is, however, smaller and in specifications (2)
and (3) the significance has vanished. However, the result for specification
(1) could be considered as the lower bound of the estimated effect, because
those people who reacted purely to the announcement of the reform would
probably have entered early retirement later on if they behave rationally.
The dummies that represent geographical location show high variation in
labor market behavior across different regions in the Czech Republic. For example,
individuals from the Karvina region (border region to Poland strongly
affected by the structural changes after fall of socialism) have a 40% higher
chance of being inactive compared to individuals from Prague, even after
controlling for all possible observable characteristics.
Our results show that the probability of being inactive (out of the labor
force) has decreased since the reform came into force. This means that people
have not started to leave the labor force by using other social programs (e.g.
disability pensions), but this leaves the possibility of becoming unemployed
and so this policy change might still have a negative impact on the fiscal
position. Therefore, we decided to run the same probit specification but with
the indicator variable of being employed. The results, available in Appendix
G, are quite similar to those obtained earlier.
The Appendix H presents additional robustness check and further extension
of our analysis. We divided the control and treatment into the three
smaller fractions of the length of one year. Further we also explore other labor
statuses—employed and unemployed. The control group is considered only
19
those who are 3–4 years before retirement. The results show that the reform
was really efficient for increasing activity on the labor market. However, we
also see that unemployment was also temporally increased. The employment
is increased, but results are not significant. This will be explored further in
the next version of this paper.
Table 5: Estimated coefficients from the probit model in three different specifications
without the first half of 2001
Model (1) (2) (3)
OLD*AFTER1 -0.0004 0.0034 0.0031
(0.0209) (0.0211) (.02104)
OLD*AFTER2 -0.0361* -0.0201 -0.0197
(0.0196) (0.0201) (0.0201)
OLD*AFTER3 -0.0308 -0.0214 -0.0193
(0.0204) (0.0206) (0.0207)
Personal characteristics X X
District dummies X
N 46,127 46,127 46,127
Pseudo R-squared 0.06 0.11 0.13
Note: Coefficients are recalculated into the probability measure (min 0, max 1). The
excluded variables are dummies for: control group, one period before policy change,
interaction of control group and all periods. Full results are presented in Appendix E.
Standard errors are in parentheses. We also performed linear probability estimation
with OLS and it does not change the significance of the results.
*** p<0.01, ** p<0.05, * p<0.1
One more approach to check the robustness of our results is reported in
Appendix I. This Annex reports results for the multinomial logit in which
we compare the relative risks between the three basic statuses on the labor
market—employed, unemployed and inactive. We find that for the treatment
group in second and third period after the policy change the risk to
be employed is higher than inactive. However we find the temporal effect
for the unemployment as well which is even robust for all three equation
specification.
We also attempted to use an explanatory variable that indicates change
in labor market status. However, as we mentioned earlier, we face a problem
with the lack of observations for people who change status during the period
they were surveyed (i.e., four or five quarters). We divided our time span
into two periods: two years before the reform and two years after the reform.
20
We observed only a few changes in labor market status for the treatment
group: 172 out of 2,541 individuals for the two years before the policy change
and 113 out of 2,587 after the policy change. We can conclude that these
numbers are in line with our hypothesis that the reduction in early retirement
benefits caused fewer workers to enter early retirement. However, the number
of observations in our sample does not allow any formal econometric analysis
in this setting.
7 Conclusions and Policy Implications
Our results confirm that the 2–3% cut in early retirement benefits due to
the 2001 reform boosted the labor participation of males eligible for early
retirement by approximately the same amount. The reform increased the
probability of being employed in the three-year period before a worker reaches
the statutory standard retirement age. These results show that the elasticity
of the extensive margin of labor supply of older Czech workers is relatively
high, although we are not able to calculate the exact value because we lack
individual data on wages. Nevertheless the policy change was not purely
fiscal improving since some of the affected people did not continue to work,
but rather switched to unemployment as a substitute to early retirement.
Our findings are generally in line with those, for example, from Germany,
where Borsch-Supan (2000) found a high sensitivity of older workers employment
to the social security system design. Our results also correspond with
Galuscak (2001), who found a substantially high sensitivity of the participation
rate to change in the earnings test for workers older than the statutory
retirement age. In this respect, our results are not fully comparable, because
we examine older workers who are eligible for early retirement and have not
reached the statutory retirement age.
In our approach, we assume that the difference in the labor supply between
older and younger cohorts was not affected by any other shock than
the policy change. This is the only possible way of empirically testing a
public policy intervention affecting the whole population of one country.
The extent of our analyses is also limited by data availability. The dataset
contains important characteristics about the retirement of males and on top
of that it does not contain wages. Therefore, our analysis does not cover
the labor supply of females and we do not directly estimate the elasticity
of the labor supply to the individual budget constraint. Our results also
indicate high differences of labor supply behavior across males with different
characteristics (education, geographic location). This could be the subject
of additional research.
21
References
Babecky, Jan, and Kamil Dybczak. 2009. “The Impact of Population Ageing
on the Czech Economy.” Working papers 1/2009, Czech National Bank.
Baker, Michael, and Dwayne Benjamin. 1999. “Early Retirement Provisions
and the Labor Force Behavior of Older Men: Evidence from Canada.”
Journal of Labor Economics 17 (4): 724–56.
Bicakova, Alena, Jiri Slacalek, and Michal Slavik. 2006. “Fiscal Implications
of Personal Tax Adjustments in the Czech Republic.” Working papers
7/2006, Czech National Bank.
Borsch-Supan, Axel. 2000. “Incentive effects of social security on labor
force participation: evidence in Germany and across Europe.” Journal
of Public Economics 78 (1-2): 25–49.
Brinch, Christian, Erik Hernaes, and Steinar Strom. 2001. “Labour Supply
Effects of an Early Retirement Programme.” Working paper series 463,
CESifo Group Munich.
Coile, Courtney, and Jonathan Gruber. 2007. “Future Social Security Entitlements
and the Retirement Decision.” Review of Economics and
Statistics 89 (2): 234–246.
Dusek, Libor. 2007. “Political Risk of Social Security: The Case of the
Indexation of Benefits in the Czech Republic.” Working papers 318,
The Center for Economic Research and Graduate Education - Economic
Institute, Prague.
Dusek, Libor, and Juraj Kopecsni. 2008. “Policy Risk in Action: Pension
Reforms and Social Security Wealth in Hungary, Czech Republic, and
Slovakia.” Czech Journal of Economics and Finance (Finance a uver)
58 (07-08): 329–357.
Galuscak, Kamil. 2001. “Retirement Decisions of Older Czech Male Workers.”
Working papers 190, The Center for Economic Research and Graduate
Education - Economic Institute, Prague.
Gruber, Jonathan, and David A. Wise. 2002. “Social Security Programs
and Retirement Around the World: Micro Estimation.” Working papers
9407, National Bureau of Economic Research, Inc.
Hurd, Michael D. 1996. “The Effect of Labor Market Rigidities on the Labor
Force Rigidities on the Labor Force.” In Advances in the Economics of
Aging, NBER Chapters, 11–60. National Bureau of Economic Research,
Inc.
22
MLSA. 2002. Actuarial Report on Social Insurance.
. 2006. Actuarial Report on Social Insurance.
Stock, James H., and David A. Wise. 1990. “Pensions, the Option Value of
Work, and Retirement.” Econometrica 58 (5): 1151–1180.
23
A Social security formulae
P = BA + PA
PA = CB · AP
CB = PAB · rp1 + max(0,PAB - rl1) · (rp1 - rp2) + max(0,PAB - rl2)·(rp2 - rp3)
AP = int((IP1 + IP2· 0.8)/365) · (PYI - 90per · rPYI)
PAB =
∑Y −1
y=Y −1−min(30,Y −1−1986) AABy
min(30, Y − 1 − 1986) − EP/365
AABy = ABy· CGGABy
CGGABy =
GABy−2 · CvCy−2
GABy
where:
P - pension benefit
BA - basic amout
PA - percentage-based assessment
CB - calculation base
AP - adjustment percentage
PAB - personal assessment base
rp1 = 100 %,
rp2 = 30 %,
rp3 = 10 %
- reduction percentage
rlj - first and second reduction limit in yearly terms
IPj, j = 1,2 insured
period (j = 1) and compensatori insured period (j = 2)
counted as 80 % of the length befor reaching the age of 18 (only
whole 365 days are included)
PYI - percentage for each year of insurance (1.5 %)
90per - number of 90-day periods
rPYI reduced
percentage for each 90-day period of early reteirement
(subject of policy change)
AAB - annual assessment base
EP - excluded period
AB - assessment base
CGGAB - coefficient of the growth of the general assessment base
GAB - general assessment base
CvC - conversion coefficient
Y - number of years worked
y - year in which the retirement benefits are claimed
24
B Appendix
Table : Monetary incentives before and after the reform (low-wage earner)
Last age
of work
Replacement
rate - before
Replacement
rate - after
SSW - before SSW - after Accrual rate
- before
Accrual rate
- after
58 1.074 1.062 648,662 640,970 -0.005 -0.005
59 1.118 1.110 606,802 601,781 -0.069 -0.065
60 1.165 1.160 561,769 559,087 -0.080 -0.076
61 1.203 1.203 515,824 515,214 -0.089 -0.085
62 1.247 1.247 469,394 468,659 -0.099 -0.099
63 1.304 1.304 428,142 427,275 -0.096 -0.097
64 1.336 1.336 377,901 376,903 -0.133 -0.134
65 1.429 1.429 347,119 345,990 -0.089 -0.089
Note: SSW – social security wealth – is defined as the sum of all discounted pension benefits and social
security contributions. The accrual rate is defined as the relative year-to-year change in SSW.
Table : Monetary incentives before and after the reform (high-wage earner)
Last age
of work
Replacement
rate - before
Replacement
rate - after
SSW - before SSW - after Accrual rate
- before
Accrual rate
- after
58 0.658 0.650 783,855 773,719 -0.005 -0.005
59 0.682 0.676 722,477 715,687 -0.069 -0.065
60 0.710 0.707 660,765 656,923 -0.080 -0.076
61 0.731 0.731 595,214 593,909 -0.077 -0.074
62 0.746 0.746 522,766 521,19 -0.089 -0.089
63 0.788 0.788 473,132 471,275 -0.087 -0.088
64 0.809 0.809 407,865 405,728 -0.123 -0.124
65 0.858 0.858 360,896 358,476 -0.085 -0.086
Note: SSW – social security wealth – is defined as the sum of all discounted pension benefits and social
security contributions. The accrual rate is defined as the relative year-to-year change in SSW.
25
C Appendix
Table : Forward-looking social security incentives — Option Value
Ret. age
Before the change After the change
low wage avg. wage high wage SD low wage avg. wage high wage SD
56 10,84 14,364 20,048 3,794 10,837 14,342 20,571 4,025
57 4,491 6,136 8,97 1,85 4,49 6,126 9,503 2,087
58 0 0 343 162 0 0 883 416
59 1,777 912 0 726 1,462 566 156 545
60 4,117 2,366 178 1,612 3,533 1,734 0 1,442
61 6,752 4,334 1,19 2,277 5,932 3,441 728 2,125
62 9,615 6,862 3,295 2,587 8,795 5,969 2,834 2,435
63 12,107 8,449 3,538 3,511 11,287 7,556 3,076 3,357
64 15,67 11,426 5,68 4,094 14,85 10,532 5,218 3,939
65 17,407 12,743 6,41 4,507 16,587 11,85 5,948 4,352
Note: SD stands for standard deviation.
Table : Forward-looking social security incentives — Peak Value
Ret. age
Before the change After the change
low wage avg. wage high wage SD low wage avg. wage high wage SD
56 -3,313 -4,652 -6,978 1,514 -3,432 -4,917 -7,327 1,605
57 -3,174 -4,574 -6,909 1,541 -3,389 -4,935 -7,356 1,633
58 -41,86 -49,189 -61,378 8,05 -39,189 -46,229 -58,032 7,774
59 -45,032 -51,238 -61,712 6,883 -42,694 -48,747 -58,765 6,627
60 -45,945 -53,334 -65,551 8,084 -43,872 -51,011 -63,014 7,898
61 -46,429 -56,17 -72,448 10,733 -46,556 -56,351 -72,719 10,793
62 -41,252 -44,409 -49,634 3,456 -41,383 -44,597 -49,915 3,518
63 -50,241 -55,864 -65,267 6,199 -50,372 -56,05 -65,547 6,26
64 -30,782 -36,873 -46,97 6,676 -30,914 -37,061 -47,252 6,738
65 -40,928 -46,833 -56,75 6,528 -41,061 -47,024 -57,036 6,591
Note: SD stands for standard deviation.
26
D Appendix
Table : Descriptive statistics
variable
control group treatment group
Mean Std. Dev. Min. Max. Mean Std. Dev. Min. Max.
inactivity status 0.17 0.38 0 1 0.42 0.49 0 1
elementary educ. 0.09 0.29 0 1 0.12 0.32 0 1
apprenticeship 0.54 0.50 0 1 0.50 0.50 0 1
high school educ. 0.24 0.43 0 1 0.25 0.43 0 1
lower tertiary educ. 0.01 0.10 0 1 0.01 0.09 0 1
upper tertiary educ. 0.11 0.32 0 1 0.12 0.32 0 1
unmarried 0.04 0.21 0 1 0.04 0.20 0 1
married 0.84 0.37 0 1 0.84 0.37 0 1
widowed 0.04 0.20 0 1 0.05 0.22 0 1
divorced 0.07 0.26 0 1 0.07 0.26 0 1
before the policy change 0.22 0.42 0 1 0.25 0.43 0 1
1-1.5 year after the policy change 0.24 0.43 0 1 0.26 0.44 0 1
1.5 - 3 years after the policy change 0.28 0.45 0 1 0.26 0.44 0 1
3 - 4.5 years after the policy change 0.26 0.44 0 1 0.23 0.42 0 1
number of household members 2.60 1.07 1 11 2.41 0.97 1 10
age 56.90 0.94 55.0 58.8 59.72 0.78 58.25 62.25
Standard errors are in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
27
E Appendix
Table : Econometric results of the full baseline model
Model 1 Model 2 Model 3
eligible age (old) 0.281*** 0.275*** 0.274***
(0.0145) (0.0146) (0.0147)
1–1.5 year after the policy change (after1) -0.0234* -0.0180 -0.0205
(0.0135) (0.0136) (0.0136)
1.5–3 years after the policy change (after2) -0.0135 -0.0110 -0.0106
(0.0143) (0.0144) (0.0144)
3–4.5 years after the policy change (after3) -0.0223 -0.0193 -0.0223
(0.0146) (0.0147) (0.0146)
interaction variable (oldxafter1) -0.0159 -0.0108 -0.00922
(0.0180) (0.0182) (0.0182)
interaction variable (oldxafter2) -0.0509*** -0.0340* -0.0318*
(0.0179) (0.0184) (0.0184)
interaction variable (oldxafter3) -0.0457** -0.0354* -0.0317
(0.0187) (0.0189) (0.0191)
apprenticeship -0.125*** -0.131***
(0.0130) (0.0131)
high school educ. -0.191*** -0.188***
(0.0108) (0.0109)
lower tertiary educ. -0.162*** -0.161***
(0.0237) (0.0224)
upper tertiary educ. -0.250*** -0.243***
(0.0076) (0.0077)
unmarried 0.109*** 0.118***
(0.0228) (0.0231)
widowed 0.0454** 0.0479**
(0.0199) (0.0199)
divorced 0.0377** 0.0369**
(0.0171) (0.0172)
number of household members -0.0157*** -0.0161***
(0.0045) (0.0046)
76 districts (not reported) X
Observations 50,152 50,152 50,152
Pseudo R-squared 0.07 0.11 0.14
Standard errors are in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
28
F Appendix
Table : Estimated coefficients from the probit model in three different specifications
(yearly periods)
Model (1) (2) (3)
OLD*AFTER1 -0.0067 -0.0033 -0.0009
(0.0195) (0.0198) (0.0199)
OLD*AFTER2 -0.0689*** -0.0573*** -0.0564***
(0.0201) (0.0204) (0.0203)
OLD*AFTER3 -0.0623*** -0.0435** -0.0366*
(0.0198) (0.0204) (0.00206)
Personal characteristics X X
District dummies X
N 33,842 33,842 33,842
Pseudo R-squared 0.07 0.11 0.14
Note: Coefficients are recalculated into the probability measure (min 0, max 1). The
excluded variables are dummies for: control group, one period before policy change,
interaction of control group and all periods. Standard errors are in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
29
G Appendix
Table : Estimated coefficients from the probit model in three different specifications
(dependent variable being employed)
Model (1) (2) (3)
OLD*AFTER1 0.0117 0.0054 0.0049
(0.0188) (0.0193) (0.0193)
OLD*AFTER2 0.0419** 0.0226 0.0196
(0.0192) (0.0198) (0.0199)
OLD*AFTER3 0.0467** 0.0351* 0.0312
(0.0197) (0.0201) (0.0203)
Personal characteristics X X
District dummies X
N 33,842 33,842 33,842
Pseudo R-squared 0.07 0.11 0.14
Note: Coefficients are recalculated into the probability measure (min 0, max 1). The
excluded variables are dummies for: control group, one period before policy change,
interaction of control group and all periods. Standard errors are in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
30
H Alternative specification with narrow treatment
and control group (one-year age win-
dow)
Table : Marginal effects probit
Variables Active Employed Unemployed
OLD1*after1 -0.000 0.011 0.009
(0.028) (0.030) (0.014)
OLD1*after2 0.072*** 0.045 0.038**
(0.023) (0.027) (0.020)
OLD1*after3 0.053** 0.040 0.013
(0.024) (0.027) (0.015)
OLD2*after1 0.030 0.024 0.002
(0.028) (0.030) (0.015)
OLD2*after2 0.079*** 0.062** 0.034*
(0.025) (0.029) (0.025)
OLD2*after3 0.081*** 0.078** 0.008
(0.026) (0.029) (0.018)
OLD3*after1 0.010 -0.019 0.003
(0.031) (0.033) (0.019)
OLD3*after2 0.069** 0.051 0.042
(0.027) (0.031) (0.035)
OLD3*after3 0.041 0.031 0.034
(0.031) (0.033) (0.035)
Note: Coefficients are recalculated into the probability measure (min 0, max 1).
Standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1
31
Table : Linear regression
Variables Active Employed Unemployed
OLD1*after1 -0.007 -0.005 0.014
(0.027) (0.028) (0.015)
OLD1*after2 -0.080*** 0.047* 0.038**
(0.0260) (0.0272) (0.015)
OLD1*after3 -0.060* 0.045* 0.015
(0.026) (0.027) (0.016)
OLD2*after1 -0.044 0.035 0.006
(0.029) (0.030) (0.015)
OLD2*after2 -0.098*** 0.074** 0.030*
(0.032) (0.032) (0.016)
OLD2*after3 -0.1056*** 0.0938*** 0.010
(0.0315) (0.0323) (0.016)
OLD3*after1 0.007 -0.015 0.011
(0.030) (0.0305) (0.017)
OLD3* after2 -0.081** 0.057* 0.034*
(0.032) (0.0331) (0.019)
OLD3* after3 -0.051 0.037 0.025
(0.0336) (0.034) (0.020)
Note: Control group is defined by the distance 3–4 years before retirement. The other
control variables are year, control and treatment group fixed effects are family status,
education and districts. *** p<0.01, ** p<0.05, * p<0.1
32
I Appendix
Table : Estimated coefficients from the multinomial logit
Model (1) (2) (3)
employed
OLD*AFTER1 1.0601 1.0279 1.0220
(0.1036) (0.1041) (0.10496)
OLD*AFTER2 1.2661** 1.1538 1.1350
(0.1311) (0.1226) (0.1227)
OLD*AFTER3 1.2434** 1.1869 1.1546
(0.1336) (0.1305) (0.1294)
unemployed
OLD*AFTER1 1.1994 1.1923 1.1943
(0.3075) (0.3062) (0.3096)
OLD*AFTER2 1.9621*** 1.9149** 1.9371***
(0.4999) (0.4888) (0.4978)
OLD*AFTER3 1.3542 1.3369 1.3278
(0.3726) (0.3679) (0.3672)
Personal characteristics X X
District dummies X
N 50,152 50,152 50,152
Pseudo R-squared 0.06 0.10 0.13
Note: Coefficients are recalculated into the probability measure (min 0, max 1).
Standard errors are in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
33