How Aging of the Labor Force A¤ects
Equilibrium Unemployment
January 17, 2006
Abstract
This paper argues that aging of the labor force a¤ects job creation
and job destruction. To analyze this, we extend a standard model
of equilibrium unemployment and search in the labor market by the
distinction between age speci…c separation risks and a productivity
di¤erential between young and elderly workers. Based on the theoretical model, we identify four regimes of changes in the Beveridge
Curve and job creation which can occur if the age structure varies.
We also present an econometric model to …nd out which country relates to which regime. According to the estimates we can identify all
four cases. For example, Spain and the US may experience positive
e¤ects on employment when the labor force grows older. In contrast
to this, the unemployment rate in Japan will rise as a consequence of
the increase in the share of elderly workers.
Keywords: Vacancies and Separations, Unemployment, Job Creation,
Aging of the Labor Force, Demographic Change
JEL classi…cation: J63, J64, J23, J21, J10
1
1
Introduction
With more and more elderly workers being employed, the labor markets of
nearly all developed economies will go through deep changes . The share of
the age group 55 to 64 years in the US is increasing by one third over the next
twenty years. Europe and Japan will experience an even more considerable
development. Besides other e¤ects, the altered age composition is supposed
to a¤ect the labor market via changes in job and labor turnover, namely job
creation, job destruction and job-search behavior. Young job seekers may
be di¤erent from old ones in their incorporated skills and the di¤erent time
horizon, for example the remaining time until retirement, which a¤ects the
separation risk of a job-worker match. We therefore expect e¤ects of the
aging working population on unemployment and vacancies.
Recent papers on aging deal in particular with the change in consumption,
savings and growth, and the impact on the pension systems.1 Labor market consequences of population aging are discussed in terms of reduced total
labor supply including feedbacks of capital intensity, the e¤ects of the age
structure on labor productivity, shifts in the aggregate or cohort wage level,
and changes in goods demand which a¤ect labor mobility (see Börsch-Supan,
2003; Johnson and Zimmermann, 1993). This paper analyzes another aspect
of aging and the labor markets, which is widely ignored in the literature to
date: The impact of the age structure on the search equilibrium on the labor
market and equilibrium unemployment. Some parts of the empirical literature on search and the matching function include demographic variables but
only as a sideline of the analysis. Coles and Smith (1996) …nd that matching
decreases with an older working population for England and Wales. Other
authors argue that separation rates are higher for younger workers as they
are more likely to undertake on-the-job search. Pissarides and Wadsworth
(1994) and Burgess (1993) …nd evidence for Great Britain. However, the
existent theoretical and empirical literature does not allow drawing any conclusions with regard to demographic e¤ects on variations in unemployment
if e.g. job separation and matching decline in equal size. Hence, the question
of how aging a¤ects search and matching on the labor market and thereby
equilibrium unemployment has not been answered yet. To our knowledge,
this paper is the …rst contribution to the literature which deals with this
issue.
The aim of this paper is to identify and estimate the e¤ects of aging on
unemployment via changes in the ‡ows on the labor market related to the
1
See, for example, Batey and Madden (1999), Bloom and Canning (2004), Bloom et al.
(2003), Breyer and Stolte (2001), Butrica et al. (2004), Ehrlich and Kim (2005), Sneddon
and Triest (2001, 2002), Miles (1999) and Sellon (2004).
2
matching function and changes in job creation. For this, we develop a model
of equilibrium unemployment which follows the standard search models of
Diamond-Mortensen-Pissarides. The results will depend on the assumption
on relative separation rates and relative productivity between young and
elderly workers. We identify four regimes with di¤erent changes in the Beveridge Curve and job creation which can occur if the age structure varies.
Only in two out of the four cases the theoretical outcome for changes in unemployment as the job seekers get older is clear-cut. This makes econometric
estimations necessary, which we undertake for nine countries (the US and
Japan in addition to selected European economies). We …nd all four cases
in our empirical results. For example, Spain and the US may experience
positive e¤ects on employment when the labor force grows older. In contrast
to this, unemployment in Japan will rise as a consequence of the increase in
the share of elderly workers. In the other countries the results depend on
whether we consider the aging of the employed workers, which includes the
e¤ects of on-the -job search, or take the unemployed as the only job seekers.
The remainder of the paper is organized as follows. In section 2 we extend
the standard model of search and equilibrium unemployment by age e¤ects.
Section 3 presents the econometric model and reports the estimation results.
Finally, we summarize our results in section 4.
2
The Model
Our modeling is a simple extension of the standard framework of search and
equilibrium unemployment (see Pissarides, 2000). The search equilibrium
of the labor market is given by the extent of job creation subject to the
optimal job posting of …rms, job destruction and the matching technology,
manifested in the Beveridge Curve. We examine the e¤ects of aging on the
search equilibrium which arise from a change in the age structure of the
labor force but we ignore size e¤ects of a decline in population.2 The way we
introduce heterogeneity into the labor force follows Acemoglu (1997), who
distinguished between high-skilled and low-skilled workers. In contrast to
this, we di¤erentiate between young and elderly workers who may be di¤erent
not only in productivity but also in their separation risk. With respect to
the separation rate one can think of two reasons for di¤erences between the
young and the elderly: On the one hand, old workers may separate because
2
Most of the existing empirical studies suggest constant returns to scale of matching
functions (see Petrongolo and Pissarides, 2001 for an overview). Therefore, no size e¤ects
on search in the labor market are expected if the population shrinks due to demographic
change.
3
they retire before the match gets unproductive. On the other hand, young
workers bring the current match to an end because they leave for better
jobs, whereas older ones stay, for example because of tenure rents or higher
mobility costs. In the model we consider di¤erences in the separation risk
and in productivity, but we make no assumption on whether they are higher
or lower for one age group.
2.1
Trade in the Labor Market
There is a continuum of workers normalized to 1 and a larger continuum
of …rms. Each …rm can decide to be inactive at zero return or can open a
vacancy at ‡ow cost . Each vacancy can employ only one worker. The
total labor supply is divided into two age groups. Workers are young at a
share p; henceforth symbolized with superscript y, and elderly at a fraction
1 p, identi…ed by superscript e. The elderly and the younger workers are
identical in all respects apart from a productivity di¤erential and a di¤erent
expected duration as part of a job-worker match. The formed matches come
to an end because of exogenous technological shocks, which a¤ect matches
with young and elderly at the same probability. However, retirement3 and
a di¤erent intensity of on-the-job search produce di¤erent separation risks
of young workers and their elderly colleagues. The probability to separate
with the job is denoted by sy and se respectively. The di¤erence in the two
separation rates is given by a positive value of :
s y = se :
(1)
Therefore, the expected duration of a match is of equal length for young and
elderly workers if = 1, but it can be di¤erent from unity, too. The average
separation rate then comes from the shares of young and old according to
s = (p + 1
p) se :
(2)
New employment relations are created through a standard matching technology which forms the number of matches from the number of unemployed
workers and the number of vacancies. With a population normalized to unity
the matching rate is given by:
(3)
mt = M (ut ; vt );
where ut is the unemployment rate, vt is the vacancy rate and M (ut ; vt ) is
the ‡ow rate of matches formed at time t. As standard, M (ut ; vt ) exhibits
3
For simplicity suppose that retirement does not change the total population because
the in‡ux of young workers exactly replaces the retired workers.
4
constant returns to scale in its two arguments4 , is continuous and di¤erentiable, and M (ut ; vt ) < 1. De…ne = v=u as a measure of the tightness
of the labor market. Then the ‡ow rate of matches for an un…lled vacancy,
q( t ), is equal to:
q( t ) =
M (ut ; vt )
, with q 0 ( t ) < 0.
vt
(4)
The share of workers who enter unemployment during a small time interval
is s(1 ut ), while t q( t ) is the transition probability for ut unemployed. The
evaluation of unemployment is given by the di¤erence between the two ‡ows,
u_ t = s(1
ut )
(5)
t q( t )ut :
We can rewrite eq. (5) as an equation determining unemployment in terms
of the two transition rates:
(p + 1 p) se
u=
(p + 1 p) se + t q( t )
(BC).
(6)
By the properties of the matching function, equation (6) represents the socalled Beveridge Curve (BC), a convex and downward-sloping curve in the
( ; u) space. For constant parameters of the model, in particular a stable
age distribution p=1 p, the value of the market tightness …xes the unemployment rate. The unknown is explained by the willingness to create vacancies
by the …rms, put down in the next section.
2.2
Job Creation
The following equilibrium of job creation will be characterized through a set
of Bellman equations, which de…ne the values of vacancies and jobs. Search
models generally assume foresighted …rms, especially when they optimize
their job posting. Compared with the standard analysis, the following equations consider the probabilities of matching the vacancy with a young and
an old worker respectively. The age of a worker a¤ects the revenues of a
match via a di¤erent length of a match and di¤erent pro…ts generated at one
point in time. This is because old-age retirement and on-the-job-search argue
for a di¤erent separation risk, while the pro…ts depend on the productivity
di¤erential between young and old workers. Let J y denote the net present
discounted value of a …rm that employs a young worker when the job is …lled,
4
Most of the empirical papers …nd constant returns to scale. See, for example, Blanchard and Diamond (1989, 1990), Burda (1993), Coles and Smith (1996), Layard et al.
(1991), Pissarides (1986), and van Ours (1991, 1995).
5
and V when the state of the vacancy is un…lled. The discount rate which
values future income streams is r. Similarly, for a …rm which employs an
elderly worker we use J e and again V . This yields:
rJ e =
rJ y =
we
+
rV =
se (J e
wy
sy (J y
+ q( t ) [p (J
y
(7)
V)
(8)
V)
y
V ) + (1
p) (J
e
e
V )]
(9)
Eq. (7) implies that the gain from a job …lled with an elderly worker is the
output of the worker less its wage cost we minus the expected value of the
capital loss. The probability of this loss is the separation risk and the value of
the …lled vacancy is then replaced by the value of an un…lled vacancy. From
similar arguments follows in eq. (8) the value of a job …lled with a young
worker, who di¤ers in his/her productivity from the old one by R 0. This
means that young workers can be equal, more or less productive than their
older colleagues.5 Assessing the value of a vacancy in eq. (9) has to consider
the ‡ow costs of the job posting faced by the possible additional value of
the asset when the state of the vacancy changes with probability q( t ) from
un…lled to …lled. In this case p is the chance to …ll it with a young worker
and 1 p is the chance to employ an elderly one.
As soon as a vacancy is …lled, workers and …rms share the value of the
match and the wage is given by the fraction of the output levels,
wy =
(10)
( + );
and
we =
(11)
:
Equilibrium requires that the value of a vacancy is zero, otherwise …rms
would open an in…nite number of v. Hence, V = 0 and from eq. (8) and eq.
(10) then follows that the value of a match with a young worker is:
Jy =
1
( + );
r + sy
(12)
If an elderly employee …lls the vacancy, eq. (7) together with eq. (11) implies
that the value becomes:
Je =
1
:
r + se
(13)
5
See Börsch-Supan et al. (2005) on the di¢ culty of the measurement of individual
productivity. Hence, even if we would apply a micro econometric approach in section 3,
it seams not to be advisable to use a proxy for productivity.
6
Wether it is more pro…table for a …rm to employ a young or an elderly
worker depends on the productivity di¤erential and the di¤erential in the
separation rates . For example, the expectation of a relative longer match
duration in the case of employing elderly workers could compensate their
lower productivity, and the other way around. The comparison between the
values of J y and J e yields that:
Jy R Je
if
Q1+
r + se
se
(14)
A di¤erential between J y and J e does not mean that vacancies are not
…lled with either young or elderly workers. A …rm will not wait for the chance
of a higher valued match in the next period if the chance is low or waiting
costs are high. This is the case if the value of a vacancy eq. (9) is equal
to zero for a given age distribution. Hence, a …rm is indi¤erent between
employing a young or elderly worker when she/he faces the open vacancy if:
Je =
1
1
p
q( t )
pJ y :
(15)
Only the market tightness is variable and guarantees the identity of eq. (15),
while J y and J e are identi…ed by parameters. Therefore we substitute J y
and J e in the equation with the expressions from eq. (12) and eq. (13). This
yields the job creation curve (JC) which has only one value of to solve the
equation.
Q( t ) =
1
1
=
q( t )
(1
p)
r+
se
+
p( + )
(1 p) (r + se )
(JC) (16)
The vacancy-matching ratio Q( t ) is an indicator for job creation. If Q( t )
increases …rms open more vacancies for a given matching technology. Q0 > 0
and the ratio increases via a rise in t . As a result, JC determines the market
tightness and establishes together with the BC the search equilibrium with
the equilibrium values ; u . Figure 1 reveals a graphical illustration with the
corresponding JC-curve and BC and the intersection of both as equilibrium.
2.3
The E¤ects of Aging
A change in the age distribution p=(1 p) has e¤ects on the search equilibrium
; u if young workers di¤er from their elderly colleagues with respect to
productivity and separation risk. The relative size of the age groups a¤ects
the amount of job creation through the willingness of …rms to open vacancies
which depends on the expected value of a match. An aging labor force may
7
change this value. Furthermore, the age distribution a¤ects the average rate
of job destruction.
The comparative static analysis yields that an increase in the productivity
t)
> 0 => t ". The di¤erdi¤erential shifts the JC-curve to the left: @Q(
@
ential measures an extra productivity whose rise also increases the average
productivity and the expected pro…ts of a …rm with an open vacancy. Higher
expected pro…ts mean more vacancies and less unemployment. Furthermore,
for a given separation risk of the elderly, an increase in the di¤erential parameter raises the mean separation rate. Accordingly, the JC-curve shifts to
t)
< 0 => t #.
the right: @Q(
@
However, the most interesting results come from the e¤ects of a labor
force that is growing old. This implies that the share of younger worker p
falls. The e¤ect on job creation yields:
+
@Q( t )
=
@p
r + se
r + se
= Jy
J e:
(17)
t)
Less young workers over time shifts the JC to the right, namely @Q(
>0
@p
r+s
y
e
if J > J ; respectively if < 1 + s . In this case, matches with young
workers are more gainful. The probability increases to …ll a vacancy with an
elderly worker and, consequently, the …rms reduce their number of vacancies.
As a result, the aging working force tends to reduce the market tightness. In
contrast to this, the number of job posting will increase with the rise in the
mean age if elderly employees are more productive or stay longer in the …rm.
t)
< 0; if J y < J e equal to > 1 + r+s
:
That is, JC shifts to the left, @Q(
@p
s
Aging does not a¤ect only job creation but has e¤ects also on job destruction. As separation rates di¤er between young and elderly employees,
aging will change the average duration of a match. Obviously the change in
the BC with respect to the share of younger workers,
@u
=(
@p
1) se
t q( t )
[(p + 1
p) se +
2
t q( t )]
;
(18)
depends on whether the young or the elderly workers have a lower probability
of bringing the match to an end. Hence, aging shifts the BC inwards @u
>
@p
0 if > 1. The increase in the mean age of the workers tends to lower
unemployment if job destruction is reduced via the low separation rate of
elderly workers. In contrast to this, if retirement is a considerable risk for
the …rms to end a productive match, more elderly workers lead to more job
destruction and ‡ows on the labor market. The BC shifts outwards, @u
< 0;
@p
in case of < 1.
8
v
τ > 1+ δ
r + se
se
τ <1+ δ
JC
r + se
se
τ <1
τ >1
BC
u
Figure 1: The e¤ects of aging on the search equilibrium
Figure 1 reveals the possible outcomes that arise from considering jointly
changes in BC and JC. The results are also summarized in Table 1. Solving
the model produces either more or less job creation and job destruction for a
labor force which grows old. Di¤erent combinations of the e¤ects of aging on
unemployment are possible. For example, more old-age retirement reduces
the mean duration of a match in regime (1). This shifts the BC outwards.
In case of > 0 this e¤ect is not compensated by a higher productivity of
the elderly and the average value of a match reduces. Hence less jobs are
created and the JC-curve shifts to the right. Both e¤ects imply an increase
in the unemployment rate due to the demographic change, but the change in
vacancies is not clear-cut. In regime (2) less jobs are destroyed if the mean
age of the working force goes up. But as elderly workers are less productive
the average value of a match still decreases and reduces the number of job
o¤ers. The two e¤ects work in di¤erent directions and in this case the e¤ect
on unemployment is ambiguous, while the vacancy rate decreases. The third
regime yields a clear reduction of unemployment as soon as the population
grows old because elderly workers have a so much lower probability to separate that this advantage outweighs their productivity disadvantage. The
value of an average match thus increases and results in more job creation
while the rate of job destruction falls coincidently. In the previous regimes
we assumed that > 0. However, if the elderly are more productive than
the younger workers, aging means that more jobs are created on average to
bene…t from the high productivity of the elderly. If the retirement risk is high
9
at the same time, job creation increases, but job destruction rises as well.
The overall e¤ect on unemployment is ambiguous, although the economy will
create more vacancies.
regime
(1)
(2)
>0
(3)
(4)
<0
8
>
>
<
>
>
:
Table 1: The e¤ects of aging
BC
JC
<1
1<
<1+
>1+
1>
r+se
se
r+se
se
>1+
r+se
se
o
r
i
r
i
l
o
l
o=outward, i=inward, l=left, r=right, +=increase, -=decrease,
3
u
v
+
+
=ambiguous e¤ect
Estimation and Results
Even with micro data it is di¢ cult to estimate the model discussed in section
2. Since our concern is to compare di¤erent economies, we decided to use
macro data and reduce the econometric model to the essential elements.
Our main objective is to use the estimation results in order to di¤erentiate
between the four regimes in Table 1.
There are several possibilities for constructing a proxy variable for the
aging of the labor force. We decide to use the ratio of young (age cohort 16
to under 30) to old (age cohort 50 to 64). Furthermore, we use this proxy
for two di¤erent groups, employed and unemployed. Let denote this proxy
in either case. Since individuals from both groups may look for a new job,
one should analyze whether …rms react di¤erently or similarly on the age
structure of the employed and the unemployed.6
Figure 2 shows the di¤erent developments of the aging proxy of the employed for the countries we are going to analyze in this section. Unfortunately,
the data is not available for the whole period between 1960 and 1999 for all
considered countries. We see as a common pattern that aging starts somewhere in the 80s or early 90s when the ratio young to old fell considerably.
6
Burgess (1994) allows in his theoretical model for job search of both employed and
unemployed, which has signi…cant e¤ects on the unemployment dynamics. Van Ours
(1995) di¤erentiates between the two groups and estimates higher ‡ow elasticities for the
unemployed.
10
2,5
2,3
2,1
1,9
1,7
1,5
1,3
1,1
0,9
0,7
0,5
1960
1965
Canada
Portugal
1970
France
Spain
1975
1980
Germany
Sweden
1985
1990
Japan
US
1995
Norway
Figure 2: Employed with age under 30 / Employed with age 50 to 64
8,0
7,0
6,0
5,0
4,0
3,0
2,0
1,0
0,0
1960
Canada
1965
1970
France
1975
1980
Germany
1985
Japan
1990
Spain
1995
US
Figure 3: Unemployed with age under 30 / Unemployed with age 50 to 64
11
Before that time, the working force got rather younger on average. The exception to this observation is Japan, where the aging process is ongoing since
the late 60s and Spain where the pattern is not fully de…nite. According
to the proxy variable the Canadian employees are the youngest among the
selected group of countries, whereas Japan experiences the strongest aging
process of all mapped countries. It is interesting to see that the Nordic countries had a comparatively high average age of the employees already in the
sixties and early seventies. This can be explained by the typical high labor
market participation in these countries.
Comparing the values of Figure 3 with 2 reveals that the unemployed
are on average younger than the employed workers. The only exception is
Germany up to 1974 and since 1988. Furthermore, Germany and Japan have
the lowest proportion of young to elderly unemployed since the middle of
the seventies. A considerably decline in this number is found for Canada,
Spain, and the US with the beginning of the 80s. The data is not available
for Portugal, Norway, and Sweden.
3.1
Econometric Model and Data
We have to estimate two equations to di¤erentiate between the four regimes.
Aging a¤ects the locus of the BC and the JC-curve as pointed out in section
2.3. Hence, the …rst equation to be estimated is the BC. In the second case we
estimate the JC-curve. Consequently, we regress the unemployment rate and
the tightness of the labor market on the proxy variable for aging according
to:7
log(ut ) =
0+
1 log(vt ) +
2
t+
I
P
i Xit
+ "1t ;
(19)
i=3
log( t ) =
0
+
1 t
+
I
P
i Xit
+ "2t :
(20)
i=2
The estimated e¤ects of aging, 2 and 1 , reveal the moves of the BC
and identify the regime according to Table 1. The set of control variables X
comprises bene…t replacement rate, bene…t duration, employment population
ratio, union density, labor costs, and the real interest rate.
Because of other e¤ects which are not considered here, it is possible that
the error terms "1 and "2 may be correlated across the equations of the
7
The use of the logarithm of the proxy for aging would estimate the wrong functional
form if the parameter is positive but less than one. In this case the relationship between
unemployment and aging is a monotonic increasing concave function. This would be
contradictory to the theoretical model with a monotonic increasing convex function.
12
system. To allow for this possible outcome, we use the following assumption
on the error terms in the system:
E("1t "2t ) =
12
E("1t "2t ) = 0 with t 6= s.
(21)
That is, the error terms are homoskedastic and independent across t, but
may be correlated across the equations. Therefore, we estimate a seemingly
unrelated regression (SUR) model with unknown covariance matrix.
We estimate several speci…cations for each country to control for the effects of multicollinearity.8 In all cases we start with the full set of variables.
We then sequentially remove a variable based on the information of the correlation matrix of the variables, and estimate the system again. We repeat this
process until the remaining simple correlations are below 0.8. To improve the
signi…cance of the remaining parameters, we …nally remove variables from the
system which are insigni…cant in both equations. This procedure has been
carried out for both proxies of aging.
The data for the unemployment rate, the vacancy rate, and the control
variables is taken from Nickell and Nuncita (2002). For aging we use two
di¤erent proxies. First, we use the ratio of the employees under 30 years old
to those 50-64 years old. Secondly, we do the same with the unemployed.
These time series are taken from the OECD online database. We undertake
the investigation for Canada, France, Germany, Japan, Norway, Portugal,
Spain, Sweden, and the US over the period from 1960 to 1995. Due to data
availability the actual period starts after 1960 for most countries.
3.2
Results
We estimated almost 60 speci…cations of the system for the nine countries.
To make sure that the illustration remains clear, we focus on the two relevant
parameters 2 and 1 only, which represent semi-elasticities. For the complete results see appendix. It should be mentioned that the o¢ cial vacancy
statistics report only a fraction of un…lled jobs in the economies. However,
it is not possible to account for this problem for each country. Therefore,
the interpretation of the estimates has to be done carefully and standardized
vacancy rates are badly needed.
First, we discuss the results for the ratio of young to old employees as the
proxy for aging. We …nd the regimes with > 0 only. Table 2 summarizes
all countries. The …rst regime implies an outward shift of the BC and a
clockwise rotation of the JC-curve. This means that unemployment will
8
In particular the considered labor market institutions have a low volatility and are
highly correlated among each other.
13
rise as a consequence of an aging labor force with ambiguous e¤ects on the
vacancies. The negative e¤ect of aging in the BC equations implies that
unemployment increases for a given number of vacancies. The coe¢ cients
for Japan and France are below unity. However, according to the used proxy,
these are the countries with the longest aging process. The proxy declined
between 1968 and 1999 from 2.02 to 0.81 in Japan and it fell from 1.75 in
1979 to 1.01 in 1999 in Germany. Therefore, the average annual growth rate
of is nearly the same for these countries (-2.6% for Japan and -2.8% for
Germany).9 However, the impact of aging on unemployment is much higher
in Germany as we will see later on. From clockwise rotation of the JC-curve
follows that vacancies decrease and unemployment increases. Again, the
estimated e¤ects of aging are low in France and Japan and high in Germany,
Portugal, and in Sweden. The total e¤ect of the change in vacancies depends
on the curvature of the BC estimated in the …rst equation.
In the second regime the BC shifts inwards and the JC-curve again rotates
clockwise. The outcome is ambiguous in terms of the expected change in
unemployment but the vacancy rate will de…nitely decrease. The positive
coe¢ cient for aging indicates the inward shifts of the BC. The consequent
e¤ect on the unemployment rate is much lower for Canada, Spain, and the
US than for Norway (third regime). The aging proxy for Spain undulates
and decreases only since the beginning of the 1990s. The average annual
growth rate of the employment proxy for Canada and the US is 2.8% and
for Norway 3.5%. Since we found the highest estimated e¤ect for Norway, it
does not surprise that the total e¤ect on unemployment is higher in Norway
than in the other countries. The positive e¤ect of aging in the JC-curve in
the countries of the second regime is the same as in the group of the …rst
regime. We cannot conclude on the basis of the two equations whether the
overall e¤ect of aging on unemployment is positive or negative because the
BC shifts inwards and the JC-curve rotates clockwise. For this reason, we
calculate the net e¤ect in the next section.
To sum up, except for Norway the estimates with the employment proxy
identify only the …rst and the second regime. They indicate a negative e¤ect
of aging on job creation but di¤erent shifts of the BC. According to the
model of section 2 this implies a productivity disadvantage of the elderly
workers, namely has a positive sign. Sneddon and Triest (2002) …nd a
signi…cant negative e¤ect of the growth rate of the working age population
on average productivity in the US. This coincides with our …ndings because
a fall in the share of the young increases their relative productivity. Beside
9
The average yearly growth rate is calculated from the peak point when the aging
process started, which is not necessarily the …rst observation.
14
15
US
Sweden
Spain
Portugal
Norway
Japan
Germany
France
Canada
(1.546)
(9.444)
0.996
(2.539)
(-4.105)
(-7.853)
-0.556 -0.682 -0.689 -0.661
(-3.486)
(2.107)
(2.104)
JC
(3.043)
3.323
(-0.667)
-0.435
(3.858)
1.561
(4.925)
3.224
(1.105)
1.273
(6.140)
1.328
(3)
(11.327)
0.934
(0.199)
0.322
(0.618)
(0.337)
0.180
0.187
(0.400)
2
1
2
1
3
1
1
1
2
Regime
The Table contains the estimates for 2 and 1 . For each country di¤erent system speci…cations are estimated.
T-statistics are in parentheses. For complete results see appendix.
(9.833)
0.900
0.102
(9.675)
(3.742)
0.909
(4.013)
(9.631)
(3.489)
0.911
(-3.275)
(3.257)
(-2.049)
14.610 15.178 17.115 7.823
(7.689)
(-2.020)
(7.463)
(-2.251)
(5.892)
(-0.912)
-0.601
(6.852)
1.992
(4.870)
3.202
(1.285)
1.516
(6.160)
1.325
(4)
-2.950 -2.869 -2.897 -1.821
(2.513)
0.803
(2.943)
2.518
(-1.706)
-1.328
(3.746)
1.537
(5.735)
3.365
(1.000)
1.175
(1.159)
0.571
(2)
10.795 10.827 10.311
0.574
0.610
(-2.809)
(4.247)
(-0.162)
3.007
(2.530)
1.242
(-3.216)
(3.210)
1.382
(-4.257)
(6.833)
(4.482)
(-4.299)
(-4.163)
-1.580 -1.175 -1.185
2.386
2.293
(-4.215)
(-4.430)
-0.160
3.561
-2.587 -1.747 -2.404 -2.377
(-4.519)
(0.999)
(-0.218)
1.178
(-0.192)
(-0.191)
(2.687)
1.309
(1)
(-2.004)
(1.548)
0.204
(4)
-0.176 -0.192 -0.218 -0.191
(2.211)
0.204
BC
(3)
(-0.444)
(2)
-0.045 0.380
(1)
Table 2: Aging of the employed
this age composition e¤ect of the work force we control for the labor force
participation rate. In this case the ratio of civilian employed to working age
population (15-64 years) has a signi…cant negative e¤ect on unemployment
in most considered countries. Since the participation rate is lower for the
elderly, a fall in the share of the young increases total unemployment. This
coincides with the …nding by Bloom and Canning (2004).
Up to now we analyzed the e¤ects of aging in the segment of on-the-job
searchers. In the remainder of the section, we look at the segment of unemployed job searchers. Unfortunately, the corresponding data is not available
for Portugal, Norway, and Sweden. If we take the ratio of young to elderly
unemployed as the proxy for aging, we …nd all possible regimes. Table 3 summarizes the results. We identify for Japan and Canada the …rst regime and
for the US the second regime. However, with respect to Germany, France,
and Spain, we now …nd a positive e¤ect on vacancies. That is, in Spain
aging of the unemployed unambiguously reduces unemployment because the
BC shifts inwards and the JC-curve rotates counterclockwise (third regime).
In contrast to this, for Germany and France the BC shifts outwards as before
but the JC-curve now rotates counterclockwise. This represents the fourth
regime and results in an ambiguous e¤ect on unemployment because of the
opposing e¤ects of the changes in the BC and the JC-curve when the age
composition alters.
The previous results imply that the e¤ects of aging on search unemployment depend in some countries on whether we consider on-the-job search or
job search of the unemployed. To be more precisely: While the direction of
the e¤ects on the BC is more or less the same, the response of …rms to aging
with a rise or cut of vacancies can be di¤erent for employed or unemployed
job-seekers. Why do …rms create more jobs when the job-seekers grow old
if they decide to hire former unemployed workers but decrease job openings
if they face on-the-job searchers? And why is this true in France, Germany,
and Spain but not in Canada, Japan, and the US? Aging could not be the
reason because the share of the elderly increases within the employed as well
as within the unemployed. Following the arguments of the model in section
2, we argue that relative productivity di¤erences can explain the story. One
indication for this can be found in Börsch-Supan et al. (2005) and Yashiro
(2001). Early retirement of the age cohort 50 to 64 years has been comparatively high in Southern Europe (Spain) as well as in France and Germany.
More precisely, early retirement plays an important role for men and women
in France and Germany, whereas in Spain the women’s share of home-makers
is nearly as high as the group of early retired man. This low participation,
in turn, tends to increase the average productivity of the elderly unemployed
because it is reasonable to say that the low skilled leave the labor force by a
16
17
US
Spain
Japan
Germany
France
Canada
(1)
(1.650)
0.081
(2.784)
(1.958)
0.103
(3.144)
(3.771)
0.097
(4.918)
(4.937)
0.377
(4)
(4.640)
0.853
(-1.624)
(2.454)
0.222
(-1.197)
(3.198)
0.236
(-2.275)
(3.484)
0.254
(-2.241)
-0.736
(4.764)
0.846
(0.021)
0.003
(-1.700)
(2.986)
0.192
(8.137)
0.931
(0.014)
0.002
(-2.073)
2
3
1
4
4
1
Regime
The Table contains the estimates for 2 and 1 . For each country di¤erent system speci…cations are estimated.
T-statistics are in parentheses. For complete results see appendix.
(2.451)
0.073
(1.627)
0.041
(-2.057)
-0.403 -0.716
0.032
0.040
(-1.342)
(-2.826)
(-3.925)
(3)
0.378
JC
-0.255 -0.334
(-2.270)
0.623
(-3.577)
(0.437)
0.057
(2)
-0.573 -0-489 -0.395 -0.468
(3.603)
(-2.863)
-0.211 -0.210
(-1.917)
(-3.469)
(0.476)
(0.345)
(-1.680)
(-3.099)
0.043
0.030
(-1.244)
-0.035 -0.047
(-1.255)
-0.275 -0.328 -0.329 -0.313
-0.026
(-1.161)
-0.010
(-0.440)
(-1.034)
0.060
(4)
(0.488)
BC
(3)
(-3.230)
(2)
-0.049 -0.047 -0.044 -0.043
(1)
Table 3: Aging of the unemployed
majority. In contrast to this, the low skilled remain in the labor force within
the group of the young unemployed. Putting things together we arrive to the
conclusion that the average productivity of the elderly unemployed may be
higher than (may be nearly equal to) the average productivity of the young
unemployed in France and Germany (in Spain). That is, from the theoretical model is smaller if we look at unemployed instead of employed job-seekers
and gets even negative in France and Germany.
An alternative explanation for Spain (switch from the second to the third
regime) is that the relative separation risk is higher if we take only the job
search of unemployed into account. Due to their high rate of unemployment
the young unemployed are apt to accept the …rst best job o¤er. The experience has shown that these matches have a comparatively shorter duration.
That is, the separation rate of the young unemployed is higher than the rate
of the young on-the-job searchers.
3.3
Calculation of Net E¤ects
In this section we summarize the estimates and, additionally, we want to
…nd out what are the total e¤ects of the estimates on unemployment. With
respect to the two estimated equations we distinguish between a direct and
an indirect e¤ect. The direct e¤ect 2 shifts the BC and the indirect e¤ect 1
leads to moves on the BC. In some of the cases, regime two and four precisely,
we get opposing e¤ects that result in ambiguous changes in unemployment
as a consequence of the aging of the labor force. However, even if both
e¤ects have the same direction it is interesting to know which of the e¤ects
dominates the other. Table 4 shows the di¤erent e¤ects of aging on the
unemployment rate. A negative (positive) sign denotes that aging decreases
(increases) search unemployment. The total e¤ect depends on the direction
of the direct and indirect e¤ects and, if they are opposing, on their relative
magnitude.
If we consider the aging of the employed, we see the rise in unemployment
as expected in regime 1. However, in some cases the direct e¤ect dominates
in other cases it is the other way around. The total e¤ect of regime 2 is
negative in all three countries. That is, the shift of the BC dominates the
move along the curve and unemployment decreases. The picture is a good
deal more mixed if we consider the age structure of the unemployed. Only
in Japan and the US the identi…ed regimes and net e¤ects are the same as
before. For Canada the identi…ed regime changes from two to one and the
net e¤ect on unemployment is now positive. The e¤ect for Spain is negative
as predicted for the third regime. In France and Germany the fourth regime
has the same total e¤ect, thus unemployment decreases with aging.
18
Table 4: E¤ects of Aging on Unemployment
aging of employed
direct
Canada
France
Germany
Japan
Norway
Portugal
Spain
Sweden
US
aging of unemployed
indirect
total
regime
direct
indirect
total
regime
-
>
+
-
2
+
>
+
+
1
+
<
+
+
1
+
<
-
-
4
+
<
+
+
1
+
<
-
-
4
+
>
+
+
1
+
>
+
+
1
-
<
-
-
3
-
>
+
-
2
-
>
-
-
3
+
<
+
+
1
-
>
+
-
2
+
>
+
+
1
-
>
+
-
2
Taking all results into consideration, we can conclude that no negative
consequences for unemployment are expected for Spain and the US when the
mean age of the labor force is going to increase continuously. With respect
to the US, the results coincide with those of Bleakley and Fuhrer (1997) and
Katz and Krueger (1999). On the other hand, the results for Japan showed
for either measure that this countries has to prepare for a further increase in
unemployment when the labor force grows older.
4
Conclusions
In this paper, we examined the relationship between the aging of the labor
force, according to the demographic change, and unemployment by means
of both a theoretical and an empirical model. The modeling applies to the
literature on search in the labor market and matching with equilibrium unemployment. We extended the standard framework by age-speci…c variables
which consider di¤erent separations risks and a di¤erent productivity/wage.
From a theoretical perspective, the e¤ect of aging on unemployment is ambiguous and divides into four possible regimes. In the case that one age group
brings strictly more pro…ts to the …rms in terms of productivity and separation risk, the …rms will respond to a change in the relative share of the age
groups with a variation in the number of o¤ered vacancies. If this e¤ect on
job creation goes in the same direction as the e¤ect of aging on job destruction, unemployment will either strictly increase or decrease. Unemployment
goes up (down), when the labor force grows older, if …rms prefer younger
19
(elderly) job seekers. In contrast to this, the total outcome is ambiguous if
the two e¤ects are opposing. The net e¤ect on employment then depends on
the magnitude of the changes in job creation and job destruction.
In the empirical part of the analysis we estimate jointly two equations:
The Beveridge Curve and the job-creation curve. Based on our proxy for aging, we are able to identify which of the regimes dominates in the considered
nine OECD countries. Furthermore, this approach allows to calculate the net
e¤ect in the theoretical ambiguous cases. Therefore, we can say what is the
expected change in search unemployment when the share of elderly workers
grows continuously as a consequence of the demographic change.
Taking all employed as job seekers, aging of the employed labor force
leads to less job creation in terms of a reduced vacancy rate in all considered
economies, with the exception of Norway. This strictly means a higher unemployment rate for France, Germany, Japan, Portugal, and Sweden because in
these countries the e¤ect which follows from changes in the Beveridge Curve
is of the same kind. In contrast to this, we …nd that aging causes a fall in the
unemployment rate in Canada, Norway, Spain, and the US. This is an interesting result because it means that less job destruction - the Beveridge Curve
shifts inwards - outweighs the loss of job creation. Furthermore, we obtain
all four regimes, which are theoretically possible if we take the age of the unemployed as a proxy. The investigation yields that aging of the unemployed
increases (decreases) the unemployment rate in Canada and Japan (France,
Germany, Spain, and the US). Taking all estimation results into account, we
suppose that aging of the labor force reduces search unemployment in Spain
and the US, whereas the opposite is true for Japan.
5
References
Acemoglu, D., 1997, Technology, Unemployment and E¢ ciency, European
Economic Review, 41, 525 - 533.
Batey, P.; Madden, M., 1999, The Employment Impact of Demographic
Change: A Regional Analysis, Papers in Regional Science, 78, 69-87.
Blanchard, O.J.; Diamond, P., 1989, The Beveridge Curve, Brookings Papers on Economic Activity, 1, 1-60.
Blanchard, O.J.; Diamond, P., 1990, The Aggregate Matching Function, in:
P. Diamond (ed.), Productivity, Growth, and Unemployment: Essays
in Honor of Robert Solow’s Birthday, MIT Press, Cambridge MA.
20
Bleakley, H.; Fuhrer, J.C., 1997, Shifts of the Beveridge Curve, Job Matching, and Labor Market Dynamics, New England Economic Review,
September/October, 3-19.
Bloom, D.E.; Canning, D., 2004, Global Demographic Change: Dimensions
and Economic Signi…cance, NBER Working Paper Series, No. 10817.
Bloom, D.E.; Canning, D.; Sevilla, J., 2003, The Demographic Dividend: A
new Perspective on the Economic Consequences of Population Change,
Rand, Santa Monica.
Börsch-Supan, A., 2003, Labor Market E¤ects of Population Aging, Labour,
17, 5 - 44.
Börsch-Supan, A.; Brugiavini, A.; Jürges, H.; Mackenbach, J.; Siegrist,
J.; Weber, G., 2005, Health, Aging and Retirement in Europe, MEA,
Mannheim.
Börsch-Supan, A.; Düzgün, I.; Weiss, M., 2005, Altern und Produktivität:
Zum Stand der Forschung, MEA, Discussion Paper 73-2005.
Breyer, F.; Stolte, K., 2001, Demographic Change, Endogenous Labor Supply and the Political Feasibility of Pension Reform, Journal of Population Economics, 14, 409-424.
Burda, M., 1993, Unemployment, Labour Market Institutions and Structural Change in Eastern Europe, Economic Policy, 16, 101-137.
Burgess, S. 1993, A Model of Competition between Unemployed and Employed Searchers: An Application to the Unemployment Out‡ows in
Britain, Economic Journal, 103, 1190 –1204.
Burgess, S.M., 1994, Matching Models and Labour Market Flows, European
Economic Review, 38, 809-816.
Butrica, B.A.; Johnson, R.W.; Smith, K.E.; Steuerle, E., 2004, Does Work
Pay at Older Ages?, Center for Retirement Research, Working Paper
2004-30.
Coles, M. and E. Smith, 1996, Cross-Section Estimation on the Matching
Function: Evidence from England and Wales, Economica, 63, 589 -597.
Ehrlich, I.; Kim, J., 2005, Social Security, Demographic Trends, and Economic Growth: Theory and Evidence from the International Experience, NBER Working Paper Series, No. 11121.
21
Johnson, P. and Zimmermann, K.F., 1993, Labour Markets in an Aging
Europe, Cambridge Press, Cambridge.
Katz, L.F.; Krueger, A.B., 1999, The High-Pressure U.S. Labor Market of
the 1990s, Brookings Papers on Economic Activity, 1, 1-65 + 82-87.
Layard, R.; Nickell, S.J.; Jackman, R., 1991, Unemployment: Macroeconomic Performance and the Labour Market, Oxford University Press,
Oxford.
Miles D., 1999, Modelling the Impact of Demographic Change upon the
Economy, Economic Journal, 109, 1-36.
Nickell, S.J.; Nunziata, L, 2002, Labour Market Institutions Database, version 2, Centre for Economic Performance.
Petrongolo, B. and Pissarides, C. A., 2001, Looking into the Black Box: A
Survey of the Matching Function, Journal of Economic Literature, 39
(2), 390 - 431.
Pissarides, C. A., 2000, Equilibrium Unemployment Theory, MIT Press,
Cambridge MA.
Pissarides, C. A., 1986, Unemployment and Vacancies in Britain, Economic
Policy, 3, 499-559.
Pissarides, C.A. and J. Wadsworth, 1994, On-the-Job Search: Some Empirical Evidence from Britain, European Economic Review, 38, 385 –
401.
Sellon, G.H. Jr., 2004, Global Demographic Change: Economic Impact and
Policy Challenges, Federal Reserve Bank of Kansas City, Kansas City.
Sneddon Little, J.; Triest, R.K., 2001, Seismic Shifts: The Economic Impact
of Demographic Change, Federal Reserve Bank of Boston, Conference
Series No. 46, Boston.
Sneddon Little, J.; Triest, R.K., 2002, The Impact of Demographic Change
on U.S. Labor Markets, New England Economic Review, First Quarter
2002, 47-68.
Van Ours, J.C., 1991, The E¢ ciency of the Dutch Labour Market in Matching Unemployment and Vacancies, De Economist, 139, 358-378.
22
Van Ours, J.C., 1995, An Empirical Note on Employed and Unemployed
Job Search, Economics Letters, 49, 447-452.
Yashiro, N., 2001, Social Implications of Demographic Change in Japan, in:
Sneddon Little, J.; Triest, R.K. (eds.), Seismic Shifts: The Economic
Impact of Demographic Change, Federal Reserve Bank of Bosten, Conference Series No. 46, Boston, 297-304.
23
Appendix
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 6a: Canada: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
-0.201
-0.226
-0.266
-0.266
(-4.073)
(-2.365)
(-2.812)
(-2.812)
JC
(3)
(4)
-0.045
0.380
0.204
0.204
1.309
0.571
1.328
1.325
(-0.444)
(2.211)
(1.546)
(1.548)
(2.687)
(1.159)
(6.140)
(6.160)
5.006
-9.114
(8.508)
(-2.879)
-3.694
-0.467
0.029
7.868
2.052
0.427
(-5.762)
(-0.439)
(0.027)
(2.268)
(0.612)
(0.124)
-2.696
1.728
0.028
0.008
-11.891
-20.656
-16.682
-17.016
(-2.012)
(0.643)
(0.011)
(0.003)
(-1.767)
(-2.895)
(-2.318)
(-2.548)
-0.045
-0.031
-0.019
-0.019
0.131
0.108
0.074
0.074
(-7.924)
(-2.868)
(-2.475)
(-2.571)
(5.209)
(3.805)
(3.433)
(3.546)
1.076
0.948
-3.350
-3.204
(3.589)
(1.521)
(-2.095)
(-1.685)
0.966
3.747
4.100
4.113
-1.224
-6.386
-7.992
-7.810
(1.788)
(3.989)
(4.368)
(5.004)
(-0.403)
(-2.193)
(-2.718)
(-3.062)
0.980
0.905
0.899
0.899
0.930
0.901
0.887
0.887
2.610
1.805
1.745
1.749
1.452
1.959
1.676
1.665
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 6b: Canada: Estimates with Aging of Unemployed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.220
-0.080
-0.042
-0.042
(-6.370)
(-0.821)
(-0.530)
(-0.532)
-0.049
-0.047
-0.044
-0.043
0.060
0.057
0.378
0.377
(-3.230)
(-1.034)
(-1.255)
(-1.244)
(0.488)
(0.437)
(4.918)
(4.937)
4.879
-4.666
(12.538)
(-1.494)
-3.444
-0.092
-0.210
6.236
2.706
0.515
(-6.833)
(-0.075)
(-0.173)
(1.561)
(0.797)
(0.131)
-2.539
3.432
4.127
4.280
-19.385
-22.701
-20.876
-21.258
(-2.331)
(1.118)
(1.416)
(1.538)
(-2.698)
(-3.151)
(-2.449)
(-2.653)
-0.046
-0.031
-0.035
-0.034
0.158
0.131
0.131
0.130
(-9.834)
(-2.620)
(-3.186)
(-3.246)
(5.772)
(5.986)
(5.027)
(5.268)
0.578
-0.225
-5.973
-4.605
(2.149)
(-0.315)
(-3.360)
(-2.866)
0.377
3.729
3.804
3.723
-2.152
-5.178
-3.216
-3.017
(0.797)
(3.140)
(3.169)
(3.364)
(-0.571)
(-1.544)
(-0.825)
(-0.839)
0.987
0.872
0.865
0.865
0.906
0.896
0.853
0.853
2.156
1.629
1.690
1.676
1.801
1.961
1.248
1.239
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 7a: France: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.172
-0.170
-0.171
-0.144
(-10.518)
(-9.686)
(-9.448)
(-6.659)
-0.176
-0.192
-0.218
-0.191
1.178
1.175
1.273
1.516
(-2.004)
(-2.044)
(-2.293)
(-1.601)
(0.999)
(1.000)
(1.105)
(1.285)
0.556
0.099
(2.055)
(0.027)
-1.115
-0.992
-0.940
-3.764
-3.744
-3.944
(-4.941)
(-4.235)
(-3.938)
(-1.285)
(-1.324)
(-1.413)
4.859
3.248
3.064
2.384
-1.322
-1.608
-0.929
-3.938
(5.569)
(7.867)
(7.595)
(5.192)
(-0.111)
(-0.308)
(-0.188)
(-0.851)
-0.070
-0.075
-0.077
-0.074
0.228
0.227
0.235
0.265
(-10.576)
(-11.349)
(-11.606)
(-8.960)
(2.683)
(2.883)
(3.087)
(3.501)
2.246
2.156
2.138
1.965
-0.587
-0.602
-0.533
-1.181
(29.000)
(31.264)
(30.541)
(28.266)
(-0.583)
(-0.732)
(-0.661)
(-1.722)
0.683
0.581
-2.110
-2.129
(1.723)
(1.376)
(-0.392)
(-0.398)
0.998
0.998
0.998
0.996
0.869
0.869
0.868
0.859
1.789
1.788
1.717
1.357
0.790
0.792
0.763
0.670
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 7b: France: Estimates with Aging of Unemployed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.180
-0.184
-0.189
-0.170
(-9.644)
(-9.477)
(-9.745)
(-7.618)
-0.010
-0.026
-0.035
-0.047
-0.573
-0.489
-0.395
-0.468
(-0.440)
(-1.161)
(-1.680)
(-1.917)
(-2.270)
(-2.057)
(-1.700)
(-2.073)
0.547
-3.314
(1.765)
(-0.897)
-1.058
-0.889
-0.813
-0.813
-1.890
-2.967
(-4.117)
(-3.543)
(-3.273)
(-0.265)
(-0.660)
(-1.055)
4.799
3.474
3.461
3.167
0.044
8.302
9.042
8.680
(5.195)
(6.077)
(5.927)
(4.611)
(0.004)
(1.328)
(1.413)
(1.332)
-0.071
-0.079
-0.082
-0.082
0.069
0.116
0.161
0.168
(-8.174)
(-9.962)
(-10.901)
(-9.132)
(0.661)
(1.272)
(1.856)
(1.906)
2.229
2.190
2.201
2.107
0.886
1.192
1.215
1.062
(22.756)
(21.793)
(21.558)
(17.975)
(0.870)
(1.224)
(1.215)
(1.052)
0.762
0.557
-7.738
-6.613
(1.679)
(1.209)
(-1.461)
(-1.267)
0.998
0.998
0.998
0.997
0.885
0.882
0.875
0.870
1.628
1.683
1.702
1.404
0.879
0.835
0.681
0.580
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 8a: Germany: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
0.573
0.252
0.628
0.625
(5.991)
(2.281)
(4.757)
(4.704)
-2.587
-1.747
-2.404
-2.377
3.561
3.365
3.224
3.202
(-7.853)
(-4.430)
(-4.163)
(-4.105)
(9.444)
(5.735)
(4.925)
(4.870)
8.282
11.960
19.946
20.396
-7.639
-13.585
-19.981
-20.504
(4.893)
(5.944)
(8.395)
(8.943)
(-2.963)
(-3.642)
(-6.404)
(-6.837)
-17.453
23.711
(-6.023)
(6.118)
17.619
4.036
2.338
1.931
-26.855
-14.531
-8.302
-7.899
(5.376)
(1.269)
(0.525)
(0.436)
(-6.917)
(-2.803)
(-1.616)
(-1.543)
-0.379
-0.319
-0.480
-0.485
0.497
0.552
0.591
0.599
(-12.677)
(-8.435)
(-10.648)
(-10.937)
(23.807)
(18.751)
(20.797)
(23.962)
1.514
1.772
-1.277
-1.426
(6.354)
(5.895)
(-3.597)
(-2.577)
-1.510
-1.573
1.782
1.000
0.402
-2.066
(-1.059)
(-0.862)
(0.642)
(0.460)
(0.119)
(-0.567)
0.992
0.987
0.966
0.966
0.995
0.987
0.984
0.983
1.932
1.705
1.384
1.400
1.605
1.248
1.304
1.305
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 8b: Germany: Estimates with Aging of Unemployed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
0.016
-0.059
-0.036
-0.033
(0.146)
(-0.559)
(-0.289)
(-0.262)
0.030
0.043
-0.211
-0.210
-0.255
-0.334
0.003
0.002
(0.345)
(0.476)
(-2.863)
(-2.826)
(-1.342)
(-1.624)
(0.021)
(0.014)
13.380
14.115
23.603
23.887
-15.624
-18.015
-30.154
-30.377
(5.089)
(5.229)
(13.433)
(14.191)
(-2.655)
(-2.825)
(-9.148)
(-9.769)
-7.901
18.698
(-2.112)
(2.341)
-5.288
-8.769
-10.750
-10.932
3.930
12.847
15.291
15.433
(-2.298)
(-5.291)
(-5.447)
(-5.594)
(0.761)
(3.349)
(3.839)
(3.933)
-0.189
-0.185
-0.291
-0.297
0.355
0.378
0.510
0.514
(-5.519)
(-5.226)
(-8.964)
(-9.523)
(5.428)
(5.327)
(12.807)
(14.863)
1.852
2.014
-2.133
-2.616
(3.718)
(3.949)
(-1.908)
(-2.164)
-2.632
-2.627
1.211
3.751
3.962
-1.028
(-1.250)
(-1.204)
(0.503)
(0.794)
(0.762)
(-0.203)
0.985
0.985
0.976
0.976
0.978
0.973
0.968
0.968
2.034
2.167
2.008
2.046
1.376
1.612
1.531
1.536
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 9a: Japan: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.242
-0.364
-0.366
-0.318
(-2.210)
(-3.423)
(-3.438)
(-3.121)
-0.556
-0.682
-0.689
-0.661
0.996
1.537
1.561
1.992
(-3.486)
(-4.215)
(-4.299)
(-4.519)
(2.539)
(3.746)
(3.858)
(6.852)
#
#
#
#
#
#
#
#
1.047
0.401
0.358
-0.396
1.867
2.003
(1.869)
(0.746)
(0.689)
(-0.280)
(1.320)
(1.479)
4.496
-0.115
-0.044
-0.119
-19.853
-9.964
-10.205
-13.367
(1.912)
(-0.066)
(-0.025)
(-0.069)
(-4.070)
(-2.299)
(-2.385)
(-3.473)
-0.011
-0.010
-0.009
-0.008
0.015
0.015
0.012
0.015
(-0.840)
(-0.754)
(-0.692)
(-0.644)
(0.482)
(0.411)
(0.327)
(0.408)
0.673
-1.795
(2.737)
(-3.142)
1.239
0.164
-3.273
-0.482
(1.940)
(0.307)
(-2.106)
(-0.325)
0.947
0.942
0.942
0.938
0.864
0.816
0.815
0.801
1.083
0.922
0.969
0.873
1.305
0.789
0.826
0.861
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration (< 1 year and therefore = 0 per definition); BRR: benefit
replacement rate; UDNET: net union density; EPOP: employment population ratio; LABC: labor costs; RIRL:
real interest rate; DW: Durbin-Watson statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 9b: Japan: Estimates with Aging of Unemployed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.126
-0.271
-0.270
-0.242
(-1.008)
(-2.140)
(-2.137)
(-1.936)
-0.275
-0.328
-0.329
-0.313
0.623
0.853
0.846
0.931
(-3.099)
(-3.469)
(-3.577)
(-3.925)
(3.603)
(4.640)
(4.764)
(8.137)
#
#
#
#
#
#
#
#
1.159
0.304
0.299
-1.125
0.879
0.841
(1.883)
(0.504)
(0.506)
(-0.844)
(0.639)
(0.625)
4.591
-1.691
-1.674
-1.765
-18.365
-7.849
-7.751
-8.660
(1.844)
(-1.018)
(-1.031)
(-1.113)
(-4.308)
(-2.474)
(-2.511)
(-3.160)
0.017
0.022
0.022
0.021
-0.058
-0.080
-0.078
-0.085
(0.916)
(1.087)
(1.180)
(1.121)
(-1.461)
(-1.795)
(-1.874)
(-2.149)
0.818
-1.656
(3.193)
(-3.200)
1.334
0.024
-2.515
0.186
(1.945)
(0.040)
(-1.723)
(0.133)
0.937
0.927
0.927
0.924
0.886
0.844
0.844
0.841
1.109
1.058
1.064
0.994
1.375
1.031
1.009
1.053
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration (< 1 year and therefore = 0 per definition); BRR: benefit
replacement rate; UDNET: net union density; EPOP: employment population ratio; LABC: labor costs; RIRL:
real interest rate; DW: Durbin-Watson statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 10: Norway: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.301
-0.304
0.279
0.346
(-1.934)
(-2.120)
(2.030)
(1.921)
2.293
2.386
1.382
1.242
-0.160
-1.328
-0.435
-0.601
(4.482)
(6.833)
(3.210)
(2.530)
(-0.162)
(-1.706)
(-0.667)
(-0.912)
6.506
6.773
8.119
-10.051
-13.663
-8.512
(3.697)
(4.154)
(5.499)
(-2.641)
(-4.004)
(-3.460)
0.333
-4.292
(0.278)
(-1.750)
2.435
3.419
12.166
24.364
-15.668
-28.943
-21.634
-33.886
(0.535)
(0.896)
(3.376)
(5.300)
(-1.649)
(-4.736)
(-4.158)
(-6.237)
-0.165
-0.163
-0.179
-0.111
0.241
0.217
0.186
0.112
(-8.885)
(-10.048)
(-8.738)
(-4.333)
(6.668)
(6.078)
(5.451)
(3.050)
-0.087
0.002
5.281
4.245
(-0.077)
(0.002)
(2.545)
(2.000)
0.710
0.774
2.599
-4.227
-5.158
-4.144
(0.705)
(0.767)
(1.979)
(-2.076)
(-2.460)
(-1.940)
0.939
0.939
0.866
0.685
0.870
0.852
0.833
0.713
1.587
1.588
1.465
1.295
1.906
1.902
1.574
1.602
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 11: Portugal: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
0.417
0.318
0.194
(5.478)
(4.372)
(2.690)
-1.580
-1.175
-1.185
3.007
2.518
3.323
(-4.257)
(-3.216)
(-2.809)
(4.247)
(2.943)
(3.043)
2.191
-5.120
(2.660)
(-3.182)
0.873
1.264
-1.502
-3.011
(2.337)
(3.449)
(-1.865)
(-3.743)
5.407
2.081
-0.027
-11.280
-4.247
1.200
(3.728)
(2.598)
(-0.045)
(-4.154)
(-2.176)
(0.699)
0.009
-0.040
-0.067
-0.094
0.009
0.077
(0.335)
(-2.077)
(-3.282)
(-1.849)
(0.190)
(1.297)
-0.372
-0.672
-1.550
0.031
0.761
3.602
(-0.820)
(-1.423)
(-3.362)
(0.031)
(0.634)
(2.933)
-0.655
-0.941
1.398
2.662
4.568
-0.775
(-0.776)
(-1.048)
(2.061)
(1.503)
(2.214)
(-0.395)
(4)
0.864
0.848
0.802
0.856
0.779
0.616
2.239
2.204
1.807
2.193
1.745
0.803
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 12a: Spain: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
0.024
0.024
0.020
(1.187)
(1.186)
(0.794)
0.609
0.574
0.803
10.795
10.827
10.311
(2.103)
(2.107)
(2.513)
(5.892)
(7.463)
(7.689)
-0.083
0.075
(-0.347)
(0.028)
0.318
0.343
0.524
-2.182
-2.204
-2.754
(2.274)
(2.830)
(3.901)
(-1.465)
(-1.740)
(-2.488)
-1.900
-1.835
0.674
-21.224
-21.281
-28.875
(-1.883)
(-1.844)
(0.824)
(-2.160)
(-2.214)
(-8.191)
-0.108
-0.106
-0.113
-0.416
-0.418
-0.404
(-7.726)
(-8.539)
(-7.622)
(-4.296)
(-6.553)
(-6.450)
1.224
1.222
-3.524
-3.522
(3.236)
(3.220)
(-0.847)
(-0.847)
1.167
1.102
1.565
-3.902
-3.843
-5.225
(3.838)
(4.589)
(6.473)
(-1.179)
(-1.492)
(-2.574)
(4)
0.998
0.998
0.998
0.836
0.836
0.830
2.804
2.757
2.477
2.350
2.347
2.078
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 12b: Spain: Estimates with Aging of Unemployed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
0.066
0.071
0.087
(5.243)
(5.990)
(5.975)
0.040
0.032
0.041
-0.403
-0.716
-0.736
(1.958)
(1.650)
(1.627)
(-1.197)
(-2.275)
(-2.241)
0.263
7.243
(1.043)
(1.826)
0.367
0.312
0.533
-1.020
-2.981
-1.718
(2.588)
(2.297)
(3.380)
(-0.420)
(-1.263)
(-0.761)
-0.397
-0.553
3.040
-24.644
-34.606
-13.072
(-0.390)
(-0.532)
(8.466)
(-1.496)
(-2.053)
(-2.823)
-0.080
-0.082
-0.079
0.128
0.078
0.098
(-25.592)
(-30.325)
(-23.595)
(2.433)
(1.599)
(2.044)
1.325
1.395
5.220
8.792
(3.438)
(3.560)
(0.812)
(1.324)
0.734
1.021
1.594
-4.186
4.604
9.436
(1.883)
(3.587)
(5.324)
(-0.631)
(0.931)
(2.706)
(4)
0.998
0.998
0.997
0.568
0.493
0.446
3.165
3.343
2.485
1.046
0.817
0.877
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 13: Sweden: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.352
-0.302
-0.303
-0.350
(-4.552)
(-3.683)
(-3.855)
(-5.161)
-2.950
-2.869
-2.897
-1.821
14.610
15.178
17.115
7.823
(-2.251)
(-2.020)
(-2.049)
(-3.275)
(3.257)
(3.489)
(4.013)
(3.742)
8.645
8.849
(1.826)
(0.481)
-1.277
0.042
-3.371
-2.102
(-1.564)
(0.094)
(-1.098)
(-1.336)
-0.542
-1.360
-1.284
16.439
16.186
12.856
(-0.336)
(-0.789)
(-0.880)
(3.102)
(3.054)
(2.643)
-0.005
-0.027
-0.026
-0.049
-0.088
-0.117
-0.213
0.019
(-0.136)
(-0.766)
(-0.803)
(-3.468)
(-0.635)
(-0.928)
(-1.975)
(0.304)
1.763
1.699
1.704
1.389
-5.459
-5.681
-6.120
-3.573
(4.610)
(4.144)
(4.153)
(7.648)
(-4.310)
(-4.792)
(-5.167)
(-5.517)
1.921
1.077
1.037
-7.998
-9.181
-7.486
(1.758)
(1.033)
(1.120)
(-1.941)
(-2.765)
(-2.346)
0.980
0.976
0.976
0.976
0.924
0.923
0.917
0.883
2.203
1.614
1.621
1.637
1.498
1.554
1.420
1.028
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 14a: USA: Estimates with Aging of Employed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.720
-0.723
-0.717
-0.715
(-13.952)
(-14.648)
(-15.106)
(-14.919)
0.911
0.909
0.900
0.934
0.102
0.322
0.180
0.187
(9.631)
(9.675)
(9.833)
(11.327)
(0.199)
(0.618)
(0.337)
(0.400)
0.132
0.151
0.244
-1.362
-2.516
0.045
(0.332)
(0.390)
(0.834)
(-0.599)
(-1.102)
(0.025)
1.096
1.106
1.227
1.126
-6.048
-7.062
-4.044
-4.063
(2.231)
(2.262)
(3.374)
(3.256)
(-2.255)
(-2.576)
(-1.839)
(-1.973)
-0.230
11.743
(-0.195)
(1.818)
-0.031
-0.029
-0.027
-0.026
0.057
-0.056
0.005
0.005
(-2.388)
(-3.916)
(-7.058)
(-6.935)
(0.773)
(-1.353)
(0.202)
(0.205)
0.074
0.081
2.346
2.200
(0.328)
(0.362)
(1.921)
(1.728)
0.469
0.575
0.557
0.350
0.569
-5.233
-6.312
-6.352
(0.608)
(1.047)
(1.013)
(0.710)
(0.129)
(-1.641)
(-1.940)
(-2.234)
0.951
0.952
0.951
0.950
0.406
0.351
0.297
0.297
1.301
1.323
1.351
1.341
0.819
1.033
0.824
0.823
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.
(1)
LOG(v)
AGING
BD
BRR
UDNET
EPOP
LABC
RIRL
R2
DW
Table 14b: USA: Estimates with Aging of Unemployed
BC
(2)
(3)
(4)
(1)
(2)
JC
(3)
(4)
-0.661
-0.651
-0.659
-0.688
(-7.189)
(-6.935)
(-6.832)
(-7.149)
0.103
0.081
0.073
0.097
0.222
0.236
0.254
0.192
(3.144)
(2.784)
(2.451)
(3.771)
(2.454)
(3.198)
(3.484)
(2.986)
1.044
1.379
0.810
-3.482
-3.704
-2.608
(1.643)
(2.316)
(1.569)
(-1.693)
(-1.970)
(-1.623)
3.092
3.427
2.818
2.444
-8.627
-8.837
-7.693
-6.725
(4.309)
(4.965)
(4.606)
(4.265)
(-4.014)
(-4.416)
(-4.461)
(-4.012)
-2.727
1.897
(-1.266)
(0.266)
-0.020
0.005
-0.011
-0.009
-0.024
-0.041
-0.010
-0.016
(-0.895)
(0.469)
(-1.955)
(-1.654)
(-0.331)
(-1.210)
(-0.543)
(-0.930)
-0.753
-0.632
1.323
1.236
(-2.088)
(-1.779)
(1.112)
(1.079)
1.416
2.545
2.825
2.516
-2.218
-3.001
-3.556
-2.563
(1.151)
(2.927)
(3.191)
(2.860)
(-0.545)
(-1.066)
(-1.265)
(-0.901)
0.869
0.862
0.852
0.847
0.490
0.489
0.473
0.434
1.055
1.215
0.964
0.959
1.102
1.157
1.015
1.001
BC and JC are estimated simultaneously with the SUR method. T-statistics are in parentheses. v: vacancy rate;
AGING: ratio of young to old; BD: benefit duration; BRR: benefit replacement rate; UDNET: net union
density; EPOP: employment population ratio; LABC: labor costs; RIRL: real interest rate; DW: Durbin-Watson
statistic.