Demographics and Unemployment
Málstofa í Seðlabanka Íslands
9. desember 2002
Tryggvi Thor Herbertsson
IoES, University of Iceland
Edmund S. Phelps
Columbia University
Gylfi Zoega
Birkbeck College, University of London
Introduction
• Current research aims at explaining decade-to-decade movements
in average unemployment as well as cross-country variation. A
consensus appears to be emerging on the importance of the
interaction between shocks and unemployment but less so on the
nature of the shocks.
• Potential shocks include:
– changes in the rate of productivity growth (Pissarides, 1990, 2000;
Hoon and Phelps, 1997),
– skill-biased technological progress (Wood, 1994; Nickell, 1996),
– the world real interest rate (Phelps, 1994; Blanchard and Wolfers,
2000, Nickell, 1999),
– the price of energy (Bruno and Sachs, 1986; Oswald, 1999),
– changes in share of profits (Blanchard and Wolfers, 2000)
– the state of stock market (Phelps, 1999; Fitoussi, Jestaz, Phelps, and
Zoega, 2000).
Introduction
• The literature has paid surprisingly little attention to demographic
factors, in particular changes in the age structure of the population.
Shimer (1999) is a novel exception.
• This paper investigates some possible channels through which a
changing age-structure might affect unemployment, in particular:
– Compositional effects:
• When the cohorts suffer different unemployment rates, their relative size
affects the aggregate unemployment rate and labor force participation.
– Macroeconomic shocks:
•
The origins. Generations differ in their saving propensities and possibly
creativity, as innovations may come more frequently at certain stages in
life. Thus both investment opportunities and the supply of available
savings may depend on the age distribution of the population.
• Responsiveness. When an adverse shock hits the economy, the
responsiveness of unemployment depends on labor-market institutions
such as the duration and level of unemployment benefits and the
organization of labor unions. However, demographics can play a role that
has so far not received much attention.
Demographic Structure
0.50
0.45
0.40
0.35
Canada
France
Germany
Italy
Japan
UK
USA
0.30
1950
1960
1970
1980
1990
2000
2010
2020
2030
2040
2050
The share of prime-aged workers in the total population (defined as
the 25-54 year old/total population ratio) in the G7
Age-structure and Unemployment
Compositional Effects
When the cohorts suffer different unemployment rates, their
relative size affects the aggregate unemployment rate and
labor force participation.
Compositional Effects
• Unemployment differs substantially between cohorts (OECD).
25
15-19
20-24
25-34
35-44
45-54
55+
20
15
10
5
0
1965
1970
1975
1980
1985
1990
1995
Compositional Effects
• Differences in unemployment across cohorts can to some degree be
traced to differences in the level of non-wage income. This has many
sources:
– Accumulated savings in the form of real wealth are normally
monotonically rising in age.
– The income of a spouse increases with age so this form of non-wage
income is also rising in age.
– The cost of raising children, which initially rises in their age and then
falls. As a result, non-wage income falls initially and then rises again as
the children leave the nest.
– Parental support, which is declining in age.
– The expected present value of anticipated inheritance which is rising in
one’s own age (becoming less important due to annuitization of wealth).
– Social spending such as health care, public education for children,
minimum income support, state pension, etc. is often an important form
of non-wage income and depends on age.
Compositional Effects
• For the calculation we define:
• ui as the unemployment rate in group i
• u as the aggregate unemployment rate
• Li as the number of people in group i
• L as the total number of people in the labor force
• Furthermore, we define vi = Li/L ,
•Then by definition:
n
ut   vit uit
i 1
•We get the unemployment rate adjusted for demographic
changes by using initial weights:
n
uht   vi1uit
i 1
Compositional Effects
• The age-group-unemployment pattern is remarkably stable (G6):
Compositional Effects
10
0.2
9
Adjusted
0.1
0.0
Unemployment
7
Actual
6
-0.1
5
-0.2
4
3
-0.3
2
Demographic
Demographic Unemployment
8
-0.4
1
0
-0.5
1965
1970
1975
1980
1985
1990
1995
• Unemployment would have been 50 basis points lower in 1998 in the
OECD if the demographic structure had remained the same as in 1965.
Participation Rates
– Unemployment is only one side of the coin.
– Changing age structures may also be reflected in changing laborforce participation rather than unemployment.
– In 13 OECD countries, the average labor force participation of 5564 year-old males fell by more than 12 percentage points between
1979 and 1998.
– The participation rate increased by approximately 5 percentage
point for females, resulting in an overall average drop in OECD
labor force participation of almost 3 percentage points, cf.
Herbertsson and Orszag (2001).
Participation Rates
Cohort nonparticipation rates in Canada
Cohort nonparticipation rates in France
100
80
80
1980-89
1970-79
1990-98
60
Cohort nonparticipation rates in the UK
80
1970-79
1980-89
1980-89
60
1990-98
1990-98
60
40
40
40
20
20
20
0
0
15-19
20-24
25-34
35-44
45-54
0
15-19
55+
Cohort nonparticipation rates in Germany
20-24
25-34
35-44
45-54
55+
Cohort nonparticipation rates in Italy
25-34
1970-79
80
1980-89
35-44
45-54
55+
80
1970-79
60
20-24
Cohort nonparticipation rates in the US
100
80
15-19
1970-79
1980-89
1990-98
1980-89
60
1990-98
60
1990-98
40
40
40
20
20
20
0
0
15-19
20-24
25-34
35-44
45-54
55+
0
15-19
20-24 25-39
40-49
50-59 60-64
65+
15-19
20-24
25-34
35-44
45-54
55+
Participation Rates
Participation rates Canada in 1979-98
Participation rates in France 1965-98
90
Participation rates in the UK 1965-98
90
90
85
85
85
A ctual
80
80
80
75
A djusted
75
75
70
Data so urce: OECD
70
70
65
1965
1970
1975
1980
1985
1990
1995
1965
1970
1975
1980
1985
1990
1965
1995
Participation rates Italy in 1965-98
Participation rates in Germany 1965-98
90
90
85
85
85
80
80
80
75
75
75
1970
1975
1980
1985
1990
1995
1980
1985
1990
1995
70
70
1965
1975
Participation rates in the US 1965-98
90
70
1970
1965
1970
1975
1980
1985
1990
1995
1965
1970
1975
1980
1985
1990
Participation would have fallen by more if the baby-boom generation had
not come of age!
1995
Macroeconomic Shocks: The Origins
Generations differ in their saving propensities and possibly creativity, as
innovations may come more frequently at certain stages in life. Thus both
investment opportunities and the supply of available savings may depend
on the age distribution of the population and thus create structural booms
as well as slumps.
Macroeconomic Shocks: The Origins
• National savings depend on the age distribution (life-cycle hypothesis).
• With perfect capital mobility a changing (global) age distribution would affect
the world interest rate.
• Some recent NAIRU models that treat labor as a (quasi) fixed asset suggest an
important role for the world interest rate. The sign of this effect goes contrary
to that implied by the Mundell-Fleming analysis.
• (age up -> savings up -> investment up -> investment in humans up ->
returns up)
10
Variable
Coefficient
t-Statistic
C
-0.06
0.10
15-34
-0.27
0.21
35-54
-2.19
1.13
55-
3.63
2.14
5
0
2
R
Adjusted R
world real
interest rates
(%)
0.81
2
0.78
S.E.
0.01
D.W.
1.37
-5
0.17
0.18
0.19
0.20
0.21
0.22
0.23
share of population
over 55 years of age
Macroeconomic Shocks: The Origins
• Feldstein and Horioka have shown that savings and investment are highly
correlated. If hiring has an investment dimension we would also expect
investment in hiring and training of new workers to move with the aggregate
saving rate.
0.20
0.15
0.09
Unemployment
Investment
0.10
0.07
0.05
0.06
0.00
0.05
-0.05
0.04
-0.10
0.03
-0.15
0.02
-0.20
0.01
-0.25
-0.30
0 - 39
40 - 54
55+
0.00
Investment
Unemployment
0.08
Macroeconomic Shocks: The Origins
• Recent structural booms (the “new economy”?) can be traced to
innovations and creativity of entrepreneurs. Here, the age distribution
may come into play.
• Casual observation suggests that people may be most creative in their
thirties and early forties (with greatest potential of turning ideas into
reality).
• A relatively young population would then have proportionately greater
number of entrepreneurs than an older one and hence a faster rate of
expected productivity growth and a higher-valued stock market.
• This may induce firms to step up their hiring, driving down the
unemployment rate.
Macroeconomic Shocks: The Origins
proportional
rise in share
prices 90-97
2.5
2.5
2.0
proportional
rise in share
prices 90-97 2.0
ne
us
ne
nz
nz
us
1.5
no
uk
ge
fisp
aut
sw fr
de
1.0
0.5
0.0
0.8
be
ca
1.5
ir
ir
1.0
sp
it
ja aus
0.5
1.0
no
1.2
1.4
1.6
1.8
2.0
number of people between 25 and 34
in 1990 relative to 1960 level
0.0
0.12
de
ca
uk aut
ge
fi
fr sw
be
it
aus
ja
0.13
0.14
0.15
0.16
0.17
0.18
number of people between 25 and 34
in 1990 as a proportion of total population
The relationship between the the rise in the stock market 1990-1997 (nomalized by
labor productivity) and the number of people between the ages of 25 and 34 in 1990
(relative to the number in 1960 in left-hand panel and the current population
in the right-hand panel)
Macroeconomic Shocks: Responses
When an adverse shock hits the economy, the responsiveness of unemployment
depends on labor-market institutions such as the duration and level of
unemployment benefits and the organisation of labor unions, but also on
the age-structure.
Macroeconomic Shocks: Responses
• The interaction of institutions and macroeconomic shocks plays a key
role in the emerging consensus on the determination of medium-term
unemployment, (Blanchard and Wolfers, 2000).
• Thus certain institutions are likely to mitigate the effect of shocks
while others exacerbate them.
• The data appear to suggest that shocks have more serious
consequences when:
–
–
–
–
–
the unemployment benefit replacement ratio is high,
the duration of such benefits is long,
there is employment-protection legislation,
there is a lack of co-ordination among unions and employers,
a low level of active labor-market spending such as retraining and
placement schemes.
Macroeconomic Shocks: Responses
• How can the age distribution affect the sensitivity of aggregate
employment to macroeconomic shocks?
– Job security is rising in tenure and hence, ceteris paribus, in age. Since it
is more difficult to dismiss an old worker, the sensitivity of employment to
shocks should be a decreasing function of the size of the older cohorts.
– Similarly, the age structure and the institutional framework may interact in
affecting employment. Thus employment-protection legislation may be
more effective for the older is the average worker. (However, firms may
opt for early retirement instead of dismissals which would cause the effect
to show up in labor-force participation instead of unemployment.)
– A transitory shock is more likely to lead to the dismissal of an older
worker because of his shorter expected post-recession tenure. Thus the
level of labor hoarding may be smaller for the older workers due to their
shorter remaining worklife. This would make the sensitivity to shocks
greater.
Macroeconomic Shocks: Responses
 Older workers may find it more difficult to find another job as their
remaining tenure is shorter. They are thus more likely to become longterm unemployed. As a result, the higher is the proportion of older
workers in the labor force the more likely is a transitory shock to
employment to have a persistent effect on employment.
 Older workers may be more resistant to real wage moderation as their
accumulated wealth reduces the dependence on employment. This raises
the possibility that real-wage cuts are less likely the higher the proportion
of older workers.
 What due the data say (rational: Fitoussi, Jestaz, Phelps, and Zoega,
2000)
uit  i  iuit 1  i X i  it


ˆi 1  ˆi  0  1Yi  
Macroeconomic Shocks: Responses
Institutions and the sensitivity of unemployment to shocks
(1)
(2)
(3)
(4)
Constant
4.86
(2.10)
47.69
(2.92)
53.16
(3.12)
63.45
(3.39)
Duration of benefits
1.10
(2.65)
0.70
(1.81)
0.66
(1.65)
0.69
(1.63)
Union density
0.11
(2.20)
0.09
(1.73)
0.08
(1.50)
0.08
(1.41)
Union coordination
-2.62
(2.34)
-2.41
(2.92)
-2.18
(2.26)
-2.3
(2.22)
Active labour market
programmes
-0.13
(1.86)
…
…
-0.18
(4.30)
-135.5
(2.74)
…
…
…
-0.18
(4.09)
-133.7
(2.69)
-24.7
(0.52)
…
-0.18
(3.92)
-172.1
(2.67)
14.3
(0.20)
-84.6
(0.84)
0.57
2.72
0.74
2.20
0.74
2.27
0.75
2.33
Pop. 15-33
Pop. 35-54
Pop 55-64
Adj. R2
S.E.
Note: The table shows regressions of the form:
 i 1   i    0   1Yi  
where Y is a vector of the explanatory variables; replacement ratio, duration
of benefits–the number of months at which benefits continue at a
reasonable level–union coverage, union density, coordination of unions and
employers (indices where “3” refers to maximum coordination)
employment protection, active labour market spending and the age structure
of the population. We use the average value for the variables 1983-88.
Macroeconomic Shocks: Responses
15
steady -state
s ens itiv ity
10
5
0
0.27
0.28
0.29
0.30
0.31
0.32
0.33
s hare of population between
15 and 34 years of age
The sensitivity to macroeconomic shocks and the proportion
of labor force between 24-34 years
Conclusions
• The paper does not spell out any deep consistent theorems, rather
it aims at investigating various sources through which changing
age-structures might affect unemployment.
• To conclude:
– We have found that the compositional effect of demographic
changes do not account for the big swings in unemployment
observed in many of the OECD countries.
– Looking at the OECD as a whole, only current unemployment is
only 50 basis points lower than it would be if the age distribution of
the labor force were the same as in the sixties.
– However, the effect varies somewhat between countries with France
showing the biggest effect of 140 basis points.
– A changing age structure is reflected more strongly in lower
participation rates than in unemployment.
– We find that if it were not for the aging of the baby-boom generation
participation rates would have fallen by more than observed.
Conclusions
– We also looked for the effect of age distribution on the origins and
responses to macroeconomic shocks.
– We found a surprising (time-series) correlation between world real
interest rates and the share of workers over 55 years of age in the OECD.
– We also found a strong (cross-sectional) correlation between the share of
workers between 25 and 34 and the rise in the stock market from 1990 to
1997.
– On both counts we might expect a relationship between the age
distribution – either locally or globally – and national unemployment
rates.
– Finally, we found that the sensitivity of unemployment to macroeconomic
shocks depends negatively on the share of the young in the population
possibly due to labor hoarding when we also take into account various
labor market institutions. A relatively young population suffers less
unemployment for given shocks.
• General conclusion:
– When analysing unemployment, investigating demographics is a path
worth taking!