After 25 years of high unemployment.
What have we learned?
Olivier Blanchard
October 20, Barcelona
Unemployment
2
How to reduce unemployment?
Many strong opinions:
•
eliminate “labor market rigidities” (list follows...)
•
use macroeconomic policy more aggressively (list follows...)
After nearly 25 years of research, I am much more agnostic. Goal of
the lecture: Summarize what we know, what we do not know, and
implications for policy.
Do it in six steps:
1.
A quick look at facts, over time and across countries
2.
A central distinction: Actual versus natural unemployment rate
3.
Determinants of the natural rate: Shocks and institutions
4.
The panel data evidence
5.
Unemployment “miracles”: Ireland, the Netherlands, Sweden
6.
Policy lessons.
Unemployment
3
1. Basic unemployment facts
Two essential facts:
•
The steady increase in the EU unemployment rate since the
early 1970s.
Earlier on, the “European miracle”...
Increase in the 1970s, the 1980s, the 1990s. Turnaround since
the mid 1990s?
Figure 1. EU-wide unemployment rate.
•
The very diverse country evolutions.
Figure 2. EU unemployment rates. (For each country: Mean,
centered over 5 years).
At the top: Spain At the bottom, typically, Luxembourg, Austria, Sweden.
Four more revealing close ups: (Figure 3)
•
Germany, France. The steady increase.
•
Spain, Portugal. Surprising different evolutions.
•
Scandinavian countries. A fever attack in the early 1990s.
•
The turnarounds. Ireland, the Netherlands
Good news: Hope for identification.
Bad news: One liners will not do.
10
U rate
8
6
4
2
1960
1970
1980
1990
Unemployment rate, EU15 average
2000
20
ESP
ESP
IRE
FIN
15
IRE
U rate
GBR
ESP
IRE
ITA
FRA
ITA
FRA
BEL
GRE
GBR
FRA
BEL
10
IRE
GRE
IRE
ITA
5
PRT
ESP
AUT
FRA
FIN
SWE
BEL
GBR
DNK
DEU
0
1960
IRE
BEL
FRA
GBR
ITA
DNK
IRE
PRT
ITA
GRE
FRA
GBR
ESP
FIN
SWE
BEL
AUT
DNK
NLD
DEU
LUX
1970
ESP
PRT
ITA
DNK
FRA
PRT
NLD
BEL
GBR
ESP
DEU
FIN
GRE
SWE
AUT
LUX
FIN
NLD
GRE
SWE
DEU
AUT
LUX
1980
NLD
PRT
ITA
GRE
DEU
DNK
FIN
AUT
SWE
LUX
DNK
GRE
BEL
NLD
GBR
PRT
DEU
AUT
FIN
SWE
DEU
PRT
NLD
DNK
AUT
LUX
GRE
ESP
ITA
FIN
FRA
DEU
BEL
GBR
SWE
AUT
DNK
IRE
PRT
NLD
LUX
SWE
LUX
1990
Unemployment rate, EU15 countries
2000
germany
france
15
FRA
FRA FRA
FRA
FRA
FRA
FRA
FRA
FRA
FRA
FRA
FRA
FRA
FRA FRA
FRA
FRA
FRA
U rate
10
FRA
DEU FRA
FRA
DEU FRA
DEU
DEU
DEU
DEU
DEU DEU
DEU
DEU
DEU
DEU
FRA DEU DEU
DEU
FRA
DEU
DEU
FRA
DEU DEU
FRA
DEU
DEU
FRA
FRA
5
FRA
DEU
FRA
FRA
FRA
FRA FRA
DEU
DEU
DEU
FRA FRA
DEU
DEU
DEU
FRA
FRA
DEU
DEU
DEU
DEU
DEU
DEU
DEU
DEU
DEU
DEU DEU
DEU
DEU
DEU
0
1960
1970
1980
1990
Unemployment rates, Germany, France
2000
portugal
spain
20
ESP
ESP
ESP
ESP
ESP
ESP
ESP
ESP
ESP
ESP
15
ESP
U rate
ESP
ESP
ESP
ESP
ESP
ESP
ESP
ESP
ESP
ESP
ESP
10
5
ESP
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
ESP
PRT
PRT
PRT
PRT
PRT
PRT
ESP
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT
PRT ESP
ESP
PRT
ESP
PRT
PRT
PRT
PRT
PRT
PRT
ESP
PRT
ESP
ESP
PRT
ESP
ESP
ESP
ESP
ESP
PRT
ESP
ESP
ESP
ESP
ESP
ESP
0
1960
1970
1980
1990
Unemployment rates, Spain, Portugal
2000
finland
sweden
20
FINFIN
FIN
FIN
15
FIN
U rate
FIN
FIN
FIN
FIN
FINFIN
10
SWE
SWE
SWE
SWE
SWE
FIN
FIN
FIN FIN
FINFINFIN FINFIN
FIN
FIN
FIN
FIN
5
SWE
FIN
FIN
SWE
SWE
FINFIN
SWE
SWE
FIN FIN
SWE
SWE
FIN
SWE
SWE
SWE
SWE
FIN
FIN
FIN
SWE
SWE
SWE
SWE SWE
SWE
SWE
SWE
FIN
SWE
FIN
SWE
SWE
SWE
SWE
SWE
SWE
SWE
FINFIN
FIN
SWE
SWE
SWE
SWE
FINFIN
FIN
SWE
SWE
SWE
SWE
SWE
SWE
0
1960
1970
1980
1990
Unemployment rates, Finland, Sweden
2000
ireland
netherlands
20
IRE
IREIRE IREIRE
IRE
15
IRE
IRE
IRE
IRE
IRE
U rate
IRE
IRE
IRE
IRE
IRE NLD
NLD
10
IRE
IRE
5
IRE
IRE
IRE
IREIREIREIRE IRE IRE
IRE
IREIRE
NLD
IREIRE
IRE
IRE
IRE
IRE
NLD
IRE
NLD
NLD
NLD
NLD
NLD
IRE
NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD IRE
NLD IREIREIRE
NLD
NLD NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD
NLD
0
1960
1970
1980
1990
Unemployment rates, Ireland, Netherlands
2000
Unemployment
4
2. Actual versus natural rate
A central distinction:
•
From one quarter to the next, movements in unemployment
determined by aggregate demand.
Correct policy measures: shift aggregate demand, through fiscal or monetary policy.
•
Over longer periods (3,10 years?), unemployment rate returns
to an equilibrium value, such that:
The wage demands of workers are consistent with the wages
firms are willing to offer.
Called the “natural rate”. Worse choice of words ever: Depends
on behavior of unions, monopoly power of firms, and so on.
Correct policy measures: Shift wage demands, or wages offered
by firms.
Essential to know whether high u is high natural u*, or large deviation from u*.
Not easy:
•
The natural rate is not constant
•
It is not observable
•
It may be affected by the actual rate itself (hysteresis)
•
It depends on some of the same factors as the actual rate (interest rates for example)
Unemployment
5
Fortunately, a strong hint from theory: The behavior of inflation as
a signal.
•
If u = u∗ , wage demands consistent with wage offers. Inflation
stable.
•
If u < u∗ , inflation increases.
•
If u > u∗ , inflation decreases.
Useful tool, if complex to use (reliability at high, and at low inflation?)
The evidence:
•
EU inflation and unemployment. Figure 4.
•
An estimated natural rate series. Figure 5.
Conclusions. For the most part, high u∗ . Deviations in the 1980s,
the 1990s, the early 2000s.
unemployment
inflation
.1
10
6
.05
4
2
0
1960
1970
1980
1990
2000
Unemployment and Inflation, EU15 average
inflation
unemployment
8
U rate
10
5
0
1960
1970
1980
1990
Actual and Natural U rates (u*=u+2 Delta pi), EU15
2000
Unemployment
6
3. Determinants of the natural rate: Shocks and Institutions
A very large amount of work over the last 25 years. Better models,
integrating shocks and detailed institutions.
•
An example of shocks: A decrease in the rate of technological
progress.
Short and medium term effects. Wage demands may continue
at the previous rate, until workers and firms understand.
Long term effects: Probably not much.
Unemployment will increase for some time, and eventually turnaround.
•
An example of institutions. Unemployment benefits.
(At least some) insurance is desirable. But higher unemployment benefits increase unemployment:
They lead to higher wages, increase costs for firms, decrease
job creation, and lead to higher unemployment duration.
They decrease search intensity, and so increase unemployment
duration.
•
An example of interactions.
Open-ended unemployment benefits in the face of an adverse
technological shock can lead to long term unemployment, unemployability, and longer lasting effects of shocks.
The need to identify the respective role of shocks, institutions, and
interactions.
Unemployment
7
4. Shocks and Institutions. The panel evidence
Explaining u over time and countries. A necessary intellectual discipline; otherwise too easy...
Shocks: A plausible list of adverse shocks.
•
The slowdown in TFP growth from the 1970s on. More generally, the end of the “30 Glorious years”. Figure 6.
•
The high cost of capital in the 1980s and early 1990s.
•
An open question. The role of globalization/reallocation in the
1990s. Very much in the news. Harder to find in the data.
Measured shocks appear very similar across countries (some exceptions, Franco in Spain, Reunification in Germany...). So cannot explain cross country differences.
Labor market institutions
•
Clear differences across countries. Form of wage bargaining,
generosity of unemployment insurance...
•
Relatively stable over time. A bit more protection in the 1960s
and the 1970s. A bit less since.
But, in general, cannot explain the time series pattern of unemployment.
Example: The maximum replacement rate over time. Figure 7.
This suggests the importance of interactions.
.08
ITA
.06
ESP
ITA
TFP growth
GBR
DEU
.04
GBR
GBR
FRA
GBR
DEU
DEU
ESP
ITA
FRA
.02
GBR
FRA
ESP
FRA
ITA
ESP
ITA
FRA
ITA
DEU
0
ITA
ESP
GBR
DEU
DEU
1960
1970
1980
TFP growth, EU5 countries
GBR
1990
FRA
GBR
DEU
ESP
ESP
DEU
FRA
ITA
2000
80
ESP
ESP
ESP
ESP
ESP
FRA
FRA
FRA
FRA
FRA
60
FRA
Replacement rate
ESP
ITA
FRA
ESP
GBR
DEU
40
DEU
GBR
GBR
DEU
GBR
DEU
DEU
DEU
DEU
GBR
GBR
GBR
20
ITA
ITA
ITA
ITA
ITA
ITA
0
1960
1970
1980
1990
Maximum replacement rate, EU5 countries
2000
Unemployment
8
Interactions. The overall evidence is consistent with the
following description:
•
Largely common shocks.
•
Countries with “worse institutions” have had higher, more persistent unemployment.
•
Which institutions? Econometric answer:
Heavier employment protection
Open ended unemployment benefits
Low coordination of wage bargaining
Low “trust” between labor and capital
A large consensus, but still fuzzy:
•
Exact channels?
•
Spain versus Portugal?
•
Difficulty of measuring institutions.
The usefulness of case studies. Looking at the “miracles”.
Unemployment
9
5.1. Ireland
Basic facts:
•
From 5% in 1973, to 17% in 1986 to 4% in 2002.
•
Mostly movements in u∗ . Figure 8.
•
For real? Yes. (Employment increase much larger)
Proximate causes:
•
Very high TFP growth
•
Much lower real wage growth
•
Lower cost of labor, and an increase in employment given capital
•
High profit rates, capital accumulation and employment
Deeper causes:
•
TFP growth. The trigger (and only the trigger) FDI, in part
due to fiscal breaks.
•
Wage moderation? A common labor market between the UK
and Ireland. Or wage moderation and responsible unions, starting in 1987? Open question.
Lessons: Cannot export the experience. But an important lesson:
Wage moderation can achieve miracles.
UNR
natunr
20
10
0
1960
1970
1980
1990
Ireland: Actual and natural rate
2000
Unemployment
10
5.2. The Netherlands
Basic facts:
•
From 1% in 1970, to 11% in 1983, to 2% in 2002. Figure 9.
•
Mostly movements in u∗
•
For real? (Part time employment, workers on disability?) Yes
Proximate causes: Same as Ireland, on a smaller scale
•
Wage moderation. Wage growth below TFP growth, from early
1980s. Lower labor costs, higher profit rates
•
More employment given capital. More capital
Deeper causes:
•
Explicit and sustained wage moderation. Wassenaar 1982 and
follow up.
•
Institutional reforms: Limited.
Employment protection: Decrease for full time, Increase for
part time.
Unemployment insurance. Effective average replacement rate,
from 71% in 1980 to 56% in 2000. (high relative to EU average)
Lessons: Wage moderation. Low unemployment consistent with high
social protection.
UNR
natunr
15
10
5
0
-5
1970
1980
1990
Netherlands: Actual and natural rate
2000
Unemployment
11
5.3. Sweden
Basic facts:
•
Below 3% until 1990. Then a sharp increase to 8% in 1993,
then back down to 4% in 2002.
•
The increase in the early 1990s is mostly due to deviation of u
from u∗ . Figure 10.
•
For real? Yes but. Evolution of public employment.
Proximate causes for the 1990s fever: A standard boom-bust.
•
Excessive demand in the late 1980s, due to liberalization of
financial markets, sharp decrease in saving rate.
•
Fixed exchange rate, inflation, appreciation. Then, from 1992
on, speculation, depreciation and return to normal.
Deeper causes for low u∗ ?
•
Wage moderation. In evidence in 1991-1993, after depreciation.
•
Generous social protection. Not much change in level. Average
effective replacement rate, from 90% in 1992 to 70% in 2000.
But change in form: Linking unemployment benefits to searce
Lessons. Low unemployment consistent with high social protection.
Can be exported?
UNR
natunr
20
10
0
-10
1960
1970
1980
1990
Sweden: Actual and natural rate
2000
Unemployment
12
6. Implications for policy
Stating them with more confidence than I do have:
•
High social protection consistent with low unemployment. But
has to be done right.
Employment protection. Layoff tax to finance unemployment
benefits, but limited judicial intervention.
Unemployment insurance. Generous, but conditional on search
and job taking.
Wage compression: A lower minimum wage, a larger role for
the negative income tax
•
Coordination, allowing for wage moderation when needed, and
in response to shocks.
Necessary but not sufficient condition (Spain in the early 1980s).
(Elusive “trust”)
•
Make sure that if u∗ goes down, u follows. In other words, more
reactive demand policies. (probably more so than today).