MIT LIBRARIES
DUPL
i!»B
DEWEY
5
5
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
THE QUALITY OF LABOR RELATIONS
AND
UNEMPLOYMENT
Olivier Blanchard
Thomas Philippon
Working Paper 04-25
June 15,2004
Room
E52-251
50 Memorial Drive
Cambridge,
MA 02142
paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
This
http://ssrn.com/abstract=559203
ASSACHUSETTS INSTITUTE
OF TECHNOLOGY
DEC
7 2004
LIBRARIES
The Quality
of Labor Relations and
Unemployment
Olivier Blanchard
June
15,
and Thomas Philippon*
2004
Abstract
In countries where wages are primarily set by collective bargaining, the
effects
on unemployment of changes
crucially
in the
economic environment depend
on the speed of learning of unions.
This speed of learning
is
likely to
depend
dialogue that unions have with firms, on
in turn
on the quality of the
what can more generally be
called
the quality of labor relations.
In this paper,
we examine the
in the evolution of
30 years.
We
role this quality of labor relations has played
unemployment across European countries over the
conclude that
it
has played an important
role:
last
Countries with
worse labor relations have experienced higher unemployment. This conclusion remains even after controlling for labor institutions.
respectively. We thank Emmanuel Farhi for excelWe thank Giuseppe Nicoletti, Gian Maria Milesi-Ferretti, Jim
data. We thank Daron Acemoglu, Roland Benabou, Xavier Gabaix,
MIT and NBER, and NYU-Stern
lent research assistance.
Snyder for help with
Ricardo Caballero, Pierre Fortin, Richard Freeman, Peter Hall, Francis Kramarz, and
Andrei Shlcifer for discussions.
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium Member Libraries
http://www.archive.org/details/qualityoflaborreOOblan
Introduction
1
In countries where wages are primarily set by collective bargaining, the
effects
to
on unemployment of changes
depend
crucially
in the
economic environment are
likely
on the speed of learning of unions.
This speed of learning
is
likely to
depend
in turn
on the quality of the
dialogue that unions have with firms, on what can more generally be called
the quality of labor relations.
we examine
In this paper,
in the evolution of
30 years.
We
We
conclude that
start, in Section 1,
increasing to
11%
reaching a low of
able to
the role this quality of labor relations has played
unemployment across European countries over the
has played an important
it
with a "case study", that of the Netherlands. After
in the early 1980s,
2%
Dutch unemployment turned around,
in the early 2000s.
This turnaround
wage moderation. This wage moderation appears
in turn to the dialogue
mitment to reestablish
last
role.
between unions and
firms,
is
directly trace-
directly traceable
1
and to the unions com-
margins so as to increase capital
sufficient profit
accumulation and employment.
We
then present, in Section
unemployment.
2,
We show how
a simple model of wage determination and
the effects on unemployment of different
changes in the economic environment depend crucially on the perceptions
of unions
We
and
their
speed of learning.
then turn, in Section
3,
to the cross country evidence.
two measures of the quality of labor relations, the
intensity in the 1960s, the other based
We
We
We
construct
based on strike
on surveys of managers
show that countries with worse labor
and longer-lasting unemployment.
first
in the 1990s.
relations have experienced higher
show that the
after controlling for labor market institutions.
effect
remains even
The Netherlands
2
The
evolution of
unemployment
often called the "Dutch
in
11%
1970 to
in the
Netherlands since the early 1980s
unemployment miracle". After increasing from 2%
in 1983, the
steadily to reach a low of
unemployment
2%
rate turned around, decreasing
in 2001. Today,
has increased back to 6%,
it
but most of the increase since 2001 appears to be
The
focus on the
is
1
unemployment
rate
is
cyclical.
not misleading.
It is
sometimes
argued that the increase in employment has come from job sharing, from
an increase
in part-time
employment. This
is
not correct:
The
increase in
part-time employment, which has indeed been substantial, has been more
than matched by an increase
an increase
in the participation rate of
women.
decrease in unemployment hides a high
While
it is
by
in the participation rate, in particular
true that disability rolls are
It is also
number
still
argued that the
of workers on disability.
very high, they have decreased
with unemployment, and thus cannot account
unem-
for the decrease in
ployment.
There
in
is
general agreement that the proximate source of the turnaround
unemployment
is
wage moderation. This can be seen
plots the
unemployment rate and the
wage
more
(or
(log)
wage
precisely the labor cost, as the
wage and non-wage
in
Figure
1,
which
in efficiency units.
The
measure includes both the
benefits) in efficiency units
is
constructed under the
assumption of Harrod neutral technological progress, following Blanchard
[1997];
it
is
normalized to equal zero in 1970-71.
On
a balanced growth
path with constant unemployment, the wage in efficiency units should be
constant. Thus, one can think of decreases in this
wage moderation. As Figure
1
indicates, the
wage measure
wage
as reflecting
in efficiency units
indeed decreased steadily since the early 1980s, standing in 2001 at
its
1970 value; this decrease
1.
Two
[2000].
is
has
83%
of
substantially larger than the decrease of the
useful references are Visser
and Hemerijck
[1997],
and Nickell and Van Ours
same
European Union
variable for the
wage moderation
as a whole. This
has led to an increase in profit, which has led to an increase in capital
accumulation, and an increase in employment over time.
Figure
1.
Unemployment and wage
the
in efficiency units, in
Netherlands, 1970-2002.
There
is
also general
agreement that the start of
this
wage moderation can
be traced to a particular event, the "Wassenaar Accord," signed at the
end of 1982 between unions, business organizations and the government.
At that time, unemployment had increased very
quickly,
and
and
profit
investment rates had collapsed. Unions, explicitly recognizing the need to
reestablish profit margins, agreed to
wage moderation. Despite tensions
along the way, this commitment to wage moderation has remained to this
day.
This short description
it
actually was.
may
give the impression of a process smoother than
The Wassenaar Accord was preceded by
nine failures in the previous nine years.
agreement, with
much
give
and
The Accord
take, including a
nine attempts and
itself
was a complex
commitment by
firms to
consider decreasing working time (at the same wage per hour). Yet,
mains the case that the turnaround
in
wage
setting
happened
it
re-
earlier in
the Netherlands than in other European countries, and observers credit
and
its
it
subsequent follow- through to a long-standing tradition of dialogue
between unions and
firms. Katzenstein [1985] for
example attributes
it
to
"corporatism," which he defines as "an ideology of partnership, expressed
at the national level
flicting objectives
state bureaucracy,
An
(
);
through continuing bargaining between interest groups,
and
political parties."
important question,
ployment,
is
and voluntary and informal coordination of con-
how much
in light of the research
of the decrease in
on the causes of unem-
Dutch unemployment can be
ajey }U8iuAo|dwaun
9
P
Z
8
o
o
o
CNJ
CM
00
G)
CD
4—
05
C\l
o
o
CM
o
i
a:
(D
+j
CD
CD
^
>>
o
N(D "o
^"~
C/)
"D
o
CD
U=
LU
CD
(D
-C
E
>^
o
CD
CD
JC
h•
Y-
CD
E
CD
C
.!?
D
to
D
o
o
o
<
c
CD
§
00
CD
CO
c
CD
CO
CO
CO
D
CO
c/5
+-
•
c
o
c
(D
'o
E
o
^
LU
_c
Q.
E
O) C
L
CD
a
1_
CD
CD
r
Z
*-•
c
(i
_c
TJ
-»—
CD
(D
Q.
C
03
c
o
T™*
c
o
"a
•*—•
c
C
4—
CD
uo
D
CD
T
o
i
CD
t—
C
o
CD
N
o
CD
*-»
CO
TJ
<>
CD
N
ro
E
L_
o
c
o
CD
T
Lsnufi Aouapijjg
in
a6e/v\
CD
O)
CD
attributed directly to this change in union attitudes given labor market
and how much can be attributed to reforms of these labor
institutions,
The question goes beyond
market
institutions.
Nickell
and Van Ours
[2000] for
the scope of this paper (see
an exploration, Schettkat
[2003] for a care-
ful characterization of institutions and institutional change), but one can
make
the following remarks:
The period
since 1982 has seen a
number
Dutch labor
of reforms of the
market. These reforms have however taken place in the late 1980s and the
1990s, thus
some time
after the start of
the decline in unemployment.
rather, they have reduced
wage moderation and the
The reforms have
some
level of social insurance, rather
typically not
start of
been
radical;
of the distortions associated with a given
than decreased that
level.
A
decrease in
replacement rates, and the conditioning of benefits on the acceptance of
"suitable jobs" have decreased the incentives to remain unemployed.
Expe-
rience rating of firms according to their history of worker disabilities have
decreased the incentives for firms and workers to use disability programs
as alternatives to
unemployment. In general, however, the Dutch system
of social insurance remains
EU average, indeed
EU countries — France,
more generous than the
more generous than the average of the four largest
Germany, Spain, and
Italy
—that today have much higher unemployment
than the Netherlands.
3
Shocks, learning, and unemployment
The purpose
of this section
is
to
show that the
effects of
ployment depend very much on whether and how
fast
shocks on unem-
they are perceived
by unions. The model we present makes no claim to generality;
is
to introduce
some
its
goal
of the major shocks which are typically thought to
have been behind movements
in
unemployment
in
Europe over the
last
30
years,
and show how,
crucially
•
in
each case, the response of unemployment depends
on union's perceptions.
Assume an aggregate production
y
where
n
y, k,
=
function of the form:
min(fc, an)
are output, capital,
and employment
captures Harrod neutral technological change.
given by:
7T
y
=
— wn
=
Assume
=
where r
is
a(l
profit rate is
then
w
a
that the supply of capital
k
The
and a
1
k
•
respectively,
is
given by:
+ 0(tt - r))
the user cost faced by firms, and the parameter 6 measures
the response of the supply of capital to net profit
of the user cost.
The presence
of a on the right
—the
hand
profit rate net
side
is
to insure
the existence of a balanced growth path, where capital and output
increase at the
The
first
same
rate.
two assumptions together imply a labor demand schedule
given by:
n
= -: = H-0(7r-r) = l + 0(l---r)
a
•
a
Assume an economy wide
maximizing the wage
bill,
union, acting as a
(1)
monopoly union, and
wn, subject to the perceived labor demand
schedule:
a
Note that we allow union perceptions of
hats, to differ
and
r, all
denoted by
from the actual values of these three parameters.
The assumption
that the union chooses the wage unilaterally, subject
to the perceived labor demand,
ever convenient,
issue, the
8, a,
is
not particularly
realistic. It is
and allows us to avoid an interesting but
nature of bargaining
when
how-
difficult
the two sides (the firms and the
union) differ in their perceptions. Here, only the union's perceptions
matter
•
for the choice of the
The wage implied by
this
wage.
maximization
W= l{
and employment
is
l
is
given by:
' f+
(2)
l)
.
given by equation (1) (firms are assumed to
know
the true parameters).
If
perceptions are correct, employment
is
given by:
n=i±^>
Keeping 8 and
economy
r constant, the
path, where capital, output,
and
real
(3)
follows a balanced growth
wages grow with
a,
and where
employment remains constant.
Now
•
consider three types of shocks.
Suppose there
is
a slowdown in productivity growth, a decrease in
the rate of growth of a, as was indeed the case in Europe starting in
the mid-1970s.
If
perceptions are correct, so 6 remains equal to
with
If
a,
wages adjust
in line
and, as equation (3) indicates, employment remains constant.
however perceptions do not
time,
a,
employment
will
fully adjust, so
be lower
a exceeds o for some
until a converges
back to
a.
Suppose there
•
is
a sharp increase in the real interest rate, as was
indeed the case in the 1980s and early 1990s in most European countries.
Even
ployment
perceptions are correct, so f remains equal to
if
will decrease as
employment.
If
r,
em-
unions prefer to cut both the wage and
perceptions do not adjust however, the wage does
not change, and the decrease in employment will be larger.
Suppose,
•
in 9.
last,
that there
is
an increase
This evolution has been
less
in capital mobility,
studied than the previous two, but
has clearly taken place in Europe over the
quite relevant:
The elimination
of other barriers to capital
an increase
last
30 years, and
may be
of capital controls and the reduction
movements have
surely increased the elas-
So has the widespread process
ticity of capital to its rate of return.
much more
sensitive to
correct, such
an increase
of privatization: Private owners are typically
the rate of return than the state.
Under the assumption that perceptions are
in 9 leads to
ital
an increase
to the profit rate,
in
employment: The higher
and by implication the higher
ployment to the wage, lead the union to moderate
and thus lead to an increase
in
employment.
If
elasticity of capelasticity of
its
perceptions do not
change however, wages do not adjust, and employment
may
decrease. This depends on whether, initially, firms were
itive or negative net profits. If net profits, 1
negative,
In
all
employment
— w/a —
as perceptions
recover and end
up ultimately
and wages
making pos-
adjust, will
moves
out.
employment
higher.
these cases, the specific dynamics of adjustment will depend on the
speed at which unions learn and adjust wages. The speed of learning
to
actually
r were initially
will initially decrease, as capital
Only over time,
em-
wage demands,
depend
in turn
on how easy these changes
in the
component from year to
likely
economic environment
are to identify. Productivity growth, measured, say, by
a large transitory
is
TFP
growth, has
year. It clearly took a while for
8
the slowdown of the
mid 1970s
to be fully perceived (not only by unions,
but by economists and policy makers as
was
largely hidden
Changes
from view by the
in interest rates should
some
well); for
effects of the
two
time, the
oil
slowdown
price increases.
may
be easier to observe, although unions
not fully perceive the implications for investment and ultimately for em-
may
ployment. Increases in capital mobility
some unions
Europe
in
still
speak as
if
also be hard to assess. Indeed,
profits
were largely rents to be ap-
propriated by workers, without implications for employment. While such
rhetoric
past,
may have been
it is
surely not a
appropriate in some parts of the economy in the
good assumption about
capital mobility today.
Formally introducing uncertainty and learning raises a number of conceptual issues. First, firms, under the maintained assumption that they perceive the changes accurately,
ting
may want
to
employment more than implied by the
be
example cut-
strategic, for
static labor
demand
parameters.
We
move
order
may want
to convince unions that the world has changed. Second, unions
to experiment, to
in
the wage in order to learn faster about the relevant
leave both of these issues aside,
and focus on what
us as the main effect of uncertainty and learning, namely that
it
strikes
takes
some
time for unions to learn, and thus to adjust the wage.
So assume that labor demand
not strategic),
is still
given by equation (1) (so firms are
and that unions maximize the expected wage
period (so they have no incentive to experiment
Consider the case where only productivity a
it
is
bill
period by
2
).
stochastic,
and assume that
follows:
log(a t )
=
log(aj)
+ et,
log(a f*)
=
log(a(_i)
+g
In fact, for the case we consider below, namely uncertainty about productivity a,
unions have no incentive to experiment since the signal to noise ratio does not depend
on the wage they set. This is no longer the case when unions are uncertain about the
2.
elasticity of capital supply, 9.
where
normal, and g can take one of two values, g
et is i.i.d.
words, actual productivity
noise.
is
in the
which
will
In
either positive or zero.
is
mid
ductivity growth decreases from g to
at a rate
0.
equal to underlying productivity plus white
Underlying productivity growth
To capture what happened
— g or g =
1970s, suppose that underlying pro-
0.
Unions
will learn
and adjust wages
depend on the tightness of their prior and the standard
deviation of the transitory component. Figure 2 shows such a path, both
for the
wage
in efficiency units
wt/at and for employment (which
mirror image of the wage), based on an
9
—
Figure
2.
subjective probability that
The response
of the path
is
given in the appendix.)
of wages in efficiency units to a decrease
productivity growth.
The main
feature of Figure 2
is
the
hump shaped
in efficiency units, which, given the static labor
a
the
9 equal to 0.98, a standard deviation of the transitory shock of 2%,
and a value of g of 1% (the derivation
in
initial
is
hump shaped
response of the wage
demand
function, implies
response of employment and of unemployment. As pro-
ductivity growth slows down,
productivity growth for
in efficiency units for
wage growth
some time. This
some time,
until
it
also decreases
leads to
but
an increase
still
exceeds
in the
wage
turns around and returns to
its
steady state value. For the parameters we have chosen, this leads to an
increase in wages in efficiency units of about
roughly consistent what
is
5%
at the peak,
an increase
observed in the data at the end of the 1970s for
the European Union as a whole.
Similar dynamics can be derived with respect to the other shocks, be
increase in interest rates, or an increase in capital mobility.
capital mobility, even
may
if it
well also generate a
initial increase,
An
it
an
increase in
ultimately leads to an increase in employment,
hump shape
response of unemployment, with an
followed by a larger decrease later on.
10
1
1
1
'
1
1
a
1
1
1
'c
3
>.
u
c
D
'o
HI
c
si
>.
Ul
0)
o
5
£
5iS
"1
j
-C
1
U
z
>
o
z:
-o
o
a.
/
o
c
J
2
—
•D
jS'
s
o
'-.
s^
^s^
to
o
1—
n
OJ
0)
cC
"v
n
—
*
IS
'
ra
c
c
g
ro
s
13
C
o
Q_
LL
v-
o
5
o
-
T3
c
It
£
;
>.
o
00
/
1
UJ
o
I
.
03
c
o
3
-»^
.
c
/
s
CD
E
,
111
>>
o
/
Q.
E
LU
ac
03
CO
QJ
-
1
1
1
1
\l
1
1
I
1
cn
03
o
CO
c
g
-+-
TO
E
CN
03
t-
3
Learning, the quality of labor relations, and
4
unemployment
What might
lead to different speeds of learning
and adjustment by unions?
Bayesian learning points to differences in the tightness of priors, and the
informativeness of signals.
ogy,
and
generally,
it
points to differences in ideol-
differences in the quality of labor relations:
common economic
firms share a
nomic implications of
is
More
The more unions and
model, or the more they discuss the eco-
different shocks, the faster learning
likely to be.
4.1
Quality of labor relations and unemployment
With
this motivation in
mind, we construct and study the
measures of the quality of labor relations across
•
and adjustment
The
first
for
theoretical
and empirical work on
looking at strike activity
of labor relations
is
example, Hibbs
1960 to 1967
we want
1970,
is
is
The
a simple one: most of the
strikes suggests that the quality
an important determinant of strike activity
[1987].)
The motivation
for
that this predates the increase in
to explain.
The reason
that, in the late 1960s,
is
countries.
a measure of strike intensity, from 1960 to 1967.
is
motivation
for
OECD
two
effects of
for
(see,
choosing the period
unemployment that
stopping in 1967 rather than, say,
many European
countries, especially
France, Germany, and Italy, were affected by social and political unrest, for
Two
reasons largely unrelated to the quality of labor relations.
measures of
OECD
countries.
strike activity are available for the 1960s for
The
first is
most
a measure of days lost in strikes (DL),
the second a measure of workers involved in strikes (WI).
We express
both as ratios to employment.
When
needed,
STRIKE, which we
define
by the maximum of normalized
measure,
we combine them
in
one
11
days
lost
and normalized workers involved:
\ std{DL)
The standard
for using the
are recorded, while
choice,
when
specification
be lower bounds on
variable, this
that, in the 1960s,
were dictatorships,
max
where
all
strikes
two
strike activity.
OECD
countries, Spain
was
illegal.
and Portugal,
We have no other
using this measure, than to drop these two countries
is
that strike activity depends on other
some
of
them observable
unemployment during the period) some
of
them unobserv-
variables than the quality of labor relations,
able.
that not
measure presents two problems. The
strike activity
from the sample. The second
(such as
is
recorded strikes actually happened, so both
all
likely to
As an explanatory
first is
J
deviations are the cross country standard deviations.
The motivation
measures are
std{WI)
,
This implies that using strike activity as a proxy for quality in
a regression entails measurement error.
We
shall return to the issue
below.
The second measure
is
a direct, survey, measure. In 1999, the World
Economic Forum asked the following question
in
to
managers of firms
a large number of countries: "Are labor relations in your firm
cooperative?"
the country
.
The answer was
mean response
labor relations, and denote
it
Clearly, this second variable
than the
tial
first. It
suffers
given on a scale from
as our second
by
is
to
7.
We
use
measure of the quality of
"COOP"
closer to
what we want
to capture
however from an obvious problem of poten-
reverse causality, as perceptions of the quality of labor relations
by firms
ment
in
1999
may
since the 1960s.
well reflect in part the evolution of
(A similar question was asked
unemploy-
in 1993.
This
is
however also quite late in the period, and the 1993 and 1999 measures
are very highly correlated).
12
If
we think
of the quality of labor relations as a stable feature of an econ-
omy, a natural question
The answer
there
is
sion of
on
how
given in Figure
is
OECD
the 18
is
correlated the two measures of quality are.
3,
COOP
which plots
against
countries for which both measures exist.
a strong negative relation between the two (The
COOP
STRIKE
on
is
STRIKE
is
STRIKE,
As can be
R
2
for
seen,
of the regres-
equal to 0.62, the t-statistic on the coefficient
equal to -5.0). This
is
good news
our implicit main-
for
tained hypothesis that the quality of labor relations
a relatively stable
is
characteristic of countries over time.
Figure
We
3.
Cooperation
in
1999 versus Strikes
1960-1967
in
can then turn to the relation between unemployment and the quality
To do
of labor relations over time.
we
so,
divide the period 1965-2002
into decades (more specifically 1965-1974, 1975-1984, 1985-1994,
2002), and for each decade,
uu
where uu
each
t,
is
the
=
at
we
+b
t
regress:
(quality measure^)
mean unemployment
the coefficient at captures the
across countries during the decade.
Table
1.
The
+ en
rate in decade
common
Our
(4)
,
t
in
evolution of
interest
which captures the extent to which differences
explain differences in
and 1995-
is
country
i.
For
unemployment
in the coefficient bj,
in the quality
measure can
unemployment across countries during the decade.
quality of labor relations
and unemployment
rates,
by decade
Basic results are reported in Table
of estimation using the
COOP
results of estimation using the
1.
The
first
regression reports the results
measure. The second regression reports the
STRIKE
measure (Both are normalized by
13
<
-
[2
CM
C/)
-I—
o
c
o
O
un
!_
o
-Q
CO
o
CD
CD
_l
CD
w
"D
0)
C
c
o
o
CD
CO
c
o
-4—
o
c
o
_CD
CD
o
_Q
o
o
0)
ro
E
o
c
_Q
o
CD
o
00
6)
X
LO
l"
0)
L0
CO
~1
1
1
1
5
L
9
AaAjns 6661 'suo!;e|ay joqei
fr
in
1
e
uouejadooQ
Table
1:
Labor relations and unemployment
in 4 decades.
Dependent variable
is
average unemployment over the decade. Three regressions are run for each decade,
with a different
RHS
variable.
Figure 4
(Cooperation, OLS). Standard errors are in
Time period
Cooperation in labor
a
o
relations
(OLS)
C)
d
R2
Strike
activity
in
1960's
Cooperation in labor
C
relations (IV)
R2
Note: Cooperation
italics.
1985-94
1995-2002
-0.904
-1.769
-2.582
-2.548
0.406
0.648
0.767
0.571
20
21
21
21
0.216
0.282
0.377
0.512
0.967
1.515
2.219
1.785
0.432
0.712
0.766
0.572
17
18
18
18
0.251
0.221
0.344
0.379
-0.9
-1.444
-2.484
-2.461
0.633
0.928
1.017
0.695
17
18
18
18
0.19
0.325
0.412
0.532
N
0)
regression
1975-84
N
R2
first
1965-74
N
i-,
corresponds to the
in labor relation is,
from 1999 survey of managers, the answer
to question:
labor relations in your firm cooperative?". Strike activity per employee from 1960 to 1967
composite measure based on employees involved and days
and employees involved
in strikes in the 1960's to
lost.
"Are
is
a
Last regression (IV) uses days lost
instrument for 1999 survey measure.
their
standard deviation, so the coefficients are directly comparable.) For
reasons discussed earlier, both of these regressions suffer from potential
bias,
measurement
in the third line,
error for the
we
first,
simultaneity bias for the second. Thus,
report the results of estimation, using days lost per
employee and workers involved as instruments
for
the measurement error in using the strike measure
COOP.
is
If,
as plausible,
uncorrelated with the
disturbance term in the above regression, this yields consistent estimates
of the effect of the quality of labor relations
on unemployment
each
for
decade.
All three regressions suggest a strong
and
significant effect of the quality
of labor relations on unemployment.
The magnitudes
across the two measures,
OLS and IV
and across
estimation. Countries
with one-standard deviation better quality had about
ment than the average
unemployment
To
OECD
in 1975-1984,
COOP, and
The
relation
Figure
4.
is
the
OLS
1%
less
unemploy-
country in 1965-1974, about 1.5 to
about 2 to 2.5%
in 1985-1994,
give a better sense of the results, Figure 4 plots the
against
are largely similar
2%
less
and 1995-2002.
unemployment rate
regression line for each of the four decades.
quite striking, especially in the last two periods.
The
quality of labor relations
and unemployment
rates,
by decade
Standard checks suggests that this relation
is
robust. Figure 4
makes
clear
that the results are not driven by any single country. Using the underlying
components
little
DL
and
WI
separately rather than combined in
difference to the results.
makes
So does the use of the 1993 answers instead of
the 1999 answers to the World Economic
for the average
CONF
unemployment
Forum
survey. So does controlling
rate in 1960-1967. Estimating equation (4)
using annual unemployment rather than decadal averages) yields a time
series for
1980s,
estimated
bt
and decreasing
which
is
hump
at the very
shaped, reaching a peak in the mid
end of the sample.
1
1
CD
CO
c
g
CM
o
o
CM
•sj-
co
03
0)
i
LO
LO
i
LO
CD
CD
cd
01
S-_
O
JD
CO
_l
CO
Dc
c
o
CO
CD
-h-»
c
a:
o
CD
CO
E
>>
o
.Q
CO
_J
g
c o
C
o
(/)
C/J
CL
E *
hp
Q)
C LO
D
'-*—
TO
03
- LO
^1"
(7)
CD
i—
CO
CD
s.1
i
LO
00
CD
CD
CD
.
O
o
Q.
c/J
- M"
CO
Q)
L_
O
o
CO
H
CD
9L
CH
9
91
(H
9
a;ey }U8UjAo|dujaun aBejaAy
QJ
CO
The quality of labor
unemployment
4.2
An
relations, Institutions,
obvious question at this point
tions affects
unemployment
is
directly,
and
whether the quality of labor
i.e.
given labor market institutions,
The
or through these institutions themselves.
earlier description of the
Netherlands suggests the presence of at least some direct
wage moderation
tions were reformed only after
rela-
started.
effect: Institu-
We now
look at
the cross country evidence.
A
natural starting point
of institutions on
is
to use a specification which allows for
an
unemployment, and then add our measure of
Such a specification was suggested and explored
the two authors [Blanchard and Wolfers 2000].
in earlier
We
effect
quality.
work by one
of
use that specification
benchmark and estimate:
as a
=
Uit
(1
+ (fe_74 + d 7 5_84 + ^85-94 + ^95-03) (» + ^PjXij) + £«
(5)
i
We
divide time in 4 decades: 1965-74, 1975-84, 1985-94, 1995-2003
use decades to
earlier.
make the
Using half decades as
ence to the results.) uu
Xij
is
is
in
more comparable
linear
common
to the results presented
Blanchard and Wolfers makes
unemployment
in
a measure of institution j in country
the impact of the
on a
results
i.
(We
country
i
little differ-
and decade
and
t,
This specification allows for
(unobservable) aggregate shock
dt
to
depend
combination of the different measures of institutions. The
stitutions (described in Blanchard
an index of the length of time
for
and Wolfers) are the replacement
in-
rate,
which unemployment benefits are paid,
union coverage, an index of employment protection, an index of active
bor market policies, union density, a measure of the tax
la-
wedge between
the cost of labor to firms and take-home pay by workers, and a measure of
15
coordination of bargaining.
Table
2:
Shocks and institutions revisited
Results of estimation of this non-linear specification are given in Table
The
column uses only COOP, instrumented by STRIKE. This spec-
first
can be thought as the same as equation
ification
constraint across time periods that bt/at
straint
2.
close to being satisfied
is
results are very similar,
=
(4),
with the additional
constant. This additional con-
by the unconstrained regressions, so the
and our measure of labor quality
is,
as in Table
1,
very significant.
The second column
uses only the variables in Blanchard and Wolfers, and
the results are, in general, similar to theirs. Institutions are defined so that
their expected sign
on unemployment
is
positive; all coefficients turn out
to have the expected sign, although only the last 4 are significant.
The
third
column uses both
institutions
and our measure of
quality.
The
coefficient on the measure of quality decreases from -0.927 to -0.638, but
remains very significant. Interestingly, the coefficients on most institutions
become more
than
significant
in
column
2.
One might have expected
the
"coordination" variable and our quality variable to capture similar dimensions of labor relations
and be very
correlation between the
two
is
0.42),
collinear; this
is
not the case (the
and both dimensions are strongly
significant.
Column
column
up
4 excludes labor market institutions that were not significant in
3.
The estimated
to -0.834,
and
is
coefficient for the quality of labor relations goes
just as significant as in
column
1.
16
~
=
ft
u
OS
3
g
•-
•*
0\
>
"*
•-)
>
ts
B>
1
r
"!J-
>o
C)
m
cr
"i
2
"->
-T
f^
co
<^
B 3
i
:c
^ —
m
in
t^
O d
^
o Bi
C>
"1 r
o
r
r
3 s
s
a
— B! d a
Bi °
Bi
SD
CQ
I
tN
1
|
d
Bi
°
Bi
m
^
d
p!
t-
<n
P:
°°
r-
C3
a
G
3
S
v.
I
(N
"*
rn
m
o
>
^
>
i—t
Bi
ts'
CO
oj
3—
> OS
oo
<N
b>
"->
>«->
Bs
CO
so
d
oo
"1
<N
<=5
— t\
O C
a
—
c>
9
in
1
d
^o
<=j
ts
ts
tN
oo
~^
r-
m
o
d o d
"-I
Oo
<N
ta
Bi
r
--,
Os
r!
q B q
d B! d
1
CO
—
o
ci o
Bi d
?n
1
B>
Bi
d
E 2
c a*
= z
cd
£
u
>
cd
00
rs
.>
&
—
m a>
a
r,
C
3
n
-J
—V
.3
|2
—
i
0-
n
°o OS
o
^
en * (N
~
~,
cs
O
>s-i
(N
OO
B)
oo
r
t\
SO
OS
in
o o ° °0
O
o s
^
CS d
i
r->
"-,
C)
^o
rs
is-j
qi
r
m
o B en
q
d Bs d
in
]
<M
•^f
B)
O
d
---
S
fS
CQ
B>
OS
Bi
OS
m
o
"o
-~<
Br
SO
SO
r-
d B d
Ov
Vo
—
(N
P-
Os
•*
TJ
—a
^
3
Bi
V]
CQ
>
J
I z
c
<U
n.
—
-
£
3
3
01)
E
'to
a>
i
2
£
to
.Si
oo
m
^
—
'
S
<u
3
©
—
a
u
^
•n
Os
'n
m
(N
-
rsi
-^
3
c
a
r^
in
q
fS
r
1
"-
c
B
in
>C
r
l
Os
O
i
r
rs
P
1
r
1
-.
"^*
so
~
"3
d
c
o
to
°
CI
,-k
—I
"5
"^
^_
o
O
3
U
'%
'
-1
~
OJ
u —5
s 3
<L>
u
3
.-
(A
3
u
3
O
OS
s
O
(2
J-l
X
i
to
2
o
1
u
OS
OS
3
o
•*
00
os
3
in
r~
Os
Eta
i
"O
u
cd
O
•a
o
60
.3
Oh
'3
00
CJ>
-
n. -a
n
33
H
CX,
to
"">
3 so
3
E 2
fe
—O tou
J3
-°cd
'C
cd
*
a--
y
m
oo
cr-
m
o
©
i
<n
3>
~
>,
>>
>>
3
3
3
3
S
E
a
Q
> a
CO
o
j^
t-i
3
=;
—
"_*
C3
J
to
13
p4
3
g
u
V.
s
D
E
u
o
3
Q
Ifl
<L>
1—
D
'^
ai
o
r-
o
3
D
m
00
U
>
o
:^
's
D
o
-o
^3
U
3
O
to
t-l
fj
Q.
S
UJ
t)
O
a
1>
,S
oi
Q
>
3
O
o
<
3
D
all
:>
CQ
-C
2
0.
o
.£
to
3
U
-J
to
3
>
o
c
D,
to
t/1
D
™
5
^
:-
O
o
U
^ 2
Conclusions and open issues
5
We
see the point of our paper as arguing that, in addition to formal in-
stitutions, the quality of labor relations
matters in the determination of
unemployment. Faced with adverse shocks, such as those Europe has faced
in
the last three to four decades, countries with better relations appear to
have been better able to limit the
unemployment and
rise of
to turnaround
faster thereafter.
This
is
both good and bad news
ployment
Europe. The good news
in
clearly important,
bad news
is
for those
good labor
is
who want
to achieve lower
that, while institutional reform
by themselves, go a long way. The
relations,
for
the 1960s and the late 1990s, the
quality of labor relations appears slow to change over time,
last
remark points to the obvious
of a theory for
and how
and we
The
it
what determines the
shall limit ourselves to
or country characteristics.
rate of firms clearly differs
Most
for firms to
limit of our paper,
socialist
a few remarks
is
The
if
at
all.
namely the lack
quality of labor relations, and whether
can be improved over time. This
quality of labor relations
unions.
is
that, based on the informal evidence as well as the correlation
between our two measures of quality
This
unem-
is
a large and
difficult topic,
3
.
clearly correlated
with a number of union
attitude of unions with respect to the profit
between communist,
socialist,
and social-democratic
and social-democratic unions now recognize the need
maintain an adequate profit
rate.
The
rhetoric of
many com-
munist unions, on the other hand, remains largely similar to what
30 years ago.
The
it
was
same holds not only for unions, but also for political par-
and Philippon [2004] study the link between labor relations and the control
They find that, across countries, bad labor relations are associated with more
family control, while good labor relations are associated with more dispersed ownership
of shares. They argue that bad labor relations and tight family control reinforce each
other. This may explain the persistence of bad and good relations that we see in the
3.
Miiller
of firms.
data.
17
ties.
of
4
Our
quality measure
communist votes
is
indeed strongly correlated with the number
Whether
in elections in the 1960s.
this
should be taken
as correlation, or as a proximate cause for the quality of labor relations,
still
is
unclear.
Some evidence
suggests, however, that the quality of labor relations can
change over time. The example of the
UK
appears relevant here. Unions
were surely made weaker by the policies of Margaret Thatcher; but there
is
wide agreement that they have also changed their attitude (although the
improvement
where the
in the quality of labor relations
UK
is
right
on the regression
is
not visible in our Figure
3,
line...)
Indeed, the best known early statement of the need for "a partnership between capital
and labor" is by Helmut Schmidt, then the Social Democratic Chancellor of Germany,
in 1976: "The profits of entrcprises today are the investments of tomorrow, and the
investments of tomorrow are the jobs of the day after"
4.
18
Appendix. Derivation of the wage under learning.
Let pt be the subjective probability held by the union at time
and
extract
fit
=
From the
po be the prior.
let
realized
.
t
{ao, 01,
..,
sets the
t
Vt+l
t is
therefore
wage one period
in
advance,
[wt+in t +i\
given by:
is
w +i
t
E
^h+i]
=
union can
bill
wt+i = argmax£
where 6
= g,
at}.
maximize the expected wage
the wage
nt, the
but not a* The information set at the end of period
a<
Under the assumption that the union
to
employment
that g
t
t
[(at+1 )-
1
]
i-r+s
2
Bayesian learning in turn implies
Pt
=
prob(g
=
prob{a t
|
g
|
fi t _i,
fit-i, g
a
=
prob(a
t )
g)prob{g
t
|
=g
|
fi f -i)
fit-i)
Pt-i
pt-i
+(1 -p -i)exp(^f
4
Along the path where e
=
and
<?
=
0,
((log at
2
-gi) -(loga,)
2
))
we have
Pt-i
Pt
Pt-l+(l-R-l)exp(^-)
1M
where we see that p decreases toward
t
0.
Along
this path, the
wage
is
a2
h -
and employment
4>e
Vt-\e~
a
=
2
+ 1-
Pt-l
is
nt =
The
9t
1
+
0(1
-w
t
-r)
values used to derive the path in Figure 2 are p(0)
.02, r
=
.1
and
6
=
=
0.98, g
=
.01,
1.
20
References
[1]
Olivier Blanchard
rise of
and Justin Wolfers. Shocks and
institutions
European unemployment. The aggregate evidence.
and the
Economic
Journal 110(1): 1-33, March 2000.
[2]
Douglas Hibbs. The
political
economy of industrial democracies. Har-
vard University Press, 1987.
[3]
Peter Katzenstein. Small states in world markets. Cornell University
Press, 1985.
[4]
Holger Miiller and
Thomas
Philippon. Corporate ownership and labor
relations (around the world),
[5]
Stephen
Nickell
mimeo NYU,
2004.
and Jan van Ours. The Netherlands and the United
Kingdom: a European unemployment
miracle.
Economic
Policy, 30,
April 2000.
[6]
Ronald Schettkat. Are institutional
rigidities at the root of
European
unemployment? Cambridge Journal of Economics, 27:771-787, 2003.
[7]
Jelle Visser
and Anton Hemerijck.
A Dutch miracle.
reform and corporatism in the Netherlands.
Press,
Job growth, welfare
Amsterdam
University
Amsterdam, 1997.
21
9
1
»-\
~
Date Due
MIT LIBRARIES
3 9080 02618 0106
T»I
lifi
'
''!i;-i
m
m