Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/macroeconomiceff00blan2
DEWET
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
Macroeconomic Effects of Regulation and
Deregulation in Goods and Labor Marlcets
Blanchard
Augustin Landier
Olivier
Working Paper
01 -14
March 2001
RoomE52-251
50 Memorial Drive
Cambridge, MA 021 42
This paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
http://papers.ssm.com/paper.taf7abstract id=262668
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
Macroeconomic Effects of Regulation and
Deregulation in Goods and Labor Markets
Blanchard
Augustin Landier
Olivier
Working Paper
01 -14
March 2001
RoomE52-251
50 Memorial Drive
Cambridge, MA 021 42
This paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
http://papers.ssm.com/paper.taf/abstract id=262668
MASSACHUSEHS
INSTITUTE
OF TECHNOLOGY
AUG 2 2 2001
LIBRARIES
The Perverse
Effects of Partial
Labor Market
Reform: Fixed Duration Contracts
Olivier Blanchard
in
France
and Augustin Landier*
March
6,
2001
Abstract
Rather than decrease
firing costs across
the board, a
number
of Eu-
ropean countries have allowed firms to hire workers on fixed-duration
contracts.
nated at
-
At the end
little
or no cost.
become subject
We
of a given duration, these contracts can be termiIf
workers are kept on however, the contracts
to regular firing costs.
argue in this paper that the effects of such a partial reform of
employment protection may be
perverse.
The main
effect
may be
high
turnover in fixed-duration jobs, leading in tiun to higher, not lower,
unemployment. And, even
if
unemployment comes down, workers may
many
spells of
unemployment and
fixed duration jobs, before obtaining a regular job.
Looking at French
actually be worse
off,
going through
data for young workers since the early 1980s, we conclude that the
reforms have substantially increased turnover, without a substantial
reduction in unemployment duration.
fare of
young workers appears
*MIT and NBER, and MIT
Kreimarz
Peter
for help,
Diamond
for
to
If
anything, their effect on wel-
have been negative.
respectively.
We thank
Larry Katz
for discussions, Francis
and Daron Acemoglu, David Autor, Daniel Cohen, Michael Piore, and
conunents.
—
Perverse effects of partial reform
There
is
now
substantial evidence that high
employment protection leads
to a sclerotic labor market, with low separation rates but long
ment duration.^ While
may
this sclerosis
unemploy-
not lead to high unemployment
because of the opposite effects of low flows and high duration on the unem-
—
ployment rate
it is
both lower productivity, lower output,
likely to lead to
and lower welfare.
Broad reductions
opposition.
litical
in
employment protection run however
The reason
see themselves as having
is
more
simple:
Those who are currently protected
to lose than to gain from such a reduction.
For this reason, governments have either done
at the margin, allowing for reduced protection,
tracts.
into strong po-
In France for example, firms
now
little,
or have tried to reform
but only for (some) new con-
can, under
some
conditions, hire
workers for a fixed duration, at the end of which separation occurs with low
separation costs.
later separation
If
workers are kept beyond this fixed duration however,
becomes subject
to
normal
firing costs.
The motivation
Are such partial reforms better than none?
for this
paper was our suspicion that the answer might actually be negative, that
the effects of such a partial reform might be perverse, leading to higher
unemployment, lower output, and lower welfare
was as
•
for workers.
Our
intuition
follows:
Think of firms as hiring workers
good the matches
are,
in entry-level jobs, finding out
how
and then deciding whether or not to keep the
workers in higher productivity, regular, jobs.
•
Now
think of reform as lowering firing costs for entry-level jobs while
keeping them the same for regular jobs. This
will
^See
make
OECD
firms
[1999],
more
willing to hire
and Blanchard and Portugal
new
have two
effects: It
workers, and see
how they
[2001].
will
Perverse effects of partial reform
more
them
reluctant to keep
it
will
Even
if
a match turns out to be quite productive, a
in regular jobs:
firm
may
still
prefer to
take a chance with a
•
make
perform. But, second,
One may
the worker while the firing cost
fire
new
firms
is
low,
and
worker.
therefore worry that the result of such a reform
may be more
low productivity entry-level jobs, fewer regular jobs, and so lower overall
productivity and output. Higher turnover in entry-level jobs
lead to higher, not lower, unemployment. And, even
comes down, workers may actually be worse
spells of
off,
unemployment and low productivity
if
may
unemployment
going through
many
entry-level jobs, before
obtaining a regular job.^
Our purpose
in this
paper
and empirically. Our
interest
contracts in France.
We
—a popular
partial labor
market reforms.
Our paper
solve
is
it
both theoretically
broader than just the effects of fixed duration
see our paper as shedding
labor market
We
is
First, the effect of labor
issues.
1.
to explore this argument,
is
some
light
on two larger
market institutions on the nature of the
but often fuzzy theme.
Second, the
pitfalls of
We develop a formal model in Section
Section 2. We further explore its properties
organized as follows:
analytically in
by use of simulations
in Section 3.
The model makes
clear that partial re-
form can indeed be perverse, increasing unemployment as well as decreasing
welfare.
We
then turn to the empirical evidence, looking at the effects of
the introduction of fixed duration contracts in France since the early 1980s.
^The French have a word
productivity jobs:
English expression
close.
They
for
such a succession of unemployment spells and low-
call this "precarite".
— although there
is
There does not seem
to
be an equivalent
an adjective, "precarious". "Insecurity" may come
Perverse
effects of partial
reform
Section 5 focuses on labor market
Section 4 shows the basic evolutions.
evolutions for 20-24 year olds, the group most affected by the increase in
fixed-duration contracts.
The
section looks at the evolution of transitions
between entry-level jobs, regular jobs, and unemployment, and also looks at
wages by contract type. The reforms appear to have substantially increased
unemployment duration.
turnover, without a substantial reduction in
If
anything, their effect on welfare of young workers appears to have been
negative. Section 6 concludes.'^
1
A
simple model
In formalizing the labor market,
we think
of
it
as a market in which match-
idiosyncratic productivity shocks lead to separations
context,
we think
of
employment protection
and new
hires. In that
as layoff costs, affecting
both
the layoff decision and the nature of bargaining between workers and firms.
In this section,
we
describe the model, derive the Bellman equations, and
characterize the equilibrium conditions.
1.1
Assumptions
The economy has
trants equal to
s,
a labor force of mass
and each individual
(Poisson parameter)
s,
1.
retires
There
is
a constant flow of en-
with instantaneous probability
so the flow of retirements
is
equal to the flow of
entrants.
Throughout, our focus
contracts, not
on
is
on the economic
their poUtical
effects of the introduction of fixed
duration
economy imphcations. These pohtical economy
issues,
which are highly relevant to the design of employment protection reforms, have been
studied by Gilles Saint-Paul in a series of contributions, in particular Saint-Paul [1996]
and Saint-Paul
[2000].
Perverse
effects of partial
Firms are
at cost k,
reform
risk neutral value
and then operate
create a position
They can always
forever.'*
it
They can
maximizers.
fill
the position
instantaneously, by hiring a worker from the pool of unemployed. In other
words, the matching technology has "workers waiting at the gate".'^
number
of positions in the
economy
by the condition that there
is
is
determined by
The
zero net profit.
free entry,
interest rate
The
and thus
equal to
is
r.
New
matches
all
start with productivity equal to yg. Productivity then
changes with instantaneous probability
is
The new
A.
level of productivity
y
drawn from a distribution with cumulative distribution function F{y) and
expected value Ey. y
Nothing
is
then constant until the worker
in the algebra
as smaller than Ey.
depends on
if
constant until the worker retires,
in the simplest
it
is
they are not laid
The assumption
productivity, "regular" jobs.
is
but
natural to think of
j/o
This captures the idea that workers start in low pro-
ductivity, "entry-level" jobs, and,
ductivity
it,
retires.
way the notion
off,
move on
to higher
that, after the first draw, prois
also inessential but captures
that regular jobs are likely to last inuch longer
than entry-level jobs.
When
productivity changes from
to lay off the worker
productivity yo
worker
— and
hire a
—or keep him
retires, at
j/o
to y, the firm can decide either
new worker
in
an entry-level job with
in a regular job, with productivity y (until the
which point the firm hires a new worker with productivity
yo-)
At the center of our model and
imposed
firing costs.
We
take
them
crucial to the firm's decisions ai'e state-
to be pure waste (think administrative
''Allowing Poisson stochastic depreciation for positions would introduce an additional
parameter, but not change anything of substance.
^The
effects of
them out makes
it
matching
frictions
easier to focus
on the equilibrium are well understood.
Leaving
on the distortions implied by employment protection.
Perverse
and
than transfers.^ The
legal costs) rather
entry-level job
(i.e.
up
to
firing cost associated
and including the time
changes from yo to y)
level
job
reform
effects of partial
is Cq.
The
at
with an
which the productivity
with a regular
firing cost associated
starting just after the change in productivity from yo to y)
(i.e.
is c.
Separations due to retirement are not subject to firing costs.
We can look at the same labor market
Workers are
ers.
risk neutral, with discount rate equal to
with instantaneous probability
unemployed
They look
X
=
equal to
is
for
from the point of view of the work-
0.
New
s.
By
r,
and they
retire
normalization, the flow utility of being
workers enter the labor market unemployed.
an entry-level job, which they find with probability
x,
where
h/u, with h being the flow of hires, and u being the unemployment rate.
Their entry-level job comes to an end with instantaneous probability
which time they are either
laid off,
laid off, or retained in a regular job. If
A, at
they are
they become unemployed, and look for another entry-level job. The
model therefore generates a work
life-cycle, in
which young workers typically
go through a succession of unemployment spells and entry-level jobs until
they obtain a regular job, which they keep until they
The
workers
flow into
who
are laid
unemployment
jobs.
unemployment
is
ofl"
at the
equal to the
is
composed
of
new
retire.
entrants and of those
job.
The
flow out of
of workers hired in
new
entry-level
end of their entry-level
number
All regular jobs are filled from within,
and
all
regular jobs end with
retirement.
The only element
What we need
tions of thinking
is
of the
that at least
about
model
left
to specify
some component
firing costs as
is
wage determination.
of firing costs be waste.
The
implica-
waste or as transfers, and the scope for bonding to
cancel the effects of the transfer component, axe well understood. See for example Lazear
[1990].
We
think that there
our assumption.
is
enough evidence of waste and fimited bonding to warrant
Perverse
We
effects of partial
reform
assume that wages, both
in entry-level
and
in regular jobs, are set
by
symmetric Nash bargaining, with continuous renegotiation. All entry-level
jobs have the
same
level of productivity
Regular jobs have different
with productivity y
will
wage
in a regular job
denoted w{y).
is
be on the
up the model, distortions
set
come only from the presence
paper
and thus pay the same wage wq.
levels of productivity; the
Given the way we have
this
j/o
of the two firing costs, c
eff'ects
and
of a decrease in cq given
economy
in this
cq-
c, i.e.
in the firing costs associated with entry-level jobs, keeping
Our
focus in
of a decrease
unchanged the
with regular jobs.^
firing costs associated
Bellman equations
1.2
Consider
first
the Bellman equations characterizing the firm.
the expected present value of profits from a position currently
entry-level job (the value of
an entry-level job
rent productivity equal to yo-
productivity equal to
y.
is
filled
as an
a job with cur-
Let V{y) be the value of a regular job with
Let y* be the threshold level of productivity above
which the firm keeps a worker, and below which
Vo
for short),
Let Vq be
it
lays
him
off'.
given by:
roo
rVo
The
first
=
{yo
-
wo)
- coXF{y*) + A
term on the right gives flow
/
{V{y)
profit.
- Vo)dF{y)
The second
gives the firing
cost associated with terminating the entry-level job, times the probability
^Note that our assumption that regular jobs are not subject to productivity shocks
implies that the only role of
wage bargaining
c,
in regular jobs,
the firing cost associated with regular jobs,
is
to affect
not layoffs from regular jobs. Allowing for productivity
shocks to regular jobs would complicate the algebra, generate a richer structure of flows,
but not change anything of substance.
Perverse
reform
effects of partial
that the worker
—
laid off
is
itself
equal to the probability of a productivity
change, times the probability that y
is less
than the threshold value y*
third term reflects the expected change in the value of the job
if
.
The
the worker
is
kept in a regular job. (Note the absence of a term reflecting the probability
that the worker retires.
firm can replace
If
the worker retires while in an entry-level job, the
him instantaneously
productivity, so this term
at
no cost by a worker with the same
equal to s(Vo
is
—
Vq)
=
0.)
The sum
of these
three terms must be equal to the annuity value of an entry level job, rVo
V{y)
given in turn by:
is
rViy)
The
first
term on the
= {y-w{y)) + s{Vo-V{y))
right gives flow profit
The second term
reflects the
firm must hire a
new worker
change in value
if
to the
Bellman equations
is
equal to
y.
the worker retires and the
at productivity level yo-
must be equal to the annuity value of a regular
Turn
productivity
if
The sum
of the
two
job, rV{y).
for a worker. Let Vq
denote the expected
present value of utility for a worker currently in an entry-level job (the value
of being in an entry-level job for short),
F"
the present value of utility for a
worker currently unemployed (the value of being unemployed
V'^{w{y))
is
is
also the expected lifetime utility of
market; for this reason,
is
rVo'
and
the value of being employed in a regular job with productivity
Note that V"
Vq
for short),
it is
y.
an entrant in the labor
a natural measure of welfare in this model.
given by:
=
wo
+ AF(y*)(F" -
Vq')
-
sVq'
+
A /
{V'{w{y))
- V^)dF{y)
Jy'
The
is
first
term on the
right
is
the wage for an entry-level job.
The second
the probability that the job ends, times the change in value from going
Perverse effects of partial reform
from employment to unemployment. The third
The
retirement.
is
reflects the loss in value
fourth reflects the expected change in value
The sum
retained in a regular job.
of these terms
is
if
from
the worker
equal to the annuity
value of the value of being in an entry-level job.
V'^{w{y))
given by:
is
rV%w{y)) = w{y) - sV'iwiy))
The worker
he
retires, in
The sum
job.
wage associated with productivity
receives the
which case he
loses the value of
of these terms
is
level y, until
being employed in a regular
equal to the annuity value of being employed
in a regular job.
Finally,
F"
is
given by:
rV = x{Vq^ - F") - sV
The
first
term
is
equal to the probability of being hired in an entry-level
job, the second the probability of retiring while
in value
from retirement.
The sum
unemployed, times the
of these terms
loss
must be equal to the
annuity value of being unemployed.
Equilibrium conditions
1.3
The model imposes
four equilibrium conditions.
condition, that the value of a
new
The
first is
the free entry
position be equal to the cost of creating
it:
Vo
The second
indiflFerent
cost,
is
=
k
(1.1)
that, at the threshold level of productivity, the firm be
between keeping the worker, or laying him
and hiring a new worker:
off,
paying the firing
Perverse
effects of partial
10
reform
y(y*)
The
who
in
third
loses
is
= Fo-co
the Nash bargaining condition for entry-level jobs.
an entry-level job
loses Vq
—
an entry-level job loses Vb
Vq
-f-
— V^. A
cq
=
cq.
yo'-^" =
The
who
in
fourth
loses
(1.2)
the
is
a regular job loses Viy)
—
Vq
worker
firm which lays off a worker
This implies:
co
(1-3)
Nash bargaining condition
a regular job loses V^{w{y))
A
— F". A
+ c. The Nash
for regular jobs.
A
worker
firm which lays off a worker
condition therefore takes the
form:
V\w{y))~V^ = V{y)-VQ +
We now
2
c
(1.4)
turn to a characterization of the equilibrium.
The equilibrium
The equilibrium
is
easiest to characterize
by focusing on two
variables,
V^,
the value of being unemployed, and y*, the threshold level of productivity
below which workers are
One can then think
first,
which we
of the equilibrium in terms of two relations.
shall call the "layoff relation"
as a function of labor
firing costs c
laid-off.
and
cq.
this
is
done,
and
all
cq.
gives threshold productivity y*
market conditions, summarized by F", and of the two
The
second, which
gives F", the value of being
firing costs c
,
The
we
shall call the "hiring relation"
unemployed as a function of
Together the two relations determine
y*
and the two
F" and
y*
.
Once
other variables can easily be derived, and so can the effects
of changes in firing costs.
Perverse
11
reform
effects of partial
The layoff relation
2. 1
The
condition determining the choice of the threshold productivity value y*
by the firm
given by (1.2). Using equation (1.4),
is
{V{f) -Vo +
+
Co)
- y") =
[V^iwiy*])
Note that the right side gives the
can be rewritten
it
c
-
CO
as:
(2.1)
total surplus (i.e the surplus to the
firm and the surplus to the worker from staying together rather than separating)
from a match with productivity y* Were the choice of the threshold
.
productivity level privately efficient, the threshold productivity level would
be chosen so that the total surplus was equal to zero. As
c
—
Co
is
equal to zero, this
positive,
is
some workers
not the case here.
is
will
be
If c
(2.1) shows, unless
exceeds
laid-off despite the fact that
co, so c
is
higher than cq, the worker,
if
Cq
keeping them
would yield a positive total surplus. The source of the distortion
c
—
is
clear: If
kept in a regular job, will be in a stronger
bargaining position and thus be able to extract a higher wage. Anticipating
this,
the firm will only keep jobs where the surplus
offset this increase in
=
the free entry condition Vq
^l±i^
r + s
k, gives
left
-V^-k =
.
The
The next two terms subtract the
The two terms on the
If
-co
V{;w{y*)), together with
relation between y*
first
+
side gives the total gross surplus
a match of productivity y*
.
the
+
(c
-
first
term
is
(i.e.
and F":
Co)
shall refer to this relation as the "layoff relation"
y". The
y*
sufficiently lai^ge to
the worker's bargaining power.
Using the Bellman equations to derive V{;y*)
We
is
(2.2)
between y* and
ignoring firing costs) of
the expected value of output.
outside options of workers and firms.
right side
show the two
roles of co in determining
the layoff decision were privately efficient, only the
first
term would
be present: The firm would choose y* so that the net surplus on a job with
.
Perverse
12
reform
effects of partial
productivity y* was equal to zero.
distortion
due to bargaining.
be (privately)
will
We
It
The second term
imphes
that,
c
if
reflects the private
higher than
is
cq,
then y*
inefficiently high.
can now look at the
effects oi Vu, c
and
on y*
cq
The
.
derivatives
are as follows:
dy*
The
higher the value of being unemployed F", the higher must be the
productivity of the marginal match.
The
lower the firing cost for entry- level jobs,
the higher the threshold
cq,
(and also the larger the deviation of the threshold y* from
efficient level,
The
relation
V^ and
derivation of the second relation between
Nash bargaining condition
for entry-level jobs,
subtracting Vq, this equation can be rewritten
Note that the
first
term
privately
thus the larger the overdestruction)
The hiring
2.2
its
-
^o)
right side
is
equal to the surplus from a
in
is
V^")
CO
is
equal to the
sum
of the outside option of
the worker and the firm. Note that, again, this condition
Firms should
to zero.
This
is
new match. The
the expected value of output from the match.
parentheses
efficient.
Adding and
as:
+
(Vo
+
(1-3).
(^o'
in parentheses
The second term
-
equation
y* starts with the
hire workers until the surplus
not the case here:
The
surplus
is
is
not (privately)
from a match was equal
only driven
down
to cq,
not to zero. Just as before, this distortion reflects the increased bargaining
.
Perverse
effects of partial
13
reform
power of workers coming from renegotiation
in the
+
Using the Bellman equations to replace Vq
entry condition Vq
yo
+
sk
+
\
=
Vq, together
r
[r
+s+
This gives the second relation between
it
set in bargaining,
free
A(l
- F(y*)))(F" +
k)
=
s
AF(y*)co
here). In effect,
with the
k gives:
r y±^dF{y)
+
Jy'
presence of firing costs.
gives the value of being
1/", y*
,
+
and
(r
+ s + A)co
cq (c
(2.3)
does not appear
unemployed such that the wages
and by implication, the present value of
profits associated
with a new position just cover the cost of creating that position and hiring
the worker.
Up
We
shall call
it
the "hiring relation"
to a discount factor (r
surplus from creating a
+
s
+
new job and
which may have to be paid
if
A), the left side gives the total gross
hiring a worker (gross of the firing cost
the productivity shock turns out to be lower
than the threshold).
Turning to the right
side,
note that there are two terms in
hiring privately efficient, then only the
first
cq.
Were
term on the right side would
be present. Hiring would take place until the total gross surplus was equal
to the expected firing cost (the probability that firing takes place times the
firing cost).
The second term
reflects the distortion
coming from the
effect
of Co on the bargaining position of workers.
We
V^
is
can now look at the
effects of y*
and
co
on V^. The
effect of y*
on
given by:
(r
+ 5 + A(l - F{y*)))-— =
dy*
The
A/(y*)(y"
+
sign of the derivative appears ambiguous:
k
-
co
-
^^)
+
r
An
s
increase in y* leads
Perverse
effects of partial
reform
I4
both to a higher expected output in continuing jobs, but also to a higher
be important
this will
is
(r
If
both
(c
(i.e.
at the intersection with the first relation, (2.2)),
given by:
+5+
-
Co)
A(l
- Fiy*)))^ = -A/(y*)(c -
and the density function f{y*) are
then an increase in y* leads to a decrease in V^.
then V"^
if
c
=
independent of
is
y*.
the layoff decision
Co,
say more, and
later on:
At the equilibrium
the derivative
we can
But, in fact,
probabihty that jobs are terminated.
The
is
intuition
is
cq)
<
from zero,
different
If either c
as follows:
=
cq or f{y*)
As we saw
=
0,
earlier,
privately efficient, so a small change in y*
has no effect on the surplus and thus no effect on the feasible I4.
If c
>
Co
however, the marginal regular job generates a positive surplus, so an increase
in y*
,
if it
leads to an increase in the layoff rate
(i.e.
if
>
f{y*)
0) leads to
a smaller total surplus, requiring a decrease in the feasible F„.
Now
consider the effect of co on Vu (given y*).
(r
+s+
A(l
- F(y*)))-— = -(r +
5
+
Prom
(2.3):
- XF{y*) <
A)
dco
An
increase in cq decreases the feasible value of being unemployed, F".
There are two separate
is
by
a direct cost
firms,
effect:
effects at
An
work
The
first,
captured by
The
Both
effects require
larger surplus. In equilibrium, this
is
new matches
to generate a
achieved through a lower value of V^.^
a familiar result from bargaining or efficiency wage models,
Shapiro and Stiglitz [1984], or more recently Caballero and
equilibrium,
in the feasible
second, captured by {r + s + A), reflects the effects of firing
costs through bargaining.
is
— AF(y*),
increase in cq increases firing costs actually paid
and therefore increases waste, leading to a decrease
value of V^.
^This
here.
unemployment plays the
role of a
Hammour
(for
example
[1996]), that, in
market "disciphne device". In these mod-
Perverse
The equilihrium
2.3
The two
we have
relations
just derived are
relation, (2.2), the "layoff relation"
higher the threshold y*
flat
or
downward
.
y*
.
,
The second
sloping
(it
around the equilibrium: V^
of,
15
effects of partial reforvi
is
is
is
drawn
upward
in
sloping:
Figure
The
1.
The higher V", the
relation, the "hiring relation",
drawn
as
downward
first
is
either
sloping here), at least
either invariant to, or a decreasing function
Together the two relations determine the threshold productivity level
and the value of being unemployed. The equilibrium
The
effects of
The
value of Co
makes
increases y*
.
given by point A.
a partial reform of employment protection,
on y* and on V^, keeping
of a decrease in cq
to derive.
is
The
more attractive to
the effects
then easy
c constant, are
layoff relation shifts to the right:
it
i.e.
For given
the lower
l^",
layoff entry-level workers,
and thus
hiring relation condition shifts up: For given y*, lower cq
and
leads to a higher value of V^, both because of the reduction in costs,
because of the decrease in the bargaining power of entry-level workers.
The new
equilibrium
unambiguously
is
given by point B.
increases, the effect
on V"
is
It
is
clear that, while y*
ambiguous.
This
is
because
there are two distortions at work, and they work in opposite directions.
•
On
the one hand, the decrease in cq leads to an increase in
and thus to an increase
(a distortion
(c
—
cq)
in the distortion affecting the layoff relation
which depends on the bargaining power
in regular jobs
relative to entry-level jobs). This tends to decrease F".
•
On
the other hand, the decrease in cq leads to a decrease in the dis-
tortion affecting the hiring relation (a distortion which
the bargaining power of workers in entry-level jobs).
depends on
This tends to
increase V^.
els,
the zero profit condition
ties
down the wage. Any
factor which increases the
given reservation utihty requires, in equilibrium, a decrease in reservation
utility.
wage
-So
0)
W
(U
£ °
Pi
ID
O
.„
E ^ ^
3 U
ri2
O"
W
i-H
.
.
''
Perverse
To
more
see the two effects
first
small changes in y* have no effect on
only effect of a decrease in cq on
relation relation:
By both
suppose
clearly,
In this case the
to zero to start.
16
reform
effects of partial
F"
is
distortion
V"
first
that (c
-
cq) is
absent and, as
is
in the hiring relation.
through
its
direct
eflFect
equal
we saw,
Thus, the
in the hiring
decreasing waste and decreasing the bargaining
power of entry-level workers, the decrease
an unambiguous
in cq leads to
increase in V".
This case
=
Co)
0,
is
represented in Figure
the hiring relation
is flat
2.
We know
from above that,
The
at the equilibrium.
if
(c
—
decrease in cq shifts
the hiring relation condition up: Lower costs and lower bargaining power by
entry- level workers lead to a higher equilibrium value of F".
For given V^, a decrease in cq
in Co shifts the layoff relation to the right:
makes
moves from
When
the
more
layoffs
first
A
(c
—
in
y.
to B, with higher F",
an increase
in y*
The equilibrium
.
and a higher threshold, y*
on
cq) is positive instead, the effect of the decrease in co
distortion
increase in the
attractive, leading to
The decrease
becomes
relevant.
first distortion,
The strength
The
decrease in (c
—
and thus, other things equal,
of this effect
is
proportional to (c
—
an
cq) leads to
to a decrease
Co)/(y*) and
is
number
thus increasing in the density evaluated at the equilibrium-in the
of entry-level jobs which are (inefficiently) terminated as a result of the
increase in y*
.
If either (c
—
effect
can dominate. Figure 3
large
around y
=
adverse
cq) or /(y*) are sufficiently large, this
is
drawn on the assvunption that /(y)
is
very
y*, so the hiring relation is (nearly) vertical. In this case,
a decrease in co does not
But, as before,
shift the hiring relation.
it
shifts
the layoff relation to the right: For given F", a decrease in cq makes layoffs
more
A
attractive, leading to an increase in y*.
to B, with lower value
V^ and an unchanged
To summarize, we have
or/and /(y*) are
The
,
a
first
answer to our
sufficiently large, a partial
equilibrium moves from
threshold, y*
initial question. If (c
reform
may
—
cq)
indeed lead to an
o
\
J3
\
\
"aj
\
\\
c
o
o
CS
"S
6
o
S*"
t+-i
o
p>.
00
CS
J
u
\
1
00
\
^
\\
Ui
\
\
X
^
\
\
A
\
^ ^
'
\
1
\\
m
^i
i[x
\
\
'
\
\
^
'
\
^
'
\
^
'
\
^
1
\
^
1
\
.
c
1
^
1
\V
o
^
\
\i\
1
1
pj .s
Bi (L.
\
\
\
\
\
1
1
1
2
\
\
\
^
\
\
^
\
\
it
\
\
-a
\
\
C3
\
\
i
<
\
s
I-
\
\
\
\
\
\
\
\
\
V
\
\
\
\
\
\
\
\
1^
1^
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
1
3
>
>
\
\
^
§\ \
<U
c
ca
^
<i)
>
r/l
—
Ou
Bi
U)
\
\
\
\
S
\
\
\
\
n.
w
\
\
\
O
ri
X
\
\
C/l
,^
\
c
o
o
c
n
%
1
Ut
60
S
O o
\
\
\
\
\\
\
\
\
^
\
\
\
\
\
\
\
\
X
tu
\
s
\
\
DD
\
\
<
C3
\
\
\
o
\
\
\
\
\
^
\
\
\
\
\
\
\
\
\
O
\
\
\
\
\
\
\
\
OJ
H
\
\
\
\
\
\
1
3
>
3
>
\
\
Perverse
effects of partial
11
reform
increase in excess turnover, and, by implication, to a decrease in the value
of being unemployed.^
Other wage setting assumptions
2.1^
We
have assumed symmetric Nash bargaining.
It
is
easy to extend the
analysis to allow for differential bargaining power, both between firms
workers, and between workers in entry-level and in regular jobs.
of this extension are straightforward.
workers
results
The higher the bargaining power
in regular jobs relative to that of
of
workers in entry-level jobs, the
stronger the effect of a decrease in cq on the
it is
The
and
first
distortion, the
more
likely
that partial reform leads to a decrease rather than an increase in welfare.
We
have also examined the
shall discuss
effects of
a
minimum wage
constraint.
As we
and explain below when presenting simulations, under the Nash
bargaining assumptions, decreases in
F"
are associated with an increase in
Wq. Thus, a constraint which prevents the wage from decreasing, such as a
minimum wage
effects of partial
from increasing
will
not rule out perverse
reform on welfare (A constraint which prevents the wage
will increase welfare;
representation of a
2.
and
constraint, will not be binding,
minimum wage
but this does not seem to be the right
constraint.)
Other implications
5
Given the equilibrium values of y* and F",
it
is
straightforward to derive
the other variables of the model. For example:
•
The
layoff rate is given
by XF{y*), so a decrease in
cq,
which, as we
have seen, unambiguously increases y*, unambiguously increases the
^Note that,
maximizes V"
for values of the
will
be
less
partial "partial reform"
than
(i.e.
parameters that give
c
rise to this effect,
the value of co that
but positive. Thus, this can be seen as an argument
some decrease
in co
from
c,
but not
all
the
way
for
to zero)...
Perverse
effects of partial
18
reform
layoff rate.
Using the condition
•
ployment X
improving
—
is
given by x
if
F"
is
ambiguous
s.
Even
(r
+
cq,
the hiring rate from unem-
s)l/"/co-
when
Thus,
cg decreases
unemployment duration
—we
is
welfare
know that x
decreases.
rate
given by u{x
is
+
s
But the
— {XF{y*)x)/{\ +
unemployment duration decreases
if
reform
if
in-
effect
in general.
The vmemployment
•
=
increases
creases, equivalently,
— F") =
{Vq
tfiat
s))
=
{x increases), higher
turnover {F{y*) increases) implies an ambiguous effect on the unem-
ployment
rate.
From the Nash bargaining
•
conditions, the values of being
an entry-level job, of being employed
in a regular job
employed
with productivity
equal to the threshold, and of being unemployed, are related by Vq
F" =
Co
and V^{w{y*)) —
entry-level jobs
more
like
VJf
=c—
2co.
in
—
Thus, a decrease in cq makes
unemployment (decreasing
level jobs less like regular jobs (increasing c
—
2cq).
cq),
and entry-
In this sense, a
reduction in cq leads to increased dualism in the labor market.
To
fully characterize the effects of the decrease in cq
dimensions of our economy,
This
is
what we do
Simulations
Our
goal in this section
life-cycle of
unemployment
We
is
more convenient
to turn to simulations.
in the next section.
3
work
it
on the different
is
to
show the
effects of partial
reform both on the
an individual worker, as well as on macro aggregates, from
to
GDP.
think of the unit time period as one month, and choose the param-
eters as follows:
Perverse
•
We
to
•
effects of partial
19
reform
normalize the level of output on an entry-level job, yo to be equal
1.
We
take k to be equal to 24, implying a ratio of capital to annual
output on an entry-level job of
•
We
2.
take the monthly real interest rate,
r,
to be equal to 1%. Together
with the two previous assumptions, this implies a share of labor in
output on entry-level jobs, of
•
We
We
=
76%.
take the monthly probability of exogenous separation ("retire-
ment")
•
(1-. 01*24)
s,
to be equal to 1.5%.
take the monthly probability of a productivity change on an entry-
level job,
A to be equal to 10%. This implies an expected duration of
an entry-level job of about a year.
•
We take the distribution of productivity on regular jobs to be uniform,
distributed on [m - 1/2/, m + 1/2/], thus with mean m, and density
/.
The
use of a uniform distribution makes particularly transparent
the influence of the density
•
effects of partial reform.
To capture the notion that regular jobs
the
mean
m
minated, the
•
/ on the
equal to
mean
1.4.
are
more productive, we
(Because jobs below the threshold are
set
ter-
of the observed distribution will be higher.)
Because our theoretical analysis
in the previous section
showed that
the density function plays a crucial role in determining the outcome,
we
look at the effects of reform for different values of /.
below show the results of reform
•
We
for values of
choose the firing cost on regular jobs,
The graphs
/ varying from
equal to 24
c,
most simulations, represent about a year and a
1
to 6.
—which,
in
half of average output.
Perverse
We
20
reform
effects of partial
and empirical evidence
shall discuss the legal
for
France
in the
next section; we believe this to be a reasonable estimate.
Our simulations then
focus on the effects of a decrease in
either too large or too small, the equilibrium
a point where y*
lies
equilibria are interesting,
The
2
results
months
at
i.e.
outside the support of the productivity distribution
rate; their effect takes place only
is
at a corner,
In those cases, changes in cq have no effect on the layoff
for regular jobs.
where there
may be
If cq is
cq.
we
through bargaining. While these corner
limit the presentation of results to the range
an interior solution, so changes in y*
below are presented
of output.
The
for the
affect the layoff rate.
range where cq decreases from 6 to
results are presented in Figures 4a, 4b,
and
5.
Figures 4a and 4b show the effects of partial reform on different aspects
of a worker's individual experience. Figure 4a plots F", the value of being
unemployed, F{y*), the probability that the worker
an entry-level
job,
is
laid-off at the
end of
x the monthly hiring rate from unemployment, and T^,
the expected time to a regular job starting from unemployment. Figure 4b
gives the behavior of
given by w{y)
=
+
a
wages
in entry-level
0.5y and
=
wq
the two constant terms a and ag
cq
and regular jobs. These wages are
+ 0.5yo
respectively. Figure
—which give the
levels of
wages
4b plots
for a given
level of productivity.
For each
as one goes
X
axis,
3D
box, the firing cost cq
away from the
origin.
The
is
plotted on the y axis, decreasing
density function /
is
plotted on the
with the density decreasing as one goes away from the origin. The
variable of interest
Start with
V"
is
plotted on the vertical axis.
in Figure 4a.
For low density
increases V^. But, for high density /,
was given
in the
previous section.
it
—low /—a decrease
decreases F".
When
/
is
The
in cq
basic intuition
low, the adverse effects of
reform on excess turnover are small, and workers are better
high, the adverse effects of excess turnover dominate.
off.
'Wlien
/
is
'~ero~'^3V
s9
—
Perverse
effects of partial
This intuition
of reform
on x
increases x,
is
reform
21
confirmed by looking at x and F{y*). While the
theoretically ambiguous, in our simulation reform always
is
and thus decreases unemployment duration.
the probability that an entry-level job will lead to a
theoretically unambiguous). This second effect
high.
For /
=
efi'ect
6,
is
It also
increases
layoff" (this effect is
stronger
the probability increases from 0.3 to
when
0.8; for
/
density
=
1,
is
the
probability increases from 0.45 to 0.75.
The
it
last
takes a
density
is
box
new
in
Figure 4a shows that reform increases the average time
entrant to get a regular job.
high.
=
For /
6,
The
effect
is
stronger
when
the
the expected time increases from two years to
nearly six years.
Figure 4b gives the behavior of wages for regular and entry-level jobs for
a given
prior:
wage
The
level of productivity.
Higher
relative level of the
firing costs lead to higher
for regular jobs.
But the
two wages
effect of a decrease in cq is less intuitive at
.
One might have guessed
relative bargaining
crease in their
wage
one's
bargaining power and thus a higher
first:
•
fits
(i.e
we had guessed)
.
that the decrease in the
power of entry-level workers would lead to a derelative to that of workers in regular jobs. This
is
not necessarily the case: In general equilibrium, the duration of unem-
ployment changes, with
4b shows that the
same
differential effects
effect of
sign as the effect
'"it
Nash bargaining
co goes
down
between reservation
utility
and the wage
set
for regular jobs.^^
can be shown analytically that,
down when
reforms on the wage in regular jobs has the
on V": Like V", the wage may go up or down;
this reflects the tight link
in
on the two wages. Figure
in regions
wq decreases with
where reform
cq.
is
perverse
— where V" goes
,
22
Perverse effects of partial reform
•
Perhaps even more surprisingly, the wage
as Co goes
down
understand
this
more a decrease
regular job.
—the
is
in
for entry level jobs goes
up
The way
to
more so the higher the
density.
terms of bonding. The higher the density, the
being kept in a
in cq decreases the probability of
Thus, the lower the "bond" workers in entry-level jobs
are willing to pay in the form of low wages, or equivalently, the higher
the wage they require to take an entry-level job.^^
There
is
another
countervailing effect at work, lower bargaining power for workers in
entry-level jobs,
which leads to a decrease
simulation, this effect
is
dominated by the
in the wage;
but in our
first.
Figure 5 shows what happens to the macroeconomic aggregates.
first
box repeats the graph
for
V^
We
in Figure 4.
The
can think here of
F"
not as the value of being unemployed, but as average lifetime utility for a
worker in the economy, thus as a mefisure of welfare.
The second box shows the
and shows these
effects of lower
in
effects to
reform on the unemployment rate,
be ambiguous.
For low density, the combined
duration and only slightly higher turnover lead to a decrease
unemployment. For high
first
effects of
density, the effect
goes up as cq decreases, then goes
there was a need, that
what happens
ambiguous. Unemployment
is
down a
to utility
(This
bit.
is
a warning,
if
and to unemployment need
not have the same sign. For high density, utility goes
down
strongly while
unemployment goes up and then down.)
The
third
box
plots the proportion of workers
or employed in entry-level jobs.
The
idea
is
who
are either
unemployed
to get at the idea of "'precarite"
the idea that the decrease in unemployment,
if
any,
may come with
a large
increase in low productivity jobs. This proportion increases with reform, for
^^From an interview of a worker on a fixed duration contract (CDD): "The only reason
I
took a
CDD
was to have a shot
at a real job later on." Liberation [2000].
Q
O
yjjt*??"
Perverse
all
effects of partial
values of /. Again,
23
reform
it is
when /
stronger
is
high.
In this sense, reform
indeed increases precarite.
The
last
graph gives the value of GDP. For low density, the decrease
unemployment
in the
with the limited increase in low pro-
rate, together
ductivity entry-level jobs, leads to an increase in output. For high density,
the larger increase in the proportion of entry-level jobs, and the roughly
constant unemployment rate, combine to lead to a decline in output
nearly
5%
under our parameter assumptions. Another warning
in order here;
What happens
be quite
can
all
4
The development
to output, to unemployment,
is
— by
therefore
and to
utility,
different.
of
CDDs
in Frcince: Basic facts
and evo-
lutions
In France, regular contracts, called "Contrats a duree indeterminee"
"GDI"
layoff
for short, are subject to
workers
for
employment protection
rules.
one of two reasons: For "personal reasons"
,
in
,
or
Firms can
which case
they have to show that the worker cannot do the job he or she was hired
or for "economic refisons", in which
for,
needs to reduce
its
Ccise,
the firm must prove that
it
employment. ^-^
Barring serious negligence on the part of the worker, the firm must give
both a notice period and a severance payment to the worker. The notice period
is
relatively short, 1 or 2
months depending on
of a specific contract between unions
set
by law
is
and
firms, the
also modest, typically 1/10 of a
seniority. In the
amount
absence
of severance pay
month per year
of work, plus
1/15 of a month for years above 10 years. But firms perceive the costs to
be much higher, because of the administrative and
legal steps required to
go through the process. The monetary equivalent of these costs (which are
^^A useful source on French labor legislation
is
the
Lamy
[2000].
Perverse
reform
effects of partial
24
indeed waste from the point of view of firms and workers)
but severance packages oiTered by firms in exchange
are typically
much more generous than
is
hard to
assess,
quick resolution
for a
the legal minimum. ^'^
Since the late 1970s, successive governments have tried to reduce these
costs
by introducing fixed-duration contracts, called "Contrats a duree de-
terminee", or
CDDs. These
contracts
still
require a severance payment, but
eliminate the need for a costly administrative and legal process. -^^
4-1
A
The history and
brief history of
With the
law
the current rules
CDDs
goes as follows:
election of a socialist
in 1982, their scope
CDDs
government
was reduced:
A
were introduced in 1979.
in 1981
list
and the passage of a
of 12 conditions was drawn,
and only under those conditions could firms use fixed-duration contracts. In
1986, the 12 conditions were replaced by a general rule:
be used to
fill
a permanent position in the firm.
dates for the most part to an agreement signed in
Under
this agreement,
four reasons:
CDDs
should not
current architecture
March
1990.
can be offered by firms
The replacement
(1)
The
CDDs
for only
of an employee on leave (2)
Temporary
increases in activity (3) Seasonal activities (4) Special contracts,
facilitating
employment
unemployed. The
government has
another;
some
at
from the young to the long term
of special contracts has
tried to
aimed
grown
in the 1990s, as
improve labor market outcomes
for
each
one group or
of these contracts require the firm to provide training,
many come with
CDDs
list
for targeted groups,
one of
and
subsidies to firms.
are subject to a very short trial period, typically one month.
They have a
fixed duration,
from 6 to 24 months depending on the
'^For a comparison of France with other
OECD
countries, see
OECD
^^Poulain [1994] gives a detailed description of the rules governing
[1999].
CDDs.
specific
Perverse
effects of partial
contract type.
Mean
be renewed, and,
worker
is
worker
is
in
duration
any
case,
kept, he or she
is
roughly one year. They typically cannot
cannot be renewed beyond 24 months.
must then be hired on a regular contract.
not kept, he or she receives a severance payment equal to
total salary received during the
consideration would raise this
Two
25
reform
life
6%
If
the
If
the
of the
of the contract (a law currently under
amount
to 10%).
other dimensions of these contracts are relevant here:
First, the
law states that the wage paid to a worker under a
CDD should
be the same as the wage which would be paid to a worker doing the same
job under a GDI. This
shall see,
it
is
obviously difficult to verify and enforce, and, as
appears not to be satisfied in practice.
Second, at the end of a
Unemployment
fits.
CDD,
workers qualify for unemployment bene-
benefits start at either
40%
of the previous gross salary,
plus a fixed sum, or 57.4% of previous gross salary, whichever
vantageous.
we
The
benefits then decrease over time; the decrease
is
more ad-
is
faster the
younger the worker, and the shorter the work experience. For example, a
worker who has been working
efits for
12
4 months; a worker
months
gets 4
for
4 out of the previous 8 months, gets ben-
who has been working
months with
full benefits,
for 6
out of the previous
then 3 months at 85%, then
nothing, and so on for workers with longer employment histories. In short,
workers can alternate between
CDDs
and unemployment
spells,
and receive
benefits while unemployed.
For our purposes, the history and the specific set of rules regulating
CDDs
•
has two main implications:
One should
think of what has happened since the 1980s primarily as
an increase in fixed-duration contracts at the extensive margin (an
increase in the
number
of eligible workers
and
jobs), rather
than as an
increase in the intensive margin (a decrease in cq)}^
^A model which
formalizes the introduction of
CDDs
at the extensive margin,
and
Perverse
The
•
26
effects of partial reform.
rather stringent rules governing
CDDs
(conditions, duration,
non
CDDs
has
renewal) imply that, while the proportion of workers under
increased over time,
will
not reach
—the
in particular Spain.
it
has not reached
levels
—and, unless
rules ai^e changed,
observed in some other European countries,
^^
Data sources
4-2
Our
data, here
and
in the next section,
come from "Enquetes Emploi", a
survey of about l/300th of the French population, conducted annually by
INSEE, the French National
versus
CDD status are only available from 1983 on,
at the evidence
from 1983 to 2000. The design of the survey
Questions about
we only look
so
and the wording
discontinuities in
clearly in
We
and
of
CDI
some
some
of the questions were
changed
in 1990, leading to
of the series in 1990; these discontinuities appear
of the figures below.
use the "Enquetes Emploi'' to look at the evolution of both stocks
flows.
•
some
Statistical Institute.
The
Measures of flows can be constructed
in
two ways:
3-year panel data structure of the survey allows to follow two
thirds of individuals across consecutive surveys,
and so
to
measure
their annual transitions. Panel-based transition probabilities ("panel
transitions" for short) can be constructed from every year since 1984
on, with one exception:
Changes
in
survey design in 1990 make
it
impossible to compute transitions for 1990.
which shares some of the features of our model (but was developed independently),
given in
^
Cahuc and Postel-Vinay
is
[2000].
For a description of the nature and the scope of fixed-duration contracts in Spain, and
in Italy, see for
[1998].
example Guell-Rotllan and Petrongolo
[2000],
and
Adam and
Canziani
Perverse
27
effects of partial reform,
• In addition,
status 12
from 1990 on, the survey includes a question asking for
months
earUer.
Thus, except
for
to the question has not yet been tabulated,
1999 when the answer
we can
also construct
retrospective transition probabilities ("retrospective transitions" for
short) for each year since 1990.^"^
For our purposes, namely assessing the evolutions (rather than the
of transition probabilities over time,
As documented by many
nates.
not clear which approach domi-
it is
researchers, transitions based
spective information are subject to systematic
memory
is
more
likely to
We
some
change over time:
of workers with short duration jobs
tion.
memory
biases. ^^
on
retro-
But these
biases are likely to be fairly stable over time. Panel based transi-
tion probabilities suffer instead from
smaller,
levels)
therefore remain agnostic
may
attrition bias.
An
This bias, while
increase in the proportion
well lead to
an increase in
and present both the numbers
based transitions from 1984 to 2000, and
for retrospective
for
attri-
panel
information based
transitions for 1991 to 2000.
4-3
As a
Basic evolutions
start,
Figure 6 plots the evolution of
CDD employment
of total (salaried) employment, since 1983.
It
shows how
as a proportion
this proportion
has increased from 1.4% of salaried employment in 1983 to 10.8% in 2000.
^''The question actually asks for status during each of the previous 12 months, thus
allowing for the construction of monthly probabiUties
— which are closer conceptually to
the instantaneous probabilities in the theoretical model.
Because of well known issues
such as rounding up by respondents, these monthly probabiUties are very noisy, and we
have not explored these data further.
^*For more on the differences between the two sets of transition probabilities in the
context of Enquetes Emploi, see
Magnac and
Visser [1999], and Philippon [2000].
Figure
6.
Proportion of
CDD
in
employment
CDD/(CDD+CDI), 1983-2000
0.12
0.10 -
0.08
0.06
0.04
0.02
0.00
1984
1986
1988
1990
1992
1994
1996
1998
2000
Perverse
effects of partial
At the same
time, the graph
which firms can
today,
oiTer
28
reform
CDDs
makes
clear that the specific conditions
have hmited their scope; by contrast,
more than 30% of salaried employment
is
in the
in
under
Spain
form of fixed-duration
contracts.
CDDs in total employment remains limited, the
development of CDDs have completely changed the nature
While the proportion of
introduction and
Figure 7 shows the evolution of the
of the labor market for the young.
who
proportions of individuals, age 20-24,
employed under a
CDD,
The
number
•
figure yields a
The proportion
cally,
employed under a CDI,
are either
or unemployed, or students, from 1983 to 2000.
of conclusions:
of students in this age group has increased dramati-
from 21% in 1983 to 49%
in 2000.
This increase
is
due
in large
part to a deliberate policy aimed at increasing the proportion of chil-
dren taking and passing the baccalaureat (the
exam
school); this proportion has increased over the
to 59%.
But
it is
also a reflection of the
at the
end of high
same period from 28%
poor labor market prospects
unemployment has decreased
faced by the young; indeed, as
since the
mid-1990s, so has the proportion of students. This indicates that, for
this age group,
unemployment numbers should be interpreted with
caution.
•
The proportion
of
unemployed
roughly constant, from
in
in 1983 to
16%
in 1999,
remained
and down to 12%
2000 (although, because of the steady decrease in participation, the
unemployment
and 24%
•
15%
in a given 5-year cohort has
Most
rate has increased from
in 1983 to
32%
in 1999,
in 2000).
relevant for our purposes, the proportion of
dropped while the proportion of
60%
20%
of a cohort (equivalently
under CDIs;
CDDs
95%
CDIs has sharply
has sharply increased. In 1983,
of those employed) were
in 2000, the proportion
employed
was down to 21% (54% of those
Q tb Q
5
Q
hO O
Q.
dO<
UJ
>
o
o
o
CNJ
I
CO
00
^
^^
CO
o
I
1^
d
CD
CD
CO
in
d
CJ
d
uo|;jodojd
CNJ
d
cz>
d
Perverse
29
effects of partial reform,
And during
employed).
ployed under
CDDs
the same period, the proportion of those em-
went from 3.0% (5% of employment) to 17% (46%
of employment).
The same quahtative
evolution
is
visible in other
quantitative effect decreases across cohorts.
10%
increased from 1.6% in 1983 to
in
1983 to
6%
2000
in
for the
in
2000
The proportion
what we do
5
Transitions, wages,
We now
CDDs
and
for the
it
20-24 cohort, and
^^
utility
look at labor market evolutions for 20-24 year olds, for the period
1983-2000, with the goal of learning something about the effects of
on the labor market. Our approach
First, there
has been
many
is
descriptive,
and
RMI), to the reduction
its limits
CDDs
are obvious:
other institutional changes in the labor mar-
minimum income
ket during that period, from the introduction of a
(the
has
30-34 cohort, and so on. For this reason,
in the next section.
is
of
its
25-29 cohort, from 1.1%
for the
makes good sense to focus on market evolutions
this
age groups, but
in social contributions
floor
on low wage workers, to
a number of other programs aimed at specific groups in the labor market."'^
We
believe however that, for the group
group, the increase in the proportion of
we focus on below, the 20-24 age
CDDs
is
indeed the dominant de-
velopment.
^®We have focused here
education.
One might have expected
of education.
This
is
same across education
be used by
at differences by age group; one can take other cuts, such as
not the case.
levels,
the proportion of
CDDs
to decrease with the level
In 2000, the proportion of
CDDs
was roughly the
probably reflecting the restrictions under which
CDDs
Ccin
firms.
^°For a description of some of the programs aimed at the youth, look for example at
Fougere et
al.
[2000].
Perverse
effects of partial
Second,
much
30
reform
of the evohition of
for the 20-24 year olds or for the
much
unemployment during the period,
population at large, has been due not so
macroeconomic
to institutional changes but to
would have raised a very serious
this
either
factors. Until recently,
From
identification issue:
the early
1980s to the late 1990s, macroeconomic factors had led to a trend increase
in
unemployment, making
it
very
difficult to
disentangle the effects of that
CDDs.
trend from those of the trend increase in
Fortunately (both for
France, and for us), unemployment has started decreasing, so there
hope of disentangling the two. To
see
why and how, we
is
now
start this section
by
looking at aggregate evolutions.
Aggregate evolutions
5.1
The top panel of Figure 8
plots the evolution of the aggregate
rate in France since 1983.
unemployment
The
unemployment
triangles give the evolution of the official
rate (which conforms to the
BIT
definition); the squares give
the evolution of the unemployment rate obtained from Enquetes Emploi. For
our purposes, the relevant series
other series we look at below)
gives a
(in the sense of
is
that from Enquetes Emploi.
more pessimistic assessment
France than the
rate of 11.7%,
general picture
is
That
series
of the evolution of the labor market in
series for the official rate.
unemployment
a series consistent with the
In 2000, the series implies an
compared to an
of a trend increase
official
rate of 9.7%.
The
from 1983 to the mid-1990s, and of a
limited decrease since then.
What
is
relevant to a worker in the labor market
unemployment
he
is
rate per se, but the probabilities of
currently employed, or of becoming employed
ployed.
The
is
not however the
becoming unemployed
if
he
is
currently
if
unem-
evolutions of these two transition probabilities are given in the
two bottom panels of Figure
8.
For each panel, the series with squares gives
o
o
"o
CM
1
CO
00
o>
^"
CM
CD
CD
"cn
CO
05
en
c
o
CD
O)
o>
4-rf
in
oc
05
o
o
.^
^o^
re
E
^_
o
n
05
Q>
^^mS
OS
UJ
LU
CT>
05
C
CO
LU
(0
re
CM
u
CJ)
St
o
o
CD
re
o
CJ5
*•>
CD
C
0)
00
CD
E
>.
o
00
CO
CD
Q.
E
0)
c
00
3
CD
00
CD
D)
O
00
CD
U>
U>
<
00
CD
CO
00
"CD
00
o
3
G)
o 00
o o
CD
r-~-
in
in
o
o in o
o
in
CO
CM
o O CO
o o
o o
O CM
O
o o o O o o o O
in
in
O
^ O
^
CD
'^
00
C3
CO
CD
C3
CM
U5
C)
CD
C)
C)
O ^
O
C) C)
CNJ
CC
00
CM
Perverse
effects of partial
31
reform
panel transitions, the series with triangles gives retrospective transitions.^^
We
draw two main conclusions from these two panels:
•
The 1980s appear
probability from
from the 1990s. In the 1980s, the transition
different
employment
to
unemployment
the transition probability from
By
increased.
contrast, in the 1990s, the
the second decreased:
dimensions.
market
•
unemployment barely
The
first
to
employment actually
transition increased,
labor market clearly
became worse
This worsening surely had a strong
for the
and
increased,
effect
in
and
both
on the labor
20-24 year olds we focus on below.
The panel
transition from
2000 than
in
employment to unemployment was lower
any previous year
in the sample.
The
in
panel transition
from unemployment to employment in 2000 was one of the highest in
the sample. In other words, despite the fact that the
was
rate
still
unemployment
high, labor market prospects were, from the point of
view of an individual in the labor market, arguably the best since
1984.
Thus a comparison
of endpoints
— 1984 with 2000—can help us
separate out the role of cyclical and structural components.
We
shall
us this below.
^^
We discussed
earlier
why 1990
for retrospective transitions.
8:
The reason
from those
is
is
missing for panel transitions, and
Note that 1995
is
why 1999
is
missing
also missing for panel transitions in Figure
that transitions computed from Enquetes Emploi are very different
in other years.
Most
of this
is
due to a program introduced
in that year
which subsidized the reemployment of the older long-term unemployed, leading to a very
different pattern of flows in 1995. Part of
data.
We
it
appears to be due to other problems with the
decided to exclude this year here and in most of the graphs below.
Perverse
effects of partial
32
reform
Transition probabilities for the 20-24 year olds
5.2
Figure 9 gives the evolution of transition probabilities between
ployment,
GDI employment, and unemployment,
1984 to 1998. Each of the nine panels plots two
for
CDD
em-
20-24 year olds, from
series.
The
first,
in black,
give panel transitions; the second in grey gives retrospective transitions.
Transitions for year
March
to
We
•
of year
refer to the
t
change in status from March of year
—
1
t.
draw three main conclusions from
The
t
three
left
panels
show
this figure:
the
transition
from
probabilities
unemployment.""
The
probability of getting a
The
1980s and the 1990s).
in
GDI
decreases in both subperiods (the
probability of getting a
GDD
increases
both subperiods. Both movements are clearly consistent with the
theory.
While the
tion of
effect is theoretically
unemployment was
increased.
The
in the 1980s.
1990s.
ambiguous, we saw that the dura-
likely to decrease as the
scope of
GDDs
probability of remaining unemployed indeed decreases
But there
is
no evidence of a further decrease in the
(Note that the retrospective measure
is
much
higher than the
panel measure, but shows the same evolution). In other words, diuring
the 1990s, the higher likelihood of getting a
did not
come with an
GDD
rather than a
GDI
overall increase in the probability of getting a
job.
•
The
three center panels
show the
transition probabilities from
GDD
employment.
^^The transition probabilities sum to
self
less
than one, as we do not report
employment, internships, military status, student status,
Eind other
trainsitions to
non participation.
o
o
O
^^ o
^J
^^ CM
^^ ^
00
U)
^
^
o
a
(0
T-
Ti
7S
CD
CD
c ^
o
^i o
(f>
Csj
33
Perverse effects of partial reform.
The
probability of moving from a
CDD
The probabiHty
the two subperiods.
to a
GDI
decreases in each of
CDD
of remaining on a
(the
same
or another one) increases throughout the period, nearly doubling in
each of the two subperiods (Recall that the level shifts between 1989
.
which are often large
to 1991,
in
in the figure, reflect largely differences
Note, again, that while panel and retrospective
measurement.)
transitions have rather different levels, their evolution
is
largely similar
over time.
The
probability of becoming unemployed decreases steadily in the
As we look
1980s.
presumably
when the
flects
the higher probability of finding another job
CDD
comes to an end. But, again, there appears to be a
across the
not exhibit
•
at year-to-year transitions, this
The
current
difference
two decades. In the 1990s, the transition probability does
much
of a trend.
three right panels of Figure 9
CDI employment. They
show
transition probabilities from
are less central to our discussion (indeed in
our formal model, these three transition probabilities were
to zero,
re-
by assumption). One evolution
One might have expected
reduced the flows from
is
all
equal
however worth mentioning.
that allowing firms to use
CDI employment. The top
CDDs
would have
panel show that this
has not been the case: The probability of keeping a
CDI
has decreased,
not increased. This suggests that other factors than changes in firing
costs have played a role in determining general trends in separations.
One can
construct similar tables for the other age groups.
tive features are the
age group.
We
for
qualita-
same, but the evolutions are more muted the older the
do not report them
To summarize: The
market
The
here.
transition probabilities give a picture of a labor
20-24 year olds where the probability of getting a
steadily increased, the probability of getting a
CDI
CDD
has
has decreased, and the
Perverse effects of partial reform
34
probability of staying or becoming
In
clear trend.
appears to be a difference across the two decades.
this last dimension, there
The
unemployed shows no
becoming unemployed when on a CDD, or remaining
probabilities of
unemployed, both decrease
in the 1980s,
but show no further trend
in the
1990s.
Expected time to a
5.3
One way
CDI
of summarizing the information from the transition matrices
compute the expected time
to a
CDI
is
to
starting from different labor market
positions.
To compute
these expected times,
we
use, for each year, the estimated
transition matrix obtained using either panel data or retrospective information,
based on eight different states (CDI,
CDD, unemployed,
self
employed,
student, intern, army, other non participation), for 20-24 year olds.
Note
that this computation assumes static expectations in two dimensions. First
it
assumes that future transition probabilities
same
as this year's. Second,
to 24
become
it
for 20-24 year olds will
ignores the fact that, as those currently 20
older, the relevant transition probabilities
relevant for the 25 to 29 year olds,
an overestimation of the
be the
level of
and so
wiU become those
This second bias leads to
on.
expected times to a CDI. But what we
care about here are changes over time,
and
this simple
approach
is
likely to
capture them.
The
from a
evolution of expected times for the 20-24 age group, starting either
CDD
or from
Starting from a
stant in the 1980s.
creases slightly.
unemployment,
CDD,
is
plotted in Figure 10.
the expected time to a
appears roughly con-
Starting from unemployment, the expected time de-
This
is
the result of two offsetting changes:
hand, a decreased probability of getting a
ployment or from a
CDI
CDD,
CDI
On
the one
starting either from
unem-
leading to an increase in the expected time.
On
Figure 10. Expected time to a regular job
8.0
7.2
6.4
5.6
4.8
4.0
3.2
2.4
1^85
8.4
-
7.8
-
7.2
-
6.6
-
6.0
-
5.4
—
4.8
—
4.2
—
3.6
-
"TSSE
"T^sg
Starting
"rasT
1^85"
1^92"
"T^95"
im^
"rag2"
T^gs"
T^gs"
fr(
~W85~
T
I
Perverse
effects of partial
35
reform
the other, an increased probability of getting a
GDI
gether with a higher probability of getting a
starting from unemployment.
CDD
when unemployed,
starting from a
CDD
to-
than
In the 1980s, the two effects roughly cancel
each other.
The
picture
is
different in the 1990s,
where the expected time increases
significantly until the late 1990s, declining partially thereafter.
expected time based on retrospective information
is
While the
higher than the expected
time based on panel data, both series go up during the period. The expected
time from unemployment based on retrospective information increases from
4.8 years in 1990 to 8.2 years in 1996, to decline to 6.5 years in 2000;
panel data counterpart goes from 4.0 to
5.
A
6.0,
down
its
to 4.7 years in 2000.
Wages
4
To do
complete picture requires looking also at wages.
so,
we run a
standard wage regression, regressing for each year, from 1983 to 2000, the
logarithm of the monthly net wage on a set of controls
egories), age (10 categories)
CDD,
if
on a CDI. Thus, we run,
\ogWi
Figure
1 1
=
equal to
for
each year:
Xil3
+ bD + a
plots the time series of estimated
wage equation
GDDs
dummy
and a
for
6's,
—education (15
cat-
the worker
on a
1 if
from estimation of the
each year from 1983 to 1998. Given age and education,
appear to pay about
20%
less
than GDIs. The evidence suggests also
that the gap between the two wages has increased over time, from
1983 to
How
29%
is
in 1993,
and to 22.5%
12%
in
in 2000.
should we interpret this decrease in the relative wage over time? In
our model, partial reform has two effects on the wage of
GDDs
relative to
GDIs: The
GDDs,
leading to
a decline in bonding, in
how low a
first is
a decrease in the bargaining power of
a decline in their wage. The second
is
-2
o
c
o
o
Q
Q
O
c
O
o
0)
0)
3
LL
o
o
CM
O
1^
o
I
o
o
CM
IT)
CM
CM
CD
O
LO
LO
CM
CM
O
O
O
00
Perverse
wage
effects of partial
36
reform
entry-level workers are willing to accept in order to have a chance at a
regular job. In our model also, a decrease in the relative
jobs,
is
is
that
necessarily associated with an increase in
if
V
.
The
the wage goes down, the decline in bonding
of reform on actual and excess turnover
is
wage
for entry-level
intuition for this
is
small, the effect
limited, so the reform increases
welfare.
Thus,
if
the economy conformed to our model, the finding that the wage
has decreased would be prima facie evidence that partial reform has been
welfare improving. There
model and
tivity.
reality: In
This
however one important difference between our
is
our model,
not the case in
is
all
reality,
what has happened over time
is
but to a change
expand on
5.5
data
in the
in the
and there
is
If this is
may be due
nature of
same produc-
a plausible argument that
the extension of the use of
to workers with lower productivity.
wage we observe
entry-level jobs have the
CDDs
the case, the decrease in the
neither to bonding or bargaining,
CDD jobs
or
CDD
workers over time.
We
this point in the conclusion.
Values
In our model, the welfare effects of partial reform are captured by
pens to y", the expected present value of utility
is
to jobs or
if
currently unemployed.
tempting to construct an empirical counterpart and see how
over time. This
cause not
all
what we do
is
average value
V
,
its
salt:
construction of
ertheless,
has evolved
specifically, be-
we construct not F", but the
evolution over time.
results of this exercise
a grain of
More
it
must obviously be interpreted with more than
There are many assumptions and many steps involved
V
It
the average expected present value of utility for a 20-24
year old, and look at
The
in this last subsection.
entrants enter as unemployed,
what hap-
,
we think
all likely
to imply substantial
measurement
this provides a rather transparent
way
of
error.
in the
Nev-
summarizing
Perverse
37
reform
effects of partial
what we have seen about the evohitions
of transition probabihties
and wages
in a single statistic.
To compute
i,
we proceed
V^, the expected present value of utility
currently in state
Let V^ be the expected present value of utility
as follows.
conditional on being in state
putation (GDI,
if
i
We
today.
CDD, unemployed,
consider five states in our com-
intern, self employed). ^^
Let
V
be the
associated vector of utilities associated with the diiferent states. Let
the transition matrix associated with these different states.
Let
vector of wages or wage equivalents associated with each state.
construct
V
w
A
be
be the
Then, we
as:
V=w+
-^AV
+r
1
Or
equivalently.
V = {I--^A) w
l
V
is
then constructed
where the
+r
as:
Pi are the proportions of individuals in state
i,
and sum to
one, and Vi are the elements of V.
We
focus on the 20-24 age group.
^''Note that
we exclude three
states: student,
states were included, our results
This
is
because,
if
wage, the increase
large
army, and out of the labor force.
would be much stronger
the flow utility of being a student
in the
downward trend
at hand,
For A, we use for each year the
namely the
in
is
(i.e
show a
these
larger decline in V.)
assumed to be low
proportion of students would dominate the
If
es-
series,
relative to the
and lead to a
V. This trend however would be largely unrelated to the issue
role of
CDDs.
Perverse
38
reform
effects of partial
timated transition matrix obtained using either panel data or retrospective
information. Just as for the construction of expected times earlier, this com-
putation assumes static expectations in two dimensions,
i.e.
an unchanged
value of the matrix for a given age group over time, and an unchanged
The
transition matrix as individuals in the group get older.
and our
simplicity,
justification
is
belief that, as evolutions are qualitatively similar across
age groups, this should capture the relevant trends.
For w, we normalize the
growth over time).
We
GDI wage
take the
to 1
CDD
we use a value
of 0.5 for the
we ignore general wage
wage to be equal to
1
minus the
Based on unemployment ben-
discount shown in Figure 11 for each year.
efit rules,
(i.e.
wage equivalent when unemployed.
Because the transition probabilities to other states are small, the other
ements of
w
play
little
employment income, a value equal
an annual
The
we assume a value
role in the results;
interest rate of
to the
V
1 for self
CDD wage for internships. We use
12%.
results are presented in the top panel of Figure 12.
gives the series for
of
el-
The
black line
using panel transitions, the grey line gives the series
using retrospective transitions.
The
general impression
a steady worsening
is
one of
little
until the late 1990s,
change
in the 1980s, followed
by
and a partial improvement at the
end. According to this measure, (and leaving aside the general increase in
real
wages over time), the average welfare of the 20-24 year old
lower in 2000 than
it
was
is
either in 1984, or (and this comparison
slightly
is
safer
given the changes in the survey in 1990) than in 1991.^"^
Can we conclude from
The answer
^^
Another
is
this that the effects of
obviously not.
Many
finding, not reported here,
is
CDDs
have been perverse?
other factors have been relevant during
how much
In that sense, the French labor market has
become
closer
Vcdd
is
to
increasingly dual.
Vy than
to Vcdi-
Figure 12. Values
1984-2000
9.00
.Average value 20-24
8.75
8.50
8.25
8.00
7.75
7.50
0.98
TO8B^
'
1989
'
1992
'
1995
'
'
1998
'
'
2001
bailie 20^4A/:aUte^0..5a
1986
'
1989
1992
199b
'
1998
2001
Perverse
39
effects of partial reform.
that period, and attributing
all
the change in
V to the introduction of CDDs
would obviously be wrong. But we can make some progress:
much
Clearly
of the decrease in V, especially in the 1990s,
been due to macroeconomic
portion of
saw
CDDs. But
earlier, in
factors, rather
here, the evidence
than to the increase
from year 2000
V
either 1984, or 1991. In short, the lower value of
is still
V
in
in the pro-
helpful.
is
terms of aggregate transition probabilities, 2000
the best year of the sample. Yet, in that year
must have
is
lower than
As we
arguably
it
was
in
2000 cannot easily be
attributed to macroeconomic factors.
We
can actually go one step further. Some of the changes
to reflect structural changes in the labor market other than
which might
in
V
are likely
CDDs, changes
In that case, attributing the decline in
affect all cohorts.
over the sample to the introduction of
CDDs
would
clearly
suggests looking not at the evolution of the average value
V
be wrong. This
V
for the 20-24
age group, but rather at the evolution of this average value relative to the
average value for the whole labor force
introduction of
With
—which
is
much
less affected
by the
CDDs.
this motivation,
we
plot the evolution of the ratio of the average
value for the 20-24 age group to the average value for the 20-59 age group
in the
bottom panel
of Figure 12
(We use the same wages
for
both groups,
thus not taking into account the age profile of wages in computing the two
values. This
time).
would change the
level,
The graph has two main
but not the evolution, of the ratio over
characteristics.
decline in the relative value from 1984 to 1997.
First,
a nearly continuous
Then an
increase, but to a
lower level than at the start of the sample.
This suggests to us two conclusions.
First,
much
of the evolution of
the relative value for the 20-24 age group reflects aggregate evolutions, the
long worsening and the recent improvement in the labor market:
suffer
more
in
remains lower
a depressed labor market.
in
The young
Second, the fact that the value
2000 suggests that more has been at work. The extension of
Perverse
effects of partial
CDDs, which
reform
40
disproportionately affects that group,
is
a plausible candidate
explanation for this underlying deterioration. Put more conservatively, there
is
no evidence that the introduction and development of
the relative welfare of those most affected by
it,
CDDs has
improved
namely the young.
Conclusions
6
We have looked
On
at the effects of the introduction of fixed-duration contracts.
the theoretical side,
we argued that the
effects of
such partial reform
may
be perverse, leading to higher turnover, and possibly lower welfare:
The
excess turnover induced by the forced coexistence of fixed-duration and
regular contracts can be high enough to offset the efficiency gains of improved
flexibility.
On
the empirical side,
lutions for
young workers
of increased turnover,
we looked
in
at the evolution of labor
FVance since 1983.
and argued
that,
if
We
market evo-
found strong evidence
anything, the effect of the fixed
duration contracts on the welfare of young workers appears to have been
negative.
If
our theoretical and empirical conclusions are valid, this suggests that,
at least from an economic viewpoint
implications), such partial reform
reform,
i.e.
Many
workers.
may be
leaving aside political
economy
a very poor substitute for broader
an across the board reduction in
firing costs for all workers.
questions remain open for future research.
may be how such a
portant
(i.e
To
us, the
most im-
reform affects the nature of the jobs offered to
We have assumed in our model that contracts had no impact on the
nature of the jobs created by firms. There are good theoretical and empirical
reasons to think they may. There are two potential effects at work (which
parallel the
two
effects at
work on
firms' decisions in
our model).
On
the one
hand, lower costs on fixed-duration contracts give more incentives for firms
to take
more
risks,
design jobs which, associated with the right worker, lead
Perverse
effects of partial
to high productivity.
may
On
reform
J,.!
the other, lower costs on fixed-duration contracts
instead induce firms to design routine, low productivity jobs, which
they can
fill
we reviewed
at work.
through the use of fixed-duration contracts. The wage evidence
in
our paper suggests that this second effect might indeed be
Perverse
reform
effects of partial
^2
References
Adam,
P.
Italy
and Canziani,
and Spain,
1998,
P.,
CEP DP
Blanchard, O. and Portugal,
Partial de-regulation: Fixed-term contracts in
386.
Comparing Portuguese and
What
2001,
P.,
U.S.
hides behind an unemployment rate.
unemployment, American Economic Review,
forthcoming.
Caballero, R. and
Hammour,
M., 1996,
The 'fundamental
transformation' in macro-
economics, American Economic Review 86{2), 181-186.
Cahuc,
P.
and Postel-Vinay,
F.,
2000,
Temporary
jobs,
and labor market performance, mimeo Cepremap,
Fougere, D., Kramarz, F., and Magnac, T., 2000,
France,
CEPR DP
Guell-Rotllan,
to
B., 2000,
permanent employment: The Spanish
Lazear, E., 1990,
Paris.
Youth employment
policies in
2394.
M. and Petrongolo,
Lamy, 2000, Lamy
employment protection,
Social, Editions
Lamy,
Workers' transitions from temporary
case.,
CEP DP
438.
Paris.
Job security provisions and employment.
Quarterly Journal of
Economics 105-3, 699-726.
Liberation, 2000, Interviews of workers
Magnac, T. and
Revieiu of
OECD,
1999,
Visser, M., 1999,
Economics and
OECD
Philippon, T., 2000,
on CDDs,
May
2,
Special Emploi.
Transition models with measurement errors,
Statistics 81-3, 466-474.
Employment
Outlook,
OECD.
Memory management, mimeo MIT.
Poulain, G., 1994, Les contrats de travail a duree determinee, Litec; Paris, France.
Saint- Paul, G., 1996,
On
the political economy of labor mai'ket
Macroeconomics Annual
8, 151-196.
flexibility,
NBER
Perverse effects of partial reform.
Saint-Paul, G., 2000,
The
Political
^3
Economy
University Press, Oxford, fortlicoming.
Shapiro, C. and Stiglitz,
J.,
Oxford
..,
1984, Equilibrium
American Economic Review 74. 433-444.
of Labor Market Reforms,
unemployment
as a discipline device,
,,6l»5
(127
Date Due
MIT LIBRARIES
3 9080 02246 2243