MSc Economics

Ec413 Macroeconomics
The Labour Market I

Christopher A Pissarides

Michaelmas Term 2009

Pissarides (MSc Econ) Labour1 2009/10 1 / 22

Walrasian failures

The biggest failure of the RBC model is in the labour market. It

cannot explain the cyclical sensitivity of employment and the lack of
sensitivity of real wages.
It does not have a theory of unemployment, or labour reallocation.
The assumption that workers choose between leisure and work freely,
with smooth adjustments of hours, lies at the heart of the problem.
It does not allow enough exibility in labour supply and yet the
assumptions of neoclassical equilibrium restrict us to be on the labour
supply curve.
To make progress we have to model choices o the labour supply
curve. We are not likely to get far by rening labour supply because
labour supply curves are steep and do not shift much.

Pissarides (MSc Econ) Labour1 2009/10 2 / 22

Indivisible labour

Abandon the assumption of smooth hours adjustments. Individuals

are either employed or unemployed, and when employed they work l0
hours each.
Can justify the assumption by xed costs of getting ready and going
to work; or by indivisible leisure; or by technology
Recall that in the RBC model we simplied the exposition by assuming
constant population, which could be normalized to 1. Therefore Lt , total
number of hours of work, is also the number of hours supplied by each one
of many representative agents.
Now each agent supplies either l0 or 0. So, of the population only a
fraction Nt will be employed, such that Nt l0 = Lt . The underlying
assumption is that if everyone worked there would be too much input, i.e.,
l0 > Lt , and so there will be unemployment in equilibrium.

Pissarides (MSc Econ) Labour1 2009/10 3 / 22

How is work allocated?
When the choice of hours is not smooth but it is between two objects
the problem is non-convex. In other words, we cannot apply the usual
maximization calculus.
However, we can convexify it by assuming random allocation of
work time and consumption insurance. Thus, one selects randomly Nt
workers and allocates l0 hours to each and the 1 Nt workers remain
unemployed. When output is produced each and every individual
consumes the same amount irrespective of unemployment status. The
representative agents choice variable is Nt .
Utility function of representative agent depends on the expected utility
from leisure
u = log Ct + b fNt log(1 l0 ) + (1 Nt ) log 1g (1)
= log C + Nt b log(1 l0 ) (2)
so the utility function is linear in the choice variable with coe cient
b log(1 l0 ). The rest of the model as before.
Pissarides (MSc Econ) Labour1 2009/10 4 / 22
So what?

With linear utility we are indierent about uctuations in

employment, we care only about the mean. So this case gives the
most favourable outcome within the conventional framework
With this utility function the standard deviation of output in model
calibrations rises from 1.30 to 1.73 (data 1.92)
The ratio of the standard deviation of hours to output rises from 0.49
to 0.76 (data 0.96).

Pissarides (MSc Econ) Labour1 2009/10 5 / 22

So far so good...

...but there is more to labour markets than lotteries. Indivisible labour

focuses on an important issue - the xed costs of going to work and the
indivisibility of labour. It is around us and it has important implications for
employment.
Is this a good model of the allocation of work between employment
and unemployment? Better than smooth work-leisure choice, but we
can do even better.
Lotteries imply that those who draw unemployment are better o:
they consume the same as everyone else yet they dont work.
The labour market involves human interchange, its where we spend
most of our awake time. Incentives, motives and information make it
dierent from goods markets and these things matter for equilibrium.

Pissarides (MSc Econ) Labour1 2009/10 6 / 22

Search and matching
Because of the importance of the labour market, there are many models
specically addressed to it. We begin with one that has been applied
extensively to business cycles - search and matching.
Its motivation comes from the study of unemployment. Empirically,
the unemployed are workers (the ILO/OECD denition)
not working for pay
looking for a job
available to take one.
Unemployment in the model is dened in the same way.
It satises indivisible labour - workers are either employed or
unemployed
But when unemployed they search for a job. Lotteries are replaced
by search, which is an economic activity that responds to incentives
(historically, search and matching came before indivisible labour but
logically it follows it)
Pissarides (MSc Econ) Labour1 2009/10 7 / 22
Key assumptions
Heterogeneity of workers and jobs. Abandon the idea that
employment is determined in a centralized market. Because of
heterogeneity and imperfect information, workers want to check out
individual jobs and job owners want to check out potential workers
before agreeing to work together.
Replace the concept of employment with that of a job - a match
between a worker and a rm.
Unemployment is a state with an economic function. The unemployed
are searching for a job, i.e., looking for ways to improve their lot in
the labour market
Unemployment is not a desired state, the unemployed worker is
getting strictly less utility than the employed and is looking to
become employed.
Assume for simplicity that total supply of labour xed at 1 (no
work/leisure choice). So all changes in employment predicted by the
model are over and above the Walrasian changes
Pissarides (MSc Econ) Labour1 2009/10 8 / 22
Heterogeneity

Assume decentralized one-on-one matching, v vacant jobs are looking

to match with u unemployed workers
Because of heterogeneity and imperfect information, some job-worker
matches are better than others
Which matches are made and how long does it take to make them?
Heterogeneity is notoriously di cult to analyse in equilibrium macro
models
Here is how we untie the Gordian knot of heterogeneity: invoke the
aggregate matching function

Pissarides (MSc Econ) Labour1 2009/10 9 / 22

The aggregate matching function

m = m (u, v ) (3)

new job matches = function of inputs into search by

unemployed workers and job vacancies
In its simplest version unemployed workers and job vacancies input 1 unit
each
m (., .) is neoclassical
mu , mv > 0, muu , mvv < 0, m (ku, kv ) = km, k > 0.
m () captures the inuence of heterogeneity on job matching
Average duration of unemployment is u/m : it is the time that an
unemployed worker needs on average to nd a suitable job
Average duration of vacancy is v /m, with similar interpretation
Given u, v higher m reduces duration. We say there is less mismatch
(or heterogeneity). Figure 1
Pissarides (MSc Econ) Labour1 2009/10 10 / 22
 Figure 1

The Aggregate Matching Function

The curves show the combinations of u

v and v that yield a given number of
matches. If each curve represents the
same number of matches in each of three
different economies, economy 3 suffers
from more mismatch then economy 2 and
economy 1 is the most efficient one in
matching workers to jobs

m3
m2
m1

u
Functional forms

For macro analysis we use representative agent assumption, despite

heterogeneity. But representative agent is constrained by the
aggregate matching function.
The matching function stands or falls on two counts
how useful is it analytically?
how successful is it empirically?
Analytically it is a gift (as you will see)
Empirically it has been very successful in many countries
A good t is the Cobb-Douglas

m = m0 u v 1
' 0.5 (4)

For calibration choose m0 from data on duration of unemployment

and average v /u ratio. Let v /u , the tightness of the market

Pissarides (MSc Econ) Labour1 2009/10 11 / 22

Transition rates

Use matching function to get transition rates for unemployed workers and
vacant jobs
Rate at which jobs get lled up
m u
= m ( , 1) q ( ) (5)
v v
q 0 ( ) < 0 and elasticity of q ( ) is < 0.
Rate at which unemployed workers nd jobs
m vm
= = q ( ), elasticity 1 >0 (6)
u uv

Pissarides (MSc Econ) Labour1 2009/10 12 / 22

Turnover

We are looking for equilibrium in three pools: employment,

unemployment and job vacancies (the endogenous variables)
Get the equilibrium from equality of entry and exit for each pool
q ( ) = rate of exit from vacancy pool, entry into employment pool
q ( ) = rate of exit from unemployment pool, entry into employment
pool
Job specic (or economy-wide, in more general models) shocks give
rise to the need to reallocate workers between activities
They hit each job-worker pair at rate s. When a pair is hit worker
leaves employment and joins unemployment. Job is destroyed
Entry of new jobs governed by prot maximization (gure 2)

Pissarides (MSc Econ) Labour1 2009/10 13 / 22

 Figure 2
Labour Flows in search equilibrium

employment
(1-u) Job
exit

q() s q()

unemployment vacancies
u v

New
entry
Evolution of unemployment

ut = unemployment
1 ut = employment (labour force xed at 1)
Statistical identity

ut +1 ut inows into u at t outows from u at t (7)

We build a theory of the labour market by modelling the inows and

outows.
In continuous time
u = s (1 u ) m (u, v ). (8)

The stationary locus u = 0 is the Beveridge curve, plotted in v , u

space.
It slopes down and it is stable (gure 3)

Pissarides (MSc Econ) Labour1 2009/10 14 / 22

 Figure 3

The Beveridge curve

Beveridge
curve

u
Empirical dynamics
Unemployment is the state of the dynamic system
Re-write in terms of transition rates
u = s (1 u) q ( )u (9)
For given s and this is a stable equation in u with stationary solution
s
u= . (10)
s + q ( )
Empirically, ows are large. With quarterly or annual data assume for
simplicity that (10) holds at all times. Its a good empirical
approximation
So changes in unemployment are driven by changes in s or .
Assume s is exogenous. is endogenous and driven by rmsprot
maximizing job creation
So job creation is the only variable input into search and the only
control that drives the dynamics of employment over the cycle (unless
we assume cyclical separations)
Pissarides (MSc Econ) Labour1 2009/10 15 / 22
Job creation

A job is an asset owned by the rm and its value is determined by

arbitrage equations. Ignore for the moment capital and write p for the
net output from a job
Let c be the cost of maintaining a vacant job (advertising, worker
screening, job maintenance)
Derive value equations. Interpret them as arbitrage conditions for the
valuation of an asset under perfect capital markets.
V = value of vacant job
J = value of occupied job

rV = c + q ( )(J V) (11)

rJ = p w sJ. (12)

Pissarides (MSc Econ) Labour1 2009/10 16 / 22

Freedom to create and destroy jobs imply exhaustion of rents from job
creation
c
V =0,J= (13)
q ( )
The job creation condition is derived from the above equations. Substitute
(12) into (13).
(r + s )c
p w = 0. (14)
q ( )

The job creation condition as a generalization of the demand for

labour
The job creation curve in w space slopes down
The cyclical variable is p. Higher p shifts the job creation curve up
(higher wages, more jobs per unemployed worker)
But the wage is endogenous, we need another equation for wages

Pissarides (MSc Econ) Labour1 2009/10 17 / 22

Wages

All job matches enjoy some local monopoly rents because of frictions.
If a rm and a worker who are together break up, their joint returns
will be strictly less than what they have now together
Wages split the rents that the pair have. Each gets the value of their
best alternative plus a share of the rent (the rent is 0 in Walrasian
markets)
So: a worker will get the value of unemployment plus her share, the
rm the value of the vacancy plus the remaining share

Pissarides (MSc Econ) Labour1 2009/10 18 / 22

The workers returns

To get the workers expected returns from employment and unemployment,

think of human capital as an asset valued by arbitrage conditions. Let
U = value of unemployed worker
W = value of employed worker
Arbitrage equations
rU = z + q ( )(W U ). (15)
rW = w + s (U W ). (16)
z = imputed and actual income during unemployment

Pissarides (MSc Econ) Labour1 2009/10 19 / 22

The Nash wage equation

J = rms reward from agreement

V = rms payo when there is no agreement
W = workers reward from agreement
U = workers payo when there is no agreement
Total rents
S = J +W U V (17)
Wage rate shares the rents, fraction goes to labour

W U = S. (18)

We can obtain the same outcome if we choose the wage rate to maximize
the Nash product
(W U ) (J V )1 (19)
So (18) is known as the Nash wage equation.

Pissarides (MSc Econ) Labour1 2009/10 20 / 22

The wage curve

Substitute the asset values from the Bellman equations into (18) and
making use of the job creation condition V = 0, you arrive at:

w = z + (p z + c ). (20)

Wages compensate for the loss of z, they then reward the worker with
a fraction of rent p z and a fraction of the average saving of
recruitment costs cv /u made when the worker agrees to a contract.
Higher z, p, raise w for obvious reasons
Higher raises wages because the workers expected returns from
search are higher and the rmslower
Equation (20) is the wage curve and in w space it slopes up.

Pissarides (MSc Econ) Labour1 2009/10 21 / 22

 Figure 4

Equilibrium wages and market tightness

w
Wage curve

Job creation

Equilibrium

Equations (14) and (20) are uniquely solved for w and (gure 4)
With knowledge of (10) gives unemployment.
Equilibrium is saddlepath-stable and the = 0 line is the saddlepath
(gure 5).
Structural factors shift the Beveridge curve; Demand and supply
factors shift the job creation curve. (In more complicated models this
clear decomposition may not hold.)
The cyclical variable is p. How does it inuence the three endogenous
variables, w , , u? This is the topic of the next set of notes.

Pissarides (MSc Econ) Labour1 2009/10 22 / 22

 Figure 5

Equilibrium vacancies and unemployment

v
Job creation

Beveridge
curve