1 Matching model for dynare example

1.1 Households
A household consists of a continuum of households, each of which is either
employed or unemployed. Employed members get a wage rate wt . Unemployed
members are searching for a new job. For simplicity there are no unemployment
bene…ts. There is no disutility of work or search and all unemployed workers
accept a job if they get o¤ered one and employment is determined by the demand
for labor by …rms. Households own the capital stock, which they rent out to
…rms.
The maximization problem of the household is then given by
c1t 1
v(kt 1; n
~t 1 ; zt ) = max + Et [v(kt ; n
~ t ; zt+1 ]
kt ;ct 1
s.t. ct + kt = n
~ t 1 wt + rt kt 1 + (1 )kt 1 + ft
Here ft stands for the pro…ts the household receives from the …rms. The other
symbols have their typical interpretation and are summarized in the table below.
Note that the Dynare timing convention is used. For example, n ~ t 1 , is the
number of workers that has a job at the beginning of period t. Households
take employment as given, since they are willing to supply whatever labor is
demanded by …rms. The tilde above a variable indicates it is chosen by the
…rm.

1.2 Firms
The production of the representative …rm is given by
y~t = k~t n
~ 1t 1 :
Capital is rented on a spot market. This means that for the individual …rm
capital is not predetermined. In equilibrium, we will get that the demand by
the representative …rm, k~t , is equal to the amount available to the household at
the beginning of period t, i.e., kt 1 .1
A …xed fraction, x , of employment separates each period. Firms can …nd
new workers by posting vacancies at cost . The representative …rm takes the
probability of …nding a worker, f;t , as given. The …rm’s maximization problem
is given by2
1
v~(~
nt 1 ; z t ; kt 1 ) = max ct kt 1 nt 1 wnt 1 rt kt 1 vt + Et [w(nt ; zt+1 )]
n ~t ;vt
~ t ;k
f
s.t. nt = (1 x )nt 1 + t vt
1 Note that the individual …rm does not treat the capital stock as a predetermined variable,

since it simply rents whatever it wants at the current spot price. Of course, the rental rate
will be such that the …rm will ask exactly what is available at the beginning of the period.
2 Note that k
t 1 is an argument of the …rm’s value function, because it a¤ects the market
prices and in particular the rental rate of capital. Since it is a state variable that the …rm
does not choose itself it is on the righthand side of the semicolon.

1
 Note that the …rm evaluates pro…ts using the marginal utility of the house-
hold (its owner) which it takes as given. The idea is that the …rm is small so
that it’s pro…ts are small and can be evaluated using the marginal utility of the
household.
An alternative way to write the …rm’s problem is as follows:
" #
c
nt 1 ; zt ; kt 1 ) = max k~t n
v~(~ ~ t 1 rt k~t
~ 1t 1 wt n vt + Et t+1 w(~ nt ; zt+1 ; kt )
n ~t ;vt
~ t ;k ct
s.t. n
~t = (1 x )nt 1 + f;t v
~t

1.3 Matching and equilibrium

Matches are equal to
1
0 ut v~t ;
1 1

where
ut = 1 nt 1

The matching probability for the …rm is, thus, given by

f ut 1

t = 0
vt

1.4 Wages
Because of the matching friction, there are positive pro…ts made by a …rm that
is matched with a worker. The question is how the revenues are divided between
the worker and the owner of the …rm. One possibility is Nash Bargaining. Here
we simply assume that wages are given by

! 1 (1 )zt k~t n
~t 1
wt = (1 !0 )
(1 ! 1 ) (1 )k n

where a bar on top of a variable indicates steady state value. If ! 1 is equal to

0, then wages are fully sticky. If ! 1 is equal to 1, then wages are proportionaly
to the marginal product of labor.
The steady state value of the wage rate is (for any value of ! 1 ) equal to

(1 ! 0 )(1 )k n1 :

In the standard RBC model, steady state wages are equal to (1 )k n . In

this model total revenues have to be divided between those that work, those
that rent out capital, and those that post vacancies, i.e., owners.

2
1.5 De…nition of variables
n
~t end-of-period t employment
kt end-of-period t capital
pt marginal pro…t of one extra worker (mp-wage)
gt value of match
f;t matching probability for the …rm
ct consumption
ut unemployed workers in period t
vt vacancies posted in t
rt rental rate of capital
yt output
zt productivity
et innovation of capital

1.6 Equations of the model

ut = 1 nt 1
nt = (1 x )nt 1+ w;t ut
1
ut 1

w;t = 0
vt
ut 1

= 0 gt
vt
" #
ct+1
gt = Et (pt+1 + (1 x )gt+1 )
ct
kt 1
pt = (1 )zt wt
nt 1

! 1 (1 )zt k~t n
~t 1
wt = (1 !0 )
(1 ! 1 ) (1 )k n
1
kt 1
rt = zt
nt 1

ct = Et ct+1 (rt+1 + 1 )
yt = ct + kt (1 )kt 1 + vt
yt = zt kt 1 n1t 1
ln(zt ) = ln(zt 1 ) + et

3
1.7 Parameter values and steady state
Most structural parameters are simply given a "typical" value. The values are
given by

= 0:99; = 1
= 0:33; = 0:025
= 0:95; = 0:01
0 = 1; 1 = 0:5

The value of ! 0 and ! 1 are key parameters and you are asked to investigate
how the properties of the model depend on them. The benchmark values are
given by
! 0 = 0:5; ! 1 = 1
The two remaining parameters are the vacancy cost, , and the exogenous de-
struction rate. It is di¢ cult to …gure out what the right value for is, especially
since posting a vacancy should be thought of as more than simply the cost of
posting an add in the newspaper. So instead we will choose the value of such
that it has sensible steady state implications. We do the same for the choice
of x . That is, instead of giving the values of and x , we give two pieces of
information with which the values of these two parameters can be solved. The
two pieces of information are that the steady state value of the matching prob-
ability for the worker should be equal to 0.7 and that the steady state value of
the unemployment rate should be equal to 5%. Thus,

f = 0:7
u = 0:05

In the parameter section of the Dynare program, and x are still parameters.
But the steady state values f and u are parameters too. The parameter section
contains the equations that determine the values of and x as a function of f
and u. To keep the program readable, I have introduced a bunch of intermediate
steps. At these intermediate steps I calculate other steady state values, which
are declared as parameters as well. These steady state values are not used in
the model section, but they come in quite handy when giving initial values for
the steady state in the "initval" section of the program.

1.8 Comments
Variables known at the beginning of t must be dated t 1
To get the dynare equations you only have to delete the conditional ex-
pectation. Dynare "knows" that there must be an expectational operator
(and prediction error) in these equations because they have "t + 1" terms

4
2 Exercises
Main exercise

1. Write down the model section. This is the only part of the program that
you have to write yourself.
2. The options of the main Dynare command are such that the program
calculates for employment and output the IRFs and standard deviations
3. Investigate the role of ! 0 and ! 1 for the volatility of employment relative
to the volatility of output. Try to understand when this type of model
can generate volatile employment?

Additional exercises

1. The program is written in levels. Rewrite the program in loglevels.

2. Program the calculation of IRFs and moments yourself. The analysis
above is based on un…ltered data, but check how the results change if you
HP-…lter the data