Business cycles and labor market search (David Andolfato)

Louis and Monica

December 3, 2015

1 The model
The model nest a labor-market-search framework into an standard RBC model. He assume that there are households
distributed uniformly on the unit interval and that they have preferences for consumption and leisure. In the model,
households face the standard consumption-saving problem, but face altogether different opportunities for exchanging
labor services. Individuals either have a job opportunity or not, and job opportunities come and go at random,
depending to some extent on individual search effort, the availability of jobs, and plain luck. Firms also face
a standard wealth maximization problem, except that, because finding new workers takes time and effort, firms
view their existing workforce as a capital asset. Given constant returns in the production technology, it may be
assumed without loss that each firm comprises a single job; in what follows, the terms firm and job will be used
interchangeably.

1.1 The search process

In order to produce output, each job requires a worker. Let nt , denote the number of jobs that are matched with
a worker at the beginning of period t. Job-worker pairs are assumed to separate at the exogenous rate 0 < σ < 1.
Replenishing this stock takes time and consumes resources.
A firm interested in filling an available job must undertake recruiting and screening activities, which are necessary
for finding a suitable employee. Let vt , denote the number of jobs vacancies during period t, each of which incurs a
flow cost equal to κ > 0, measured in units of physical output. In the version of the model studied here, workers are
assumed to search passively. Letting e denote search effort per worker seeking employment, aggregate search effort
by workers is given by (1 − nt )e. The rate at which new job matches form is governed by an aggregate matching
technology, M (v, (1 − n)e), so that employment evolves according to the following dynamic equation:

nt+1 = (1 − σ)nt + M (vt , (1 − nt )e) (1)

The matching technology is assumed to be a non-decreasing, concave function of aggregate search and recruiting
effort and is assumed to display constant returns to scale.

1.2 The Social Welfare Problem

The representative household has preferences represented by a utility function of the following form:
∞
X
E0 β t [U (ct ) + φ(t)H(1 − xt )] (2)
t=0

where ct denotes consumption, xt , denotes the fraction of time spent in nonleisure activities and 0 < β < 1 is a
discount factor. The functions U and H are increasing and concave. The value of the parameter φ(t) > 0 depends
on a household’s employment status: φ(t) is equal to φ1 if the household is employed and is equal to φ2 if the
household is unemployed.
Output is produced according to a standard neoclassical production technology, yt = F (k, nt lt ; zt ), where kt
is the aggregate capital input; lt is average hours worked by those employed; and zt is a parameter reflecting the
current state of technology, which evolves stochastically according to the transition function G(z 0 , z) = P r[zt+1 ≤
z 0 | zt = z]. The capital stock depreciates at rate 0 < δ < 1, so that the economy-wide resource constraint is then
given by:

1
 ct + kt+1 + κvt = yt + (1 − δ)kt (3)
s ≡ (k, n , z) denote the vector of the state variables of the economic system, EG denote the expectation
operator associated with the transition function G, W (s0 ) denote the maximum value of (2), given an arbitrary
initial condition (s0 ), which is obtained by solving the welfare problem stated above. The primed variables denote
’next period values’.

1.3 Equations of the model to log-linearize

Uc (c) = βEG [Fk (k 0 , n0 l0 ; z 0 ) + (1 − δ)] Uc (c0 ) (4)

φ1 Hl (1 − l) = Fl (k, nl; z)Uc (c) (5)

κvUc (c) = µαM (v, (1 − n)e) (6)

µ = βEG {φ1 H(1 − l0 ) − φ2 H(1 − e) + Uc (c0 )Fn0 (k 0 , n0 l0 ; z 0 )l0 + µ0 [1 − σ − (1 − α)p(v 0 , n0 )]} (7)

c + k 0 + κv = F (k, nl; z) + (1 − δ)k (8)

n0 = (1 − σ)n + M (v, (1 − n)e) (9)

 
[φ2 H(1 − e) − φ1 H(1 − l)] + p(v, n)(1 − α)µ 1
w = (1 − α)F2 (k, nl; z) + α (10)
Uc (c) l
The functional forms for U , H, F , M and G are as follows:
• U (c) = log(c)
(1−x)(1−η)
• H(1 − x) = (1−η)

• F (k, nl; z) = exp(z)ζk θ (nl)(1−θ)


• M (v, (1 − n)e) = min v, 1 − n, χv α ((1 − n)e)(1−α)
χv α ((1−n)e)(1−α)
• p(v, n) = 1−n

Additionally, the productivity shock is assumed to be governed by the following stochastic process:
• z 0 = ρz + ε̃
where 0 < ρ < 1 and ε̃ is an independently and identically distributed random variable. In particular, assume that
ε̃ ∈ {−ε, ε},ε > 0, and prob(ε) = prob(−ε)= 12 .
Then equations (4) to (10) take the form:
h 0 0
i
• (4’) 1c = βEG θexp(z 0 )ζ( nk0l )(1−θ) + (1 − δ) c10
−η
• (5’) φ1 (1 − l) = (1 − θ)exp(z)ζk θ (nl)−θ 1c

• (6’) κv 1c = µαχv α ((1 − n)e)(1−α)

 (1−η) i
(1−l0 ) 0α 0
(1−η)
h
)e)(1−α)
• (7’) µ = βEG φ1 (1−η) − φ2 (1−e) (1−η) + 1
c0 (1 − θ)exp(z)ζk 0θ (n0 l0 )−θ l0 + µ0 1 − σ − (1 − α) χv ((1−n
1−n 0

• (8’) c + k 0 + κv = exp(z)ζk θ (nl)(1−θ) + (1 − δ)k

• (9’) n0 = (1 − σ)n + χv α ((1 − n)e)(1−α)
h (1−η) (1−η)
 α
i
((1−n)e)(1−α)
• (10’) w = (1 − α)(1 − θ)exp(z)ζkθ (nl)−θ + α φ2 (1−e)
(1−η)
− φ1 (1−l)
(1−η)
+ (1 − α) χv 1−n
µ c
l

2
1.4 Equations at the steady state
• (4”) 1 = β θζ( n∗l∗
 (1−θ)

k∗ ) + (1 − δ)
−η
• (5”) φ1 (1 − l∗) = (1 − θ)ζ(k∗)θ (n ∗ l∗)−θ c∗
1

κv∗
• (6”) = µ ∗ αχ(v∗)α ((1 − n∗)e)(1−α)
c∗
n (1−η) (1−η)
h α (1−α)
io
• (7”) µ∗ = β φ1 (1−l∗)
(1−η) − φ2 (1−e)
(1−η) + (1 − θ)ζ(k∗)θ (n ∗ l∗)−θ c∗
l∗
+ µ ∗ 1 − σ − (1 − α) χ(v∗) ((1−n∗)e)
1−n∗

• (8”) c ∗ +δk ∗ +κv∗ = ζ(k∗)θ (n ∗ l∗)(1−θ)

• (9”) σn∗ = χ(v∗)α ((1 − n∗)e)(1−α)
h i
1−(1−σ)β
• (10”) w ∗ n ∗ l∗ = (1 − θ) − κv∗
σ β

1.5 Log-linearization around the steady state

∂Xt
we denote X∗ =X
et , and X the value of X at the steady state while Xt is the value of X at time t.

1.5.1 Equation (4)’

From (4’) and (10”) we have :
(  (1−θ) )
1 nt+1 lt+1 1
= βEt θexp(zt+1 )ζ + (1 − δ)
ct kt+1 ct+1

  
c˜t nl nl −θ ln ∂nt+1 nl ln ∂lt+1 nl −∂kt+1 1
− = βEt θζ( )(1−θ) zg
t+1 + (1 − θ)θζ( ) + (1 − θ)θζ( )−θ + (1 − θ)θζ( )−θ nl 2
c k k k n k k l k k c
(  (1−θ) ! )
nl ∂ct+1
+ βEt ζθ + 1 − δ (− 2 )
k c

(  (1−θ)  )  (1−θ) !
nl  nl
−cet = βEt ζ ∗θ∗ (1 − θ)(lg t+1 − kt+1 ) + zg
t+1 + ng
g t+1 −β ζ ∗θ∗ + 1 − δ Et cg
t+1
k k
| {z }
=1
(  (1−θ)  )
nl 
t+1 − βEt
cet = Et cg ζ ∗θ∗ (1 − θ)(lg t+1 − kt+1 ) + zg
t+1 + ng t+1
g
k

1.5.2 Equation (5)’

−η 1
φ1 (1 − l) = (1 − θ)exp(z)ζk θ (nl)−θ
c

−η−1 −∂lt l 1 1 1
− ηφ1 (1 − l) ( ) = (1 − θ)ζk θ (nl)−θ zet + θ(1 − θ)ζk θ−1 (nl)−θ ∂kt − θ(1 − θ)ζk θ (nl)−θ (∂nt n + ∂lt n)
l c c c
∂ct
− (1 − θ)ζk θ (nl)−θ 2
c

−η−1 e 1h   i
ηφ1 (1 − l) lt l = (1 − θ)ζk θ (nl)−θ zet + θ e kt − n
et − e
lt − e
ct
c
lη e h  i
lt = ect − zet + θ net + elt − e
kt
1−l

3
1.5.3 Equation (6)’

1
κv = µαχv α ((1 − n)e)(1−α)
c
1 ∂nt ne
κv (e ct ) = αµαχv α ((1 − n)e)−α (e
vt − e µt + vet ) + (1 − α)µαχv α ((1 − n)e)−α (− )
c   n
1 ne
κv (e ct ) = αµαχv α ((1 − n)e)1−α µ
vt − e et + αe vt − (1 − α) n
et
c (1 − n)e
n
(1 − α)e
vt − e
ct = µet − (1 − α) n
et
(1 − n)

1.5.4 Equation (7)’

( )
(1−η) (1−η)
(1 − l0 ) χv 0α ((1 − n0 )e)(1−α)

(1 − e) 1 0 0θ 0 0 −θ 0 0
µ = βEG φ1 − φ2 + 0 (1 − θ)exp(z )ζk (n l ) l + µ 1 − σ − (1 − α)
(1 − η) (1 − η) c 1 − n0

(
(−η)
(1 − l) 1 h   i
µ
et µ = β −(1 − η)φ1 lt+1 l + (1 − θ)ζk θ (nl)−θ lEt zet+1 + θ e
e kt+1 − n
et+1 − e
lt+1 − e ct+1
(1 − η) c
χvα((1 − n)e)(1−α) χvα((1 − n)e)(1−α)
 
et+1 1 − σ − (1 − α)
+ Et µ − µ(1 − α) (αEt vet+1 )
1−n 1−n
−α−1
−µ(−α)(1 − α)χv α e1−α (−ng

t+1 n) ∗ (1 − n)

   
l 
˜ ˜ σn
µ
et µ = β φ1 ˜ + θ(kt+1 − nt+1
Et zt+1 ˜ − lt+1 ) − ct+1
˜ + 1 − σ − (1 − α) Et µt+1
˜
(1 − l)eta 1−n

σnα(1 − α) nt+1
˜ n
− Et vt+1
˜ + )}
1−n 1−n

1.5.5 Equation (8)’

c + k 0 + κv = exp(z)ζk θ (nl)(1−θ) + (1 − δ)k

 
ce
ct + k e
kt+1 + κvevt = ζk θ (nl)(1−θ) zet + θe
kt + (1 − θ)(e lt ) + (1 − δ)ke
nt + e kt
 
ce kt+1 + κve
ct + k e vt = ζk θ (nl)(1−θ) zet + θe
kt + (1 − θ)(e lt ) + (1 − δ)ke
nt + e kt

1.5.6 Equation (9)’

n0 = (1 − σ)n + χv α ((1 − n)e)(1−α)

 
n
nt+1 = (1 − σ)ne
ne nt + σn αe vt − (1 − α) net
(1 − n)
 
n
et+1 = (1 − σ)e
n nt + σ αe vt − (1 − α) n
et
(1 − n)
(1 − n) − σ(1 − αn)
n
et+1 = σαevt + n
et
(1 − n)

4
1.5.7 Equation (10)’

" ! #
(1−η) (1−η)
θ −θ (1 − e) (1 − l) χv α ((1 − n)e)(1−α) c
w = (1 − α)(1 − θ)exp(z)ζk (nl) +α φ2 − φ1 + (1 − α) µ
(1 − η) (1 − η) 1−n l

h  i
et = (1 − α)(1 − θ)ζk θ (nl)−θ zet + θ e
ww kt − n
et − e
lt
χv α ((1 − n)e)(1−α)
 
−η e c
+ α φ1 (1 − l) lt l + (1 − α) µ [µe
µt + αe
vt ]
1−n l
" ! #
(1−η) (1−η)
(1 − e) (1 − l) χv ((1 − n)e)(1−α)
α
c 
+α φ2 − φ1 + (1 − α) µ ct − e
lt
(1 − η) (1 − η) 1−n
e
l

h  i
−η
et = (1 − α)φ1 (1 − l)
ww c zet + θ ekt − n
et − e
lt
 
−η e (1 − α)σn c
+ α φ1 (1 − l) lt l + µ [e
µt + αe
vt ]
(1 − n) l
    
1 −η c

+ α µ (1 − σ) − + φ1 (1 − l) ect − e
lt
β l

h  i
−η
et = (1 − α)φ1 (1 − l)
ww c zet + θ ekt − n
et − e
lt
 
−η e (1 − α)σn c
+ α φ1 (1 − l) lt l + µ [e
µt + αe
vt ]
(1 − n) l
    
1 −η c

+ α µ (1 − σ) − + φ1 (1 − l) ect − e
lt
β l

1.6 Matrix Form

The model has eight variables including three state variables (k, n, z). We start by the three static equations (5),(6)
and (10). We write them in matrix form, with M1 & M2 3x5 matrixes.
 
cet  
 let  ket
 
M1  v
 t
e  = M 2
 net  (11)
 µet  zet
w
ft

We have five dynamic equations, we write them in matrix form, with M3I , M3L 5x3 matrices, M4L , M4I 5x5 matrices
and M5 a 5x6 :
 
    ek t+1
cgt+1 cet en 
     t+1 
kt+1
g kt
e  lt+1 
g   let   ez 
I L I
 L
    t+1 
M3 ng t+1
 + M3 net  = M4  vg
 t+1  + M4  vet  + M5 
   (12)
 ec t+1 

zgt+1 zet  µg t+1
  µet  e 
 l t+1 
wg t+1 w
ft
ev t+1

5
Annex
Wage equation at the steady state
In this section, we prove that the wage equation at the steady state equals (10”).

" ! #
(1−η) (1−η)
θ −θ (1 − e) (1 − l∗) χ(v∗)α ((1 − n∗)e)(1−α) c∗
w∗ = (1 − α)(1 − θ)ζ(k∗) (n ∗ l∗) +α φ2 − φ1 − (1 − α) µ∗
(1 − η) (1 − η) 1−n l∗
" ! #
(1−η) (1−η)
θ −θ (1 − e) (1 − l∗) χ(v∗)α ((1 − n∗)e)(1−α) c∗
w∗ = (1 − α)(1 − θ)ζ(k∗) (n ∗ l∗) +α φ2 − φ1 − (1 − α) µ∗
(1 − η) (1 − η) 1 − n∗ l∗

From (7”) we have

(1−η) (1−η)
µ∗ (1 − l∗) (1 − e)
− µ ∗ (1 − σ) = φ1 − φ2
β (1 − η) (1 − η)
χ(v∗)α ((1 − n∗)e)(1−α)
 
1 θ −θ
+ (1 − θ)ζ(k∗) (n ∗ l∗) l ∗ +µ ∗ −(1 − α)
c∗ 1 − n∗

(1−η) (1−η)
µ∗ 1 (1 − e) (1 − l∗)
µ ∗ (1 − σ) − + (1 − θ)ζ(k∗)θ (n ∗ l∗)−θ l∗ = φ2 − φ1
β c∗ (1 − η) (1 − η)
χ(v∗)α ((1 − n∗)e)(1−α)
 
− µ ∗ (1 − α)
1 − n∗

  (1−η) (1−η)
1 1 (1 − e) (1 − l∗)
µ ∗ (1 − σ) − + (1 − θ)ζ(k∗)θ (n ∗ l∗)−θ l∗ = φ2 − φ1
β c∗ (1 − η) (1 − η)
χ(v∗)α ((1 − n∗)e)(1−α)
 
− µ ∗ (1 − α)
1 − n∗
Replacing in the previous eq. we have

   
θ −θ 1 1 θ −θ c∗
w∗ = (1 − α)(1 − θ)ζ(k∗) (n ∗ l∗) + α µ ∗ (1 − σ) − + (1 − θ)ζ(k∗) (n ∗ l∗) l∗
β c∗ l∗
  
1 c∗
w∗ = (1 − α)(1 − θ)ζ(k∗)θ (n ∗ l∗)−θ + α(1 − θ)ζ(k∗)θ (n ∗ l∗)−θ + α µ ∗ (1 − σ) −
β l∗
  
(1 − σ) β − 1 c∗
w∗ = (1 − θ)ζ(k∗)θ (n ∗ l∗)−θ + α µ ∗
β l∗
  
(1 − σ) β − 1
w ∗ n ∗ l∗ = (1 − θ)ζ(k∗)θ (n ∗ l∗)1−θ + α µ ∗ c ∗ n∗
β
From :
κv∗
= µ ∗ αχ(v∗)α ((1 − n∗)e)(1−α)
c∗
We have :
 
θ 1−θ κv∗ (1 − σ) β − 1
w ∗ n ∗ l∗ = (1 − θ)ζ(k∗) (n ∗ l∗) +   c ∗ n∗
c ∗ χ(v∗) ((1 − n∗)e)(1−α) )
α β
 
θ 1−θ κv∗ (1 − σ) β − 1
w ∗ n ∗ l∗ = (1 − θ)ζ(k∗) (n ∗ l∗) + n∗
[σn∗)] β
 
κv∗ 1 − (1 − σ) β
w ∗ n ∗ l∗ = (1 − θ)ζ(k∗)θ (n ∗ l∗)1−θ −
σ β

6
 With the technology parameter ζ chosen to normalize the steady-state level of output to unity, labor share can
be computed as :
 
κv∗ 1 − (1 − σ) β
w ∗ n ∗ l∗ = (1 − θ) −
σ β