Macroeconomics II: Search and Matching

Part 2

Luiz Brotherhood

Universitat de Barcelona

1 / 18
Idiosyncratic shocks

• Production of a match can assume different values:

y ∈ Y ≡ {y1 , y2 , . . . , yN }, where 0 < y1 < y2 < · · · < yN .
• A match that produces y today faces probability T (y , y 0 ) of
transitioning to the state where production tomorrow is y 0 .
• You can think of T as a N by N matrix.
• In the first period in which a firm and a worker meet, they face a
probability Q(y ) of starting as a match that produces y .
• No more exogenous destruction: s = 0 (it will be endogenous now).
• w (y ): wage depends on the production of match.

2 / 18
Idiosyncratic shocks
• Value functions:
 

T (y , y 0 ) max E (y 0 ), U 
X 
E (y ) = w (y ) + β 
y 0 ∈Y

   
 
Q(y 0 ) max E (y 0 ), U  + (1 − p(θ))U
X 
U =b+β p(θ) 
y 0 ∈Y
 

   
 
Q(y 0 ) max J(y 0 ), V  + (1 − q(θ))V
X 
V = −c + β q(θ) 
y 0 ∈Y
 

 

T (y , y 0 ) max J(y 0 ), V 
X 
J(y ) = y − w (y ) + β 
y 0 ∈Y

3 / 18
Idiosyncratic shocks

• Free-entry: V = 0 implies
c
Q(y 0 ) max J(y 0 ), 0 =
X 
.
y 0 ∈Y
βq(θ)

• Nash bargaining for each state:

E (y ) − U = φS(y ) and J(y ) = (1 − φ)S(y )

S(y ) = E (y ) − U + J(y )
• There is agreement between worker and firm with respect to beginning
and terminating a match (intuition?):

J(y ) > 0 ⇐⇒ (1 − φ)S(y ) > 0 ⇐⇒ φS(y ) > 0 ⇐⇒ E (y ) − U > 0

4 / 18
Idiosyncratic shocks
Aggregate distribution of agents across states

• In each period, there is an amount of unemployed workers, u, and an

amount of workers in each productivity level, µ(y ).
• Law of motion of distribution of agents in steady state:

 
X
ut+1 = ut [1 − p(θ)] + ut p(θ)  I {S(y ) ≤ 0} Q(y )
y ∈Y

I {S(y ) > 0} I S(y 0 ) ≤ 0 µt (y )T (y , y 0 )

X X 
+
y ∈Y y 0 ∈Y

(
0 if S(y ) ≤ 0
µt+1 (y ) =
ut p(θ)Q(y ) + y 0 ∈Y T (y 0 , y )µt (y 0 )
P
otherwise

5 / 18
Idiosyncratic shocks
Solving the model

• We need to characterize S(y ).

• We start with E (y ) − U:

 
 X
T (y , y 0 ) max E (y 0 ) − U, 0 

E (y ) − U = w (y ) − b + β 
 0
y ∈Y
 

Q(y 0 ) max E (y 0 ) − U, 0 
X 
−p(θ) 
0

y ∈Y

• J(y ) is given by:

 

T (y , y 0 ) max J(y 0 ), 0 
X 
J(y ) = y − w (y ) + β 
y 0 ∈Y

6 / 18
Idiosyncratic shocks
• Using E (y ) − U = φS(y ) and J(y ) = (1 − φ)S(y ),
 
 X
T (y , y 0 ) max S(y 0 ), 0 

S(y ) = y − b + β 
 0
y ∈Y
 

Q(y 0 ) max φS(y 0 ), 0 
X 
−p(θ) 
0

y ∈Y

• Free-entry condition implies that:

φ c
Q(y 0 ) max φS(y 0 ), 0 =
X 

y 0 ∈Y
1 − φ βq(θ)

• So that  
φ
T (y , y 0 ) max S(y 0 ), 0 
X 
S(y ) = y − b − c θ+β
1−φ y 0 ∈Y

7 / 18
Blackwell’s conditions

• If V ∈ RN is a N-dimensional vector, denote by Vi its i-th entry.

• Let F be a function so that F : RN → RN .
• That is, F is a function that receives an input vector and returns an
output vector.
• If the following conditions hold:
1 (Monotonicity) For any V 1 , V 2 ∈ RN such that Vi1 ≤ Vi2 for all
i = 1, . . . , N, then Fi (V 1 ) ≤ Fi (V 2 ) for all i.
2 (Discounting) There exists β ∈ (0, 1) such that

Fi (V + a) ≤ Fi (V ) + βa for all V ∈ RN , a ≥ 0, i = 1, . . . , N.

• Then:
1 There exists only one V ∗ ∈ RN such that F (V ∗ ) = V ∗ .
∞
2 For any V 0 ∈ RN , the sequence of vectors {V n }n=1 defined by
n
V = F (V n−1
) converges to V . That is, limn→∞ Vin = Vi∗ for all i.
∗

8 / 18
Idiosyncratic shocks

 
φ
T (y , y 0 ) max S(y 0 ), 0 
X 
S(y ) = y − b − c θ+β
1−φ y 0 ∈Y

• How to find S(y )?

• The function S(y ) is also a vector: S = [S(y1 ), S(y2 ), . . . , S(yN )],
S ∈ RN .
• Define function F : RN → RN by
 
N
φ X
Fi (S) = yi −b−c θ+β  T (yi , yj ) max {S(yj ), 0} , i = 1, . . . , N.
1−φ j=1

• Prove that F satisfies Blackwell’s conditions.

9 / 18
Idiosyncratic shocks

• There is only one function S : Y → R that satisfies the equation

φ
T (y , y 0 ) max S(y 0 ), 0
X 
S(y ) = y − b − c θ+β
1−φ y 0 ∈Y

• For any S 0 : Y → R, if we define a sequence of functions {S n }∞

n=1 as

φ n o
T (y , y 0 ) max S n−1 (y 0 ), 0 ,
X
S n (y ) = y − b − c θ+β
1−φ y 0 ∈Y

then limn→∞ S n (y ) = S(y ) for all y ∈ Y .

10 / 18
Idiosyncratic shocks
Value function iteration

Computer algorithm to find S.

1 Set a small and positive tolerance level ε. For example: ε = 10−6 .
Guess any S in = S in (y1 ), S in (y2 ), . . . , S in (yN ) . For example:
 
2
S in = [0, 0, . . . , 0].
3 Obtain S out using
φ n o
T (y , y 0 ) max S in (y 0 ), 0 .
X
S out (y ) = y − b − c θ+β
1−φ y 0 ∈Y

4 Compute the distance between S in and S out :

distance = maxy ∈Y S (y ) − S out (y ) .
 in

5 Check if distance is lower than tolerance level, distance < ε. If it is

not, redefine S in as S out and go back to 3.

11 / 18
Idiosyncratic shocks
Algorithm to find equilibrium

1 Guess θ > 0.
2 Find S(y ).
3 Verify if free-entry condition holds:
φ c
Q(y 0 ) max φS(y 0 ), 0 =
X 
.
y 0 ∈Y
1 − φ βq(θ)

If it doesn’t hold, update θ and go back to 2. If it holds, we found an

equilibrium.

12 / 18
On-the-job search

• Production of a match can assume different values:

y ∈ Y ≡ {y1 , y2 , . . . , yN }, where 0 < y1 < y2 < · · · < yN .
• The productivity of a match doesn’t change over time.
• In the first period in which a firm and a worker meet, they face a
probability Q(y ) of starting as a match that produces y .
• Back to exogenous destruction: s > 0.
• With probability λ, in each period an employed worker has the
possibility to look for another job.

13 / 18
On-the-job search

• Worker’s value functions

    

Q(y 0 ) max E (y 0 ), E (y ) 
X 
E (y ) = w (y ) + β sU + (1 − s) λ p(θ) 
y 0 ∈Y

 

+(1 − p(θ))E (y ) + (1 − λ)E (y )


   
 
Q(y 0 ) max E (y 0 ), U  + (1 − p(θ))U
X 
U =b+β p(θ) 
y 0 ∈Y
 

14 / 18
On-the-job search
• Firm’s value functions
   
 
Q(y 0 ) max J(y 0 ), V  + (1 − q(θ))V
X 
V = −c + β q(θ) 
0
 
y ∈Y

   

Q(y 0 )
X
J(y ) = y − w (y ) + β sV + (1 − s) λ p(θ) 
y 0 ∈Y

 
I E (y 0 ) > E (y ) V + 1 − I E (y 0 ) > E (y )
 

J(y )
 

+(1 − p(θ))J(y ) + (1 − λ)J(y )


(
 0 1 if E (y 0 ) > E (y )
I E (y ) > E (y ) =
0 otherwise.
15 / 18
On-the-job search

• Free-entry: V = 0 implies
c
Q(y 0 ) max J(y 0 ), 0 =
X 
.
y 0 ∈Y
βq(θ)

• Nash bargaining for each state:

E (y ) − U = φS(y ) and J(y ) = (1 − φ)S(y )

S(y ) = E (y ) − U + J(y )
• There is agreement between worker and firm with respect to beginning
a match:

J(y ) > 0 ⇐⇒ (1 − φ)S(y ) > 0 ⇐⇒ φS(y ) > 0 ⇐⇒ E (y ) − U > 0

16 / 18
On-the-job search
Aggregate distribution of agents across states

 

Q(y 0 )I S(y 0 ) ≤ 0  + (1 − ut )s
X 
ut+1 = ut [1 − p(θ)] + ut p(θ) 
y 0 ∈Y
(
0 if S(y ) ≤ 0
µt+1 (y ) =
f (y , µt , ut ) otherwise

f (y , µt , ut ) = ut p(θ)Q(y ) + µt (y )(1 − s)(1 − λ)

 

Q(y 0 )I E (y ) ≥ E (y 0 ) 
X 
+ µt (y )(1 − s)λp(θ) 
y 0 ∈Y

+ µt (y )(1 − s)λ(1 − p(θ))

 

µt (y 0 )(1 − s)λp(θ)Q(y )I E (y ) > E (y 0 ) 

X 
+
y 0 6=y

17 / 18
On-the-job search

• It is possible to show that

φ
S(y ) = y − b − θc + β(1 − s) {[1 − λp(θ)] S(y )
1−φ
Q(y 0 ) φ max S(y 0 ), S(y )
X  
+ λp(θ)
y 0 ∈Y

+(1 − φ)I S(y 0 ) ≤ S(y ) S(y )

 

18 / 18