Macro II

Professor Griffy

UAlbany

Spring 2025
Announcements

I Today: frictional labor markets:

1. summarize regularities about labor markets;
2. give simple partial equilibrium model of labor market
I Next Tuesday (4/22): please be here on time.
I Having class eval for my tenure case.
Why are Similar Workers Paid Differently?

I Posed by Dale Mortensen in his book “Wage Dispersion”

I Abowd, Kramarz, and Margolis (1999): “That... observably
equivalent individuals earn markedly different compensation
and have markedly different employment histories–is one of
the enduring features of empirical analyses of labor markets...”
I What are some possible reasons?
1. Ability
2. Selectivity
Residual Wage Dispersion

I We will look to theory to understand residual wage dispersion:

wage/income dispersion left over after we condition on
observables.
I There’s a lot:
1. Mortensen (2005): 70% of wage dispersion is unexplained.
I Understanding where this comes from is (one of) the goal of
labor economics.
Unconditional Wage Dispersion across Industries
Unexplained Variation
Abowd, Kramarz, and Margolis (1999)
I Famous paper for estimating the size of worker and firm
effects on residual wage dispersion.
I Longitudinal panel of matched employer-employee
observations in France.
I Empirical specification:

ln(yit ) = µy + ✓i + j,t + (xit µx ) + ✏it (1)

yit : income (2)
µy : average income in year t (3)
✓i : individual FEs (4)
j,t : firm FEs (5)
I Key findings:
1. Individual FEs explain more than Firm FEs.
2. Ind. FEs: 90% of inter-industry wage differentials.
3. 75% of the firm-size wage differentials.
Abowd, Kramarz, and Margolis (1999)

I Ind. FEs strongly correlated with income, Firm FEs not as

much.
Abowd, Kramarz, and Margolis (1999)

I These are estimates of the size of firm and worker effects.

I But they are still reduced-form.
I We haven’t identified the underlying causes of the size of each.
I What are some possible heterogeneities among workers?
I What are some possible heterogeneities among firms and
industries?
Other Interesting Regularities

I Davis and Haltiwanger (1991, 1996) on the level and growth

in wage-size effects and wage dispersion between plants:
1. Plants with > 5, 000 employees: $3.14/hour more than plants
with 25-49 in 1967.
2. Between 1967 and 1986, real wage grew by $1.00, but
differential grew to $6.31.
3. Explains 40% of the between-plant wage dispersion.
4. between-plant accounts for 59% of the total variance;
within-plant accounts for 2%.
5. Mean wage grows as plant size grows; wage dispersion falls!
I So is there wage dispersion in the economy?
I Why?
Perfectly Competitive Labor Markets

I We typically think of markets as being perfectly

competitive/walrasian, etc.
I Prices are determined by the point where supply = demand,
and there is no excess.
I Implications for labor market:
1. Workers are paid w = FL (K , L), i.e., their marginal product.
2. Zero profits in equilibrium.
I Wage dispersion can exist:
1. Dispersion directly proportional to dispersion in
productivity/ability/human capital, etc.
Frictional Labor Market

I But perfect competition is an approximation, both for

analytical and computational simplicity.
I Things we observe:
1. Price dispersion among identical workers/goods.
2. Failure of markets to clear: unemployment.
3. Profits.
I Market imperfections (frictions): agents are profit maximizing,
but lack of information and randomness prevent markets from
perfectly clearing.
I w⇢=FL (K , L).
I Here: explore job search as explanation for (some) wage
dispersion.
Outline: Frictional Labor Markets

I We’ll explore the following:

1. Partial equilibrium job search models: there is some wage
distribution and workers optimize by specifying a reservation
threshold.
2. General equilibrium job search: introduce an entry decision on
the firm’s side and endogenize the matching rate.
3. Efficiency and Directed search.
I Failings of the search framework:
1. Shimer (2005): can’t account for business cycle regularities.
2. Hornstein, Krusell, Violante (2011): can’t account for wage
dispersion.
A Model of Sequential Search

I The first model we’ll look at is called the “McCall Model”

(McCall, 1970).
I Basic idea:
1. Workers can be in one of two states: employed or unemployed,
with value functions V , U.
2. Receive job offers at exogenous rate ↵, no information about
meeting prior.
3. Once employed, workers remain at current job until
unexogenously
1/2 separated (no OTJS) at rate .
4. Exogenous distribution of wages, w 2 [w, w̄], w ⇠ F (.).
5. Linear utility: u(c) = b or u(c) = w.
(tro) (4)
I Optimal policy is a “reservation strategy,” i.e., a lower bound
on the wages a worker will accept out of unemployment.
I Why is wR > w?
I What is the source of wage dispersion?
Discrete Time Formulation
I Each agent wants to maximize his discounted present value of
consumption:
1
X
t
max ct (6)
t=0
(7)
I Some simplifying assumptions: ↵ = 1, = 0.
I Unemployed Bellman:
U = b + E [max{V , U}] (8)
Z w̄
U=b+ max{V , U}dF (w) (9)
w
I Employed Bellman:
V (w) = w + V (w) (10)
w
V (w) = (11)
1
Reservation Strategy
I The reservation strategy is the lowest wage a worker will
accept to leave unemployment.
I i.e., V (wR ) = U.
I Unemployed Bellman:
wR
! V (wR ) = U = (12)
1
Z w̄
wR
! =b+ max{V , U}dF (w) (13)
1 w
Z w̄
wR w wR
! =b+ max{ , }dF (w) (14)
1 w 1 1
Z w̄
! (1 )wR = (1 )b + max{w wR , 0}dF (w)
w
(15)
Z w̄
! wR = b + max{w wR , 0}dF (w) (16)
1 w
Reservation Strategy
I Reservation strategy:
Z w̄
wR = b + max{w wR , 0}dF (w) (17)
1 w
I Integrate by parts:
Z Z
udv = uv vdu.

Zw
u = w wR v = F (w)
(w wR )dF (w) =)
du = dw dv = dF (w)
wR
Zw w Zw
(w wR )dF (w) = (w wR )F (w) F (w)dw
wR
wR wR
Zw
E-WR)
=

IwFwd e = [1 F (w)]dw
wR
Reservation Strategy

I Reservation strategy:

Zw
wR = b + [1 F (w)]dw (18)
1
wR

I How would we solve for this?

I Assume a functional form for the distribution.
I Use root-finding algorithm to find wR st:

Zw
wR b+ [1 F (w)]dw = 0 (19)
1
wR

I Sounds like a good homework assignment!

Discrete Time Formulation
I Search models typically written in continuous time.
I Easier to work with analytically.
I Discrete time Bellman equation for Unemployment:

(1 + rdt)U = bdt + ↵dtE [max{V , U}] + (1 ↵dt)U (20)

(r + ↵)dtU = bdt + ↵dtE [max{V , U}] (21)
bdt + ↵dtE [max V , U]
U= (22)
(r + ↵)dt
I Taking limit as dt ! 0:

@Num.
= b + ↵E [max{V , U}] (23)
@dt
@Denom.
= (r + ↵) (24)
@dt
b + ↵E [max{V , U}]
)U= (25)
r +↵
Existence and Uniqueness

I For simplicity, assume V = w

r , i.e. = 0. Then,

b ↵ w
U= + E [max{ , U}] (26)
r +↵ r +↵ r
I U = T (U) is a contraction:
1. Discounting: ( r +↵
↵
< 1).
2. Monotonicity: T (U) is nondecreasing in U.
I By Blackwell’s Sufficient Conditions, this is a contraction with
a unique fixed-point.
Continuous Time Formulation

I Generally, we will use the continuous time Bellman in its

“asset value” formulation:
b + ↵E [max{V , U}]
U= (27)
r +↵
(r + ↵)U = b + ↵E [max{V , U}] (28)
rU = b + ↵E [max{V U, 0}] (29)
Z w̄
rU = b + ↵ max{V U, 0}dF (w) (30)
w

I Employment:

rV (w) = w (V (w) U) (31)

I Jobs lost at rate .

Reservation Strategy
I Reservation wage: V (wR ) = U:

rV (wR ) = wR (V (wR ) U) (32)

wR
V (wR ) = U = (33)
r
Z w̄
) wR = b + ↵ max{V U, 0}dF (w) (34)
w
Z w̄
w+ U wR
=b+↵ max{ , 0}dF (w) (35)
w r+ r
Z w̄ wR
w+ r wR
=b+↵ max{ , 0}dF (w) (36)
w r+ r
Z w̄
↵
=b+ max{w wR , 0}dF (w) (37)
r+ w

I Note: if = 0, identical to discrete time formulation.

Reservation Strategy II
I Truncating and integrating by parts:
Z w̄
↵
wR = b + max{w wR , 0}dF (w) (38)
r+ w
Z w̄
↵
wR = b + (w wR )dF (w) (39)
r+ wR
Z w̄ Z w̄
(w wR )dF (w) = (w wR )F (w)|w̄
wR F (w)dw
wR wR
(40)
⇠⇠⇠
= (w̄ wR )F (w̄) R ⇠ wR )F (wR )
(w⇠
⇠
(41)
Z w̄
F (w)dw (42)
wR
Z w̄
↵
! wR = b + [1 F (w)]dw (43)
r+ wR
Hazard Rate
I What is the hazard rate of unemployment?
I Rate of leaving unemployment at time t.
Z W̄
Hu (t) = ↵ dF (w) (44)
wR
= ↵(F (w̄) F (wR )) (45)
= ↵
|{z} (1 F (wR )) (46)
| {z }
MeetingRate Selectivity

I Note, almost every search model generates a hazard composed

of the product of a meeting probability and worker selectivity.
I This is important to remember.
I Hazard rate of employment (leaving employment for
unemployment)?

He (t) = (47)
I Because separations are independent of state.
Dynamics of Unemployment
I Use hazard rates to understand dynamics and steady-state.
I What does the model predict about employment and
unemployment?

u̇ = (1 u) ↵(1 F (wR ))u (48)

ė = ↵(1 F (wR ))(1 e) e (49)

I Steady-state: u̇ = 0, ė = 0:

0 = (1 u) ↵(1 F (wR ))u (50)

!u= (51)
+ ↵(1 F (wR ))
0 = ↵(1 F (wR ))(1 e) e (52)
↵(1 F (wR ))
!e= (53)
↵(1 F (wR )) +
Next Time

I General equilibrium search model.

I Next Tuesday (4/22): please be here on time.
I Having class eval for my tenure case.