Advanced Macroeconomics I:

Extensions of the RBC Model

Krzysztof Makarski

1 Extensions of the RBC Model

1.1 Introduction

Labor market issues associated with the RBC models

• In the standard RBC model labor uctuates much more than productivity, which implies that short
run labor elasticity must be large. Hansen (1985) modies the standard (Prescott) RBC model by
introducing labor indivisibility.

• Frisch elasticity εF with logarithmic preferences u(c, n) = log c + φ log(1 − n) from un = uc w we get

φc
n=1−
w
∂n w φc w φc 1 1−n
εF = = 2 = =
∂w n w n w n n
which given that n ≈ 0.3 gives the estimate of Frisch elasticity ≈ 2.33.

• Most microstudies nd the Frisch elasticity below 0.7.

• Some macroeconomists insist that macroeconomic elasticity is larger than microeconomic elasticity
(due to e.g. adjustments on extensive margin).

Issues with technology shocks in the RBC models

• Typical estimates of the Solow residual imply a probability of technical regress on the order of 40%,
which seems implausible to many economists.

• Additionally, recent studies show that correcting the Solow residual estimates for mismeasurements of
inputs and inappropriate assumptions about market structure lowers volatility of technology shocks.

• The Solow residual measure from simple Cobb-Douglas technology very spurious.

• Since the RBC model requires large and volatile technology shocks this ndings could be interpreted
as negative falsication of the RBC model.

• We will see that they are not.

Assumptions
• Hansen (1985) modies the standard (Prescott) RBC model by introducing labor indivisibility.

• Economy is populated by innitely many households (of measure 1) indexed by i ∈ [0, 1].
• The representative rms operates a CRS technology.

• Markets are complete.

• He focuses on the business cycle properties (therefore he ignores growth).

1
Indivisible vs.Divisible
Divisible labor
Labor Indivisible Labor

All individuals employed Not everyone is employed

Individuals can work any number of Individuals work either a given
hours number of hours (e.g. 8 hours per
day) or they do not work at all
Everyone works the same # of hours Individuals choose probability and
lottery determines who actually
works
Each agent chooses average # of Each agent choose the same
hours worked probability of working, but # of
actual hours worked diers
Positive technology shock with Positive technology shock with
greater returns to working, provides greater returns to working, provides
incentives to work more hours incentives to increase probability of
work (more people will work but
hours do not change).

Why Indivisible labor?

• Data.

 Fluctuations in aggregate hours worked in the economy are due to changes in both the number
of hours people choose to work (intensive margin) and the number of people entering and leaving
the work force (extensive margin)

var(log Ht ) = var(log ht ) + var(log Nt ) + 2cov(log ht , log Nt )

where Ht − total hours worked, ht −average hours worked, and Nt − # of people at work.

 Hansen uses quarterly, post war data from the Bureau of Labor Statistics and nds that 55% of
the variance of total hours worked is due to variation in the number of people at work, while 20%
is due to variation in the average hours worked.

 Since most of the variation in total hours worked is due to individuals either working or not
working, this supports modeling of RBC with indivisible labor.

• Intertemporal Elasticity of Substitution.

 Fitting a standard RBC model with divisible labor requires very high elasticity of substitution
between leisure across time periods for households, which is inconsistent with panel data.

 In a RBC model with indivisible labor elasticity of substitution for households is low, while the
elasticity for the  representative agent is innite since, it is independent of the utility function
implied by the individual household.

• Note. We still should model both intensive and extensive margin (as uctuations along both are
observed in the economy)

1.2 Divisible labor

Firms
• Producers hire labor nt and capital kt to produce output yt . They have access to the CRS technology
yt = zt ktα nt 1−α . Real wage is denoted as wt and the rental rate for capital rk,t (it is not the interest
rate)

• Problem of the rm in period t

Πt = max yt − wt nt − rk,t kt−1 (1.1)
(yt ,nt ,kt )

subject to yt ≤ zt ktα n1−α

t

2
Households
• We assume that consumers earn income from labor wt nt and capital rental rk,t kt which they spend
either on consumption ct or investments xt . Additionally, household receive dividends from rms
(which are zero in equilibrium) Πt . Hence the consumer budget constraint has the following form

ct + xt ≤ wt nt + rk,t kt + Πt (1.2)

• Capital accumulates according to the standard law of motion for capital

kt+1 = xt + (1 − δ)kt (1.3)

• From (1.2) and (1.3) we can eliminate xt

ct + kt+1 ≤ wt nt + rk,t kt + (1 − δ)kt + Πt

• Note kt > 0 gives us a NPG condition.

• Instantaneous utility function (felicity function) is given by the following function u(ct , nt ) and each
consumer maximizes the lifetime utility subject to the sequence of budget constraints

∞
hX i
max ∞ E0 β t u(ct , nt ) (1.4)
{ct ,nt ,kt }t=0
t=0
subject to ct + xt ≤ wt nt + rk,t kt + Πt
kt+1 = xt + (1 − δ)kt (1.5)

nt ∈ [0, 1] , kt ≥ 0

Market clearing
• Markets clear when demand and supply are equal.

• Goods market clearing

ct + xt = yt (1.6)

• Labor market and capital rental market clearing conditions are introduced notationally (the same n in
the rm problem as in the consumer problem).

Denition of a competitive equilibrium.

A competitive equilibrium is an allocation {ct , nt , xt , kt , yt }∞
t=0 and prices {rk,t , wt }∞
t=0 such that

• {ct , xt , nt , kt }∞
t=0 solves the consumer problem given prices

∞
" #
X
t
max E0 β (log ct + φ log(1 − nt ))
{ct ,xt ,kt+1 ,nt }∞
t=0
t=0
sb. to ct + xt = wt nt + rk,t kt + Πt
kt+1 = xt + (1 − δ) kt
nt ∈ [0, 1]

• for each t, (nt , kt , yt ) solves the producer problem given prices

Πt = max yt − rk,t kt − wt nt
(yt ,nt ,kt )

sb. to yt = zt ktα n1−α

t

• markets clear
ct + xt = yt

3
First Order Conditions
Maximization of

∞
X
Et β t+τ (log ct+τ + φ log(1 − nt+τ ))
τ =0

subject to the budget constraint

ct+τ + xt+τ = wt+τ nt+τ + rk,t+τ kt+τ + Πt+τ

and the capital accumulation equation

kt+1+τ = (1 − δ) kt+τ + xt+τ

Substituting we get
ct+τ + kt+1+τ = wt+τ nt+τ + (rk,t+τ + (1 − δ))kt+τ + Πt+τ
To solve we construct the following Lagrangian

∞ 
X
L = Et β t+τ (log ct+τ + φ log(1 − nt+τ ))
τ =0

− λt+τ (ct+τ + kt+1+τ − wt+τ nt+τ − (rk,t+τ + (1 − δ))kt+τ − Πt+τ )

First order conditions

ct : Et [β t uc,t − λt ] = 0
nt : Et [β t un,t − λt wt ] = 0
kt+1 : Et [λt − λt+1 (rk,t+1 + 1 − δ)] = 0

plus the TVC

lim β t+τ Et [λt+τ kt+1+τ ] = 0
τ →∞

Simplifying

ct : β t uc,t = λt
nt : β t un,t = λt wt
kt+1 : λt = Et [λt+1 (rk,t+1 + 1 − δ)]

Solving

un,t = uc,t wt
uc,t = βEt [uc,t+1 (rk,t+1 + 1 − δ)]

Substituting for uc,t and un,t we get

φct = wt (1 − nt )
 
ct
1 = βEt (rk,t+1 + (1 − δ))
ct+1

The producer problem is the same as earlier.

wt = At (1 − α) ktα n−α
t
rk,t = At αktα−1 n1−α
t
yt = At ktα n1−α
t

4
The market clearing condition
ct + xt = yt
with
kt+1 = xt + (1 − δ) kt
Summarizing the following equations characterize the economy

• The consumer problem Labor

un,t = uc,t wt (1.7)

Euler
uc,t = βEt [uc,t+1 (rk,t+1 + 1 − δ)] (1.8)

and the transversality condition

lim β t+τ Et [uc,t+τ kt+1+τ ] = 0 (1.9)

τ →∞

• The producer problem

wt = At (1 − α) ktα n−α
t (1.10)

rk,t = At αktα−1 n1−α

t (1.11)

yt = At ktα n1−α
t (1.12)

• Market clearing
ct + kt+1 = At ktα n1−α
t + (1 − δ)kt (1.13)

Summarizing the optimal allocation is characterized by (1.7) - (1.13).

Properties
Next, we check whether a competitive allocation is ecient. First, we need to dene what ecient means.

Denition 1. An allocation {ct , nt , kt }∞

t=0 is ecient if it solves the following benevolent social planner
problem

∞
X
max E0 β t ln ct
{ct ,kt+1 ,nt }∞
t=0
t=0
p.w. ct + kt+1 = At ktα n1−α
t + (1 − δ)kt
nt ∈ [0, 1] , given k0

In any period t social planner maximizes

∞
X
Et β t+τ (log ct+τ + φ log(1 − nt+τ ))
τ =0

subject to the feasibility constraint

α
ct+τ + kt+1+τ = At+τ kt+τ n1−α
t+τ + (1 − δ)kt+τ )

To solve we construct the following Lagrangian

∞ 
X
L = Et β t+τ (log ct+τ + φ log(1 − nt+τ ))
τ =0

α
− λt+τ (ct+τ + kt+1+τ − At+τ kt+τ n1−α
t+τ − (1 − δ)kt+τ )

First order conditions

5
 ct : Et [β t uc,t − λt ] = 0
nt : Et [β t un,t − λt (1 − α)At+1 kt+1
α
n−α
t+1 ] = 0
α−1 1−α
kt+1 : Et [λt − λt+1 (αAt+1 kt+1 nt+1 + 1 − δ)] = 0

Simplifying

ct : β t uc,t = λt
nt : β t un,t = λt (1 − α)At+1 kt+1
α
n−α
t+1
α−1 1−α
kt+1 : λt = Et [λt+1 (αAt+1 kt+1 nt+1 + 1 − δ)]

Solving

un,t = α
uc,t (1 − α)At+1 kt+1 n−α
t+1
α−1 1−α
uc,t = βEt [uc,t+1 (αAt+1 kt+1 nt+1 + 1 − δ)]

Substituting for uc,t and un,t we get

φct = α
(1 − α)At+1 kt+1 n−α
t+1 (1 − nt )
 
ct α−1 1−α


1 = βEt αAt+1 kt+1 nt+1 + (1 − δ)
ct+1
Furthermore, we have the transversality condition

lim β t+τ Et [uc,t+τ kt+1+τ ] = 0

τ →∞

Summarizing the optimal allocation is characterized by

un,t = uc,t (1 − α)At ktα n−α

t (1.14)
α−1 1−α
uc,t = βEt [uc,t+1 (αAt+1 kt+1 nt+1 + 1 − δ)] (1.15)

ct + kt+1 = yt + (1 − δ)kt (1.16)

yt = Aktα n1−α
t (1.17)

and the TVC

lim β t+τ Et [uc,t+τ kt+1+τ ] = 0 (1.18)
τ →∞

Next we have

Theorem 2. First Welfare Theorem: Under certain conditions any competitive allocation is ecient.
Proof. First we simplify equations (1.7) − (1.13). Substituting from (1.10) − (1.11) into (1.7) − (1.8) we get

un,t = uc,t (1 − α)At ktα n−α

t (1.19)
α−1 1−α
uc,t = βEt [uc,t+1 (αAt+1 kt+1 nt+1 + 1 − δ)] (1.20)

Thus competitive allocation satises equations (1.14) - (1.18). Note, that those equations are the same as
equations characterizing an ecient allocation. Hence, any competitive allocation is ecient.

and

Theorem 3. Second Welfare Theorem: Under certain conditions any ecient allocation can be decentralized
as a competitive allocation.
Proof. {c , n , k , y }
Consider an ecient allocation
∞
t
∞
(1.14) − (1.18)
t t t t=0 that satises equations . Construct
prices {wt , rk,t }t=0 using (1.10) − (1.11). Then the ecient allocation and the prices is a competitive equi-
librium, since they satisfy equations (1.7) − (1.13).

6
 Both theorems are important. The rst one tells us that no resources are wasted in competitive equi-
librium. And the second tells us that any ecient allocation can be decentralized as an ecient allocation.
Furthermore, they are technically useful, since one does not need to nd an equilibrium directly, instead one
can nd an ecient allocation rst, and then prices using equations (1.10)− (1.11). Thus, one can just solve
the benevolent social planner problem to get an equilibrium.
Furthermore they reveal the informative role of the prices in the competitive economy. And any time the
price system is disturbed, it leads to an inecient outcome.

1.3 Indivisible labor

Assumptions
• Indivisibility of labor is modeled by restricting the labor supply possibilities set so that individuals can
work either full time, denoted by h̄ or not work at all.

• Each period, instead of choosing labor hours, households choose a probability of working ηt .
• A lottery then determines whether or not the household actually works.

• Markets are complete (households can trade state contingent bonds in order to insure themselves
completely from the risk of being unemployed, we will not include that into equations to simplify
notation).

• Since all households are the same they will choose the same probability of working h̄, ηt .

• Even though everyone ex ante chooses the same ηt the outcomes of each lottery will be dierent, so
each household will dier ex post.

 The lottery guarantees that the consumption possibilities set will is convex (important from the
technical point of view).

 The lottery also makes individuals better o because it increases the available choices. (Otherwise
they could only choose = 0 or 1).

• Therefore, only the utility function changes (the rest remains the same).

Labor indivisibility
• ηt - probability of working in period t by any agent (lottery). If she works, she works h̄ hours, with
0 ≤ n̄ ≤ 1.

• Consider u(·) = ln ct + ηt φ ln(1 − h̄) + (1 − ηt )φ ln 1. By law of large numbers nt = ηt h̄, substituting

for ηt
nt
u(·) = ln ct + φ ln(1 − h̄)
h̄
Rewriting
φ
u(ct, nt ) = ln ct + ln(1 − h̄)nt
h̄
Let φ̃ = − φh̄ ln(1 − h̄) > 0 then

u(ct, nt ) = ln ct − φ̃nt

• Note that while individual Frisch elasticity is zero the macro one is innite (linear utility).

7
The new household problem
• Instantaneous utility function (felicity function) is given by the following function u(ct , nt ) = ln ct − φ̃nt
and each consumer maximizes the lifetime utility subject to the sequence of budget constraints

∞
hX i
max E0 β t u(ct , nt ) (1.21)
{ct ,xt ,nt ,kt }∞
t=0
t=0
subject to ct + xt ≤ wt nt + rk,t kt + Πt
kt+1 = xt + (1 − δ)kt (1.22)

nt ∈ [0, 1] , kt ≥ 0

Denition of a competitive equilibrium.

A competitive equilibrium is an allocation {ct , nt , xt , kt , yt }∞
t=0 and prices {rk,t , wt }∞
t=0 such that

• {ct , xt , nt , kt }∞
t=0 solves the consumer problem given prices

∞
" #
X
t
max E0 β (ln ct − φ̃nt )
{ct ,xt ,kt+1 ,nt }∞
t=0
t=0
sb. to ct + xt = wt nt + rk,t kt + Πt
kt+1 = xt + (1 − δ) kt
nt ∈ [0, 1]

• for each t, (nt , kt , yt ) solves the producer problem given prices

Πt = max yt − rk,t kt − wt nt
(yt ,nt ,kt )

sb. to yt = zt ktα n1−α

t

• markets clear
ct + xt = yt

First Order Conditions

Maximization of

∞
X
Et β t+τ (log ct+τ − φ̃nt+τ ))
τ =0

subject to the budget constraint

ct+τ + xt+τ = wt+τ nt+τ + rk,t+τ kt+τ + Πt+τ

and the capital accumulation equation

kt+1+τ = (1 − δ) kt+τ + xt+τ

Substituting we get
ct+τ + kt+1+τ = wt+τ nt+τ + (rk,t+τ + (1 − δ))kt+τ + Πt+τ
To solve we construct the following Lagrangian

∞ 
X
L = Et β t+τ (log ct+τ − φ̃nt+τ ))
τ =0

− λt+τ (ct+τ + kt+1+τ − wt+τ nt+τ − (rk,t+τ + (1 − δ))kt+τ − Πt+τ )

First order conditions

8
 1
ct : Et [β t − λt ] = 0
ct
nt : Et [−β t φ̃ − λt wt ] = 0
kt+1 : Et [λt − λt+1 (rk,t+1 + 1 − δ)] = 0
plus the TVC
lim β t+τ Et [λt+τ kt+1+τ ] = 0
τ →∞
Simplifying

1
ct : βt = λt
ct
nt : β t φ̃ = λt wt
kt+1 : λt = Et [λt+1 (rk,t+1 + 1 − δ)]
Solving

1
β t φ̃ = βt wt
ct
1 1
βt = β t+1 Et [ (rk,t+1 + 1 − δ)]
ct ct+1

φ̃ct = wt
 
ct
1 = βEt (rk,t+1 + (1 − δ))
ct+1
The producer problem is the same as earlier.

wt = At (1 − α) ktα n−α
t
rk,t = At αktα−1 n1−α
t
yt = At ktα n1−α
t

The market clearing condition

ct + x t = y t
with
kt+1 = xt + (1 − δ) kt
Summarizing the following equations characterize the economy

• The consumer problem Labor

φ̃ct = wt (1.23)

Euler  
ct
1 = βEt (rk,t+1 + (1 − δ)) (1.24)
ct+1
and the transversality condition

1
lim β t+τ Et [β t+τ kt+1+τ ] = 0 (1.25)
τ →∞ ct+τ
• The producer problem

wt = At (1 − α) ktα n−α
t (1.26)

rk,t = At αktα−1 n1−α

t (1.27)

yt = At ktα n1−α
t (1.28)

• Market clearing
ct + kt+1 = At ktα n1−α
t + (1 − δ)kt (1.29)

Summarizing the optimal allocation is characterized by (1.23) - (1.29).

9
Log-linearization a a xt
xt
It is based on xat = xa xat x−a = xa elog xt −log x = xa ea log x = xa eax̂t ≈ xa (1 + ax̂t ), where x̂t = log x ).
Equation for capital accumulation
From
kt+1 = (1 − δ) kt + xt
In the steady state

δk = x
Log-linearizing

   
k 1 + k̂t+1 = (1 − δ) k 1 + k̂t + x (1 + x̂t )
   
k 1 + k̂t+1 = (1 − δ) k 1 + k̂t + δk (1 + x̂t )

k̂t+1 = (1 − δ) k̂t + δ x̂t

Euler
From
βEt ct c−1
 
t+1 (rk,t+1 + (1 − δ)) = 1

In the steady state

1 = β (rk + (1 − δ))
or
rk = β −1 − (1 − δ)
Log-linearizing we get

βEt c(1 + ĉt )c−1 (1 − ĉt+1 ) (rk (1 + r̂k,t+1 ) + (1 − δ)) = 1

 

(Since it is rst order approximation we have (1 + x̂) (1 + ŷ) = 1 + x̂ + ŷ )

βEt (1 + ĉt − ĉt+1 ) (β −1 − (1 − δ))(1 + r̂k,t+1 ) + (1 − δ) = 1




Et [(1 + (ĉt − ĉt+1 )) (1 + (1 − β (1 − δ))r̂k,t+1 )] = 1

Et [1 + ĉt − ĉt+1 + (1 − β (1 − δ))r̂k,t+1 ] = 1
1 + (1 − β (1 − δ)) Et [r̂k,t+1 ] + Et [ĉt − ĉt+1 ] = 1
(1 − β (1 − δ)) Et [r̂k,t+1 ] + Et [ĉt − ĉt+1 ] = 0
Consumption leisure choice
From
wt = φ̃ct
we get in the steady state
w = φ̃c
Log-linearizing

w(1 + ŵt ) = φ̃c(1 + ĉt )

(1 + ŵt ) = (1 + ĉt )
ŵt = ĉt

Wage equation
From
wt = At (1 − α) ktα n−α
t

In the steady state

w = A (1 − α) k α n−α

10
Log-linearizing
w(1 + ŵt ) = A(1 + Ât ) (1 − α) k α (1 + αk̂t )n−α (1 − αn̂t )
Simplifying

(1 + ŵt ) = (1 + Ât )(1 + αk̂t )(1 − αn̂t )

1 + ŵt = 1 + Ât + αk̂t − αn̂t
ŵt = Ât + αk̂t − αn̂t

Rental rate equation

From
rk,t = At αktα−1 n1−α
t

In the steady state

rk = Aαk α−1 n1−α
Log-linearizing

   
rk (1 + r̂k,t ) = A 1 + Ât αk α−1 1 + (α − 1) k̂t n1−α (1 + (1 − α) n̂t )

Simplifying

  
(1 + r̂k,t ) = 1 + Ât 1 + (α − 1) k̂t (1 + (1 − α) n̂t )

1 + r̂k,t = 1 + Ât + (α − 1) k̂t + (1 − α) n̂t

r̂k,t = Ât + (α − 1) k̂t − (α − 1) n̂t

Production function
From
yt = At ktα n1−α
t

In the steady state

y = Ak α n1−α
Log-linearizing
   
y (1 + ŷt ) = A 1 + Ât k α 1 + αk̂t n1−α (1 + (1 − α) n̂t )

Simplifying

  
(1 + ŷt ) = 1 + Ât 1 + αk̂t (1 + (1 − α) n̂t )

1 + ŷt = 1 + Ât + αk̂t + (1 − α) n̂t

ŷt = Ât + αk̂t − (α − 1) n̂t

Feasibility
Market clearing conditions replace balance of payments equation. Feasibility

ct + xt = yt
In the steady state
c+x=y
Log-linearizing
c (1 + ĉt ) + x (1 + x̂t ) = y (1 + ŷt )
Simplifying
c x
ĉt + x̂t = ŷt
y y

11
Steady state
First  
ct
βEt (rk,t+1 + (1 − δ)) = 1
ct+1
gives
rk = β −1 − (1 − δ)
furthermore since rk = Aαk α−1 n1−α
k
α−1
β −1 − (1 − δ) = Aα
n
k  β −1 − (1 − δ)  α−1
1  βAα 1
 1−α
= =
n Aα 1 − β (1 − δ)

Substituting for
rk,t kt = At αktα−1 n1−α
t · kt = αAt ktα n1−α
t = αyt
we get

rk k = αy
k α α
= = −1
y rk β − (1 − δ)

Next, from
kt+1 = (1 − δ) kt + xt
we have
δk = x
thus
x k αδ βαδ
= δ = −1 =
y y β − (1 − δ) 1 − β (1 − δ)
and since
ct + xt = yt
we have

c x βαδ
= 1− =1−
y y 1 − β (1 − δ)
1 − β (1 − δ) − βαδ
=
1 − β (1 − δ)
(1 − β) + βδ (1 − α)
=
1 − β (1 − δ)

Finally, substituting w = A(1 − α)k α n−α into

k
α
φ̃c = w = A(1 − α)
n
α
A(1 − α)  βAα  1−α
c=
φ̃ 1 − β (1 − δ)
From feasibility
c xk k
α
+ =A
n kn n
c k
α k

=A −δ
n n n
c
n=
α
A nk − δ nk

12
 k
Substituting for c and
n
 α
 1−α
A(1−α) βAα
φ̃ 1−β(1−δ)
n=  α
 1−α  1
 1−α
βAα βAα
A 1−β(1−δ) −δ 1−β(1−δ)

The rest is straightforward.

Calibration Parameter Description

α = 0.36 Capital share (Cooley and

Prescott,1995)
δ = 0.025 Rate of capital deprecation (0.1
annual rate)
β = 0.99 0.99 (steady state annual real
interest rate = 4%)
φ=2 Disutility from labor (work 1/3 of
time)
h̄ = 0.53 Hours worked
(ρ, σ) = (0.95, 0.007) Technology shock (Cooley and
Prescott,1995)

Solution method
• Models is solved in dynare (codes attached) by local approximation around the steady state.

Impulse Response Function of a positive technology shock

-3 Output -3 Consumption Investment

x 10 x 10
14 6 0.05

12 0.04
5
10 0.03

8 4 0.02

6 0.01
3
4 0

2 2 -0.01
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

-3 Capital -3 Labor productivity -3 Labor

x 10 x 10 x 10
8 6 15

6 5 10

4 4 5

2 3 0

0 2 -5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

-3 Wages -3 Capital Rental Rate -3 Technology

x 10 x 10 x 10
6 15 8
divisible labor
indivisible labor
5 10 6

4 5 4

3 0 2

2 -5 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

13
Model data t
US data Divisible labor Indivisible labor
st.dev. corr(·,y) st.dev. corr(·,y) st.dev. corr(·,y)

y 1.76 1.00 1.35 1.00 1.76 1.00

c 1.29 0.85 0.42 0.89 0.51 0.87
x 8.6 0.92 4.24 0.99 5.71 0.99
k 0.63 0.04 0.36 0.06 0.47 0.05
n 1.66 0.76 0.70 0.98 1.35 0.98
y/n 1.18 0.42 0.68 0.98 0.50 0.87

Results of the indivisible labor model

• Larger uctuations (in the indivisible labor economy than the divisible labor economy).

• Indivisible labor increases the volatility of the stochastic growth model for a given stochastic process
for the technology shock.

• The indivisible labor model generates improves the model data t (hours and to lesser extent invest-
ments).

• Indivisible labor generates a highly procyclical productivity, while U.S. data shows that productivity
is only moderately procyclical.

Summary
• We highlighted the issue.

• Solved the model

 Log-linearization and steady state

 Dynare (moments and irf )

• Results.

1.4 Capacity utilization

Assumptions
• In the standard RBC there is weak amplication, therefore model almost behaves as technology shocks.

• After a positive technology shock the initial increase in output is only due to better technology and
higher labor supply (wages go up as well).

• Including,capacity utilization allows to intensify the use of capital in response to TFP shocks.

• It is costly, since high capacity utilization makes capital to depreciate faster.

δ(ut ) = δuω
t

Capacity utilization
• δ - depreciation rate of capital.

• Depreciation rate of capital is a function of capital utilization ut , δ(ut ) is strictly increasing and strictly
0 00
convex, δ (·), δ (·) > 0. The law of motion of capital becomes

kt+1 = xt + (1 − δ(ut ))kt

where δ(ut ) = δuω
t, δ > 0, ω > 1.
• Denote the capital services κt = ut kt .
• Output depends on the capital services rather then capital stock
1−α
yt = zt κα
t nt

14
The new household problem
• Instantaneous utility function (felicity function) is given by the following function u(ct , nt ) = ln ct +
log(1 − nt ) and each consumer maximizes the lifetime utility subject to the sequence of budget con-
straints
∞
hX i
max E0 β t u(ct , nt ) (1.30)
{ct ,xt ,ut ,nt ,kt }∞
t=0
t=0
subject to ct + xt ≤ wt nt + rk,t ut kt + Πt
kt+1 = xt + (1 − δ(ut ))kt (1.31)

nt ∈ [0, 1] , kt ≥ 0

Denition of a competitive equilibrium.

A competitive equilibrium is an allocation {ct , xt , nt , ut , kt , κt , yt }∞
t=0 and prices {rk,t , wt }∞
t=0 such that

• {ct , xt , nt , ut , kt }∞
t=0 solves the consumer problem given prices
"∞ #
X
t
max E0 β (ln ct + φ ln(1 − nt )
{ct xt ,ut ,kt+1 ,nt }∞
t=0
t=0
sb. to ct + xt = wt nt + rk,t ut kt + Πt
kt+1 = xt + (1 − δ(ut )) kt
nt ∈ [0, 1]

• for each t, (nt , κt , yt ) solves the producer problem given prices

Πt = max yt − rk,t κt − wt nt
(yt ,nt ,κt )
1−α
sb. to yt = zt κα
t nt

• markets clear

ct + xt = yt
κt = ut kt

First Order Conditions

Maximization of

∞
X
Et β t+τ (log ct+τ + φ log(1 − nt+τ ))
τ =0
subject to the budget constraint

ct+τ + xt+τ = wt+τ nt+τ + rk,t+τ ut+τ kt+τ + Πt+τ

and the capital accumulation equation

kt+1+τ = (1 − δ(ut+τ )) kt+τ + xt+τ

Substituting we get

ct+τ + kt+1+τ = wt+τ nt+τ + rk,t+τ ut+τ kt+τ + (1 − δ(ut+τ ))kt+τ + Πt+τ
To solve we construct the following Lagrangian

∞ 
X
L = Et β t+τ (log ct+τ + φ log(1 − nt+τ )))
τ =0

− λt+τ (ct+τ + kt+1+τ − wt+τ nt+τ − (rk,t+τ ut+τ kt+τ + (1 − δ(ut+τ ))kt+τ − Πt+τ )

15
First order conditions

1
ct : Et [β t − λt ] = 0
ct
1
nt : Et [−β t φ + λt wt ] = 0
1 − nt
kt+1 : Et [λt − λt+1 (rk,t+1 ut+1 + 1 − δ(ut+1 )] = 0
0
ut : λt (rk,t kt − δ (ut )kt ) = 0

plus the TVC

lim β t+τ Et [λt+τ kt+1+τ ] = 0
τ →∞

Simplifying

1
ct : βt = λt
ct
1
nt : βtφ = λt wt
1 − nt
kt+1 : λt = Et [λt+1 (rk,t+1 ut+1 + 1 − δ(ut+1 ))]
0
ut : rk,t = δ (ut )

Solving

1 1
βtφ = βt wt
1 − nt ct
1 1
βt = β t+1 Et [ (rk,t+1 ut+1 + 1 − δ(ut+1 )]
ct ct+1
0
rk,t = δ (ut )

and
kt+1 = xt + (1 − δ(ut )) kt
Beware E(XY ) 6= E(X)E(Y ).

φct
= wt
1 − nt
ct
1 = βEt [ (rk,t+1 ut+1 + 1 − δ(ut+1 )]
ct+1
0
rk,t = δ (ut )
kt+1 = xt + (1 − δ(ut )) kt

The producer problem is the same as earlier.

−α
wt = At (1 − α) κα
t nt
rk,t = At ακα−1
t n1−α
t
1−α
yt = At κα
t nt

The market clearing conditions

ct + x t = y t
and
κt = ut kt
Summarizing the following equations characterize the economy

16
 • The consumer problem Labor
φct
= wt (1.32)
1 − nt
Euler
ct
1 = βEt [ (rk,t+1 ut+1 + 1 − δ(ut+1 )] (1.33)
ct+1
Rental rate
0
rk,t = δ (ut ) (1.34)

Law of motion for capital

kt+1 = xt + (1 − δ(ut )) kt (1.35)

2
where δ(u) = δ0 + δ1 (u − 1) + δ2 (u − 1) and the transversality condition

1
lim β t+τ Et [β t+τ kt+1+τ ] = 0 (1.36)
τ →∞ ct+τ

• The producer problem

−α
wt = At (1 − α) κα
t nt (1.37)

rk,t = At ακtα−1 n1−α

t (1.38)
1−α
yt = At κα
t nt (1.39)

• Market clearing in goods market

ct + xt = yt (1.40)

and capital services

κt = ut kt (1.41)

Summarizing competitive equilibrium is characterized by (1.32) - (1.41).

Steady state 0
Since rk,t = δ (ut ) and δ(ut ) = δuω
t we get in the steady state

rk = δωuω−1

Next, from
ct
1 = βEt [ (rk,t+1 ut+1 + 1 − δ(ut+1 ))]
ct+1
in the steady state
1 = β (rk u + (1 − δuω ))
Substituting rk = δωuω−1
1 = β δωuω−1 u + 1 − δuω

β −1 − 1
uω =
δ(ω − 1)
Multiplying denominator and numerator by β
 1 − β  ω1
u=
βδ(ω − 1)

So from rk = δωuω−1
ω−1
 1−β  ω
rk = δω
βδ(ω − 1)

17
furthermore since rk = Aακα−1 n1−α = Aα(uk)α−1 n1−α

k
α−1
rk = Aαuα−1
n
1
k  r u  α−1
k
=
n Aαuα
k
If we have
n we can get

k
α
w = A(1 − α)(uk)α n−α = A(1 − α)uα
n
Substituting for
rk uk = Aα(uk)α−1 n1−α · (uk) = Aα(uk)α n1−α = αyt
we get

rk uk = αy
k α
=
y rk u
Next, from
kt+1 = (1 − δuω
t ) kt + xt

we have
x = δuω k
thus
x k αδuω
= δuω =
y y rk
and since
ct + xt = yt
we have

c x αδuω rk − αδuω
= 1− =1− =
y y rk rk
φct
Finally, from
1−nt = wt we get
c 1 wn
φ = (1 − n)
y n y
Since w · n = A(1 − α)(uk)α n−α · n = (1 − α)A(uk)α n1−α = (1 − α)y
c 1
φ = (1 − α)(1 − n)
y n
Solving for n we get
c
φ n + (1 − α)n = (1 − α)
y
(1 − α) (1 − α)
n= c

=   <1
φ y + (1 − α) (1 − α) + φ rk −αδu
ω
rk

k w(1−n) y 1
Which allows to nd k (using
n ), c= φ , y= c ·c= c ·c and κ = uk .
y

18
Log-linearization a a xt
xt
It is based on xat = xa xat x−a = xa elog xt −log x = xa ea log x = xa eax̂t ≈ xa (1 + ax̂t ), where x̂t ≡ log x ).
Rental rate for capital
From
0
rk,t = δ (ut )
Since δ(ut ) = δuω
t we get
rk,t = δωuω−1
t

In the steady state

rk = δωuω−1
Log-linearizing
rk (1 + r̂k,t ) = δωuω−1 (1 + (ω − 1)ût )
r̂k,t = (ω − 1)ût
Equation for capital accumulation
From
kt+1 = (1 − δ(ut )) kt + xt
kt+1 = (1 − δuω
t ) kt + xt

In the steady state

δuω k = x
Log-linearizing

k(1 + k̂t+1 ) = 1 − δuω (1 + ωût ) k(1 + k̂t ) + x(1 + x̂t )

k(1 + k̂t+1 ) = 1 − δuω − δωuω ût k(1 + k̂t ) + δuω k(1 + x̂t )

(Since it is rst order approximation we have (1 + x̂) (1 + ŷ) = 1 + x̂ + ŷ )

1 + k̂t+1 = 1 − δuω − δωuω ût + (1 − δuω )k̂t + δuω + δuω x̂t )

k̂t+1 = (1 − δuω )k̂t + δuω x̂t − δωuω ût

Euler
From
ct
1 = βEt [ (rk,t+1 ut+1 + 1 − δ(ut+1 ))]
ct+1
1 = βEt [ct c−1 ω
t+1 (rk,t+1 ut+1 + 1 − δut+1 ]

In the steady state

1 = β (rk u + (1 − δuω ))
Log-linearizing we get

1 = βEt [ c(1 + ĉt )c−1 (1 − ĉt+1 ) (rk (1 + r̂k,t+1 )u(1 + ût+1 )+

1 − δuω (1 + ωût+1 )]

1 = βEt [(1 + ĉt − ĉt+1 )(rk u(1 + r̂k,t+1 + ût+1 ) + 1 − δuω − δuω ωût+1 )]
Substituting rk = δωuω−1

1 = βEt [(1 + ĉt − ĉt+1 )(δωuω + δωuω r̂k,t+1 + δωuω ût+1 + 1 − δuω − δuω ωût+1 )]

1 = βEt [(1 + ĉt − ĉt+1 )(δωuω−1 u + 1 − δuω + δωuω r̂k,t+1 )]

1 = βEt [(1 + ĉt − ĉt+1 )(rk u + 1 − δuω + δωuω r̂k,t+1 )]

19
Since β −1 = rk u + (1 − δuω )

1 = βEt [(1 + ĉt − ĉt+1 )(β −1 + δωuω r̂k,t+1 )]

1 = βEt [β −1 + β −1 (ĉt − ĉt+1 ) + δωuω r̂k,t+1 ]

Et [ĉt − ĉt+1 ] + βδωuω Et [r̂k,t+1 ] = 0
Consumption leisure choice
From
wt (1 − nt ) = φbct
in the steady state we get
w (1 − n) = φc
Log-linearizing

w (1 + ŵt ) (1 − n(1 + n̂t )) = φc (1 + ĉt )

φc
(1 + ŵt ) (1 − n − nn̂t )) = (1 + ĉt )
w
(1 − n) − nn̂t + (1 − n)ŵt = (1 − n) (1 + ĉt )
(1 − n) − nn̂t + (1 − n)ŵt = (1 − n) + (1 − n) ĉt
ŵt (1 − n) − nn̂t = (1 − n) ĉt

Wage equation
From
−α
wt = At (1 − α) κα
t nt

In the steady state

w = A (1 − α) κα n−α
Log-linearizing
w(1 + ŵt ) = A(1 + Ât ) (1 − α) κα (1 + ακ̂t )n−α (1 − αn̂t )
Simplifying

(1 + ŵt ) = (1 + Ât )(1 + ακ̂t )(1 − αn̂t )

1 + ŵt = 1 + Ât + ακ̂t − αn̂t
ŵt = Ât + ακ̂t − αn̂t

Rental rate equation

From
rk,t = At ακtα−1 n1−α
t

In the steady state

rk = Aακα−1 n1−α
Log-linearizing
 
rk (1 + r̂k,t ) = A 1 + Ât ακα−1 (1 + (α − 1) κ̂t ) n1−α (1 + (1 − α) n̂t )

Simplifying
 
(1 + r̂k,t ) = 1 + Ât (1 + (α − 1) κ̂t ) (1 + (1 − α) n̂t )
1 + r̂k,t = 1 + Ât + (α − 1) κ̂t + (1 − α) n̂t
r̂k,t = Ât + (α − 1) κ̂t + (1 − α) n̂t

Production function
From
1−α
yt = At κα
t nt

20
In the steady state
y = Aκα n1−α
Log-linearizing
 
y (1 + ŷt ) = A 1 + Ât κα (1 + ακ̂t ) n1−α (1 + (1 − α) n̂t )
Simplifying
 
(1 + ŷt ) = 1 + Ât (1 + ακ̂t ) (1 + (1 − α) n̂t )
1 + ŷt = 1 + Ât + ακ̂t + (1 − α) n̂t
ŷt = Ât + ακ̂t + (1 − α) n̂t
Feasibility
Market clearing conditions replace balance of payments equation. Feasibility

ct + xt = yt
In the steady state
c+x=y
Log-linearizing
c (1 + ĉt ) + x (1 + x̂t ) = y (1 + ŷt )
Simplifying
c x
ĉt + x̂t = ŷt
y y
Capital services
Since
κt = ut kt
in the steady state κ = uk and

κ(1 + κ̂t ) = u(1 + ût )k(1 + k̂t )

κ(1 + κ̂t ) = uk(1 + ût + k̂t )
Finally we get
κ̂t = ût + k̂t

Calibration Parameter Description

α = 0.36 Capital share (Cooley and

Prescott,1995)
δ = 0.025 Rate of capital deprecation (0.1
annual rate)
β = 0.99 0.99 (steady state annual real
interest rate = 4%)
φ=2 Disutility from labor (work 0.3 of
time)
ω = 1.45 In line with literature.
(ρ, σ) = (0.95, 0.0041) Technology shock (with capacity
utilization)
(ρ, σ) = (0.95, 0.007) Technology shock (standard RBC,
Cooley and Prescott,1995)

Dynare
• Code

• IRF

• Moments

21
Impulse Response Function of a positive technology shock

Output -3 Consumption Investment

x 10
0.012 5 0.04

0.01 0.03
4
0.008
0.02
0.006 3
0.01
0.004
2
0.002 0

0 1 -0.01
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

-3 Capital -3 Capital Services -3 Labor

x 10 x 10 x 10
8 8 6

6 6 4

4 4 2

2 2 0

0 0 -2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

-3 Wages -3 Capital Rental Rate -3 Technology

x 10 x 10 x 10
6 15 8
basic RBC
5 capacity utilization
10 6
4
5 4
3
0 2
2

1 -5 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

Model data t
US data Standard RBC Capacity Utilization
st.dev. corr(·,y) st.dev. corr(·,y) st.dev. corr(·,y)

y 1.76 1.00 1.35 1.00 1.36 1.00

c 1.29 0.85 0.42 0.89 0.30 0.94
x 8.6 0.92 4.24 0.99 4.97 0.99
k 0.63 0.04 0.36 0.06 0.29 -0.05
n 1.66 0.76 0.70 0.98 0.76 0.99
y/n 1.18 0.42 0.68 0.98 0.60 0.99

Results of the capacity utilization model

• Larger uctuations for the same parameters of technology shocks.

• Similar volatility for the lower value of the standard deviations of technological innovations.

• Increases internal propagation.

• Reduces substantially the chance of technological regress.

Summary
• We raised two (of number of ) issues:

22
  labor elasticity

 what are technology shocks

• We showed that with some modications to the original model those issues can be resolved.

• In each case:

 we dene equilibrium

 we compute FOCs

 we calibrate the model

 we simulate the moments to compare against data (evaluate model data t).

23