Hansen’s basic RBC model

George McCandless
UCEMA
Spring 2007

1 Hansen’s RBC model

Hansen’s RBC model

First RBC model was Kydland and Prescott

– (1982) "Time to build and aggregate ‡uctuations," Econometrica

– Complicated
lagged cumulative investment
strange utility function
Lots added to look for presistence

Hansen’s model much simpler

– (1985) "Indivisible labor and the business cycle," Journal of Monetary

Economics
– Simple
– Added indivisible labor to gain persistence and covariance with out-
put
– Set rules for RBC game
Match second moments
Newer rule: match impulse response functions

Hansen’s basic model

Robinson Crusoe maximizes the discounted utility function

1
X
t
max u(ct ; lt )
t=0

1
 The speci…c utility functions

u(ct ; 1 ht ) = ln ct + A ln(1 ht )

with A > 0.

The production function is

1
f ( t ; kt ; h t ) = t kt ht

t is a random technology variable that follows the process

t+1 = t + "t+1

for 0 < < 1. "t iid, positive, bounded above, E"t = 1 .

– =) E t is 1 and t+1 > 0.

Hansen’s basic model (continued)

Capital accumulation follows the process

kt+1 = (1 )kt + it

The feasibility constraint is

f ( t ; kt ; ht ) ct + it

Bellmans equation
The basic Bellmans equation

V (kt ; t) = max [ln ct + A ln(1 ht ) + Et [V (kt+1 ; t+1 ) j t ]]

ct ;ht

subject to
1
t kt ht ct + it ;
t+1 = t + "t+1 ; and
kt+1 = (1 )kt + it :

Simpler to write as
1
V (kt ; t) = max ln t kt ht + (1 )kt kt+1
kt+1 ;ht

+A ln(1 ht ) + Et [V (kt+1 ; t+1 ) j t ]]

kt+1 and ht are control variables

First order conditions

2
 First order conditions are
@V (kt ; t ) 1
= 0= 1
@kt+1 t kt ht
+ (1 )kt kt+1
+ Et [Vk (kt+1 ; t+1 ) j t ]
and
@V (kt ; t) 1
= 0 = (1 ) 1 t kt ht
@ht t kt ht + (1 )kt kt+1
1
A
1 ht

The Benveniste-Scheinkman envelope theorem condition is

@V (kt ; t) 1 1 1
= 1 t kt ht + (1 )
@kt t kt ht + (1 )kt kt+1

Simplifying the …rst order conditions

First order conditions can be written as
1
1
t kt ht + (1 )kt kt+1
" #
1 1
t+1 kt+1 ht+1 + (1 )
= Et 1
j t
t+1 kt+1 ht+1 + (1 )kt+1 kt+2

and
1
(1 ) (1 ht ) t kt ht =A t kt ht + (1 )kt kt+1

In equilibrium,
1
ct = t kt ht + (1 )kt kt+1

Simplifying the …rst order conditions (continued)

Factor markets give
1 1
rt = t kt ht
and
wt = (1 ) t kt ht

First order conditions are simply

1 rt+1 + (1 )
= Et j t
ct ct+1
and
(1 ht ) wt = Act

Stationary states

3
 Stationary state value of h = ht = ht+1 is
1
h= h i;
A
1+ (1 ) 1 1 (1 )

Stationary state value of k = kt = kt+1 = kt+2 is

" #11
k=h 1 :
(1 )

How to study dynamics

1. Find the approximate Value function and Plan

(a) These will describe the dynamics within the precision of the approx-
imation
(b) Can be complicated to …nd
i. Especially if the domain of stochastic variable is large
(c) Can be impossible
i. If the model is not single agent
ii. If the model can not be approximated by social planner

2. Alternative approachs

(a) Log linear approximation of the model

i. After the optimization has been done
ii. After equilibrium conditions have been imposed
(b) Quadratic linear appoximation of the problem

Log-linearization techniques

Consider a function of the form

G(xt )
F (xt ) =
H(xt )

Taking logs of both side gives

ln(F (xt )) = ln(G(xt )) ln(H(xt ))

The …rst order Taylor series expansion

– around the stationary state values x

4
 – gives
F 0 (x) G0 (x)
ln(F (x)) + (xt x) ln(G(x)) + (xt x)
F (x) G(x)
H 0 (x)
ln(H(x)) (xt x)
H(x)

Log-linearization techniques (direct method)

In the stationary state

ln(F (x)) = ln(G(x)) ln(H(x)))

So the …rst order Taylor expansion can be written as

F 0 (x) G0 (x) H 0 (x)
(xt x) (xt x) (xt x)
F (x) G(x) H(x)

Remember that this holds only near x

An example using a Cobb-Douglas production function

Yt = t Kt Ht1

Take logs
ln Yt = ln t + ln Kt + (1 ) ln Ht

…rst order Taylor expansion gives

1 1
ln Y + Yt Y ln + t + ln K + Kt K
Y K
(1 )
+ (1 ) ln H + Ht H
H

Since in a stationary state

ln Y = ln + ln K + (1 ) ln H

get
1 1 (1 )
Yt Y t + Kt K + Ht H
Y K H

That reduces to
Yt t Kt (1 ) Ht
+1 + +
Y K H

5
Log-linearization techniques (Uhlig’s method)
Write the original variable as
f
Xt = XeXt
or
et = ln Xt
X ln X
bring together all the exponential terms that you can
e et
At Bt AeAt B e B
=
Ct C e Cet
becomes
AB e et et
eAt + B C

C
Reference:Uhlig, Harald, (1999) "A toolkit for analysing nonlinear dy-
namic stochastic models easily", in Ramon Marimon and Andrew Scott,
Eds., Computational Methods for the Study of Dynamic Economies, Ox-
ford University Press, Oxford, p.30-61.

Log-linearization techniques (Uhlig’s method)

The Taylor series expansion (linear) gives
e et et e e e e e e et e
eAt + B C
eA+ B C
+ eA+ B C
A A
e e e et e e e e et e
+ eA+ B C
B B eA+ B C
C C

= et + B
1+A et et ;
C

So
e et et et + B
et et
eAt + B C
1+A C
The approximation is
At Bt AB et + B
et et
1+A C
Ct C
Log-linearization techniques (Uhlig’s method)
Some rules from Uhlig
e e et + aYet ;
eXt +aYt 1+X
X et Yet 0;
h i h i
e et+1
Et aeXt+1 a + aEt X
h i
Et [Xt+1 ] = X 1 + Et X et+1

6
Log linear version of Hansen’s model

The …ve equations of the Hansen model are (adjusted)

Ct
1 = Et (rt+1 + (1 ))
Ct+1
Yt
ACt = (1 ) (1 Ht )
Ht
Ct = Yt + (1 )Kt Kt+1
1
Yt = K
t t H t
Yt
rt =
Kt

We will do the log-linearization equation by equation

Log linear version of Hansen’s model

First equation
Ct
1 = Et (rt+1 + (1 ))
Ct+1
" e e
#
CeCt et+1
r CeCt
1 = Et re + (1 )
CeCet+1 CeCet+1
h i
e e e e
= Et reCt Ct+1 +ert+1 + (1 ) eCt Ct+1
h i h i
rEt 1 + C et C et+1 + ret+1 + (1 et
) 1+C et+1
C
h i
= Et 1 + Cet C et+1 + re rt+1 ;

or (after cancelling the 1’s and cleaning up the expections)

0 et
C et+1 + rEt ret+1
Et C

Log linear version of Hansen’s model

Second equation
Yt
ACt = (1 ) (1 Ht )
Ht
e Y Yet et e
ACeCt = (1 ) e H
(1 ) Y eYt
H
et Y
AC 1 + C (1 ) 1 + Yet
Het (1 ) Y 1 + Yet
H
" #
et 1 H Y e Y e
AC C (1 ) Yt (1 ) Ht
H H

7
 given that in the stationary state

1 H Y
AC = (1 )
H

Log linear version of Hansen’s model

This becomes
Y et
(1 )H H
et
C Yet e t = Yet
H
(1 H )Y 1 H
(1 ) H

so
et
H
et
0=C Yet +
1 H
Log linear version of Hansen’s model

The next three equations (in their Log-linear form) are

h i
0 Y Yet C C et + K (1 )Ket K e t+1

0 et + K
e t + (1 et
)H Yet
0 Yet et
K ret

where r = Y =K

Log linear version of Hansen’s model

The stochastic process is

t+1 = t + "t+1

putting in the log di¤erence of the ’s

et+1 et
e = e + "t+1

the linerar approximation is

1 + et+1 = 1 + et + "t+1

So the simple version is

et+1 = et +
t+1

The log-linear version of the model

8
 The equations of the full log-linear model are

0 et
= C et+1 + rEt ret+1
Et C
e
et Yet + Ht
0 = C
1 H
h i
0 = Y Yet C Cet + K (1 )Ket e t+1
K

0 = et + Ke t + (1 et
)H Yet
0 = Yet Ke t ret

and
et+1 = et +
t+1

Solving the log-linear version of the model

n o
The variables of the model are K e t+1 Yet et
C et
H ret plus the sto-
chastic variables t
De…ne the state variables as
h i
et
xt = K

De…ne the "jump" variables as

2 3
Yt
6 Ct 7
yt = 6 7
4 Ht 5
rt

De…ne the stochastic variable as

zt = [ t ]

Solving the log-linear version of the model

The model can be written as

0 = Axt + Bxt 1 + Cyt + Dzt ;

0 = Et [F xt+1 + Gxt + Hxt 1 + Jyt+1 + Kyt + Lzt+1 + M zt ] ;
zt+1 = N zt + "t+1 , Et ("t+1 ) = 0:

Where 2 3 2 3
0 0
6 K 7 6 K( + 1) 7
A=6
4
7 B=6 7
0 5 4 5
0 1

9
 2 1 3
1 1 1 H
0
6 Y C 0 0 7
C=6
4 1
7
0 1 0 5
1 0 0 1
Solving the linear version of the model
2 3
0
6 0 7
D=4 6 7
1 5
0

F = [0] ; G = [0] ; H = [0] ;

J= 0 1 0 r ;
K= 0 1 0 0 ;

L = [0]
M = [0]
N = [ ]

Solving the linear version of the model

We look for a solution of the form

xt = P xt + Qzt
1
yt = Rxt 1 + Szt

1
Note that here C is of full rank and has a well de…ned inverse C
The solutions can be found from

0 = (F JC 1 A)P 2 (JC 1
B G + KC 1
A)P KC 1
B+H
R = C 1 (AP + B);

N0 (F JC 1
A) + Ik (JR + F P + G KC 1
A) vec(Q)
1 1
= vec JC D L)N + KC D M

and
1
S= C (AQ + D)

Explaining the solution

We look for the laws of motion of the model

xt = P xt 1 + Qzt ;
yt = Rxt 1 + Szt :

10
 We begin by substituting the laws of motion into the two equations of the
model
Reduce each equation to one in which there are only two variables:
xt 1 and zt :

Use the stochastic process in the expectational equation to replace

zt+1 = N zt + "t+1

Taking expectations, the "t+1 = 0 disappear

Explaining the solution
Begin with the model
0 = Axt + Bxt 1 + Cyt + Dzt
0 = Et [F xt+1 + Gxt + Hxt 1 + Jyt+1 + Kyt + Lzt+1 + M zt ]

Substitute in
xt = P xt 1 + Qzt ;
yt = Rxt 1 + Szt :

In the …rst equation this gives

0 = A [P xt 1 + Qzt ] + Bxt 1 + C [Rxt 1 + Szt ] + Dzt

Explaining the solution

In the second equation
0 = Et [F [P xt + Qzt+1 ] + G [P xt 1 + Qzt ] + Hxt 1
+J [Rxt + Szt+1 ] + K [Rxt 1 + Szt ] + Lzt+1 + M zt ]

Substitute one more time in the second equation

0 = Et [F [P [P xt 1 + Qzt ] + Q [N zt + "t+1 ]] + G [P xt 1 + Qzt ]
+Hxt 1 + J [R [P xt 1 + Qzt ] + S [N zt + "t+1 ]]
+K [Rxt 1 + Szt ] + L [N zt + "t+1 ] + M zt ]

This simpli…es to (because Et "t+1 = 0) and we remove the expectations

operator
0 = F [P [P xt 1 + Qzt ] + QN zt ] + G [P xt 1 + Qzt ]
+Hxt 1 + J [R [P xt 1 + Qzt ] + SN zt ]
+K [Rxt 1 + Szt ] + LN zt + M zt

11
Explaining the solution
The two equations can be rearranged to give
0 = [AP + B + CR] xt 1 + [AQ + CS + D] zt ;
and
0 = [F P P + GP + H + JRP + KR] xt 1
+ [F P Q + F QN + GQ + JRQ + JSN + KS + LN + M ] zt :

Since these equations need to hold for all xt 1 and zt , it must be that
0 = AP + B + CR
0 = AQ + CS + D
0 = F P P + GP + H + JRP + KR
0 = F P Q + F QN + GQ + JRQ + JSN + KS + LN + M

Explaining the solution

The third equation is
0 = F P 2 + GP + JRP + H + KR
and the …rst is (if the inverse of C exists)
1 1
R= C AP C B

Combining these one gets

0 = F P 2 + GP J C 1
AP + C 1
B P
1 1
+H K C AP + C B
2
0 = FP JC AP + GP JC 1 AP 2 JC 1 BP
1 2

+H KC 1 AP KC 1 B
0 = F JC 1 A P 2 JC 1 B + KC 1 A G P
1
KC B+H

Explaining the solution

Here F is a 1 1 matrix (a scalar)
Finding the solution to the quadratic equation
1
0 = F JC A P2 JC 1
B + KC 1
A G P
1
KC B+H
can be done using
0 = aP 2 + bP + c

12
 The solution to this equation is found from
p
b b2 4ac
P =
2a

There are usually two di¤erent solutions to this problem. We use jP j < 1
in order to choose the stable root.
Once P is known, …nding R is simple using
1 1
R= C AP C B

Explaining the solution

Finding Q (with P and R already known, from above)

Use the equations

0 = F P Q + F QN + GQ + JRQ + JSN + KS + LN + M

and
0 = AQ + CS + D

S can be written as
1 1
S= C AQ C D

Substitute this into the …rst equation

1 1
0 = F P Q + F QN + GQ + JRQ JC AQN JC DN
KC 1 AQ KC 1 D + LN + M

Rearrange to get
1 1
F P + G + JR KC A Q+ F JC A QN
1 1
= JC DN + KC D LN + M

Explaining the solution

This equation
1 1
F P + G + JR KC A Q+ F JC A QN
1 1
= JC DN + KC D LN + M

has Q in two di¤erent places on the left hand side

1
– Q in the …nal position in F P + G + JR KC A Q
1
– Q in the second to the last position in F JC A QN

13
 Need to use a theorem from advanced matrix algebra

Theorem 1 Let A, B, and C be matrices whose dimensions are such that the
product ABC exists. Then

vec(ABC) = (C0 A) vec(B)

where the symbol denotes the Kronecker product.

Explaining the solution

Think of
1 1
F P + G + JR KC A Q+ F JC A QN
1 1
= JC DN + KC D LN + M

as
W QI + XQN = Z
(notice that we added I) where
1
W = F P + G + JR KC A
X = F JC 1 A
Z = JC 1 DN + KC 1 D LN + M

Take vec of both sides of the equation, so

vec (W QI) + vec (XQN ) = vec (Z)

This equals

(I 0 W ) vec (Q) + (N 0 X) vec (Q) = vec (Z)

or
(I 0 W + N0 X) vec (Q) = vec (Z)

If (I 0 W + N0 X) is invertible
1
vec (Q) = (I 0 W + N0 X) vec (Z)

Explaining the solution

What are vec and (the Kronecker product)

14
 First vec 2 3
a11
6 a21 7
6 7
a11 a12 a13 6 a12 7
vec =6
6
7:
7
a21 a22 a23 6 a22 7
4 a13 5
a23

the columns are made into a vector

Explaining the solution

The Kronecker product is

2 3
b11 b12
a11 a12 4 b21 a11 B a12 B
A B = b22 5 =
a21 a22 a21 B a22 B
b31 b32
2 3
a11 b11 a11 b12 a12 b11 a12 b12
6 a11 b21 a11 b22 a12 b21 a12 b22 7
6 7
6 a11 b31 a11 b32 a12 b31 a12 b32 7
= 6
6
7:
7
6 a21 b11 a21 b12 a22 b11 a22 b12 7
4 a21 b21 a21 b22 a22 b21 a22 b22 5
a21 b31 a21 b32 a22 b31 a22 b32

Calibration

Solution to model is numerical

Need values for parameters
Some we borrow from literature (quarterly)

– = :99
– = :025
– = :36

Need a value for A

– Choose A so that H = 1=3

– Use stationary state equation for H
1
H= h i
A
1+ (1 ) 1 1 (1 )

– A = 1:72 for H = :3335

K = 12:6695 and using the production function, Y = 1:2353

15
 r = 1= = 1:0101
From data for US use = :95

Matices for Calibrated model

2 3 3 2
0 0
6 12: 670 7 6 12: 353 7
A=6 4
7
5 B=64 0:36 5
7
0
0 1
2 3
1 1 1:5004 0
6 1:2353 0:9186 0 0 7
C=6 4
7
1 0 :64 0 5
1 0 0 1
2 3
0
6 0 7
D=6 7
4 1 5
0
Matices for Calibrated model

F = [0]
G = [0]
H = [0]

J= 0 1 0 :0348
K= 0 1 0 0

L = [0]
M = [0]
N = [:95]

Numerical solution for model

The quadratic equation gives the solutions

P = 1:0592 and P = 0:9537

The stable value is

P = 0:9537

The value for Q is

Q = 0:1132

16
 The matrices R and S are
2 3 2 3
0:204 5 1: 452 3
6 0:569 1 7 6 0:392 7
R=6 4
7
5 and S= 6
4 0:706 7 5
7
0:243
0:795 5 1: 452 3

Numerical solution for model

The laws of motion are

e t+1
K = 0:9537Ke t + 0:1132et ;
Yet = 0:2045Ke t + 1:4523et ;
Cet = 0:5691Ke t + 0:3920et ;
et
H = 0:2430K e t + 0:7067et ;
ret = 0:7955K e t + 1:4523et :

Recall that et follows the process

et = :95et 1 + t

Two ways of …nding the variances of the variables of the model

Simulations

– Run lots of simulated economies

– Calculate the variances from this "data"

Calculate variances from laws of motion

– See book for detains

Need to calibrate var( t ) so that var(Yet ) = 1:76%

– gets standard error of t = :0032

Tables of second moments

Standard errors as fraction of output

Yet Cet et
H ret Iet
Standard error 5:484 " 4:065 " 1:640 " 3:492 " 11:742 "
As % of output 100% 74:1 2% 29:9 0% 63:6 7% 214:1%
Standard errors from the data
Yet Cet et
H Iet
As % of output 100% 73:30% 94:3 2% 4 88:6 4%

17
 Does well for consumption
Badly for hours worked and investment

% stationary state values are found in another program

A=[0 -kbar 0 0]’;
B=[0 (1-delta)*kbar theta -1]’;
C=[1 -1 -1/(1-hbar) 0
ybar -cbar 0 0
-1 0 1-theta 0
1 0 0 -1];
D=[0 0 1 0]’;
F=[0];
G=F;
H=F;
J=[0 -1 0 beta*rbar];
K=[0 1 0 0];
L=F;
M=F;
N=[.95];
Cinv=inv(C);
a=F-J*Cinv*A;
b=-(J*Cinv*B-G+K*Cinv*A);
c=-K*Cinv*B+H;
P1=(-b+sqrt(b^2-4*a*c))/(2*a);
P2=(-b-sqrt(b^2-4*a*c))/(2*a);
if abs(P1)<1
P=P1;
else
P=P2;
end
R=-Cinv*(A*P+B);
Q=(J*Cinv*D-L)*N+K*Cinv*D-M;
QD=kron(N’,(F-J*Cinv*A))+(J*R+F*P+G-K*Cinv*A);
Q=Q/QD;
S=-Cinv*(A*Q+D);
Hansen’s model with indivisible labor

Objective: increase variance of hours worked

Make labor indivisible

– one works X hours per week or not at all

Add unemployment

– since some fraction of the population will not be working

18
Problem of non-convexity of consumption set

In general, maximization is only valid over convex sets

Def of a convex set

– straight lines between any two points in set are also in set

Example of a non-convex set

How non-convexity is …xed in Hansen’s model

The problem is the jump in income

– between working and not working

Hansen invented an "unemployment insurance"

Lump sum transfers that make income equal for all

– solves non-convexity problem

– consumption increases smoothly with wage
– since all receive same income (based on wages)
– solve problem of too much heterogenity

Household problem

maximize
1
X
t
max u(ct ; t)
t=0

subject to
ct + it = wt ht + rt kt
t = probability in time t of supplying h0 units of labor

19
 Expected utility

A ln(1 h0 ) ht
u(ct ; t) = ln ct + ht + A(1 ) ln(1)
h0 h0
A ln(1 h0 )
u(ct ; t) = ln ct + ht
h0

Household problem

Maximization problem becomes

1
X
t
max [ln ct + Bht ]
t=0

with
A ln(1 h0 )
B=
h0
subject to constraints
1
t kt ht = ct + kt+1 (1 )kt

and
ln t+1 = ln t + "t+1

Household problem

First order conditions

1
0 = (1 ) t kt ht + B;
ct
1 1 1 1
0 = + Et t+1 kt ht + (1 )
ct ct+1

With equilibrium condition, these simplify to

Ct
1 = Et (rt+1 + (1 )) ;
Ct+1
(1 ) Yt
Ct = :
BHt

Full model

Ct
1 = Et (rt+1 + (1 ))
Ct+1
(1 ) Yt
Ct =
BHt

20
 Ct + Kt+1 = Yt + (1 )Kt
1 1
rt = K
t t H t
1
Yt = t K t Ht

Stationary state

Equations

1
= r + (1 )

(1 )Y
C =
BH
1 1
r = K H
1
Y = K H
C = Y K

solve to give
1
(1 ) 1
H= and K= H
B 1 1 + (1 )
1 (1 )

Stationary state
Comparing to basic Hansen model

To get same stationary state, need H the same in both cases

Then other variables will be the same
Old stationary state equation
1
H= h i
A
1+ (1 ) 1 1 (1 )

Set the two equal

1 (1 )
h i= ;
A A ln(1 h0 )
1+ (1 ) 1 1 (1 ) h0 1 1 (1 )

A ln(1 h0 )
We replaced B with h0

Need to determine h0 that make the two SS the same

21
Stationary state

Solve to get
h i
A
h0 (1 ) 1 1 (1 )
= h i =G
ln(1 h0 ) 1+ A
1
(1 ) 1 (1 )

G is a constant
To …nd h0
Get h0 = :583, = :573, and H = :3335

Log-linear model

Taking the log-linear approximation of the model gives

0 Cet Et C et+1 + rEt ret+1

0 Cet + H
e t Yet
0 Y Yet C C et + (1 )K Ket K K
e t+1
0 Yet et Ke t (1 )Het
0 Yet et
K ret

Solution method

Use Uhlig’s method

0 = Axt + Bxt 1 + Cyt + Dzt ;

0 = Et [F xt+1 + Gxt + Hxt 1 + Jyt+1 + Kyt + Lzt+1 + M zt ] ;
zt+1 = N zt + "t+1 , Et ("t+1 ) = 0;
h i h i0 h i
e t , yt = Yet ; C
where, xt = K e t ; ret , and zt = et
et ; H

22
 Solve for

xt = P xt 1 + Qzt
yt = Rxt 1 + Szt

2 3 2 3
0 0
6 K 7 6 K( + 1) 7
A=6
4
7 B=6 7
0 5 4 5
0 1
2 3 2 3
1 1 1 0 0
6 Y C 0 0 7 6 0 7
C=6
4 1
7 D=6 7
0 (1 ) 0 5 4 1 5
1 0 0 1 0
F = [0] ; G = [0] ; H = [0]
J= 0 1 0 r K= 0 1 0 0
L = [0] ; M = [0] ; and N = [ ] :
Results

The linear policy functions are

e t+1 = :9418K
K e t + :1552 t

and
et + S
yt = R K t

where

2 3 2 3
0:055 1: 941 8
6 0:531 6 7 6 0:470 3 7
R=6 7
4 0:476 6 5 S=6 7
4 1: 471 5 5
0:945 1: 941 7
Results

Using this model, we calculate the variances of the variables

Yet et
C et
H ret Iet
Standard errors 6:431 " 4:081 " 3:444 " 4:514 " 15:722 "
As % of output 100% 63:4 6% 53:5 5% 70:1 9% 244:5%
Increased variance in hours worked
Slight increase in investment
Lower variance in consumption (compared to data)

23
 0.01

0.008

technology
0.006

0.004

0.002

0
0 10 20 30 40 50 60 70 80 90 100
periods

Impulse response functions

How does the economy respond to a one time shock to technology

"t = 0 except "2 = :01 Recall that = :95
et = et 1 + "t

Response of technology (path of et )

Impulse response functions

e 1 = 0 using
Then calculate the time path of capital with K
e t + Qet
e t+1 = P K
K

e t . Use this to …nd path of other variables using

get path of K
et + S
yt = R K t

Impulse response functions: Basic Hansen model

Impulse response functions: Hansen with indivisible labor
Comparing impulse responses

Both models get same impulse

Put each set of responses on di¤erent axis
Get

Comparing impulse response

Rotating so that we don’t see the time axis

24
 -3
x 10
20

15
Y

K
10
C

r
-5
0 10 20 30 40 50 60 70 80 90 100
periods

Figure 1: Responses of Hansen’s basic model

-3
x 10
20

15 Y

10
C

-5
0 10 20 30 40 50 60 70 80 90 100
periods

Figure 2: Responses for Hansen’s model with indivisible labor

25
 0.02
K C
0.015 Y

Indivisible labor model

0.01

0.005

100 r
-0.005 H
50 0.015 0.02
0.005 0.01
-0.01 -0.005 0
-0.01
0
Basic model

Figure 3: Responses for both Hansen models

-3
x 10
20

15
response of model with indivisible laobr

H
45°
Y
10

K
5

-5
-5 0 5 10 15 20
response of model w ith divisible labor -3
x 10

Figure 4: Comparing the response of the two models

26