The solution strategy

Hansens benchmark Real Business Cycle Model

The solution steps
Representations

Solving DSGE models: an example.

Hansens Real Business Cycle Model
IAMA, Lecture 5

Prof. H. Uhlig1

1 Humboldt Universität zu Berlin

[email protected]

Winter 2006/07

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
The solution steps
Representations

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 The solution strategy
Hansens benchmark Real Business Cycle Model
Overview
The solution steps
Representations

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 The solution strategy
Hansens benchmark Real Business Cycle Model
Overview
The solution steps
Representations

The solution strategy

The solution strategy for a model works as follows:

1. Find the first order necessary conditions

2. Calculate the steady state
3. Loglinearize around the steady state
4. Solve for the recursive law of motion
5. Calculate impulse responses and
(HP-filtered) moments

We will execute this strategy, using Hansens real business

cycle model as particular example.

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Hansens benchmark Real Business Cycle Model

" ∞
#
X
max E β t (log ct − Ant )
t=0

s.t.
ct + kt = γ̄ezt kt−1
θ
nt1−θ + (1 − δ)kt−1
and
zt = ρzt−1 + ǫt , ǫt ∼ N(0, σ 2 ) i.i.d.
where ct is consumption, nt is labor, kt is capital, γt = γ̄ezt is
total factor productivity (TFP).

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Hansens benchmark Real Business Cycle Model

Define, for convenience;

output: yt = γ̄ezt kt−1

θ
nt1−θ
yt
return: Rt = θ +1−δ
kt−1

See:
1 Hansen, G., “Indivisible Labor and the Business Cycle,”
Journal of Monetary Economics, 1985, 16, 309-27.
2 Cooley, editor, Frontiers of Business Cycle Research,
Princeton University Press, 1995.

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Rational expectations

We assume that the social planner chooses ct , kt , nt etc.,

using all available information at date t, and forming
rational expectations about the future.

Rational expectations are the mathematical

expectations, using all available information
Rational expectations only ”live in” a model, in which the
stochastic nature of all variables is clearly spelled out.

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Rational expectations

Example: dice role.

Dice 1, date t: Xt . Dice 2, date t + 1: Yt+1 . Sum:
St+1 = Xt + Yt+1 .
Et−1 [St+1 ] = 7. Et [St+1 ] = 3.5 + Xt .
Et+1 [St+1 ] = Xt + Yt+1 .
E.g. Xt = 2, Yt+1 = 1. Then Et−1 [St+1 ] = 7, Et [St+1 ] = 5.5,
Et+1 [St+1 ] = 3.
Example: AR(1)
zt+1 = ρzt + ǫt+1 , Et [ǫt+1 ] = 0.
Then: Et [zt+1 ] = ρzt .

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Labor lotteries and labor supply

We assume a very elastic labor supply for aggregate labor

nt ,
ut = log(ct ) − Ant
... which turns out to be needed in order to quantitatively
explain observed employment fluctuations.
However, we typically imagine individual labor elasticity to
be small.
This can be true simultaneously by considering labor
lotteries.
Source: Richard Rogerson, ”Indivisible Labor, Lotteries
and Equilibrium,” Journal of Monetary Economics; 21(1),
January 1988, 3-16.
Prof. H. Uhlig IAMA: Lecture 5
 The solution strategy
The model
Hansens benchmark Real Business Cycle Model
Rational expectations
The solution steps
Labor supply
Representations

Labor lotteries and labor supply

Individual labor supply ñt may be based on some utility
function u(ct ) + v (ñt ).
Suppose that
labor is indivisible: agents either have a job or do not,
ñt = 0 or ñt = n∗ .
Agents are assigned to jobs according to a lottery, with
probability πt .
Shirking, moral hazard etc. are not possible.
Unemployment insurance is perfect, and consumption ct is
independent of job status.
Total labor supplied: nt = πt n∗
Normalization: v (0) = 0, v (n∗ )/n∗ =: −A < 0.
Expected utility:
E[u(ct ) + v (ñt )] = u(ct ) + πt v (n∗ ) = u(ct ) − Ant
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Step 1:
Find the first-order necessary conditions (FONCS)

Form the Lagrangian

"∞
X
L = max E β t ((log ct − Ant )
t=0
  i
−λt ct + kt − γ̄ezt kt−1
θ
nt1−θ − (1 − δ)kt−1 )

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Find the first-order necessary conditions...

Differentiate:
∂L 1
: = λt
∂ct ct
∂L yt
: A = λt (1 − θ)
∂nt nt
∂L
: ct + kt = γ̄ezt kt−1
θ
nt1−θ + (1 − δ)kt−1
∂λt
∂L
: λt = βEt [λt+1 Rt+1 ]
∂kt
The last equation needs explanation.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Differentiating with respect to kt

Write out the objective at date t: for the future, one can
only form conditional expectations Et [·]. ”Telescope” out
the Lagrangian:
L = . . . + β t ((log ct − Ant )
 
−λt ct + kt − γ̄ezt kt−1 θ
nt1−θ − (1 − δ)kt−1
h
+Et β t+1 ((log ct+1 − Ant+1 )
 i
1−θ
−λt+1 ct+1 + kt+1 − γ̄ezt+1 ktθ nt+1 − (1 − δ)kt + ...
Differentiate with respect to kt :
  
t t+1 yt+1
0 = β λt − Et β λt+1 θ +1−δ
kt
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Differentiating with respect to kt

Sort terms and use

yt+1
Rt+1 = θ +1−δ
kt
to find
λt = βEt [λt+1 Rt+1 ]
This equation is called an Euler equation and also the
Lucas asset pricing equation.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Collecting equations
1 First order conditions and a definition:
1
= λt
ct
yt
A = λt (1 − θ)
nt
yt
Rt = θ +1−δ
kt−1
λt = βEt [λt+1 Rt+1 ]
2 Technology and Feasibility constraints:
yt = γ̄ezt kt−1
θ
nt1−θ
ct + kt = yt + (1 − δ)kt−1
zt = ρzt−1 + ǫt , ǫt ∼ N(0, σ 2 ) i.i.d.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Step 2: Calculate the steady state

At the steady state, all variables are constant.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Take all equations ...

1 First order conditions and a definition:
1
= λt
ct
yt
A = λt (1 − θ)
nt
yt
Rt = θ +1−δ
kt−1
λt = βEt [λt+1 Rt+1 ]
2 Technology and Feasibility constraints:
yt = γ̄ezt kt−1
θ
nt1−θ
ct + kt = yt + (1 − δ)kt−1
zt = ρzt−1 + ǫt , ǫt ∼ N(0, σ 2 ) i.i.d.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

... and drop the time subscripts.

1 First order conditions and a definition:
1
= λ̄
c̄
ȳ
A = λ̄(1 − θ)
n̄
ȳ
R̄ = θ +1−δ
k̄
λ̄ = β λ̄R̄
2 Technology and Feasibility constraints:
ȳ = γ̄ez̄ k̄ θ n̄1−θ
c̄ + k̄ = ȳ + (1 − δ)k̄
z̄ = ρz̄
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Parameters

1 Calibration: θ = 0.4, δ = 0.012, ρ = 0.95, σǫ = 0.007,

β = 0.987, γ̄ = 1, A so that n̄ = 1/3 (see Cooley,
Frontiers...).
2 Estimation:
1 GMM: mimics calibration, see Christiano and Eichenbaum,
“Current Real-Business Cycle Theories and Aggregate
Labor Market Fluctuations,” American Economic Review,
vol 82, no. 3, 430 - 450.
2 Maximum Likelihood: see e.g. Leeper and Sims, “Toward a
Modern Macroeconomic Model Usable for Policy Analysis,”
NBER Macroeconomics Annual, 1994, 81 - 177.
With numbers for the parameters, the steady state can be
calculated explicitely.
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Explicit calculation

From the production function,

ȳ = γ̄ez̄ k̄ θ n̄1−θ

we get
1
 −θ ! 1−θ
ȳ
z̄
ȳ = γ̄e n̄
k̄

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Explicit calculation: n̄ given, solve for A.

1
1. R̄ = β
5. c̄ = ȳ − δ k̄

ȳ R̄−1+δ 1
2. k̄
= θ
6. λ̄ = c̄

 1
 −θ  1−θ
ȳ
3. ȳ = γ̄e z̄
k̄
n̄ 7. A = λ̄(1 − θ) ȳn̄

 −1
ȳ
4. k̄ = k̄
ȳ

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Explicit calculation alternative: A given, solve for n̄.

1 c̄ ȳ
1. R̄ = β
5. k̄
= k̄
−δ

ȳ R̄−1+δ 1
2. k̄
= θ
6. c̄ = λ̄

 1
 −θ  1−θ
ȳ ȳ c̄
3. = γ̄ez̄ 7. k̄ =
n̄ k̄ ( k̄c̄ )
 
A ȳ
4. λ̄ = (1−θ)( ȳn̄ )
8. ȳ = k̄
k̄

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Step 3: Loglinearize around the steady state

Replace the dynamic nonlinear equations by dynamic

linear equations.
Interpretation and calculation are made easier, if the
equations are linear in percent deviations from the steady
state.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

The Principle of Loglinearization

For x ≈ 0,
ex ≈ 1 + x
For xt , let x̂t = log(xt /x̄) be the log-deviation of xt from its
steady state. Thus, 100 ∗ x̂t is (approximately) the percent
deviation of xt from x̄. Then,

xt = x̄ex̂t ≈ x̄(1 + x̂t )

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Application of Loglinearization

Application: The equation

xt + ct = yt

together with its steady state version

x̄ + c̄ = ȳ

deliver the dynamic relationship

x̄ x̂t + c̄ ĉt = ȳ ŷt

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Example: RBC

Do it slowly for two equations:

The resource constraint:

ct + kt = yt + (1 − δ)kt−1
c̄eĉt + k̄ ek̂t = ȳ eŷt + (1 − δ)k̄ ek̂t−1
c̄(1 + ĉt ) + k̄ (1 + k̂t ) ≈ ȳ (1 + ŷt ) + (1 − δ)k̄ (1 + k̂t−1 )
(Note: c̄ + δ k̄ = ȳ )
c̄ ĉt + k̄ k̂t ≈ ȳ ŷt + (1 − δ)k̄ k̂t−1

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Example: RBC

The asset pricing equation:

λt = βEt [λt+1 Rt+1 ]

h i
λ̄eλ̂t = βEt λ̄R̄eλ̂t+1 +R̂t+1
h i
1 + λ̂t ≈ β R̄Et 1 + λ̂t+1 + R̂t+1
( Note: 1 = β R̄)
h i
λ̂t ≈ Et λ̂t+1 + R̂t+1

On “ignored” Jensen terms: can also assume joint

normality of logdeviations insteady. This changes the
steady state, not the dynamics.
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

All loglinearized equations

# Equation Loglinearized
1
(i) ct = λt 0 = ĉt + λ̂t

(ii) A = λt (1 − θ) nytt 0 = λ̂t + ŷt − n̂t

 
yt
(iii) Rt = θ kt−1 +1−δ 0 = −R̄ R̂t + θ ȳk̄ ŷt − k̂t−1

(iv) yt = γ̄ezt kt−1

θ
nt1−θ 0 = −ŷt + zt + θk̂t−1 + (1 − θ)n̂t

(v) ct + kt = yt + (1 − δ)kt−1 0 = −c̄ ĉt − k̄ k̂t + ȳ ŷt + (1 − δ)k̄ k̂t−1

(vi) λt = βEt [λt+1 Rt+1 ] 0 = −λ̂t + Et [λ̂t+1 + R̂t+1 ]

(vii) zt+1 = ρzt + ǫt+1 zt+1 = ρzt + ǫt+1

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Recursivity

State variables are: kt−1 , zt (or, alternatively, kt−1 and

zt−1 ).
The dynamics of the model should be describable by a
recursive law of motion (RLOM),

λt = f(λ) (kt−1 , zt )
kt = f(k ) (kt−1 , zt )
yt = f(y ) (kt−1 , zt )

etc.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Recursivity

Assume that the RLOM is linear in the log-deviations,

λ̂t = ηλk k̂t−1 + ηλz zt

k̂t = ηkk k̂t−1 + ηkz zt
ŷt = ηyk k̂t−1 + ηyz zt

etc. for coefficients ηλk , ηλz , etc.

To make life simpler here, we shall try to reduce the system
to only k and λ (one doesn’t have to).

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Simplify:
Note: ŷt = 1θ zt + k̂t−1 + 1−θ
θ λ̂t
Abbreviations:
ȳ
α1 = + (1 − δ)
k̄
c̄ 1 − θ ȳ
α2 = +
k̄ θ k̄
ȳ
α3 =
θk̄
α4 = 0
ȳ
α5 = 1 + (1 − θ)
R̄ k̄
ȳ
α6 =
R̄ k̄
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Obtaining the solution

We obtain the following first-order two-dimensional

stochastic difference equation:

0 = −k̂t + α1 k̂t−1 + α2 λ̂t + α3 zt (1)

0 = Et [−λ̂t + α4 kt + α5 λ̂t+1 + α6 zt+1 ] (2)
zt = ρzt−1 + ǫt (3)

where zt is an exogenous stochastic process.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Obtaining the solution

Compare to the following first-order two-dimensional

stochastic difference equation to be studied in the lecture
on difference equations:

0 = −xt + α1 xt−1 + α2 yt + α3 zt (4)

0 = Et [−yt + α4 xt + α5 yt+1 + α6 zt+1 ] (5)
zt = ρzt−1 + ǫt (6)

They are the same with xt = k̂t , yt = λ̂t .

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

The Method of Undetermined Coefficients

Postulate the recursive law of motion
λ̂t = ηλk k̂t−1 + ηλz zt (7)
k̂t = ηkk k̂t−1 + ηkz zt (8)
Plug this into equations (1) once and (2) “twice” and exploit
Et [zt+1 ] = ρzt , so that only the date-t-states k̂t−1 and zt
remain,
0 = (−ηkk + α1 + α2 ηλk ) k̂t−1
+ (−ηkz + α2 ηλz + α3 ) zt
0 = (−ηλk + α4 ηkk + α5 ηλk ηkk ) k̂t−1
+ (−ηλz + α4 ηkz + α5 ηλk ηkz + (α5 ηλz + α6 ) ρ) zt
Compare coefficients
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

On plugging in twice...
Plugging

λ̂t = ηλk k̂t−1 + ηλz zt , k̂t = ηkk k̂t−1 + ηkz zt ρzt = Et [zt+1 ]

twice into

0 = Et [−λ̂t + α4 k̂t + α5 λ̂t+1 + α6 zt+1 ]

= Et [−(ηλk k̂t−1 + ηλz zt ) + α4 (ηkk k̂t−1 + ηkz zt )
+α5 (ηλk k̂t + ηλz zt+1 ) + α6 zt+1 ]
= Et [−ηλk k̂t−1 − ηλz zt + α4 ηkk k̂t−1 + α4 ηkz zt )
+α5 ηλk (ηkk k̂t−1 + ηkz zt ) + (α5 ηλz + α6 )zt+1 ]
= (−ηλk + α4 ηkk + α5 ηλk ηkk ) k̂t−1
+ (−ηλz + α4 ηkz + α5 ηλk ηkz + (α5 ηλz + α6 ) ρ) zt

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Comparing coefficients

On k̂t−1 :

0 = −ηkk + α1 + α2 ηλk
0 = −ηλk + α4 ηkk + α5 ηλk ηkk

One gets the characteristic quadratic equation

 
2 α2 1 α1
0 = p(ηkk ) = ηkk − α1 − α4 + ηkk + (9)
α5 α5 α5

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Solving the characteristic equation

Solutions:
 
1 α 1
ηkk = α1 − 2 α4 + (10)
2 α5 α5
s 
 2
α 1 α
± α1 − 2 α4 + −4 1 
α5 α5 α5

Choose the stable root | ηkk |< 1. There is at most one stable
root, if
α
| ηkk ,1 ηkk ,2 |=| 1 |> 1
α5
With ηkk , calculate
ηλλ − α1
ηλk =
α2
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Comparing coefficients

On zt :

0 = −ηkz + α2 ηλz + α3
0 = −ηλz + α4 ηkz + α5 ηλk ηkz + (α5 ηλz + α6 ) ρ

Solution:
α4 α3 + α5 ηλk α3 + α6 ρ
ηλz =
1 − α4 α2 − α5 ηλk α2 − α5 ρ
ηkz = α2 ηλz + α3

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Step 5: Calculate impulse responses

and (HP-filtered) moments

Impulse responses: will be explained now.

HP-filtered moments: will be discussed later.

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Impulse Response Functions: response to a shock in

zt

1 Set z0 = 0, ǫ1 = 1, ǫt = 0, t > 1
2 Calculate zt = ρt
3 Set k̂0 = 0.
4 Calculate recursively

k̂t = ηkk k̂t−1 + ηkz zt

5 With that, calculate

λ̂t = ηλk k̂t−1 + ηλz zt

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Results: Impulse Responses to shocks

Impulse responses to a shock in technology
2
output
Percent deviation from steady state

1.5 labor

1 technology capital
consumption
0.5

0 interest

−0.5
−2 0 2 4 6 8
Years after shock

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Impulse Response Functions: response to an initial

deviation of the state kt from its steady state.

1 Set zt = 0, t ≥ 1
2 Set k̂0 = 1.
3 Calculate recursively

k̂t = ηkk k̂t−1

4 With that, calculate

λ̂t = ηλk k̂t−1

Prof. H. Uhlig IAMA: Lecture 5

 Step 1: find the FONCs
The solution strategy
Step 2: Calculate the steady state
Hansens benchmark Real Business Cycle Model
Step 3: Loglinearize
The solution steps
Step 4: Solve for the RLOM
Representations
Step 5: Calculate impulse responses

Results: Impulse Responses to capital deviations

Impulse responses to a one percent deviation in capital
1
capital
Percent deviation from steady state

0.5 consumption

0
output
interest

−0.5 labor
−2 0 2 4 6 8
Years after shock

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Outline
1 The solution strategy
Overview
2 Hansens benchmark Real Business Cycle Model
The model
Rational expectations
Labor supply
3 The solution steps
Step 1: find the FONCs
Step 2: Calculate the steady state
Step 3: Loglinearize
Step 4: Solve for the RLOM
Step 5: Calculate impulse responses
4 Representations
Alternative representations
Prof. H. Uhlig IAMA: Lecture 5
 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Recall: the loglinearized equations

# Equation Loglinearized
1
(i) ct = λt 0 = ĉt + λ̂t

(ii) A = λt (1 − θ) nytt 0 = λ̂t + ŷt − n̂t

 
yt
(iii) Rt = θ kt−1 +1−δ 0 = −R̄ R̂t + θ ȳk̄ ŷt − k̂t−1

(iv) yt = γ̄ezt kt−1

θ
nt1−θ 0 = −ŷt + zt + θk̂t−1 + (1 − θ)n̂t

(v) ct + kt = yt + (1 − δ)kt−1 0 = −c̄ ĉt − k̄ k̂t + ȳ ŷt + (1 − δ)k̄ k̂t−1

(vi) λt = βEt [λt+1 Rt+1 ] 0 = −λ̂t + Et [λ̂t+1 + R̂t+1 ]

(vii) zt+1 = ρzt + ǫt+1 zt+1 = ρzt + ǫt+1

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

A representation of the problem

There is an endogenous state vector xt , size m × 1, a list of

other endogenous variables yt , size n × 1, and a list of
exogenous stochastic processes zt , size k × 1. The equilibrium
relationships between these variables are

0 = Axt + Bxt−1 + Cyt + Dzt (11)

0 = Et [Fxt+1 + Gxt + Hxt−1 + Jyt+1 + Kyt + Lzt+1 + Mzt ]
zt+1 = Nzt + ǫt+1 ; Et [ǫt+1 ] = 0,

where it is assumed that C is of size l × n, l ≥ n and of rank n,

that F is of size (m + n − l) × n, and that N has only stable
eigenvalues.

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Example: RBC

Variables:
   
Lagrangian λ̂t

 consumption  
  ĉt 

xt = [capital] = [k̂t ], yt = 
 output =
  ŷt 

 labor   n̂t 
interest R̂t

and
zt = [technology] = [zt ]

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Example: RBC

Matrices:
0 1 1 0 0 0
     
0
 0   0   1 0 1 −1 0 
−θ ȳk̄ 0 0 θ ȳk̄
     
A=  0
,B = 
 
,C = 
  0 −R̄ 

 0   θ   0 0 −1 (1 − θ) 0 
−k̄ (1 − δ)k̄ 0 −c̄ ȳ 0 0

and
F = [0], G = [0], H = [0], J = [1, 0, 0, 0, 1],
K = [−1, 0, 0, 0, 0], L = [0], M = [0], N = [ρ]

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Alternative representations 1

Redefine the system as

     
xt 0 0 A C
x̃t = , F̃ = , G̃ = ,
yt F J G K
     
B 0 0 D
H̃ = , L̃ = , M̃ = ,
H 0 L M

The system can then be rewritten as a second-order

stochastic matrix difference equation,
h i
0 = Et F x̃t+1 + G̃x̃t + H̃ x̃t−1 + L̃zt+1 + M̃zt
zt = Nzt−1 + ǫt ; Et−1 [ǫt ] = 0 + ǫ̃t

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Alternative representations 2
Redefine the system as
   
xt 0
x̃t =  yt , ǫ̃t =  0 ,
zt ǫt
     
0 0 0 A C D B 0 0
F̃ =  F J L , G̃ =  G K M , H̃ =  H 0 0 
0 0 0 0 0 −Ik 0 0 N

The system can then be rewritten as a second-order

stochastic matrix difference equation,
h i
0 = Et F x̃t+1 + G̃x̃t + H̃ x̃t−1 + ǫ̃t

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Alternative Representations 3

E.g. per stacking,  

x̃t
x̌t =
x̃t−1
etc., one can even rewrite the system as a first-order
stochastic matrix difference equation,

0 = Et [F x̌t+1 + Gx̌t ] + ǫ̌t

Here, one needs to keep in mind, that some entries in xt

are predetermined, i.e. already fixed as of t − 1.
This representation is often used, e.g. in Blanchard-Kahn,
Farmer, many others.

Prof. H. Uhlig IAMA: Lecture 5

 The solution strategy
Hansens benchmark Real Business Cycle Model
Alternative representations
The solution steps
Representations

Various Representations 4

Various representations appear in the literature.

Which representation is most convenient? That depends
on the solution approach.
The “complicated” first representation has the advantage
of focussing on a small numer of state variables.

Prof. H. Uhlig IAMA: Lecture 5