Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Advanced Macroeconomics
Real Business Cycle Models

Francesco Lamperti1,2

Email: [email protected]

1 Sant’Anna School of Advanced Studies

2 RFF-CMCC European Institute on Economics and the Environment

April 21, 2021

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Readings

Romer (2018), Advanced Macroenomics, 5th edition, McGraw-Hill, ch. 5.4-5.7

King, R. and Rebelo, S. (1999), Resuscitating Real Business Cycles, in Taylor, J.

and Woodford, M., (eds), Handbook of Macroeconomics, Elsevier Science:
Amsterdam.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Outline

1 Introduction

2 A Simple RBC Model

3 The RBC framework

4 Solving the full model

5 Discussion
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Real Business Cycle Theory

Real Business Cycle (RBC) theory was initiated by Kydland and Prescott (1982)1 , who received the
“Nobel” prize in 2004.

Edward Prescott (left) and Finn Kydland (right)

“I look back on this paper [Time to Build and Aggregate Fluctuations] with amazement: How
did [Kydland and Prescott] even think to put all these pieces together in just this way?” [Lucas, 2005]
1
Kydland, F. E., & Prescott, E. C. (1982). Time to build and aggregate fluctuations. Econometrica:
Journal of the Econometric Society, 1345-1370.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Real Business Cycle Theory

RBC models depart from traditional macro in three ways
1 Business cycle fluctuations of real variables can largely be accounted for by real
shocks (e.g. innovations, wars, oil crisis)
1 In the extreme case without nominal frictions, there is no role for nominal shocks
(i.e. not need of nominal variables - such as money - in models of business cycles)
2 This is in contrast to traditional Keynesian macroeconomics in which nominal
shocks have real effects due to nominal frictions

2 A model used to explain growth also explains cycles (in traditional

macroeconomics, growth and cycles are studied using different models)

3 Fluctuations (even recessions) are optimal (i.e. they results from the optimal
reaction of agents to unanticipated shocks in a Walrasian world)

RBC models remain very controversial (see Rebelo, 20052 )

2
Rebelo, S. (2005). Real Business Cycle Models: Past, Present and Future. The Scandinavian
Journal of Economics, 107(2), 217-238.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Real Business Cycle Theory

RBC models depart from traditional macro in three ways
1 Business cycle fluctuations of real variables can largely be accounted for by real
shocks (e.g. innovations, wars, oil crisis)
1 In the extreme case without nominal frictions, there is no role for nominal shocks
(i.e. not need of nominal variables - such as money - in models of business cycles)
2 This is in contrast to traditional Keynesian macroeconomics in which nominal
shocks have real effects due to nominal frictions

2 A model used to explain growth also explains cycles (in traditional

macroeconomics, growth and cycles are studied using different models)

3 Fluctuations (even recessions) are optimal (i.e. they results from the optimal
reaction of agents to unanticipated shocks in a Walrasian world)

RBC models remain very controversial (see Rebelo, 20052 )

2
Rebelo, S. (2005). Real Business Cycle Models: Past, Present and Future. The Scandinavian
Journal of Economics, 107(2), 217-238.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Real Business Cycle Theory

RBC models depart from traditional macro in three ways
1 Business cycle fluctuations of real variables can largely be accounted for by real
shocks (e.g. innovations, wars, oil crisis)
1 In the extreme case without nominal frictions, there is no role for nominal shocks
(i.e. not need of nominal variables - such as money - in models of business cycles)
2 This is in contrast to traditional Keynesian macroeconomics in which nominal
shocks have real effects due to nominal frictions

2 A model used to explain growth also explains cycles (in traditional

macroeconomics, growth and cycles are studied using different models)

3 Fluctuations (even recessions) are optimal (i.e. they results from the optimal
reaction of agents to unanticipated shocks in a Walrasian world)

RBC models remain very controversial (see Rebelo, 20052 )

2
Rebelo, S. (2005). Real Business Cycle Models: Past, Present and Future. The Scandinavian
Journal of Economics, 107(2), 217-238.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Real Business Cycle Theory

RBC models depart from traditional macro in three ways
1 Business cycle fluctuations of real variables can largely be accounted for by real
shocks (e.g. innovations, wars, oil crisis)
1 In the extreme case without nominal frictions, there is no role for nominal shocks
(i.e. not need of nominal variables - such as money - in models of business cycles)
2 This is in contrast to traditional Keynesian macroeconomics in which nominal
shocks have real effects due to nominal frictions

2 A model used to explain growth also explains cycles (in traditional

macroeconomics, growth and cycles are studied using different models)

3 Fluctuations (even recessions) are optimal (i.e. they results from the optimal
reaction of agents to unanticipated shocks in a Walrasian world)

RBC models remain very controversial (see Rebelo, 20052 )

2
Rebelo, S. (2005). Real Business Cycle Models: Past, Present and Future. The Scandinavian
Journal of Economics, 107(2), 217-238.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Real Business Cycle (RBC) Approach

They share the same main ingredients of the new “Classical” school

However, business cycles are originated by exogenous, supply (technological)

shocks

Growth and cycles are inseparably interrelated (deterministic vs. stochastic

trends)

Business cycles are efficient: they emerge as the aggregate outcome of

maximizing decisions made by all the agents populating an economy.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Real Business Cycle (RBC) Approach

RBC models are simply stochastic neoclassical growth (i.e. Ramsey) models
with endogenous labor supply

The model is Walrasian (no imperfections, no heterogeneity): fluctuations are

optimal. Expansions, booms and even crises and recessions reflect no market
failures, each stage of the business cycle is an equilibrium.

The model is real: money is not even present

Methodological innovations: calibration, simulation, stylized-fact replication

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Main issue: can growth be separated by fluctuations?

The conventional approach (Keynesians, monetarists, new classicals) has been to
imagine that the economy fluctuates along a path reflecting an underlying trend rate
of growth described by the Solow model → output deviations from trend are
temporary
Nelson and Plosser (1982): real factors are behind aggregate fluctuations → business
cycles should not be viewed as temporary events → recessions might have permanent
effect on GDP
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Nelson and Plosser reached their important conclusion because in their research into US
data they were unable to reject the hypothesis that GNP follows a random walk.

Trend stationary model Random walk model

Yt = g + bYt−1 + zt Yt = g + Yt−1 + zt

g > 0 is a deterministic trend g > 0 is the drift of output

0 < b < 1 imposes mean-reversal b=1
The impact of a shock zt on output dies The impact of the shock on output
out. does no die out.

The identification of unit roots is assumed to be a manifestation of shocks to the

production function
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Nelson and Plosser reached their important conclusion because in their research into US
data they were unable to reject the hypothesis that GNP follows a random walk.

Trend stationary model Random walk model

Yt = g + bYt−1 + zt Yt = g + Yt−1 + zt

g > 0 is a deterministic trend g > 0 is the drift of output

0 < b < 1 imposes mean-reversal b=1
The impact of a shock zt on output dies The impact of the shock on output
out. does no die out.

The identification of unit roots is assumed to be a manifestation of shocks to the

production function
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Implications:
If shocks to productivity growth due to technological change are frequent and
random, then the path of output following a random walk will exhibit features that
resemble a business cycle

→ The observed fluctuations in GNP are fluctuations in the natural (trend) rate of
output, not deviations of output from a smooth deterministic trend

What looks like output fluctuating around a smooth trend is in fact fluctuations in
the natural rate of output induced by a series of permanent shocks, with each
permanent productivity shock determining a new growth path.

→ The economic forces determining the trend are not different from those causing
fluctuations
The main source of fluctuations are real disturbances!
By ending the distinction between trend and cycle, real business cycle theorists began to integrate
the theory of growth and fluctuations
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Implications:
If shocks to productivity growth due to technological change are frequent and
random, then the path of output following a random walk will exhibit features that
resemble a business cycle

→ The observed fluctuations in GNP are fluctuations in the natural (trend) rate of
output, not deviations of output from a smooth deterministic trend

What looks like output fluctuating around a smooth trend is in fact fluctuations in
the natural rate of output induced by a series of permanent shocks, with each
permanent productivity shock determining a new growth path.

→ The economic forces determining the trend are not different from those causing
fluctuations
The main source of fluctuations are real disturbances!
By ending the distinction between trend and cycle, real business cycle theorists began to integrate
the theory of growth and fluctuations
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Implications:
If shocks to productivity growth due to technological change are frequent and
random, then the path of output following a random walk will exhibit features that
resemble a business cycle

→ The observed fluctuations in GNP are fluctuations in the natural (trend) rate of
output, not deviations of output from a smooth deterministic trend

What looks like output fluctuating around a smooth trend is in fact fluctuations in
the natural rate of output induced by a series of permanent shocks, with each
permanent productivity shock determining a new growth path.

→ The economic forces determining the trend are not different from those causing
fluctuations
The main source of fluctuations are real disturbances!
By ending the distinction between trend and cycle, real business cycle theorists began to integrate
the theory of growth and fluctuations
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Implications:
If shocks to productivity growth due to technological change are frequent and
random, then the path of output following a random walk will exhibit features that
resemble a business cycle

→ The observed fluctuations in GNP are fluctuations in the natural (trend) rate of
output, not deviations of output from a smooth deterministic trend

What looks like output fluctuating around a smooth trend is in fact fluctuations in
the natural rate of output induced by a series of permanent shocks, with each
permanent productivity shock determining a new growth path.

→ The economic forces determining the trend are not different from those causing
fluctuations
The main source of fluctuations are real disturbances!
By ending the distinction between trend and cycle, real business cycle theorists began to integrate
the theory of growth and fluctuations
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Problem of Explaining Fluctuations

Implications:
If shocks to productivity growth due to technological change are frequent and
random, then the path of output following a random walk will exhibit features that
resemble a business cycle

→ The observed fluctuations in GNP are fluctuations in the natural (trend) rate of
output, not deviations of output from a smooth deterministic trend

What looks like output fluctuating around a smooth trend is in fact fluctuations in
the natural rate of output induced by a series of permanent shocks, with each
permanent productivity shock determining a new growth path.

→ The economic forces determining the trend are not different from those causing
fluctuations
The main source of fluctuations are real disturbances!
By ending the distinction between trend and cycle, real business cycle theorists began to integrate
the theory of growth and fluctuations
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Outline

1 Introduction

2 A Simple RBC Model

3 The RBC framework

4 Solving the full model

5 Discussion
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Production

Production function: Yt = Ktα (At Lt )1−α , with α ∈ (0, 1)

Law of motion of K: Kt+1 = Kt + It − δKt

Yt = Ct + It + Gt , so Kt+1 = (1 − δ)Kt + Yt − Ct − Gt

Real wage (wt ) and interest rate (rt ) are equal to their MPL and MPK
respectively:
 α
wt = (1 − α)Ktα (At Lt )−α At = (1 − α) AKt Lt t At
 1−α
rt = α AKt Lt t −δ
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Representative Household

The representative household maximize the expected value of:

∞
X Nt
U= e−ρt u(ct , 1 − `t )
H
t=0

Exogenous population growth: ln Nt = N + nt, (n < ρ);

with no loss of generality we assume n = 0
Log-linear istant. utility: ut = ln ct + b ln(1 − `t ), b > 0
Law of motion of technology and government purchases:
ln At = A + gt + Ãt Ãt = ρA Ãt−1 + A,t , −1 < ρA < 1
ln Gt = G + (n + g)t + G̃t G̃t = ρG G̃t−1 + G,t , −1 < ρG < 1

A and G white noise shocks and A ⊥G

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Intertemporal Substitution of Labor

Case 1: one-period framework

Agent lives for one period and wealth is equal to zero

Utility: U = ln c + b ln(1 − `); budget constaint: c = w`

Lagrangian: L = ln c + b ln(1 − `) + λ(w` − c)

FOC for c and `:

1 b
−λ=0 − + λw = 0
c 1−`
Substituting λ = 1/(w`) in the ` FOC we get:

b 1
− + =0
1−` `
Labor supply is independent of w because with log-utility income and
substitution effects offset each other
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Intertemporal Substitution of Labor

Case 2: two-periods framework

Lagrangian:
 
1 1
L = ln c1 +b ln(1−`1 )+e−ρ [ln c2 +b ln(1−`2 )]+λ w1 `1 + 1+r w2 `2 − c1 − 1+r c2

FOC for `1 and `2 :

b b
1−`1 = λw1 =⇒ w1 (1−`1 ) = λ
e−ρ b 1 e−ρ b 1+r
1−`2 = 1+r λw2 =⇒ 1−`2 w2 = λ

Hence:
b e−ρ b 1+r 1−`1 1 w2
w1 (1−`1 ) = 1−`2 w2 =⇒ 1−`2 = e−ρ (1+r) w1
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Intertemporal Substitution of Labor

Case 2: two-periods framework

In a multi-period framework there’s intertemporal substitution in labor

supply

LS responds to the relative wage (w2 /w1 ): if w1 rises, households provide

more labor

LS responds to the interest rate: a rise in r increases the attractiveness of

working more today and saving relative to working tomorrow
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Intertemporal Substitution of Consumption

Consider the following utility function:

∞
X
U= e−ρt (ln ct + b ln(1 − `t ))
t=0

We derive Euler equation (−∆ct ⇐⇒ +∆ct+1 ):

 
−ρt ∆ct −ρ(t+1) ∆ct
e = Et e (1 + rt+1 )
ct ct+1
We get the Ramsey model result:
   
1 −ρ 1 + rt+1 −ρ 1 1
= e Et = e Et (1+rt+1 )Et +e−ρ Cov( , 1 + rt+1 ) (1)
ct ct+1 ct+1 ct+1
1
Note that +∆rt+1 implies +∆ct+1 , but Cov( ct+1 , 1 + rt+1 ) < 0
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Trade-off Between Consumption and Labor

Consider again the utility functions:

∞
X Nt
U= e−ρt (ln ct + b ln(1 − `t ))
H
t=0

Household supplies more labor in order to consume more at time t leaving its
utility constant:
Nt b Nt 1
e−ρt ∆` = e−ρ(t) wt ∆`
H 1 − `t H ct
Relation between current leisure and consumption:
ct wt
= (2)
1 − `t b
Note: current variables, no uncertainty, no expectations
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Simplifying the Model

In order to solve the model analytically, we assume that δ = 1 and we

eliminate government purchases

The evolution of capital stock and real interest rate equations become:
Kt+1 = Kt + Yt − Ct − Gt − δKt = Yt − Ct
 1−α  1−α
1 + rt = α AKt Lt t − δ + 1 = α AKt Lt t

Solution strategy: competitive equilibrium; solve the model for ` and s,

log-linearizing and substituting (1 − s)Y for C

We focus on the two conditions for household optimization

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving the Model I

Let’s rewrite the Euler equation of consumption with ct = (1 − st )Yt /Nt and
taking logs:
 
1 −ρ E 1+rt+1
ct = e t ct+1 =⇒
h i h i
Yt 1+rt+1
=⇒ − ln (1 − st ) N t
= −ρ + ln Et (1−st+1 )Yt+1 /Nt+1

Note that Kt+1 = Yt − Ct = st Yt and that

A Lt+1 1−α
 
Yt+1
1 + rt+1 = α t+1 Kt+1 = α Kt+1

Substituting these equations in the Euler equation we get:

h i
αY
− ln(1 − st ) − ln Yt + ln Nt = −ρ + ln Et Kt+1 (1−st+1t+1
)Yt+1 /Nt+1
h i
αNt+1
= −ρ + ln Et st (1−st+1 )Yt
h i
= −ρ + ln α + ln Nt + n − ln st − ln Yt + ln Et 1−s1t+1
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving the Model II

Simplifying we get:
h i
1
ln st − ln(1 − st ) = −ρ + n + ln α + ln Et 1−st+1

Note that A and K aren’t present in the last equation. There’s no uncertainty
and s is constant over time:  
1
ln s − ln(1 − s) = −ρ + n + ln α + ln 1−s =⇒
=⇒ ln s = ln α + n − ρ
=⇒ s = αen−ρ
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving the Model III

Now we try to find `. Since ct = (1 − s)Yt /Nt we get:

h i
ct
1−`t = b
wt
=⇒ ln (1−s)Y Nt
t
− ln(1 − `t ) = ln wt − ln b

The production function is Cobb-Douglas, so:

 α  α  1−α
wt = (1 − α) AKt Lt t At = (1 − α) AKt Lt t At Lt
At Lt At = (1 − α) `Yt Nt t

Substituting this fact into the first equation of the slide:

ln(1 − s) + ln Yt − ln Nt − ln(1 − `t ) = ln(1 − α) + ln Yt − ln `t − ln Nt − ln b

Simplifying and removing logs we find ` which is constant

`t 1−α 1−α
1−`t = b(1−s) =⇒ `t = (1−α)+b(1−s) ≡`
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Discussion

s is constant because it’s not affected by K and A

` is constant because K and A movements have offsetting impacts on the

relative-wage and interest-rate effects on labor supply

s and ` constitute the solution of the model. The remaining equations don’t
require optimization but follow from technology, accounting and competition

The representative agent assumption implies that the competitive equilibrium

just derived is unique
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Business Cycles

Fluctuations are driven by real shocks

The economy is Walrasian: output fluctuations are optimal responses to

shocks

Output movements over time represent time-varying Pareto optimun

This implies that government interventions always reduce welfare

The dynamics of output fluctuations stems from the law of motion of K and
from the assumptions about technology dynamics
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Output Fluctuations I

The Cobb-Douglas production function implies:

ln Yt = α ln Kt + (1 − α)(ln At + ln Lt )

Since Kt = sYt−1 and Lt = `Nt we get

ln Yt = α ln s + α ln Yt−1 + (1 − α)(ln At + ln ` + ln Nt )

Indicating with Ŷt the log-deviation of output from its long-run path, one can
infer [recall that Nt is constant]:

Ŷt = αŶt−1 + (1 − α)Ât

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Output Fluctuations II
Naturally Ŷt−1 = αŶt−2 + (1 − α)Ât−1 . This implies that
Ât−1 = 1/(1 − α)(Ŷt−1 − αŶt−2 )
Since Ât = ρA Ât−1 + A,t we get:
Ŷt = αŶt−1 + (1 − α)(ρA Ât−1 + A,t )
Ŷt = αŶt−1 + ρA (Ŷt−1 − αŶt−2 ) + (1 − α)A,t
Ŷt = (α + ρA )Ŷt−1 − αρA Ŷt−2 + (1 − α)A,t

(Log) output deviation from its normal path follow a second-order

autoregressive process AR(2)
Note that if α + ρA > 0 and αρA < 0, output has an hump-shaped
impulse-response function
Note also that since α is “small”, the dynamics of output is driven by the
persistence of technology shock, ρA
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Problems

The toy RBC model isn’t able to replicate many business cycle stylized facts:
s constant: consumption and investment have the same volatility
l constant: employment is acyclical
w = (1 − α)Y/L: highly procyclical instead of almost uncorrelated

The model must be extended in order to improve its empirical performance

If δ < 1, investment and employment respond more to shock

If G > 0, real wage becomes moderately procyclical

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Outline

1 Introduction

2 A Simple RBC Model

3 The RBC framework

4 Solving the full model

5 Discussion
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Digression

The simple RBC model satisfies the two fundamental welfare theorems (it is
from this fact that follows that recessions are optimal)
Every competitive equilibrium corresponds to a social optimum
Every social optimum can be supported by a competitive equilibrium subject to
an appropriate distribution of resources

Because the simple RBC model satisfies the second fundamental welfare
theorem, it can be set up and solved as a social planner’s problem which
does not involve prices (and - obviously - there is no a priori need of a
decentralized solution)
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Digression

The simple RBC model satisfies the two fundamental welfare theorems (it is
from this fact that follows that recessions are optimal)
Every competitive equilibrium corresponds to a social optimum
Every social optimum can be supported by a competitive equilibrium subject to
an appropriate distribution of resources

Because the simple RBC model satisfies the second fundamental welfare
theorem, it can be set up and solved as a social planner’s problem which
does not involve prices (and - obviously - there is no a priori need of a
decentralized solution)
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Social Planner’s Problem

The social planner’s problem
(∞ )
X
max E0 β t u(ct , lt ) s.t.
{ct ,lt ,kt+1 }
t=0
ct + kt+1 = (1 − δ)kt + yt , y = At f (kt , lt ), and k0 > 0

How do we obtain solutions to the social planner’s problem above? More

generally, how do we solve RBC models which can be set up as social
planner’s problems?
For most RBC models, under reasonable assumptions on technology and
preferences, analytical solutions to their social planner’s problems cannot be
found!
→ computational methods!
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Social Planner’s Problem

The social planner’s problem
(∞ )
X
max E0 β t u(ct , lt ) s.t.
{ct ,lt ,kt+1 }
t=0
ct + kt+1 = (1 − δ)kt + yt , y = At f (kt , lt ), and k0 > 0

How do we obtain solutions to the social planner’s problem above? More

generally, how do we solve RBC models which can be set up as social
planner’s problems?
For most RBC models, under reasonable assumptions on technology and
preferences, analytical solutions to their social planner’s problems cannot be
found!
→ computational methods!
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Calibration

A major methodological innovation

1 Assign numerical values to the parameters of a model under consideration

2 Solve and simulate the model using computational methods

3 Use the output from numerical simulations of the model to assess whether the
model can account for summary statistics of the actual data
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Assign Numerical Values to Parameters

Both micro and macro data are used to pin down parameters of RBC models

As an example, let us work with At and assume a standard Cobb-Douglas:

yt = At f (ct , lt ) = At ktα l1−α
t
Hence,
ln At = ln yt − α ln kt − (1 − α) ln lt
Assuming factors are paid their marginal products, α becomes the share of
capital income in output. Using US data, α ≈ 0.4

Using data on y, k, l, we can recover At (the Solow residual)

Fitting an AR(1) on At using US data, we would obtain ln At = 0.95 ln At−1 + t

and σ =0.007
One can do similar things for all the parameters (see Cooley and Prescott, 19953 )
3
Cooley, T. F., & Prescott, E. C. (1995). Economic growth and business cycles. In Frontiers of
business cycle research, ch. 1.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving and Simulating

The most common approach to solving RBC models is the use of a log-linear approximation
around the steady state of the model (known as perturbation approach)

This results in a system of linear difference equations which is easy to solve numerically

But these solutions may be misleading when the model is sufficiently away from the steady
state (→ we always assume the economy is close to it; this is a strong assumption!)

There are other approaches (but less interesting for us)

Second order approximation (e.g. Schmitt-Grohe and Uribe, 2004 4 ; very
technical)
Simplifying assumptions (e.g. Brock and Mirman, 19725 ; excessively strict
assumptions)
4
Schmitt-Grohe, S., & Uribe, M. (2004). Solving dynamic general equilibrium models using a
second-order approximation to the policy function. Journal of economic dynamics and control,
28(4), 755-775)
5
Brock, W. A., & Mirman, L. J. (1972). Optimal economic growth and uncertainty: The discounted
case. Journal of Economic Theory, 4(3), 479-513.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving and Simulating

Once approximate solutions are obtained, they are used for simulation.

Simulations are obtained by studying the system’s reaction to real shocks

(most often, shocks to At )

Simulation is often repeated thousands of times using a variant of the Monte

Carlo method and results of each simulation round are averaged across all
simulation rounds for assessment

Assessment - in this context - means checking whether business cycle

statistics have been matched by the simulations of the model
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Assessment of the Simple RBC

Business statistics from US data (top) and model (bottom)

Volatility of investment and of output are considerably large, but not as large as the ones from
the US data
Volatility of hours worked is about half of what is observed in the US data
Consumption is too smooth compared to what is the case for the US data
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Kydland and Prescott (1982)

To fix these problems, Kydland and Prescott (1982) introduce three

modifications:
1 Non-time-separability of preference. They specify the instantaneous utility
function in such a way that the current utility depends on the weighted average
of leisure up to today, not just on today’s.
This increases the intertemporal substitutability of leisure → H becomes more
volatile

2 Time to build. the idea that an investment project takes several time periods
between its commissioning and its completion and requires continuous funding
in-between → extra persistence to the model’s dynamics

3 Inventories enter the production function.

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Kydland and Prescott (1982)

To fix these problems, Kydland and Prescott (1982) introduce three

modifications:
1 Non-time-separability of preference. They specify the instantaneous utility
function in such a way that the current utility depends on the weighted average
of leisure up to today, not just on today’s.
This increases the intertemporal substitutability of leisure → H becomes more
volatile

2 Time to build. the idea that an investment project takes several time periods
between its commissioning and its completion and requires continuous funding
in-between → extra persistence to the model’s dynamics

3 Inventories enter the production function.

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Kydland and Prescott (1982)

To fix these problems, Kydland and Prescott (1982) introduce three

modifications:
1 Non-time-separability of preference. They specify the instantaneous utility
function in such a way that the current utility depends on the weighted average
of leisure up to today, not just on today’s.
This increases the intertemporal substitutability of leisure → H becomes more
volatile

2 Time to build. the idea that an investment project takes several time periods
between its commissioning and its completion and requires continuous funding
in-between → extra persistence to the model’s dynamics

3 Inventories enter the production function.

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Assessment of Kydland and Prescott (1982)

Business statistics from US data (top) and model (bottom)

The model is better than the simple RBC model

Time to build doesn’t do much, so subsequent literature doesn’t keep it
The non-time-separability does a lot, but its form is rather arbitrary
Labour market requires further amendments
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Assessment: a Digression

Prescott believes that econometric methods reject theoretical models too easily

He believes that theoretical models should be assessed based on whether they can replicate
volatility and serial correlation properties of the actual data closely
But there is no standard metric on closeness
And why is he concerned with only standard deviations and serial correlations;
why not other statistics?

Calibration has been heavily criticized by econometricians on both theoretical and empirical
grounds

Del Negro and Schorfheide (20086 ) provide a statistically rigorous way of doing calibration, but
this is often not practiced

6
Del Negro, M., & Schorfheide, F. (2008). Forming priors for DSGE models (and how it affects the
assessment of nominal rigidities). Journal of Monetary Economics, 55(7), 1191-1208.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Hansen (1985)
Hansen (1985) introduces indivisible labor to improve the labor market
performance
This makes the issue of intertemporal elasticity of substitution irrelevant for the
aggregate fluctuations

Assume u(ct , lt ) = ln ct + ψ ln(1 − lt )

The economy is occupied by a measure one of identical individuals

Individuals can work either h hours or none at all

There is a fixed cost to working

Hansen’s model treats labor market fluctuations as happening at the extensive

margin (this is the case for the US data)
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Outline

1 Introduction

2 A Simple RBC Model

3 The RBC framework

4 Solving the full model

5 Discussion
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The framework
Recap on notation
yt := output in period t
At := total factor productivity in period t
lt := hours worked in period t (N.B. l is not leisure here!)
rt := real interest rate during period t

Technology
yt = At f (kt , lt )
At is stochastic and will give impulse to the model
f is homogeneous of degree 1 [f (αx, αy) = αf (x, y)]
capital accumulates according to: kt+1 = yt − ct + (1 − δ)kt

Preferences
u(ct , 1 − lt ) is well behaved (see intro set of slides)

Markets
production factors and goods markets are competitive
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The framework
Recap on notation
yt := output in period t
At := total factor productivity in period t
lt := hours worked in period t (N.B. l is not leisure here!)
rt := real interest rate during period t

Technology
yt = At f (kt , lt )
At is stochastic and will give impulse to the model
f is homogeneous of degree 1 [f (αx, αy) = αf (x, y)]
capital accumulates according to: kt+1 = yt − ct + (1 − δ)kt

Preferences
u(ct , 1 − lt ) is well behaved (see intro set of slides)

Markets
production factors and goods markets are competitive
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The framework
Recap on notation
yt := output in period t
At := total factor productivity in period t
lt := hours worked in period t (N.B. l is not leisure here!)
rt := real interest rate during period t

Technology
yt = At f (kt , lt )
At is stochastic and will give impulse to the model
f is homogeneous of degree 1 [f (αx, αy) = αf (x, y)]
capital accumulates according to: kt+1 = yt − ct + (1 − δ)kt

Preferences
u(ct , 1 − lt ) is well behaved (see intro set of slides)

Markets
production factors and goods markets are competitive
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The framework
Recap on notation
yt := output in period t
At := total factor productivity in period t
lt := hours worked in period t (N.B. l is not leisure here!)
rt := real interest rate during period t

Technology
yt = At f (kt , lt )
At is stochastic and will give impulse to the model
f is homogeneous of degree 1 [f (αx, αy) = αf (x, y)]
capital accumulates according to: kt+1 = yt − ct + (1 − δ)kt

Preferences
u(ct , 1 − lt ) is well behaved (see intro set of slides)

Markets
production factors and goods markets are competitive
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Hansen (1985)

An individual chooses a job-lottery (e.g. insurance market) such that with a

probability πt s/he works h hours and with a probability 1 − πt none, rather
than choosing hours worked

Regardless of the outcome, individuals receive the same income (this is

controversial!)

Hence, the expected utility of an individual is

E[u(ct , lt )] = πt (ln ct + ψ ln(1 − h)) + (1 − πt )(ln ct + ψ ln(1)

Which boils down to

E[u(ct , lt )] = ln ct + ψπt ln(1 − h)

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Hansen (1985)

The total working time in the economy is then lt = πt h

Hence, for a representative agent of the economy

ln(1 − h)
E[u(ct , lt )] = ln ct + ψ lt = ln ct − χlt
h
letting χ = −ψ ln(1 − h)/h.

The representative agent can have an infinite intertemporal elasticity, though

individuals may have low intertemporal elasticities
This makes hours worked very volatile
But is this a good representation of the labour market? Probably no
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Hansen (1985)

The total working time in the economy is then lt = πt h

Hence, for a representative agent of the economy

ln(1 − h)
E[u(ct , lt )] = ln ct + ψ lt = ln ct − χlt
h
letting χ = −ψ ln(1 − h)/h.

The representative agent can have an infinite intertemporal elasticity, though

individuals may have low intertemporal elasticities
This makes hours worked very volatile
But is this a good representation of the labour market? Probably no
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Hansen (1985)

The total working time in the economy is then lt = πt h

Hence, for a representative agent of the economy

ln(1 − h)
E[u(ct , lt )] = ln ct + ψ lt = ln ct − χlt
h
letting χ = −ψ ln(1 − h)/h.

The representative agent can have an infinite intertemporal elasticity, though

individuals may have low intertemporal elasticities
This makes hours worked very volatile
But is this a good representation of the labour market? Probably no
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Hansen (1985)

The social planner’s problem for Hansen’s model is

(assuming now CRRA utility for consumption)
"∞ !#
X
t ct1−σ − 1
max E0 β − χlt
ct ,lt ,kt+1 1−σ
t=0

subject to kt+1 = At ktα l1−α

t − ct + (1 − δ)kt and ln At = ρ ln At−1 + t

The FOCs are [check at home]

ct−σ = βEt [ct+1

−σ α−1 1−α
(αAt+1 kt+1 lt+1 + 1 − δ)]

χ = ct−σ (1 − α)At ktα l−α

t
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Hansen (1985)

The social planner’s problem for Hansen’s model is

(assuming now CRRA utility for consumption)
"∞ !#
X
t ct1−σ − 1
max E0 β − χlt
ct ,lt ,kt+1 1−σ
t=0

subject to kt+1 = At ktα l1−α

t − ct + (1 − δ)kt and ln At = ρ ln At−1 + t

The FOCs are [check at home]

ct−σ = βEt [ct+1

−σ α−1 1−α
(αAt+1 kt+1 lt+1 + 1 − δ)]

χ = ct−σ (1 − α)At ktα l−α

t
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985)

We will solve Hansen’s model by taking log-linear approximations of the

model’s first order conditions and constraints around their steady state values
and solving the resulting linear expressions

We will take four steps to accomplish this:

1 Calculate the model’s steady state
2 Log-linearize the model around the steady state
3 Assign numerical values to parameters (calibrate)
4 Solve for decision rules which link endogenous variables to predetermined and
exogenous variables
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985)

We will solve Hansen’s model by taking log-linear approximations of the

model’s first order conditions and constraints around their steady state values
and solving the resulting linear expressions

We will take four steps to accomplish this:

1 Calculate the model’s steady state
2 Log-linearize the model around the steady state
3 Assign numerical values to parameters (calibrate)
4 Solve for decision rules which link endogenous variables to predetermined and
exogenous variables
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the equations

1 ln At = ρ ln At−1 + t

2 kt+1 = At ktα l1−α

t − ct + (1 − δ)kt

3 ct−σ = βEt [ct+1

−σ α−1 1−α
(αAt+1 kt+1 lt+1 + 1 − δ)]

4 χ = ct−σ (1 − α)At ktα l−α

t
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

In the steady state, all variables in the model are constant

(i.e. t = 0 ∀t)

Let x denote the steady state value of xt

(i.e. remove dependence on time)

Hence,
1 ln A = ρ ln A →A = 1
2 k = k α l1−α − c + (1 − δ)k → c = k α l1−α − δk
3 c−σ = βc−σ (αk α−1 l1−α + 1 − δ)] → 1 = β(αk α−1 l1−α + 1 − δ)
4 χ = c−σ (1 − α)k α l−α
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

In the steady state, all variables in the model are constant

(i.e. t = 0 ∀t)

Let x denote the steady state value of xt

(i.e. remove dependence on time)

Hence,
1 ln A = ρ ln A →A = 1
2 k = k α l1−α − c + (1 − δ)k → c = k α l1−α − δk
3 c−σ = βc−σ (αk α−1 l1−α + 1 − δ)] → 1 = β(αk α−1 l1−α + 1 − δ)
4 χ = c−σ (1 − α)k α l−α
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

In the steady state, all variables in the model are constant

(i.e. t = 0 ∀t)

Let x denote the steady state value of xt

(i.e. remove dependence on time)

Hence,
1 ln A = ρ ln A →A = 1
2 k = k α l1−α − c + (1 − δ)k → c = k α l1−α − δk
3 c−σ = βc−σ (αk α−1 l1−α + 1 − δ)] → 1 = β(αk α−1 l1−α + 1 − δ)
4 χ = c−σ (1 − α)k α l−α
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

In the steady state, all variables in the model are constant

(i.e. t = 0 ∀t)

Let x denote the steady state value of xt

(i.e. remove dependence on time)

Hence,
1 ln A = ρ ln A →A = 1
2 k = k α l1−α − c + (1 − δ)k → c = k α l1−α − δk
3 c−σ = βc−σ (αk α−1 l1−α + 1 − δ)] → 1 = β(αk α−1 l1−α + 1 − δ)
4 χ = c−σ (1 − α)k α l−α
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

From 3,
  1
k 1 − β(1 − δ) α−1
=
l αβ
From 4,
  1  α
1−α σ k σ
c=
χ l
From 2, !−1
 α−1
k
k=c −δ
l
By identity
 −1
k
l=k
l
By the production function
y = k α l1−α

Solved: we have found c, k, l and y as functions of just parameters!

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

From 3,
  1
k 1 − β(1 − δ) α−1
=
l αβ
From 4,
  1  α
1−α σ k σ
c=
χ l
From 2, !−1
 α−1
k
k=c −δ
l
By identity
 −1
k
l=k
l
By the production function
y = k α l1−α

Solved: we have found c, k, l and y as functions of just parameters!

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

From 3,
  1
k 1 − β(1 − δ) α−1
=
l αβ
From 4,
  1  α
1−α σ k σ
c=
χ l
From 2, !−1
 α−1
k
k=c −δ
l
By identity
 −1
k
l=k
l
By the production function
y = k α l1−α

Solved: we have found c, k, l and y as functions of just parameters!

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

From 3,
  1
k 1 − β(1 − δ) α−1
=
l αβ
From 4,
  1  α
1−α σ k σ
c=
χ l
From 2, !−1
 α−1
k
k=c −δ
l
By identity
 −1
k
l=k
l
By the production function
y = k α l1−α

Solved: we have found c, k, l and y as functions of just parameters!

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

From 3,
  1
k 1 − β(1 − δ) α−1
=
l αβ
From 4,
  1  α
1−α σ k σ
c=
χ l
From 2, !−1
 α−1
k
k=c −δ
l
By identity
 −1
k
l=k
l
By the production function
y = k α l1−α

Solved: we have found c, k, l and y as functions of just parameters!

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): the steady state

From 3,
  1
k 1 − β(1 − δ) α−1
=
l αβ
From 4,
  1  α
1−α σ k σ
c=
χ l
From 2, !−1
 α−1
k
k=c −δ
l
By identity
 −1
k
l=k
l
By the production function
y = k α l1−α

Solved: we have found c, k, l and y as functions of just parameters!

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Log-linearization: using Uhlig’s toolkit

Example 1
axt = axex̂t ' ax(1 + x̂t ) = axx̂t + ax
(application of rule 1)

Example 2

(xt + a)yt = (xex̂t + a)yeŷt = xyex̂t eŷt + ayex̂t

' xy(1 + x̂)(1 + ŷ) + ay(1 + ŷt )
= xy + xyx̂t + xyŷt + xyx̂t ŷt + ay + ayŷt
= xyx̂t + (x + a)yŷt + (x + a)y

(application of rule 1 and rule 2)

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Log-linearization: using Uhlig’s toolkit

Example 1
axt = axex̂t ' ax(1 + x̂t ) = axx̂t + ax
(application of rule 1)

Example 2

(xt + a)yt = (xex̂t + a)yeŷt = xyex̂t eŷt + ayex̂t

' xy(1 + x̂)(1 + ŷ) + ay(1 + ŷt )
= xy + xyx̂t + xyŷt + xyx̂t ŷt + ay + ayŷt
= xyx̂t + (x + a)yŷt + (x + a)y

(application of rule 1 and rule 2)

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Log-linearization: using Uhlig’s toolkit

Example 1
axt = axex̂t ' ax(1 + x̂t ) = axx̂t + ax
(application of rule 1)

Example 2

(xt + a)yt = (xex̂t + a)yeŷt = xyex̂t eŷt + ayex̂t

' xy(1 + x̂)(1 + ŷ) + ay(1 + ŷt )
= xy + xyx̂t + xyŷt + xyx̂t ŷt + ay + ayŷt
= xyx̂t + (x + a)yŷt + (x + a)y

(application of rule 1 and rule 2)

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Log-linear approximation: an example

The dynamics of productivity

ln At = ρ ln At−1 + t
ln AeÂt = ρ ln AeÂt−1 + t
ln A + Ât = ρ ln A + ρÂt−1 + t

Because A = 1 in steady state

Ât = ρÂt−1 + t
[Note of caution: even though we use (for convention and simplicity) the equality sign, never forget these are
approximations around the steady state!]
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Log-linear approximation: an example

The dynamics of productivity

ln At = ρ ln At−1 + t
ln AeÂt = ρ ln AeÂt−1 + t
ln A + Ât = ρ ln A + ρÂt−1 + t

Because A = 1 in steady state

Ât = ρÂt−1 + t
[Note of caution: even though we use (for convention and simplicity) the equality sign, never forget these are
approximations around the steady state!]
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): log-linear approximation

The equilibrium of the model is characterized by 4 equations

1 ln At = ρ ln At−1 + t
2 kt+1 = At ktα l1−α
t − ct + (1 − δ)kt
−σ α−1 1−α
3 ct = βEt [ct+1 (αAt+1 kt+1 lt+1 + 1 − δ)]
4 χ = ct−σ (1 − α)At ktα l−α
t

The steady state of the model satisfies

1 A=1
2 c = k α l1−α − δk
3 1 = β(αk α−1 l1−α + 1 − δ)
4 χ = c−σ (1 − α)k α l−α
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The log-linear approximation around the SS

1

Ât = ρÂt−1 + t

cĉ + k k̂t+1 = (αy + (1 − δ)k)k̂t + (1 − α)yl̂t + yÂt

−σĉt = (α − 1)(1 − β(1 − δ))k̂t+1 + (1 − α)(1 − β(1 − δ))Et l̂t+1

σĉt = Ât + αk̂t − αl̂t

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): state-space representation

Let us put the approximation of the model in the state space form
  
(1 − β(1 − δ))ρ (α − 1)(1 − β(1 − δ)) (1 − α)(1 − β(1 − δ)) −σ Ât

 −y k 0 0    k̂t+1
 

 1 0 0 0   Et (l̂t+1 ) 
−1 0 0 0 Et (ĉt+1 )
    
0 0 0 −σ Ât−1 0
 0 (αy + (1 − δ)k) (1 − α)y −c   k̂t   0 
= + 
0   l̂t   1  t

 ρ 0 0
0 α −α −σ ĉt 0

Or, more concisely A0 xt+1 = A1 xt + B0 t , where xt and t are the corresponding

column vectors
Assuming that A1 is invertible, we can pre-multiply each side by A−1
1 :

Axt+1 = xt + Bt
where A ≡ A−1 −1
1 A0 and B ≡ A1 B0
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): state-space representation

Let us put the approximation of the model in the state space form
  
(1 − β(1 − δ))ρ (α − 1)(1 − β(1 − δ)) (1 − α)(1 − β(1 − δ)) −σ Ât

 −y k 0 0    k̂t+1
 

 1 0 0 0   Et (l̂t+1 ) 
−1 0 0 0 Et (ĉt+1 )
    
0 0 0 −σ Ât−1 0
 0 (αy + (1 − δ)k) (1 − α)y −c   k̂t   0 
= + 
0   l̂t   1  t

 ρ 0 0
0 α −α −σ ĉt 0

Or, more concisely A0 xt+1 = A1 xt + B0 t , where xt and t are the corresponding

column vectors
Assuming that A1 is invertible, we can pre-multiply each side by A−1
1 :

Axt+1 = xt + Bt
where A ≡ A−1 −1
1 A0 and B ≡ A1 B0
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving Hansen (1985): state-space representation

Let us put the approximation of the model in the state space form
  
(1 − β(1 − δ))ρ (α − 1)(1 − β(1 − δ)) (1 − α)(1 − β(1 − δ)) −σ Ât

 −y k 0 0    k̂t+1
 

 1 0 0 0   Et (l̂t+1 ) 
−1 0 0 0 Et (ĉt+1 )
    
0 0 0 −σ Ât−1 0
 0 (αy + (1 − δ)k) (1 − α)y −c   k̂t   0 
= + 
0   l̂t   1  t

 ρ 0 0
0 α −α −σ ĉt 0

Or, more concisely A0 xt+1 = A1 xt + B0 t , where xt and t are the corresponding

column vectors
Assuming that A1 is invertible, we can pre-multiply each side by A−1
1 :

Axt+1 = xt + Bt
where A ≡ A−1 −1
1 A0 and B ≡ A1 B0
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solution Methods
State of the art: the method of Blanchard and Kahn (19807 ) to find the saddle
path of the system of dynamic linear Rational Expectations equations. You
will see this solution method in the tutorial!

[Notice: if A1 is singular, different methods must be applied; see Klein (20008 )]

7
Blanchard, O. J., & Kahn, C. M. (1980). The solution of linear difference models under rational
expectations. Econometrica, 1305-1311.
8
Klein, P. (2000). Using the generalized Schur form to solve a multivariate linear rational
expectations model. Journal of economic dynamics and control, 24(10), 1405-1423.
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Final solution

l̂t = 0.7185Ât−1 − 0.3473k̂t + 0.7563t

ĉt = 0.6626Ât−1 + 0.5389k̂t + 0.6975t
Ât = 0.95Ât−1 + t
k̂t+1 = 0.3404Ât−1 + 0.8214k̂t + 0.3584t
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions

Impulse response for the Hansen model (dashed) vs “log-log” (solid)

∂yt+h
Recall that IRFs correspond to: ∂xt
The figure shows the reaction of the economy to a shock to A
Hansen model has utility u(ct , lt ) = ln ct − χlt
Log-log model has utility u(ct , lt ) = ln ct + ψ ln(1 − lt )
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions: interpretation

significantly more amplification in the indivisible labour case than in the log-log case
output and hours both increase by significantly more on impact
as a result, consumption and investment both go up by more initially
labour supply is horizontal (because of indivisible labour) → we get a bigger increase in labor
hours (and a smaller increase in wages) after the productivity shock
better match the output volatility in US data with smaller TFP shocks
indivisible labor increases the relative volatility of hours substantially
indivisible labor makes wages somewhat less volatile and less procyclical
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions: interpretation

significantly more amplification in the indivisible labour case than in the log-log case
output and hours both increase by significantly more on impact
as a result, consumption and investment both go up by more initially
labour supply is horizontal (because of indivisible labour) → we get a bigger increase in labor
hours (and a smaller increase in wages) after the productivity shock
better match the output volatility in US data with smaller TFP shocks
indivisible labor increases the relative volatility of hours substantially
indivisible labor makes wages somewhat less volatile and less procyclical
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions: interpretation

significantly more amplification in the indivisible labour case than in the log-log case
output and hours both increase by significantly more on impact
as a result, consumption and investment both go up by more initially
labour supply is horizontal (because of indivisible labour) → we get a bigger increase in labor
hours (and a smaller increase in wages) after the productivity shock
better match the output volatility in US data with smaller TFP shocks
indivisible labor increases the relative volatility of hours substantially
indivisible labor makes wages somewhat less volatile and less procyclical
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions: interpretation

significantly more amplification in the indivisible labour case than in the log-log case
output and hours both increase by significantly more on impact
as a result, consumption and investment both go up by more initially
labour supply is horizontal (because of indivisible labour) → we get a bigger increase in labor
hours (and a smaller increase in wages) after the productivity shock
better match the output volatility in US data with smaller TFP shocks
indivisible labor increases the relative volatility of hours substantially
indivisible labor makes wages somewhat less volatile and less procyclical
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions: interpretation

significantly more amplification in the indivisible labour case than in the log-log case
output and hours both increase by significantly more on impact
as a result, consumption and investment both go up by more initially
labour supply is horizontal (because of indivisible labour) → we get a bigger increase in labor
hours (and a smaller increase in wages) after the productivity shock
better match the output volatility in US data with smaller TFP shocks
indivisible labor increases the relative volatility of hours substantially
indivisible labor makes wages somewhat less volatile and less procyclical
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions: interpretation

significantly more amplification in the indivisible labour case than in the log-log case
output and hours both increase by significantly more on impact
as a result, consumption and investment both go up by more initially
labour supply is horizontal (because of indivisible labour) → we get a bigger increase in labor
hours (and a smaller increase in wages) after the productivity shock
better match the output volatility in US data with smaller TFP shocks
indivisible labor increases the relative volatility of hours substantially
indivisible labor makes wages somewhat less volatile and less procyclical
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Impulse Response Functions: interpretation

significantly more amplification in the indivisible labour case than in the log-log case
output and hours both increase by significantly more on impact
as a result, consumption and investment both go up by more initially
labour supply is horizontal (because of indivisible labour) → we get a bigger increase in labor
hours (and a smaller increase in wages) after the productivity shock
better match the output volatility in US data with smaller TFP shocks
indivisible labor increases the relative volatility of hours substantially
indivisible labor makes wages somewhat less volatile and less procyclical
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Alternative Solution Strategy - applied to the Simple RBC

State variables: K, A, G. Control variables: C and L

Log-linearizing the model around the deterministic balanced growth path:

C̃t ' αCK K̃t + αCA Ãt + αCG G̃t
L̃t ' αLK K̃t + αLA Ãt + αLG G̃t

To solve the model, one has to find the values of a’s, which depend on the
underlying parameters of the model. We apply this to the simple model
analysed above.

The set of a’s which solve the model must satisfy household optimization
conditions (Eqs. 1 and 2)

This is the solution method of undetermined coefficients

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Intratemporal First-Order Condition I

Let’s start from the FOC ct /(1 − `t ) = wt /b. Knowing that

wt = (1 − α)[Kt /(At Lt )]α At and taking logs we get:

ln ct − ln(1 − `t ) = ln 1−α


b + (1 − α) ln At + α ln Kt − α ln Lt

The first-order Taylor approximation around the balanced growth path (in
logs) is (note that C̃t = c̃t , `˜t = L̃t ):
`∗
C̃t + 1−`∗ L̃t = (1 − α)Ãt + αK̃t + αL̃t

Since C̃t and L̃t are linear function of K̃t , Ãt , G̃t we get:
 ∗ 
`
αCK K̃t +αCA Ãt +αCG G̃t + 1−` ∗ + α (αLK K̃t +αLA Ãt +αLG G̃t ) = (1−α)Ãt +αK̃t
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Intratemporal First-Order Condition II

The last equation must hold for all values of K̃t , Ãt , G̃t , because it’s the
solution of an optimization problem

This implies that the coefficients of K̃t , Ãt , G̃t on the two sides must be equal:
 ∗ 
`
αCK + 1−` ∗ + α αLK = α
 ∗ 
`
αCA + 1−` ∗ + α αLA = 1 − α
 ∗ 
`
αCG + 1−` ∗ + α αLG = 0

G doesn’t affect w. L and C respond to ∆G going in opposite directions

An increase in A or K raise w (1 − α or α): households can increase either C or

L or both
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Intertemporal First-Order Condition I

Now consider the other FOC: 1/ct = e−ρ Et [(1 + rt+1 )/ct+1 ]

Z̃t+1 is the difference between the log of (1 + rt+1 )/ct+1 and the log of its
balanced growth path value

The expression C̃t ' αCK K̃t + αCA Ãt + αCG G̃t implies
C̃t+1 ' αCK K̃t+1 + αCA Ãt+1 + αCG G̃t+1

The last expression and rt+1 = α(At+1 Lt+1 /Kt+1 )1−α − δ can be used to
express Z̃t+1 in terms of K̃t+1 , Ãt+1 and G̃t+1

The endogenous variable K̃t+1 can be removed log-linearizing

Kt+1 = (1 − δ)Kt + Yt − Ct − Gt
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Intertemporal First-Order Condition II

This allows to write K̃t+1 in terms of K̃t , Ãt , G̃t , L̃t and C̃t

Since C̃t ' αCK K̃t + αCA Ãt + αCG G̃t and L̃t ' αLK K̃t + αLA Ãt + αLG G̃t we get:
K̃t+1 ' βCK K̃t + βKA Ãt + βKG G̃t

This last result allows to express Z̃t+1 in terms of Ãt , Ãt+1 , G̃t , G̃t+1 and K̃t

One can then compute Et [Z̃t+1 ] in terms of Ãt , K̃t , G̃t

Finally, one can substitute the just derived Et [Z̃t+1 ] in

1/ct = e−ρ Et [(1 + rt+1 )/ct+1 ]

This allows to get 3 additional restrictions on the α’s which allow to express
the α in terms of underlying parameters
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Solving the Model Numerically

Even after log-linearization, the model is still sufficiently complicated to get

economic intuitions (sic)

Solution strategy: employ numerical methods

choose a parameterization (i.e. calibrate the model)
study how the chosen parameters affect the α’s and the β’s
analyze how C, L and K respond to government and technology shock
study the response of the other variable of the model (e.g. Y, I) to the shocks
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Effects of Technology Shocks I

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Effects of Technology Shocks II

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Effects of Technology Shocks III

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Effects of Government Purchases Shocks I

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Effects of Government Purchases Shocks II

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

The Effects of Government Purchases Shocks III

Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Outline

1 Introduction

2 A Simple RBC Model

3 The RBC framework

4 Solving the full model

5 Discussion
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Extensions

Distortionary taxes

Multiple sectors and industry-specific shocks

Credit

Variable capital utilization

Labor hoarding
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Policy Implications

Since fluctuations are optimal, stabilization policies reduce welfare

Monetary policy has a limited (if any) role

The economy is always in full-employment and unemployment is always

voluntary
Introduction A Simple RBC Model The RBC framework Solving the full model Discussion

Recap: The RBC Legacy

Three ingredients:
1) Walrasian benchmark
2) Dynamic stochastic general equilibrium framework
3) Calibration

These characteristics impose discipline to modelers, but increase the

complexity of the models

The same discipline has been kept in New Keynesian Theories