Real Business Cycle Models

Daniel Vernazza
[email protected]

1 Intro

RBC models belong to a class of models called Dynamic Stochastic General

Equilibrium (DSGE) Models. They are dynamic because investment is a key
endogenous variable, which facilitates the smoothing of household consumption
over time. They are stochastic because an exogenous shock process is intro-
duced to trigger ‡uctuations in the endogenous variables around their balanced
growth paths (that is, a business cycle). RBC models consider exogenous shocks
to productivity and/ or government expenditure (i.e. real shocks). The model
is written in discrete time because business cycles are ‡uctuations with a peri-
odicity of between 2 and 8 years.

It is a general equilibrium model, that is, households and …rms are maximis-
ing utility and pro…ts respectively, and markets clear.

2 Baseline RBC Model

De…ne output Y , capital K, TFP A, labour supply N , consumption C: real

interest rate R, and the real wage W . The economy is characterised by the
following set of 7 equations:

Production function:

Yt = Kt1 (At Nt ) (1)

where 2 (0; 1).
Capital accumulation:

Kt+1 = (1 )Kt + Yt Ct (2)

NB: there is no government so Yt = Ct + It .

Technology:

ln At = ln G + ln At+1 + ln A~t 1 + "t (3)

FOCs of Firms:

1
 At Nt
Rt = (1 ) + (1 ) (4)
Kt
1
Kt
W t = At (5)
Nt
where Rt ( (1 + rt )) is the gross interest rate.

FOCs of the representative household:

1 Rt+1
= Et (6)
Ct Ct+1
Wt
= (1 Nt ) (7)
Ct

3 Deriving the Household and Firm FOCs

We will now derive the FOCs of the household: equations (6) and (7). Because
this is a Walrasian model, the …rst and second welfare theorems hold and hence
the solution to the decentralised problem is identical to that of a social plan-
ner. Nonetheless, I solve both cases here, …rst the decentralised competitive
equilibrium and second the social planner’s problem.

3.1 The Decentralised Competitive Equilibrium

The general competitive equilibrium is a stochastic process for the endogenous

1
variables fYt ; Ct ; Nt ; Kt ; Rt ; Wt ; At gt=0 such that: (1) households are maximis-
ing their utility subject to their budget constraint; (2) …rms are maximising
pro…ts subject to the production function; and …nally (3) markets clear.

3.1.1 Households
1
The representative household (taking fRt ; Wt gt=0 as given) solves:
1
X
t
max1 E0 U (Ct ; 1 lt )
fCt ;Nt gt=0
t=0

subject to the period budget constraint:

at+1 = Rt at + Wt lt Ct
where lt 2 (0; 1) is the proportion of time the representative household
spends working; and at is the stock of assets (savings) held by the represen-
tative household.

2
 State: at .
Controls: Ct ; lt .
The Bellman equation is:

V (at ; At ) = max fU (Ct ; 1 lt ) + E [V (at+1 ; At+1 ) j At ]g

Ct ;lt

subject to:

at+1 = Rt at + Wt lt Ct
FOCs:

@V (at+1 ; At+1 )
U1 (Ct ; 1 lt ) = :E j At (8)
@at+1
@V (at+1 ; At+1 )
U2 (Ct ; 1 lt ) = :E :Wt j At (9)
@at+1
E.T:

@V (at ; At )
= U1 (Ct ; 1 lt ):Rt
@at
Iterating forward to t+1:

@V (at+1 ; At+1 )
= U1 (Ct+1 ; 1 lt+1 ):Rt+1 (10)
@at+1
Substituting eq. (10) into the F.O.C. in eq. (8) we get the INTERTEMPO-
RAL FOC (or Euler equation):

U1 (Ct ; 1 lt ) = :E [U1 (Ct+1 ; 1 lt+1 ):Rt+1 j At ] (11)

Substituting eq. (10) into the F.O.C. in eq. (9) we get the INTRATEMPO-
RAL FOC:

U2 (Ct ; 1 lt ) = U1 (Ct ; 1 lt ):Wt (12)

where we have substituted the Euler equation (11).

3.1.2 Firms
1
Taking the stochastc prices fRt ; Wt gt=0 as given, the representative …rm solves
8t:

max = Yt Wt Nt (Rt 1 + )Kt

fKt ;Nt g1
t=0

subject to the production function:

Yt = Kt1 (At Nt )

3
 Note: there are no intertemporal elements here so the representative …rm’s
problem is a simple static one.

FOCs (although should use perturbation argument b/c hessian is not nega-
tive de…nite w/ Cobb Douglas):

At Nt
Rt = (1 ) + (1 )
Kt
1
Kt
W t = At
Nt
These are equations (4) and (5) above.

3.1.3 Markets Clear

For all t:
wt clears the labour market: lt = Nt
rt clears the capital market: at = Kt

Equations (6) and (7) are the result of evaluating equations (11) and (12)
for the admissible utility function:

(1 lt )1
U (Ct ; 1 lt ) = log(Ct ) +
1
The above utility function is one of a limited set of functional forms for
which the model exhibits a balanced growth path (see King, Plosser and Rebelo
(1988)).1 A non-stochastic BGP must feature constant N . On the BGP we
know from the Ramsey (and Solow) model that the wage Wt grows at rate G.
However, because of our choice of utility function the household FOC governing
the choice of Nt (equation (7)) depends only on the ratio W t
Ct which is constant
on the BGP because both Wt and Ct grow at rate G on the BGP.

3.2 The Social Planner’s Problem

The social planner’s problem is:
1
X
t
max1 E0 U (Ct ; 1 Nt )
fCt ;Nt gt=0
t=0

1 Alternatively C1
one could use U (Ct ; 1 Nt ) = 1t (1 Nt ). In which case the inter-
h i
temporal FOC becomes Ct (1 Nt ) = Et Rt+1 Ct+1 (1 Nt+1 ) and the intra-temporal
0
(1 Nt )
FOC is (1 )W
C
t
= (1 Nt )
. This latter FOC exhibits a constant N on the BGP.
t

4
 subject to equations (1), (2), (3), (4) and (5).
State: Kt .
Controls: Ct ; Nt .
The Bellman equation is:

V (Kt ; At ) = max fU (Ct ; 1 Nt ) + E [V (Kt+1 ; At+1 ) j At ]g

Ct ;Nt

subject to:

Kt+1 = (1 )Kt + Yt Ct

Yt = Kt1 (At Nt )

ln At = ln G + ln At+1 + ln A~t 1 + "t

2
FOCs:

@V (Kt+1 ; At+1 )
U1 (Ct ; 1 Nt ) = E j At (13)
@Kt+1

@V (Kt+1 ; At+1 )
U2 (Ct ; 1 Nt ) = E Kt1 At Nt 1
j At (14)
@Kt+1

E.T:

@V (Kt ; At )
= U1 (Ct ; 1 Nt ) (1 ) + (1 )Kt (At Nt )
@Kt
Iterating forward to t+1:

@V (Kt+1 ; At+1 )
= U1 (Ct+1 ; 1 Nt+1 ) (1 ) + (1 )Kt+1 (At+1 Nt+1 )
@Kt+1
(15)

Substituting eq. (15) into the F.O.C. in eq. (13) we get the INTERTEM-
PORAL FOC (or Euler equation):

U1 (Ct ; 1 Nt ) = E [U1 (Ct+1 ; 1 Nt+1 ):Rt+1 j At ] (16)

At Nt
where we have substituted the gross interest rate Rt = (1 ) Kt +
(1 ).
2 It is important that you choose (C ; N ) as opposed to (K
t t t+1 ; Nt ) as controls. The
reason is that there are two FOCs - one for each control - and each FOC is derived as the
partial derivative, that is, holding the other control constant. The transition law tells us the
dependence between Kt+1 ; Ct ; and Nt . Equation (14) is only true if we hold Ct constant as
opposed to Kt+1 .

5
 Substituting eq. (15) into the F.O.C. in eq. (14) we get the INTRATEM-
PORAL FOC:

U2 (Ct ; 1 Nt ) = U1 (Ct ; 1 Nt ):Wt (17)

where we have substituted the Euler equation (16) and the wage rate Wt =
1
At K Nt
t
.
Equations (6) and (7) are the result of evaluating equations (16) and (17)
for the admissible utility function:

(1 Nt )1
U (Ct ; 1 Nt ) = log(Ct ) +
1

4 Solving the Model

1
The goal: solve for the evolution of the endogenous variables fYt ; Ct ; Nt ; Kt ; Rt ; Wt ; At gt=0
1
as a function of any realization of the exogenous technology shocks f"t gt=0 and
the parameters of the model ( ; ; ; ; ; ; G). The impulse response function
for an endogenous variable is de…ned as the evolution of this variable follow-
ing a one-time-only 1% shock to the productivity parameter ("t = 1; "t+s = 0
8s 1).

This is not easy because the 7 equations that characterise the model are
a mixture of linear and log-linear equations so no closed-form solution exists.
There are various methods that can be used. One is to take a …rst-order Taylor
approximation about the BGP for each one of the 7 equations and then use the
method of undetermined coe¢ cients.

5 Evaluating the Model

The model does a good job if it can match the stylised facts of business cycles
observed in postwar U.S. quarterly data, both qualitative (co-movement) and
quantitative (volatility, persistence). Typically this is done by eyeballing the im-
pulse responses. To derive the impulse responses, you will have to pick numbers
for the parameters of the model ( ; ; ; ; ; ; G), this is called calibration.

The big failure of the RBC model is matching the ‡uctuations in employment
under reasonable parameter values. To see why, look at the household intra-
temporal FOC (equation (12)):

U2 (Ct ; 1 lt ) = U1 (Ct ; 1 lt ):Wt

6
 The above equation tells us that a necessary condition for both Ct and lt
to be procyclical is that Wt is procyclical. Why? Consider a boom: C " and
l ". C " =) U1 # and l " =) U2 ". These two e¤ects imply W " since
U2 = U1 :W . In the data, C and l are strongly procyclical but W is only mildly
procyclical. So we need to assume a very strong substitution away from leisure
and into consumption from a change in the wage. That is, we need a very low
, because 1= is the intertemporal elasticity of substitution for leisure.

6 A Special Case: Brock-Mirman model

There is a special case of the above model which displays a closed-form solution.
The strong assumptions are:
- log utility: U (Ct ; 1 lt ) = ln(Ct ) + b ln(1 lt )
- full depreciation: = 1
- (no government) -> which I left out of the above baseline model.

Applying these changes to the 7 equations we have:

Production function:

Yt = Kt1 (At Nt )
Capital accumulation:

Kt+1 = Yt Ct
Technology:

ln At = ln G + ln At+1 + ln A~t 1 + "t

FOCs of Firms:

At Nt
Rt = (1 ) (18)
Kt
where Rt ( (1 + rt )) is the gross interest rate.
1
Kt
Wt = At (19)
Nt
FOCs of the representative household:

1 Rt+1
= Et (20)
Ct Ct+1
Wt b
= (21)
Ct 1 Nt

7
 Using Ct = (1 st )Yt , Kt+1 = st Yt , and equation (18), equation (20) be-
comes:
2 3
1 (1 ) At+1 Nt+1
Kt+1
= Et 4 5
Ct (1 st+1 )Yt+1
" #
1 (1 )
= Et 1
Ct (1 st+1 )Kt+1 Kt+1

1 (1 )
= Et
Ct (1 st+1 )st Yt
1 (1 ) 1
= Et
(1 st )Yt st Yt (1 st+1 )
1 (1 ) 1
= Et
(1 st ) st (1 st+1 )
Assume st = st+1 s. Then s = (1 ) . Why? If st < (1 ) then
lim Et [st ] = 1. If st > (1 ) then lim Et [st ] = 1. None of these cases
t!1 t!1
can be optimal.
Substituting the constant saving rate s = (1 ) and equation (19) into
equation (21) yields:
Wt b
=
Ct 1 Nt
Yt
Nt b
=
(1 s)Yt 1 Nt
b:Nt
=
1 s 1 Nt
1 Nt b (1 s)
=
Nt

Nt =
b (1 s) +

N=
b (1 (1 ) )+
So Nt is also constant. The saving and labour supply decisions are una¤ected
by changes in the interest rate and wage rate. This is the result of log utility
and full depreciation, that is, the income and substitution e¤ects exactly cancel
each other.

8
6.1 Evaluating the Special Case

- The wage is too procyclical: Wt = NYt .

- No ‡uctuations in employment: N constant.
- Consumption is too procyclical: Ct = (1 st )Yt .
- Investment is not procyclical enough: It = Yt Ct = sYt = (1 ) Yt .

So the model gets most of the comovements WRONG!.