Lecture Notes 10: New Keynesian DSGE

Zhiwei Xu ([email protected])

New Keynesian framework has emerged as the workhorse for the analysis of monetary policy and
its implications for in‡ation, economic ‡uctuations, and welfare. It constitutes the backbone of new
generation of medium-scale models under develpment at various central banks. The methodology
underlying the New Keynesian models is similar to the RBC framework. The key features (frictions)
that depart the former from the latter is the nominal rigidities and monopolistic competition. That
is, the prices (wage rate, price of products) cannot be adjusted ‡exibly and thus the monetary policy
is no longer netural. In this note, we will …rst introduce money into a standard RBC model via
Cash-in-advance speci…cation. We then relax the ‡exible-price assumption by introducing Calvo-
type price setting problem. Finally, we will derive the New Keynesian Phillips curve, which plays
a central role in monetary DSGE literature.

1 Cash-in-advance (CIA) Model

There is a …nal good, which can be either used for consumption or investment. The price of the
…nal good is Pt : The timing for the representative household is following. At the beginning of the
period t; the household has wealth Ht which is carried from last period (t 1). And the monetary
authority injects Xt amount of money to the households. The total wealth avaiable to the household
at the beginning of the period t is thus Ht + Xt : Then the household decides to hold Mt amount of
wealth in the form of cash, and Bt amount of wealth in the form of risk-free bond. The constraint
is given by
Mt + Bt Ht + Xt : (1)

The bond will pay interest rate Rbt at the end of period t: The household then chooses consumption,
which can be only purchased by the cash Mt . That is, the expenditure on consumption is subject
to cash-in-advance (CIA) assumption:
Pt ct Mt : (2)

Note that as holding cash does not earn interest rate, without the CIA constraint, the households
have no incentive to hold cash, they would strictly prefer the bond. After the consumption decision,
the households choose working hours nt with norminal wage rate Wt ; and investments kt+1
(1 ) kt . At the end of period, they receive labor income Wt nt and capital income Rt kt : The
wealth that the housholds carry to the next period (t + 1) is de…ned as

Ht+1 = Mt + (1 + Rbt ) Bt + Wt nt + Rt kt Pt [ct + kt+1 (1 ) kt ] : (3)

 2

The household’s problem is to maximize life-time utility

1
X
t
max (log ct an nt ) (4)
t=0

subject to (1), CIA constraint (2) and (3). To make things easy, we de…ne real variables as:
mt = Mt =Pt ; bt = Bt =Pt ; xt = Xt =Pt ; wt = Wt =Pt ; rt = Rt =Pt ; ht+1 = Ht+1 =Pt : Then the
constraints (1), (2) and (3) can be reduced to
ht
mt + bt + xt ; (5)
1+ t

ct mt ; (6)

ht+1 = mt + (1 + Rbt ) bt + wt nt + rt kt [ct + kt+1 (1 ) kt ] : (7)

Denote the Lagrangian multiplies for the above constraints as f t ; t; tg respectively. FOCs for
fmt ; bt ; ht+1 ; ct ; nt ; kt+1 g are given by
t = t + t; (8)

t = t (1 + Rbt ) ; (9)
1
t = Et t+1 ; (10)
1+ t+1

1=ct = t + t; (11)

an = t wt ; (12)

t = Et [ t+1 (rt+1 +1 )] : (13)

The …rm side is standard, pro…t maximization implies that

rt = yt =kt ; (14)

wt = (1 ) yt =nt : (15)

1.1 Competitive Equilibrium

The competitive equilibrium is de…ned as, each individual achieves the optimization and following
markets clear.

Good market:
ct + kt+1 (1 ) kt = yt ; (16)

Bond market:
bt = bt 1 = 0; (17)
 3

Money market:
mt 1
mt = + xt : (18)
1+ t
Note that if two market clearing conditions satisfy, the remaining one automatically satis…es.

The full system consists of three constraints (5) to (7), eight FOCs (8) to (15), and two market
clearing conditions (16) and (18). And we have 13 endogenous variables

f t; t; t ; mt ; bt ; ht+1 ; ct ; nt ; kt+1 ; rt ; wt ; t ; Rbt g :

One important implication of the CIA model is that the model only gives the dynamics of real
money balance mt and in‡ation rate. Therefore, the dynamics is independent with the initial level
nominal money stock M0 : An economy with higher money stock may have the same dynamics of
the one with lower money stock. The only di¤erence between two economies is the price level.
Higher money stock implies higher price level. We call this property as neutrality of money.

1.2 Steady State

We …rst derive the steady state. Equations (9) and (10) implies the nominal interest rate Rb is
given by
1+
Rb = 1: (19)

The in‡ation rate is determined by (18)

x
= = g; (20)
1+ m
where g is the steady state growth rate of money supply. The Euler equation of capital decision
(13) and capital demand (14) imply
k
= : (21)
y 1= 1+
With resource constraint (16), we can obtain yc . Given the steady state labor n; we can obtain
solve the fk; y; cg : Moreover, (8), (9) and (11) imply
1
= : (22)
c (1 + Rb )
Then from (8), (9), and can be solved. Note that, as Rb > 0; (9) implies > . Furthermore,
(8) implies > 0: That is, the CIA constraint in the steady state is always binding as long as
the interest rate is greater than zero. The economic intuition is that, because holding extra cash
(which is larger than consumption) does not earn interest rate, therefore the household only hold
the amount of cash that just meets the consumption need. It is easy to show that, the steady state
values of other endogenous variables are easy to solve.
 4

2 Basic New Keynesian Model

As we have already seen, monetary policy plays little role in the basic monetary model. We now
turn to the New Keynesian (NK) model. The NK model has two basic features: (i) monoplistic
competition; (ii) price is sticky. The former speci…cation ensures that …rms make positive pro…t,
and the latter one implies that the price is not ‡exible and thus the money injection may play a
non-trivial role.

2.1 Basic Economic Environment

The economy has one …nal good which can be used either for consumption or investment. The …nal
good is produced by the …nal good …rms. In particular, the …nal good …rms use the intermediate
goods as input. The …nal good market is competitive. The intermediate goods …rms are monop-
olistic competitive, they have monopoly power to set their prices. However, the prices cannot be
‡exibly set. Therefore, the prices present some extent of stickness. Households are standard, just
as those in the CIA model. We now start with the household problem.

2.2 Household Problem

The problem of representative household is similar to the CIA model. In particular, their optimiza-
tion problem is de…ned as
X1
t
max (log ct an nt ) (23)
t=0
subject to
Mt + Bt Ht + X t ; (24)
Pt ct Mt ; (25)
Ht+1 = Mt + (1 + Rbt ) Bt + Wt nt + Rt kt + t Pt [ct + kt+1 (1 ) kt ] : (26)
The only di¤erence is that the pro…t of intermediate good …rms, t; appears in the above constraint.
This is because monopolistic …rms make positive pro…ts.

2.3 Final Good Sector

Final good market is competitive, the …rm combines a continuum intermediate goods yit as inputs
to produce …nal good yt : The production function is assumed to be
Z 1 1 1
yt = yit di : (27)
0
 5

The pro…t maximization problem is

Z 1
max Pt yt Pit yit di; (28)
yit 0

subject to (27). The optimal yit implies that the demand function of yit is

Pit
yit = yt : (29)
Pt

Putting last equation into the production function, we get the price indexation function
Z 1
Pt = Pit di: (30)
0

Note that as the …nal good market is competitive, with the CRS production function, the …rm
earns zero pro…t.

2.4 Intermediate Goods Sector

The intermediate goods sector is monopolistic. Firm i produces good i with Cobb-Douglas tech-
nology
1
yit = At kit nit : (31)

The real pro…t of …rm i is de…ned as

Pit
it = yit wt nit rt kit : (32)
Pt

where yit = PPitt yt : As the labor and capital decisions are static, the above pro…t function can
be reduced to (through a cost-minimization problem)

Pit Pit
it = t yt ; (33)
Pt Pt

where t is the marginal cost

1
1 wt rt
t = : (34)
At 1
The optimization problem for the …rm i is to set price pit to maximize the discounted pro…t ‡ows.
To model the price stickness, we follow Calvo (1982) assuming that in the period t, the …rm, with
probability 1 ; can set its price ‡exibly. With probability ; the …rm cannot set price and thus
the price remains the same as the previous period (t 1). The Bellman equation for the …rm that
can adjust its price is

t+1
V0;t = max it + Et [(1 ) V0;t+1 + V1;t+1 ] ; (35)
Pit t
 6

where V0;t is the value of the …rm that can adjust its price, and Vj;t is the value of …rm that adjust
its price j period ago and still cannot adjust its price in current period (t). For instance, V1;t+1 is
the value of …rm that adjust its price one period ago and still cannot adjust in period t+1. The
optimal price Pit for an active …rm is set to satisfy
@ it t+1 @V1;t+1
+ Et = 0: (36)
@Pit t @Pit
According to the pro…t function (33), the …rst term is
!
@ it 1 1 Pit Pit
= t yt : (37)
@Pit Pt PPit t
Pit
P
Pt Pt
t t

@V1;t+1
To derive the @Pit ; we need to specify the value function of inactive …rm who adjust the price j
periods ago.
t+1
Vj;t (Pit j) = it (Pit j ) + Et [(1 ) V0;t+1 + Vj+1;t+1 ] : (38)
t
Note that there is no max operator in the above Bellman equation because the …rm is inactive, it
just takes the previous price as today’s price. Taking derivative w.r.t. Pit j ; we have
@Vj;t @ it (Pit j ) t+1 @Vj+1;t+1
= + Et : (39)
@Pit j @Pit j t @Pit j
@Vj;t
From the above recursive structure of @Pit j ; we can derive

@Vj;t @ it (Pit j ) t+1 @ it+1 (Pit j ) t+2 @ it+2 (Pit j )

= + Et +( )2 Et + :::
@Pit j @Pit j t @Pit j t @Pit j
X t+ @ it+ (Pit j )
= ( ) Et :
t @Pit j
=0
@V1;t+1
For the @Pit ; we then have

@V1;t+1 X t+1+ @ it+1+ (Pit )

= ( ) Et+1 : (40)
@Pit t+1 @Pit
=0

Plugging last equation into (36), we have

X t+ @ it+ (Pit )
Et ( ) =0 (41)
t @Pit
=0
!
@ it+ (Pit ) 1 1 Pit Pit
where @Pit = Pt+ Pit Pit Pt+ t+ Pt+ yt+ : With some algebra, we
Pt+ t+ Pt+

obtain the optimal pricing rule:

X
Et ( ) t+ Pt+ yt+ t+
=0
Pit = Pt = X : (42)
1 Et ( ) t+ Pt+ 1 yt+
=0
 7

Note that the optimal price Pit is identical across …rm index i; that is, once …rms can adjust their
price, they set the same optimal level. As a result, by law of large number, the price indexation
function (30) implies
Pt = Pt 1 + (1 ) (Pt ) : (43)

Now, we start to derive the so-called New Keynesian Phillips Curve (NKPC). First, in the
steady state, (42) implies
P = : (44)
1
And (43) implies
P =P : (45)

Log-linearizing (42) gives

X
P^t = Et (1 ) ( ) ^ t+ + P^t+ + y^t+ + ^
t+
=0
X
Et (1 )( ) ^ t+ + ( 1) P^t+ + y^t+
=0
X
= Et (1 )( ) P^t+ + ^ t+
=0
X
= (1 ) L 1
P^t + ^ t
=0
1
= 1
P^t + ^ t (46)
1 L
or
1 L 1
P^t = (1 ) P^t + ^ t ; (47)

where L 1 (x ) = Et (xt+1 ) : In addition, from (43), we have

t

P^t = P^t 1 + (1 ) P^t : (48)

Combining last two equations, we have

(1 ) (1 )^
^ t = Et ^ t+1 + t; (49)

where t = P^t P^t 1 is the in‡ation rate.