Lecture 3: Monetary Policy Analysis with New

Keynesian Models

Prof. Engin Kara

2018-2019

Prof. E. Kara MRes Macroeconomics 2018-2019 1 / 45

Key reading:

Walsh (2010). Monetary Policy and Theory, Third edition, Chapter 8

(p. 352-364, 375-379)
Mankiw (2005). Commentary: Separating the Business Cycle from
other Economic Fluctuations.
I http://www.kansascityfed.org/
publicat/sympos/2005/pdf/Mankiw2005.pdf
Optional reading:
B. Sack (2000). Does the fed act gradually? A VAR analysis. Journal
of Monetary Economics 46 (2000) 229-256.

Prof. E. Kara MRes Macroeconomics 2018-2019 2 / 45

Introduction

Have focussed on constructing DSGE model that match key features

of aggregate time series.
To some extent this has been successfully achieved.
Can proceed to consider how monetary policy should be designed.
New Keynesian Phillips Curve

Prof. E. Kara MRes Macroeconomics 2018-2019 3 / 45

 πt = βEt πt+1 + κxt (1)
ω2
where κ = 1−ω γ
Euler Equation

xt = Et xt+1 − (it − Et πt+1 ) + ut + dt (2)

xt = yt − yt∗ and ut = Et yt+1
∗ − y∗
t

yt∗ = zt ,
where zt is a productivity shock. dt is a demand shock.

Prof. E. Kara MRes Macroeconomics 2018-2019 4 / 45

The Loss Function

We need to define some target for what monetary policy is trying to

achieve.
Traditionally, it has been assumed that monetary policy should
minimize some quadratic loss function which is given by

L = πt 2 + λxt 2 (3)
where λ is the relative weight placed on output variability relative to
inflation variability and xt = yt − yt∗ .

Prof. E. Kara MRes Macroeconomics 2018-2019 5 / 45

Not completely satisfactory objective for a central bank.
If the model is microfounded on the basis of a representative
optimizing agent then maximizing the welfare of that agent should be
the appropriate target for central bank policy.
This is what Woodford(2003) does.
Derive the central bank’s objective function directly from the utility of
the representative agent

Prof. E. Kara MRes Macroeconomics 2018-2019 6 / 45

 Z 1
Ut = u (Ct ) − V (1 − Hht ) dh (4)
0

Take a second order approximation of this to give an objective

function:

L = πt 2 + λxt 2 (5)

Prof. E. Kara MRes Macroeconomics 2018-2019 7 / 45

This is similar in structure to the ad hoc quadratic loss function,
although the match is not perfect.
λ = κθ . κ is coefficient on the output gap in the New Keynesian
Phillips curve

Prof. E. Kara MRes Macroeconomics 2018-2019 8 / 45

Policy Trade-offs

Solving (1) forward yields

∞
X
πt = κ βEt xt+i
i=0

By keeping current and expected future output zero (xt+i = 0, for all
i) , inflation remains equal to zero.
Blanchard described this as ”divine coincidence”.(see Mankiw(2005,
p. 191-192)

Prof. E. Kara MRes Macroeconomics 2018-2019 9 / 45

Demand shocks

prices and output move in the same direction.

Expansionary demand shocks: p ↑, (y − y ∗ ) ↑
Contractionary demand shocks: p ↓, (y − y ∗ ) ↓
Stabilizing prices stabilizes output and vice versa.

Prof. E. Kara MRes Macroeconomics 2018-2019 10 / 45

Productivity shocks

A positive productivity shock: p ↓, y ↑, y ∗ ↑

The central bank responds with more expansionary policy.
y = y∗
stabilizing prices stabilizes output at potential.

Prof. E. Kara MRes Macroeconomics 2018-2019 11 / 45

Cost push shocks

if an error term is added to the inflation adjustment equation so that

πt = βEt πt+1 + κxt + et (6)

∞
X ∞
X
πt = κ βEt xt+i + βEt et+i
i=0 i=0

Prof. E. Kara MRes Macroeconomics 2018-2019 12 / 45

 ∞
X ∞
X
As long as βEt et+i 6= 0, maintaining βEt xt+i = 0 is not sufficient
i=0 i=0
to ensure that inflation is remains equal to zero.
The divine coincidence does not arise.
There is a trade-off between stabilizing inflation and stabilizing
output.

Prof. E. Kara MRes Macroeconomics 2018-2019 13 / 45

Optimal Policy under Commitment

Assume that the central bank could commit today to a

(state-contingent) strategy that it can adopt now and in the future

∞
X
β t πt2 + λxt2 + µt (πt − βEt πt+1 − κxt − et )


L=
t=0

Prof. E. Kara MRes Macroeconomics 2018-2019 14 / 45

First-order conditions

λxt − κµt = 0
πt + µt − µt−1 = 0

for t=0,1,2,..where µ−1 = 0 (no time t=-1 constraint).

Woodford(2003): the timeless perspective.
We have

λ
µt = xt (7)
κ

Prof. E. Kara MRes Macroeconomics 2018-2019 15 / 45

And

λ
πt = − (xt − xt−1 ) (8)
κ
Change in the output gap!

Prof. E. Kara MRes Macroeconomics 2018-2019 16 / 45

Optimal Policy under Discretion

The central bank cannot commit to taking a particular course of

action.
The central bank adopts the optimal strategy for what to do today
and then tomorrow adopt the optimal strategy for what to do
tomorrow when it arrives and so on.
The central bank is unable to affect the period sector’s expectations
about future inflation.
Thus, the decision problem of the central bank becomes the
single-period problem of minimizing πt 2 + λxt 2 subject to the new
Keynesian Phillips curve.
Under discretion, we always have κπt + λxt = 0.
It can be shown that commitment policy produces a superior welfare
outcome to discretionary policy.

Prof. E. Kara MRes Macroeconomics 2018-2019 17 / 45

Discretion versus Commitment

The response of the output gap and inflation to a one-off cost-push

shock.
Under commitment: the output gap adjust sluggishly.
Et πt+1 falls below zero.
Improves the trade-off between output gap and inflation faced by the
central bank.
Under discretion: no inertia

Prof. E. Kara MRes Macroeconomics 2018-2019 18 / 45

Intuition:
I under discretion the central bank’s only tool is the output gap.
I Under commitment: two instruments: it can affect: xt and Et πt+1 .
F the central bank promises a deflation. Thus, the fall in the output gap
does not need to be as large.

Prof. E. Kara MRes Macroeconomics 2018-2019 19 / 45

 F ig 8 . 2 : O u t p u t g a p : c o m m it m e n t ( s o lid ) a n d d is c r e t io n ( d o t te d )
0 .5

- 0 .5
0 2 4 6 8 10 12 14 16

– the central bank promises a de‡ation. Thus, the fall in the output gap
does not need to be as large.

Prof. E. Kara MRes Macroeconomics 2018-2019 20 / 45

 F i g 8 . 3 : I n f l a t i o n : c o m m i t m e n t ( s o l i d ) a n d d i s c r e ti o n ( d o t t e d )
1 .5

0 .5

- 0 .5
0 2 4 6 8 10 12 14 16

Prof. E. Kara MRes Macroeconomics 2018-2019 21 / 45

 P r i c e : c o m m i t m e n t ( s o l i d ) a n d d i s c r e ti o n ( d o tte d )

1 .4

1 .2

0 .8

0 .6

0 .4

0 .2
0 2 4 6 8 10 12 14 16

Prof. E. Kara MRes Macroeconomics 2018-2019 22 / 45

Optimal simple rules

The full optimal rule as calculated by Rotemberg and Woodford

rt = 0.29rt−1 + 0.29rt−2 + 1.28rt−3 + 0.22yt−1 − 0.25yt−2 + ...

+ 0.16πt + 1.00πt−1 + 2.45πt−2 − 1.45πt−3 + 0.74πt−4
− 0.08πt−5 + 0.25πt−6 + 0.33πt−7 + 0.23πt−8 + 0.25πt−9
+ 0.19πt−10 + ....

Prof. E. Kara MRes Macroeconomics 2018-2019 23 / 45

18 lags of output and 49 lags of inflation.
raises the question about the operational feasibility.
The optimal monetary policy problem is to find the coefficients µ, φ, θ
and the correct lags ηr , ηπ , ηy .
The central bank is assumed to set interest rates as a according to a
rule of the form:
ηr ηπ ηy
X X X
rt = r ∗ + µk (rt−k ) + φk (πt−k ) + θk (yt−k )
k=1 k=1 k=1

Prof. E. Kara MRes Macroeconomics 2018-2019 24 / 45

 What is the best monetary policy?
What the µ, φ, θ coefficients in the monetary policy should be?
Simple monetary policy rules
Taylor(1993) suggested a very simple rule for setting interest rate

rt = 1.5πt + 0.5yt

Prof. E. Kara MRes Macroeconomics 2018-2019 25 / 45

Prof. E. Kara MRes Macroeconomics 2018-2019 26 / 45
Taylor:”simple enough to put on the back of a business card”
Such a rule may not be optimal.
Why would a central bank ever want to follow this rule
may be a good approximation to the optimal rule.

Prof. E. Kara MRes Macroeconomics 2018-2019 27 / 45

certainly it is a good approximation to the behaviour of the FED over
the 40 years.
This suggest that the Taylor rule is a good approximation of the
behaviour of the FED.
Whether this implies optimality depends on whether the FED is
believed to have acted optimally over the period.

Prof. E. Kara MRes Macroeconomics 2018-2019 28 / 45

Monetary Policy with Uncertainty

In recent times it has been common for central banks to acknowledge

that they do not have perfect information about the economy.
”Uncertainty is not just an important feature of the monetary policy
landscape; it is the defining characteristic of that landscape.”

Alan Greenspan (2003)

Prof. E. Kara MRes Macroeconomics 2018-2019 29 / 45

uncertainty surrounding the precise value of the key parameters and
fundamental uncertainties regarding the correct model.
We will focus on parameter uncertainty.
discuss how the nature of monetary policy changes when such
uncertainty is accounted for.

Prof. E. Kara MRes Macroeconomics 2018-2019 30 / 45

Brainard conservatism

Brainard(1967) shows when there is uncertainty about the impact a

policy instrument has on the economy, it would be optimal to respond
more cautionsly than would be the case in the absence of uncertainty.
If there is uncertainty about the parameters of the model then the
central bank should not behave as if the uncertainty does not exist.
”Brainard uncertainty principle”
This conclusion can illustrated with a simple example.

Prof. E. Kara MRes Macroeconomics 2018-2019 31 / 45

 πt = βEt πt+1 + κt xt + et (9)
where κt = κ + υt and υt is a white noise stochastic process.
The central bank is uncertain about the true impact of the gap xt on
inflation.
e.g. they may be great deal of uncertainty about the estimates of κt .

Prof. E. Kara MRes Macroeconomics 2018-2019 32 / 45

Loss function is the same as before

L = πt 2 + λxt 2 (10)

Prof. E. Kara MRes Macroeconomics 2018-2019 33 / 45

Under discretion, the FOC for the optimal; choice of xt is

Et (κt πt + λxt ) = 0 (11)

All stochastic shocks are assumed to be serially uncorrelated.

Therefore, Et πt+1 = 0.

πt = κt xt + et

Prof. E. Kara MRes Macroeconomics 2018-2019 34 / 45

Using this, can be written as

Et (κt ((κt ) xt + et ) + λxt ) = 0

Et ((κ + υt ) ((κ + υt ) xt + et ) + λxt ) = 0
κ̄2 + συ2 xt + κ̄et + λxt = 0

Note that Et (κ) = κ̄ and Et υt2 = σu2 . The two shocks are assumed to

be uncorrelated.
Solving for xt
 
κ̄
xt = − et (12)
κ̄2 + συ2 + λ

Prof. E. Kara MRes Macroeconomics 2018-2019 35 / 45

Compare this to the case when there is no parameter uncertainty (συ2 = 0)
 
κ̄
xt = − et
κ̄2 + λ
   
κ̄ κ̄
<
κ̄2 + συ2 + λ κ̄2 + λ
Thus, the presence of multiplicative parameter uncertainty reduces the
impact of et on xt .
As uncertainty increases, it becomes optimal to respond less to et (i.e.
behave more cautiously in setting policy).
Using (12) in (9)

Prof. E. Kara MRes Macroeconomics 2018-2019 36 / 45

 πt = κt xt + et
 
κ̄
=− et κt + et
κ̄2 + συ2 + λ
−κ̄κt + κ̄2 + συ2 + λ
 
= et
κ̄2 + συ2 + λ
 2
κ̄ + συ2 + λ − κ̄ (κt − κ̄)

= et
κ̄2 + συ2 + λ
 2 
συ + λ − κ̄ ((κ + υt ) − κ̄)
= et
κ̄2 + συ2 + λ
 2 
συ + λ − κ̄υt
= et
κ̄2 + συ2 + λ

Prof. E. Kara MRes Macroeconomics 2018-2019 37 / 45

In the presence of multiplicative uncertainty output is stabilized more
and inflation less.

Prof. E. Kara MRes Macroeconomics 2018-2019 38 / 45

Proof.
(not for exam)
From ((12) , the variance of x is
 2
κ̄
σx2 =− et2
κ̄ + συ2 + λ
2

which is clearly decreasing in σx2 .

Hence, greater uncertainty in this case leads output to be stabilized more.

Prof. E. Kara MRes Macroeconomics 2018-2019 39 / 45

Proof.
(not for exam) Inflation is given by
 2 
συ + λ − κ̄υt
πt = et = Bet
κ̄2 + συ2 + λ
and the variance of inflation is E B 2 e 2 = E B 2 σe2 as υ and e are

independent.

σπ2 = E B 2 σe2 is increasing in συ2 implying inflation is stabilized less.

Prof. E. Kara MRes Macroeconomics 2018-2019 40 / 45

The Brainard basic result: multiplicative uncertainty leads to caution.
his is often referred to as Brainard conservatism-parameter
uncertainty introduces a motive for caution in optimal policy.
Such a policy means that the central bank does not expect to achieve
its inflation target.
The reason is that aiming to hit the target exactly involves large
potential losses.
Is this a general result?

Prof. E. Kara MRes Macroeconomics 2018-2019 41 / 45

The answer is no.
Soderstrom(2002) examined a model in which there are lagged
variable whose coefficient are subject to random shocks.
He showed that optimal policy reacts more aggressively.
Suppose πt depends on πt−1 but the impact of πt−1 on πt is
uncertain.
If the central bank fails to stabilize, it increases the variance of
inflation in the following period.
Can be optimal to respond more aggressively to stabilize inflation.

Prof. E. Kara MRes Macroeconomics 2018-2019 42 / 45

Is Brainard uncertainty empirically relevant?

Sack(2000, p. 251) calculates the impulse response functions implied

by the optimal rules with and without parameter uncertainty and
compares these to those estimated purely from the data.

Prof. E. Kara MRes Macroeconomics 2018-2019 43 / 45

Prof. E. Kara MRes Macroeconomics 2018-2019 44 / 45
The figure shows how the optimal policy with parameter uncertainty
tracks the federal funds rate better,
suggests that caution induced by Brainard uncertainty is
quantitatively important.
Optimal policy rules exhibit considerable persistence in interest
rate-interest rate (interest rate smoothing: a central bank should
adjust interest rates gradually rather than aggressively.)

Prof. E. Kara MRes Macroeconomics 2018-2019 45 / 45