Monetary Economics

Problem Set #1

Monetary Economics: Problem Set #1

Due Thursday August 7th in class

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. A classical real economy. A representative household maximises utility

C1 N 1+'
U (C, N ) = , ,' > 0
1 1+'
subject to the budget constraint
PC  WN
A representative perfectly competitive firm has the linear production function

Y = AN

(a) Solve for the equilibrium levels of consumption C, labor N , output Y and the real wage
W/P in terms of productivity A and the other parameters. (20 points)
(b) Suppose log productivity a = log A fluctuates randomly with variance Var{a} = 1. Cal-
culate the variances of log consumption c, log labor n, log output y, and the log real wage
w p. Which of these variables is more volatile than productivity? Which is less volatile?
Which of these variables is positively correlated with productivity? Which is negatively
correlated? How do your answers depend on the preference and parameters and '? Give
economic intuition for all your answers. (20 points)

2. Real interest rates in the classical model. Consider a classical model with the following
(log-linearised) household optimality conditions:
1
ct = (rt ⇢) + Et [ct+1 ], ⇢, >0 (1)

and
ct + 'nt = wt pt , '>0 (2)
Perfectly competitive firms choose labor demand to maximise profits subject to the (log-linear)
production function
y t = at + n t
where {at } is log productivity which follows an AR(1) process

at+1 = ⇢a at + ✏t+1 , 0  ⇢a < 1

where {✏t } is an IID white noise shock.

Monetary Economics: Problem Set #1 2

(a) Explain in words the economic interpretation of equations (1) and (2). (15 points)
(b) Solve for the equilibrium levels of (log) consumption ct , employment nt and output yt in
terms of productivity at and the exogenous parameters. Briefly explain the e↵ects of a
positive productivity shock on each of these endogenous variables. Give intuition for all
your answers. (15 points)
(c) Solve for the equilibrium real interest rate rt in terms of the productivity process and other
parameters. Does an increase in productivity increase or decrease the real interest rate?
Does a higher value of increase or decrease the sensitivity of the real interest rate to a
productivity shock? Explain. (15 points)
(d) Suppose the productivity process is instead a random walk with drift >0

at+1 = + at + ✏t+1

Does an increase in productivity increase or decrease the real interest rate? Does a higher
value of increase or decrease the sensitivity of the real interest rate to a productiv-
ity shock? Explain the di↵erences, if any, between your answers for parts (c) and (d).
(15 points)
 Monetary Economics
Problem Set #2

Monetary Economics: Problem Set #2

Due Thursday August 28th in class

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. Inflation targeting with noisy data. Consider a new Keynesian model with output gap and
inflation dynamics governed by
1
ỹt = (it Et {⇡t+1 } rtn ) + Et {ỹt+1 }

and
⇡t = Et {⇡t+1 } + ỹt
where all variables have their usual meanings. The natural rate of interest follows an exogenous
AR(1) process
n
rt+1 ⇢ = ⇢r (rtn ⇢) + ✏t+1
with 0  ⇢r < 1 and where {✏t } is IID white noise with mean zero.
Now suppose that inflation is observed with measurement error so that the central bank sees
only a noisy signal of actual inflation

⇡t0 = ⇡t + ⇠t

where ⇡t0 denotes the observed inflation rate, ⇡t the actual inflation rate, and where {⇠t } is IID
white noise with mean zero. Assume the central bank follows the feedback rule
0
it = ⇢ + ⇡ ⇡t (1)

in terms of observed inflation.

(a) Use the method of undetermined coefficients to solve for the equilibrium processes for
inflation and the output gap under the interest rate rule. (20 points)
(b) Describe the behavior of inflation, the output gap, and the nominal interest rate when
⇡ ! 1. Give economic intuition for your answers. (15 points)

To simplify the algebra, assume for the rest of this question that ⇢r = 0.

(c) Determine the value of the feedback coefficient ⇡ that minimises the variance of actual
inflation. Give economic intuition for your answer. (15 points)
Monetary Economics: Problem Set #2 2

2. Monetary policy and the e↵ects of productivity shocks. Consider a new Keynesian
model with equilibrium conditions
1
yt = (it Et {⇡t+1 } ⇢) + Et {yt+1 } (2)

and
⇡t = Et {⇡t+1 } + (yt ytn ) (3)
where all variables have their usual meanings. Monetary policy is given by the feedback rule

it = ⇢ + ⇡ ⇡t

where ⇡ > 1. The production function (in logs) is

y t = at + n t

where at is an exogenous labor productivity process that follows an AR(1) process

at+1 = ⇢a at + ✏t+1

with 0  ⇢a < 1 and where {✏t } is IID white noise with mean zero. Natural output is propor-
tional to productivity
ytn = y at
where y > 0.

(a) Describe in words the economic interpretation of equations (2) and (3). (10 points)
(b) Use the method of undetermined coefficients to solve for the equilibrium response of output,
employment, and inflation to a productivity shock. (20 points)
(c) Describe how these responses depend on the values of the parameters ⇡ and . What
happens when ⇡ ! 1? What happens as the degree of price rigidities changes? Provide
economic intuition for your answers. (10 points)
(d) Discuss with as much detail as you can the joint response of employment and output to
a productivity shock and discuss the implications for assessing the role of productivity
shocks as a source of business cycle fluctuations in this model. (10 points)
Hint: you may want to skim Gali’s 1999 paper “Technology, Employment, and the Business
Cycle: Do Technology Shocks Explain Aggregate Fluctuations?” (posted on the LMS).
 Monetary Economics
Problem Set #3

Monetary Economics: Problem Set #3

Due Thursday September 11th in class

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. Policy tradeo↵s in the new Keynesian model. Consider a new Keynesian model with
output gap and inflation given by
1
ỹt = (it Et [⇡t+1 ] ⇢) + Et [ỹt+1 ] (1)

and
⇡t = Et [⇡t+1 ] + ỹt + xt (2)
where {xt } is an exogenous shock. Monetary policy is given by the interest rate rule

it = ⇢ + ⇡ ⇡t + vt

where {vt } is an exogenous monetary policy shock and is independent of {xt }.

(a) Explain in words the economic interpretation of equations (1) and (2). How could you
interpret the xt shock? (10 points)

To simplify the algebra, assume for the rest of this question that = 1 and that {xt } and {vt }
are both IID white noise.

(b) Solve for equilibrium inflation ⇡t and the output gap ỹt in terms of the shocks vt and xt .
(20 points)
(c) Explain how inflation, interest rates and the output gap respond to the vt and xt shocks.
Give economic intuition for all your answers. (10 points)
(d) Suppose the central bank chooses the feedback coefficient ⇡ to minimise the loss function

L = Var[⇡t ] + Var[ỹt ] (3)

Solve for the value of ⇡ that minimises this loss function. Is there a policy “trade-o↵”
here? Explain how your answer depends on the parameter  and on the variances of the
shocks, Var[vt ] and Var[xt ]. Give economic intuition for your answers. (15 points)
(e) Does the value of ⇡ that minimises the loss function (3) satisfy the Taylor principle? Why
or why not? (5 points)
Monetary Economics: Problem Set #3 2

2. Interest rate versus money supply rules. Consider an economy again described by the
equilibrium conditions
1
ỹt = (it Et [⇡t+1 ] rtn ) + Et [ỹt+1 ]
⇡t = Et [⇡t+1 ] + ỹ

and now also a money demand equation of the form

mt pt = yt ⌘it , ⌘>0

where all variables are defined as usual. Both ytn and rtn evolve exogenously, independent of
monetary policy. The central bank seeks to minimize a loss function of the form

L = Var[⇡t ] + Var[ỹt ]

(a) Explain how the optimal monetary policy outcomes can be implemented by an interest
rate feedback rule. (10 points)
(b) Show that a constant money supply will generally not be optimal. (15 points)
(c) Derive a money supply rule that would implement the optimal monetary policy. (15 points)
 Monetary Economics
Problem Set #4

Monetary Economics: Problem Set #4

Due Thursday September 25th in class

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. Government purchases in the new Keynesian model. Consider a basic new Keynesian
model with the following (log-linearised) equilibrium conditions: for consumption, a dynamic
Euler equation
1
ct = (it Et [⇡t+1 ] ⇢) + Et [ct+1 ]
for labor supply
wt pt = ct + 'nt
the production function for firms
yt = nt
If prices were fully flexible, firms would set a constant markup over marginal cost. With sticky
prices, there is a new Keynesian Phillips curve in terms of the output gap

⇡t = Et [⇡t+1 ] + ỹt

The government purchases a fraction ⌧t of output each period with ⌧t varying exogenously.

(a) Derive a log-linear version of the goods market clearing equation of the form yt = ct + gt
where gt ⌘ log(1 ⌧t ). (5 points)
(b) Show that natural output is proportional to government purchases, ytn = gt , and give an
explicit formula for the coefficient . Does a fiscal expansion increase or decrease natural
output when prices are fully flexible? By how much? Explain. (10 points)

Now assume that monetary policy is set according to the simple interest rate rule

it = ⇢ + ⇡ ⇡t + y ỹt

and that government purchases gt follow an AR(1) process with persistence 0  ⇢g < 1.

(c) Show that the output gap ỹt satisfies a dynamic IS curve of the form
1
ỹt = (it Et [⇡t+1 ] rtn ) + Et [ỹt+1 ]

and derive a formula for the natural real rate rtn in terms of gt . (5 points)
(d) Use the method of undetermined coefficients to solve for the response of the key endogenous
variables —output, natural output, employment, inflation, interest rates— to an exogenous
increase in government purchases gt . Give economic intuition for your answers. (20 points)
Monetary Economics: Problem Set #4 2

(e) Explain how the response of output to government purchases depends on the monetary
policy coefficient ⇡ . Does a higher ⇡ increase or decrease the impact e↵ect of govern-
ment purchases on output? Similarly, explain how the response of output to government
purchases depends on the AR(1) persistence ⇢g . Does a more persistent process increase
or decrease the impact e↵ect of government purchases on output? Give economic intuition
for your answers. (10 points)

2. Multipliers, the ZLB, and the duration of fiscal stimulus. Consider a new Keynesian
model with government purchases gt and shocks t to the interest rate facing households
1
ỹt = (it + t Et [⇡t+1 ] rtn ) + Et [ỹt+1 ]

⇡t = Et [⇡t+1 ] + ỹt
Monetary policy is set according to an interest rate rule but also faces a zero lower bound

it = max[0, ⇢ + ⇡ ⇡t + y ỹt ]

The natural real rate and natural output are given by

rtn = ⇢ (1 )Et [ gt+1 ], ytn = gt

The interest spread shock t can take on two values L , H with H = 0 and L > 0. The
economy starts in the L state. With probability ↵ it stays in the L state. With probability
1 ↵ it transitions to the H state. Once it enters the H (“normal”) state it stays there forever.
Suppose that gH = 0. We are interested in calculating economic outcomes as a function of fiscal
policy gL in the L (“crisis”) state.

(a) To begin with, suppose that gL = 0. Solve for the equilibrium values of inflation, the
output gap, and the nominal interest rate in the L state. (5 points)
(b) Now explain how an increase in government purchases to some gL > 0 would a↵ect in-
flation, expected inflation, output and the nominal and real interest rates in the L state.
Explain how your answers depend on the size of the interest spread L . How large is
the government purchases multiplier? Does this value depend on the the size of the fiscal
stimulus? (15 points)

Now consider the possibility that government purchases gL persist at an elevated level after the
crisis has abated. To be specific, imagine that there are three states, L, S, H. The economy
starts in the L state with L > 0. Suppose the initial interest spread L is sufficiently high that
the ZLB is binding in the L state. With probability ↵ the economy transitions to the S state.
In the S (“transitional”) state, the crisis is over S = H = 0. In the S state, with probability
the fiscal stimulus continues with government purchases gS = gL > 0. With probability 1
the fiscal stimulus ends with gS = gH = 0.

(c) Let ⇡S , ỹS , iS denote the equilibrium values of inflation, the output gap and the nominal
interest rate in the S state. Following the same logic as in part (a), solve for these
equilibrium values as a function of the size of the fiscal stimulus gL > 0. Explain how
your answers depend on whether the fiscal stimulus continues gS = gL , or not gS = gH .
(15 points)
Monetary Economics: Problem Set #4 3

(d) Using your results from part (c), now solve for the equilibrium values of inflation, the
output gap and the nominal interest rate in the initial L state given that with probability
the fiscal stimulus continues after the crisis has abated. How does the government
purchases multiplier compare to the one you found in part (b)? How does the multiplier
vary with the probability of the stimulus continuing? Explain. (15 points)

[Hint: this question is based on Woodford’s article “Simple Analytics of the Government Ex-
penditure Multiplier” American Economic Journal: Macroeconomics. 3(1): 1–35, especially
Section IV.B]
 Monetary Economics
Problem Set #1

Monetary Economics: Problem Set #1

Solutions

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. A classical real economy. A representative household maximises utility

C1 N 1+'
U (C, N ) = , ,' > 0
1 1+'
subject to the budget constraint
PC  WN
A representative perfectly competitive firm has the linear production function

Y = AN

(a) Solve for the equilibrium levels of consumption C, labor N , output Y and the real wage
W/P in terms of productivity A and the other parameters. (20 points)
(b) Suppose log productivity a = log A fluctuates randomly with variance Var{a} = 1. Cal-
culate the variances of log consumption c, log labor n, log output y, and the log real wage
w p. Which of these variables is more volatile than productivity? Which is less volatile?
Which of these variables is positively correlated with productivity? Which is negatively
correlated? How do your answers depend on the preference parameters and '? Give
economic intuition for all your answers. (20 points)

Solutions:

(a) An equilibrium is characterized by household labor supply

Un W
=
Uc P
firm labor demand
W
=A
P
and market clearing
C = Y = AN
With these preferences, the household marginal rate of substitution is Un /Uc = N ' C
and so combining the household labor supply and firm labor demand conditions we get
W
N 'C = =A
P
Monetary Economics: Problem Set #1 2

Then substituting for C using the budget constraint we get one equation in one unknown,
equilibrium employment N ,
N ' (AN ) = A
Solving this for equilibrium employment in terms of the exogenous level of productivity
1
N = A '+

and hence equilibrium consumption and output are, from market clearing,
1 '+1
Y = C = AN = AA '+ = A '+

and of course the real wage is just W/P = A.

(b) Taking logs of all the endogenous variables, we have
1
n= a
'+
and
'+1
c=y= a
'+
and the log real wage w p = a. Now taking variances of all the endogenous variables
✓ ◆2 ✓ ◆2
1 1
Var[n] = Var[a] =
'+ '+

(since Var[a] = 1, by assumption) and similarly

✓ ◆2
'+1
Var[c] = Var[y] =
'+

and the variance of the log real wage is 1 since the real wage is equal to productivity.
Employment is more volatile than productivity if and only if
1 '
<
2
and otherwise employment is less volatile than productivity. To understand this condition,
first observe that with the linear production function labor demand is perfectly horizontal
(i.e., firms are willing to hire any amount of labor at real wage w p = a), so the
relative volatility of employment is determined by the steepness of the labor supply curve.
In particular, if the labor supply curve is sufficiently vertical (a sufficiently low Frisch
elasticity of labor supply, or equivalently a sufficiently high '), then employment does not
vary much in response to productivity. What makes for ' sufficiently high? Since > 0,
any ' > 1 will make employment less volatile than productivity.
Employment is perfectly positively correlated with productivity if < 1 and perfectly
negatively correlated with productivity if > 1. If = 1, then employment is constant and
hence uncorrelated with productivity. The intuition here is that if < 1, the substitution
e↵ect in labor supply dominated and employment rises with the real wage which rises with
productivity. If > 1, the income e↵ect of a higher real wage dominates and employment
Monetary Economics: Problem Set #1 3

falls when the real wage rises. If = 1 (i.e., log utility) then the income and substitution
e↵ects exactly balance.
Consumption and output are always perfectly positively correlated with productivity. They
are more volatile than productivity if
'+1
>1
'+
equivalently, if
1>
Intuitively, if high productivity calls forth more labor supply (the substitution e↵ect dom-
inates) then output and consumption will fluctuate more than productivity.

2. Real interest rates in the classical model. Consider a classical model with the following
(log-linearised) household optimality conditions:
1
ct = (rt ⇢) + Et [ct+1 ], ⇢, >0 (1)

and
ct + 'nt = wt pt , '>0 (2)
Perfectly competitive firms choose labor demand to maximise profits subject to the (log-linear)
production function
y t = at + n t
where {at } is log productivity which follows an AR(1) process

at+1 = ⇢a at + ✏t+1 , 0  ⇢a < 1

where {✏t } is an IID white noise shock.

(a) Explain in words the economic interpretation of equations (1) and (2). (15 points)
(b) Solve for the equilibrium levels of (log) consumption ct , employment nt and output yt in
terms of productivity at and the exogenous parameters. Briefly explain the e↵ects of a
positive productivity shock on each of these endogenous variables. Give intuition for all
your answers. (15 points)
(c) Solve for the equilibrium real interest rate rt in terms of the productivity process and other
parameters. Does an increase in productivity increase or decrease the real interest rate?
Does a higher value of increase or decrease the sensitivity of the real interest rate to a
productivity shock? Explain. (15 points)
(d) Suppose the productivity process is instead a random walk with drift >0

at+1 = + at + ✏t+1

Does an increase in productivity increase or decrease the real interest rate? Does a higher
value of increase or decrease the sensitivity of the real interest rate to a productiv-
ity shock? Explain the di↵erences, if any, between your answers for parts (c) and (d).
(15 points)
Monetary Economics: Problem Set #1 4

Solutions:

(a) Equation (1) is the representative household’s log-linearised intertemporal Euler equation
that governs consumption smoothing. Implicitly the representative household is being
assumed to have a separable period utility function, since only consumption (and not, say,
labor) enters the Euler equation. Moreover utility from consumption is of the CRRA form
with coefficient . Notice that a higher real rate rt induces a lower level of consumption
ct (other things equal), while a real interest rate greater than the constant rate of time
preference, rt > ⇢, induces consumption that is growing in expectation
rt ⇢
Et [ ct+1 ] =

with interest sensitivity given by 1/ , i.e., by the constant intertemporal elasticity of sub-
stitution. Equation (2) is the representative household’s labor supply condition, equating
the marginal rate of substitution between labor and consumption to the real wage. The
particular log-linear form here again indicates a period utility function that is separable
between consumption and labor with curvature over consumption again given by and cur-
vature over labor given by '. In this context, 1/' is the so-called Frisch (or “ -constant”)
elasticity of labor supply. This terminology comes from writing the labor supply condition
1
nt = (wt pt ) ct
' '
so that 1/' is the elasticity of labor supply with respect to the real wage holding fixed the
marginal utility of consumption, i.e., ignoring the wealth e↵ect of changing real wages.
(b) Firm labor demand is governed by

wt p t = at

so that the real wage is equated to the marginal product of labor. Then using the household
labor supply condition
ct + 'nt = wt pt = at
The market clearing condition is ct = yt and from the production function yt = at + nt so
that employment solves
(at + nt ) + 'nt = at
or
1
nt = at
'+
As usual, employment responds positively to an increase in productivity if the substitution
e↵ect dominates the income e↵ect, < 1, and responds negatively to a productivity
shock if the income e↵ect dominates the substitution e↵ect, > 1. With the solution for
employment in hand, output and consumption are
'+1
c t = y t = at + n t = at
'+
Both consumption and output respond positively to an increase in productivity, the re-
sponse is more than 1-for-1 if employment increases and less than 1-for-1 if employment
decreases but the e↵ect is always positive.
Monetary Economics: Problem Set #1 5

(c) Let denote the elasticity of output with respect to productivity, that is
'+1
⌘
'+
so that ct = yt = at . Then from the consumption Euler equation the real interest rate rt
satisfies
1
at = (rt ⇢) + Et [ at+1 ]

Solving for the real interest rate rt we get

rt = ⇢ + Et [ at+1 ]

where at+1 = at+1 at is productivity growth. Using the AR(1) process for productivity,
the conditional expectation of productivity is Et [at+1 ] = ⇢a at so that

rt = ⇢ + Et [ at+1 ] = ⇢ + (⇢a 1)at

Since ⇢a < 1, the real interest rate falls when productivity increases. The intuition for this
is that when productivity increases above average, it is then expected to fall back towards
its mean value (the AR(1) is mean-reverting) so that expected consumption growth is
negative. In order for consumption growth to slow, the real interest rate falls. For a given
, a higher value of makes real interest rates more sensitive to consumption growth,
precisely because higher makes consumers less interest sensitive. An increase in also
has the e↵ect of reducing , making the level of consumption less sensitive to productivity,
but the net e↵ect is given by the product

= (' + 1)
'+
which is clearly increasing in .
(d) The real interest rate is still given by

rt = ⇢ + Et [ at+1 ]

But now expected productivity growth is Et [ at+1 ] = , so that the real interest rate is a
constant
r =⇢+
The real interest rate is a constant because the current level of productivity at does not
change the forecast of productivity growth between t and t + 1, i.e., there is no mean-
reversion in productivity. The e↵ects of are essentially the same as in part (b) in that
the response of the real rate r to growth is given by the product = (' + 1) /(' + )
which is increasing in so that the real interest rate responds more to growth when
consumers are less interest sensitive.
 Monetary Economics
Problem Set #2

Monetary Economics: Problem Set #2

Solutions

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. Inflation targeting with noisy data. Consider a new Keynesian model with output gap and
inflation dynamics governed by
1
ỹt = (it Et {⇡t+1 } rtn ) + Et {ỹt+1 }

and
⇡t = Et {⇡t+1 } + ỹt
where all variables have their usual meanings. The natural rate of interest follows an exogenous
AR(1) process
n
rt+1 ⇢ = ⇢r (rtn ⇢) + ✏t+1
with 0  ⇢r < 1 and where {✏t } is IID white noise with mean zero.
Now suppose that inflation is observed with measurement error so that the central bank sees
only a noisy signal of actual inflation

⇡t0 = ⇡t + ⇠t

where ⇡t0 denotes the observed inflation rate, ⇡t the actual inflation rate, and where {⇠t } is IID
white noise with mean zero. Assume the central bank follows the feedback rule
0
it = ⇢ + ⇡ ⇡t (1)

in terms of observed inflation.

(a) Use the method of undetermined coefficients to solve for the equilibrium processes for
inflation and the output gap under the interest rate rule. (20 points)
(b) Describe the behavior of inflation, the output gap, and the nominal interest rate when
⇡ ! 1. Give economic intuition for your answers. (15 points)

To simplify the algebra, assume for the rest of this question that ⇢r = 0.

(c) Determine the value of the feedback coefficient ⇡ that minimises the variance of actual
inflation. Give economic intuition for your answer. (15 points)
Monetary Economics: Problem Set #2 2

Solutions:

(a) Let r̂tn ⌘ rtn ⇢ and substitute in the interest rate rule so that we can reduce the system
to
1
ỹt = ( ⇡ (⇡t + ⇠t ) Et {⇡t+1 } r̂tn ) + Et {ỹt+1 }

and
⇡t = Et {⇡t+1 } + ỹ
Now guess solutions of the form

ỹt = 'yr r̂tn + 'y⇠ ⇠t , and ⇡t = '⇡r r̂tn + '⇡⇠ ⇠t

for some as-yet-unknown coefficients 'yr , 'y⇠ , '⇡r , '⇡⇠ to be determined. Since the natural
rate follows an AR(1) with coefficient ⇢r and ⇠t is IID white noise with mean zero, these
solutions imply the conditional expectations

Et {ỹt+1 } = 'yr ⇢r r̂tn , and Et {⇡t+1 } = '⇡r ⇢r r̂tn

So we can write the system of two equations in terms of the four unknown coefficients and
the shocks r̂tn , ⇠t , specifically
1
'yr r̂tn + 'y⇠ ⇠t = ( n
⇡ ('⇡r r̂t + '⇡⇠ ⇠t + ⇠t ) '⇡r ⇢r r̂tn r̂tn ) + 'yr ⇢r r̂tn

and
'⇡r r̂tn + '⇡⇠ ⇠t = '⇡r ⇢r r̂tn + ('yr r̂tn + 'y⇠ ⇠t )
Collecting terms in the shocks we have
 
1 1
0 = 'yr + ( ⇡ '⇡r '⇡r ⇢r 1) 'yr ⇢r r̂tn + 'y⇠ + ⇡ ('⇡⇠ + 1) ⇠t

and
0 = ['⇡r '⇡r ⇢r 'yr ] r̂tn + ['⇡⇠ 'y⇠ ] ⇠t
Since these conditions have to hold for any realizations of the shock r̂tn , ⇠t the terms in
square brackets must each be zero so that we are left with four conditions in the four
unknown coefficients. Doing the algebra gives the solutions
1 ⇢r
'yr = >0
(1 ⇢r )(1 ⇢r ) + ( ⇡ ⇢r )

'⇡r = >0
(1 ⇢r )(1 ⇢r ) + ( ⇡ ⇢r )

(assuming ⇡ > 1, as usual) and

⇡
'y⇠ = <0
+ ⇡
 ⇡
'⇡⇠ = <0
+ ⇡
Monetary Economics: Problem Set #2 3

(b) When ⇡ ! 1, these coefficients simplify to

'yr = 0
'⇡r = 0

and, using l’Hopital’s rule, also

1
'y⇠ = <0

'⇡⇠ = 1

Therefore in this limit, equilibrium inflation is

⇡t = '⇡r r̂tn + '⇡⇠ ⇠t = 0 r̂tn 1 ⇠t = ⇠t

hence equilibrium inflation is perfectly negatively correlated with the measurement error.
This is because observed the monetary authority is very reactive and inflation is ⇡t0 ⌘
⇡t + ⇠t = 0 in this limit (i.e., the central bank is successfully keeping observed inflation at
zero). Similarly, in this limit the equilibrium output gap is
1 1
ỹt = 'yr r̂tn + 'y⇠ ⇠t = 0 r̂tn ⇠t = ⇠t
 
Calculating the equilibrium nominal interest rate is slightly more tricky. From the interest
rule evaluated at the equilibrium observed inflation rate
0
it = ⇢ + ⇡ ⇡t =⇢+ ⇡ (⇡t + ⇠t )

Plugging in for equilibrium inflation

it = ⇢ + ⇡ (⇡t + ⇠t )
n
=⇢+ ⇡ '⇡r r̂t + ⇡ ('⇡⇠ + 1)⇠t

Now the coefficient on the natural real rate is


⇡ '⇡r = ⇡ ! 1 as ⇡ !1
(1 ⇢r )(1 ⇢r ) + ( ⇡ ⇢r )
And the coefficient on the measurement error is
✓ ◆
 ⇡
⇡ ('⇡⇠ + 1) = ⇡ 1 = ⇡ ! as ⇡ !1
+ ⇡ + ⇡ 
Therefore in this limit, the nominal interest rate is

it = ⇢ + r̂tn + ⇠t = ⇢ + (rtn ⇢) + ⇠t = rtn + ⇠t

  
In the absence of measurement error, this would just be the natural real rate as usual.
(c) When the natural rate follows an AR(1), as in parts (a)-(b) above, the variance of inflation
is ✓ ◆2 ✓ ◆2
 n  ⇡
Var[⇡t ] = Var[r̂t ] + Var[⇠t ]
(1 ⇢r )(1 ⇢r ) + ( ⇡ ⇢r ) + ⇡
Monetary Economics: Problem Set #2 4

(using the fact that the shocks are independent). Trying to mimimize this leads to a bit
of a mess. But you were told to make life easier by setting ⇢r = 0, so that the natural real
rate is IID over time, in which case we have the simpler expression
✓ ◆2 ✓ ◆2
 n  ⇡
Var[⇡t ] = Var[r̂t ] + Var[⇠t ]
+ ⇡ + ⇡

The first order condition for a minimum is

d 23 n 22 ⇡
Var[⇡t ] = Var[r̂ t ] + Var[⇠t ] = 0
d ⇡ ( + k ⇡ )3 ( + ⇡)
3

Solving for ⇡ gives

 Var[r̂tn ]
⇡ =
Var[⇠t ]
and you can check the second order condition is satisfied at this point, so this is a global
minimum. Intuitively, the more noisy is the observation of inflation (the bigger is Var[⇠t ]),
the smaller the weight the central bank should attach to inflation (if its objective is to
minimize the variance of inflation).

2. Monetary policy and the e↵ects of productivity shocks. Consider a new Keynesian
model with equilibrium conditions
1
yt = (it Et {⇡t+1 } ⇢) + Et {yt+1 } (2)

and
⇡t = Et {⇡t+1 } + (yt ytn ) (3)
where all variables have their usual meanings. Monetary policy is given by the feedback rule

it = ⇢ + ⇡ ⇡t

where ⇡ > 1. The production function (in logs) is

y t = at + n t

where at is an exogenous labor productivity process that follows an AR(1) process

at+1 = ⇢a at + ✏t+1

with 0  ⇢a < 1 and where {✏t } is IID white noise with mean zero. Natural output is propor-
tional to productivity
ytn = y at
where y > 0.

(a) Describe in words the economic interpretation of equations (2) and (3). (10 points)
(b) Use the method of undetermined coefficients to solve for the equilibrium response of output,
employment, and inflation to a productivity shock. (20 points)
Monetary Economics: Problem Set #2 5

(c) Describe how these responses depend on the values of the parameters ⇡ and . What
happens when ⇡ ! 1? What happens as the degree of price rigidities changes? Provide
economic intuition for your answers. (10 points)
(d) Discuss with as much detail as you can the joint response of employment and output to
a productivity shock and discuss the implications for assessing the role of productivity
shocks as a source of business cycle fluctuations in this model. (10 points)
Hint: you may want to skim Gali’s 1999 paper “Technology, Employment, and the Business
Cycle: Do Technology Shocks Explain Aggregate Fluctuations?” (posted on the LMS).

Solutions:

(a) Equation (2) is the standard (log-linear) intertemporal consumption Euler equation plus
a simple goods market clearing condition of the form ct = yt . Equation (3) is the new
Keynesian Phillips curve. This starts with an imperfectly competitive firm’s optimal price-
setting behavior subject to a Calvo-style price-setting rigidity. This is then log-linearized
and the terms reflecting real marginal cost are eliminated using the household’s labor
supply condition and approximate resource constraint so that it can be written in terms
of the output gap yt ytn where natural output ytn is the level of output that would obtain
in the same model but with perfectly flexible price-setting.
(b) Eliminating the nominal interest rate using the monetary policy rule it = ⇢ + ⇡ ⇡t and
eliminating natural output using productivity ytn = 'y at gives the system of two equations
1
yt = ( ⇡ ⇡t Et {⇡t+1 }) + Et {yt+1 }

and
⇡t = Et {⇡t+1 } + (yt ' y at )
We now guess solutions of the form

yt = 'ya at , and ⇡t = '⇡a at

for some as-yet-unknown coefficients 'ya , '⇡a to be determined. Since productivity follows
an AR(1) with coefficient ⇢a , these solutions imply the conditional expectations

Et {yt+1 } = 'ya ⇢a at , and Et {⇡t+1 } = '⇡a ⇢a at

So we can write the system of two equations in terms of these two unknown coefficients
and the shock at , specifically
1
'ya at = ( ⇡ '⇡a '⇡a ⇢a ) at + 'ya ⇢a at

and
'⇡a at = '⇡a ⇢a at + ('ya 'y )at
Since these conditions have to hold for any realization of the shock at we are left with two
conditions in the two unknown coefficients, that is
1
'ya = ( ⇡ '⇡a '⇡a ⇢a ) + 'ya ⇢a
Monetary Economics: Problem Set #2 6

and
'⇡a = '⇡a ⇢a + ('ya 'y )
Solving these two equations in the two unknown coefficients gives

( ⇡ ⇢a )
'ya = 'y > 0
(1 ⇢a )(1 ⇢a ) + ( ⇡ ⇢a )

and
 (1 ⇢a )
'⇡a = 'y < 0
(1 ⇢a )(1 ⇢a ) + ( ⇡ ⇢a )
To sign these unambiguously, we’ve used the assumption made in the question that ⇡ > 1
so that ⇡ ⇢a > 0. Therefore the equilibrium response of output and inflation to a
productivity shock at are simply
@yt
= 'ya > 0
@at
and
@⇡t
= '⇡a < 0
@at
Now from the approximate resource constraint (ignoring price dispersion, as usual) yt =
at + nt so in equilibrium
@nt
= 'ya 1
@at
This is not unambiguously signed. Substituting in the formula for 'ya we have

@nt ( ⇡ ⇢a )
>0, 'y > 1
@at (1 ⇢a )(1 ⇢a ) + ( ⇡ ⇢a )

or equivalently
(1 ⇢a )(1 ⇢a )
'y > +1
( ⇡ ⇢a )
Hence the sensitivity of natural output y n to productivity a has to be sufficiently large in
order for employment n to rise.
(c) The response of output to productivity, 'ya , is increasing in both  and ⇡ (since ⇡ > 1,
by assumption). The magnitude of the response of inflation to productivity, |'⇡a |, is also
increasing in  but is decreasing in ⇡ . As ⇡ ! 1, the response of inflation to productivity
'⇡a ! 0 so that inflation is also driven to zero for all realizations of productivity at and
so of course the variance of inflation is zero too. Similarly, when ⇡ ! 1 the response of
output 'ya ! 'y , i.e., the underlying response of natural output to productivity. Since
the output gap is ỹt ⌘ yt ytn = ('ya 'y )at , as 'ya ! 'y the output gap is driven to zero
for all realizations of productivity at so that the variance of the output gap is zero too.
The degree of price stickiness is captured by  (it is strictly decreasing in the Calvo
parameter ✓ – high ✓ (lots of price stickiness) means low  – etc). As  ! 0 we have
'⇡a ! 0 and 'ya ! 0 so that inflation is zero (lots of price stickiness!) and the output
gap is 'y at . As  ! 1, we have 'ya ! 'y (so no output gap) and '⇡a ! 'y (inflation
negatively correlated with productivity).
Monetary Economics: Problem Set #2 7

(d) The key property of the joint response of employment and output to a productivity shock
is the covariance of employment and output. Since

yt = 'ya at

and
nt = yt at = ('ya 1)at
are both linear functions of a single underlying shock, the covariance between them is just

Cov[yt , nt ] = Cov['ya at , ('ya 1)at ] = 'ya ('ya 1)Var[at ]

Since 'ya and the variance Var[at ] are positive, the covariance is positive or negative de-
pending on whether employment nt rises or falls when productivity increases. For example,
if 'ya > 1 so that employment rises when productivity rises, then employment and output
will have positive covariance (indeed a correlation coefficient of +1) while if 'ya < 1 so that
employment falls when productivity rises, then employment and output will have negative
covariance (indeed a correlation coefficient of 1).
A real business cycle model, like the classical economy discussed in earlier lectures, usually
has the implication that employment and output are positively correlated with each other
and with productivity. Gali (AER, 1999 and subsequent papers) uses a VAR approach to
argue that labor productivity and employment are negatively correlated in the data and
that this should be taken as evidence against RBC models.
 Monetary Economics
Problem Set #3

Monetary Economics: Problem Set #3

Solutions

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. Policy tradeo↵s in the new Keynesian model. Consider a new Keynesian model with
output gap and inflation given by
1
ỹt = (it Et [⇡t+1 ] ⇢) + Et [ỹt+1 ] (1)

and
⇡t = Et [⇡t+1 ] + ỹt + xt (2)
where {xt } is an exogenous shock. Monetary policy is given by the interest rate rule

it = ⇢ + ⇡ ⇡t + vt

where {vt } is an exogenous monetary policy shock and is independent of {xt }.

(a) Explain in words the economic interpretation of equations (1) and (2). How could you
interpret the xt shock? (10 points)

To simplify the algebra, assume for the rest of this question that = 1 and that {xt } and {vt }
are both IID white noise.

(b) Solve for equilibrium inflation ⇡t and the output gap ỹt in terms of the shocks vt and xt .
(20 points)
(c) Explain how inflation, interest rates and the output gap respond to the vt and xt shocks.
Give economic intuition for all your answers. (10 points)
(d) Suppose the central bank chooses the feedback coefficient ⇡ to minimise the loss function

L = Var[⇡t ] + Var[ỹt ] (3)

Solve for the value of ⇡ that minimises this loss function. Is there a policy “trade-o↵”
here? Explain how your answer depends on the parameter  and on the variances of the
shocks, Var[vt ] and Var[xt ]. Give economic intuition for your answers. (15 points)
(e) Does the value of ⇡ that minimises the loss function (3) satisfy the Taylor principle? Why
or why not? (5 points)
Monetary Economics: Problem Set #3 2

Solutions:

(a) Putting aside the xt shock, these equations are quite standard. Equation (1) is the (log-
linear) intertemporal consumption Euler equation that results from household optimisation
plus a simple goods market clearing condition of the form ct = yt . Equation (2) is the new
Keynesian Phillips curve. This starts with an imperfectly competitive firm’s optimal price-
setting behaviour subject to a Calvo-style price-setting rigidity. This is then log-linearised
and the terms reflecting real marginal cost are eliminated using the household’s labor
supply condition and approximate resource constraint so that it can be written in terms
of the output gap ỹtn ⌘ yt ytn where natural output ytn is the level of output that would
obtain in the same model but with perfectly flexible price-setting.
The xt shock is a supply shock. An xt > 0 represents higher inflation at any given level
of the output gap, i.e., an adverse supply shock. In terms of micro-foundations, this could
come from a shock to marginal cost, e.g., a shock to the relative price of an imported
intermediate input like oil that the economy takes as given. (Such a shock increases costs
but without lowering productivity).
(b) With the IID shocks, all the conditional expectations are zero and the system of equations
dramatically simplifies. The new Keynesian Phillips curve reduces to a static aggregate
supply relationship
⇡t = ỹt + xt (AS)
And using = 1 and rearranging, the Euler equation reduces to a static aggregate demand
relationship
1
⇡t = (ỹt + vt ) (AD)
⇡
Notice that the policy coefficient ⇡ determines the slope of the AD curve. The more
reactive is policy (the larger is ⇡ ), the flatter is the AD curve (see Figure 1).
These two equations can easily be solved for the output gap and inflation. Eliminating
inflation between them
1
ỹt + xt = (ỹt + vt )
⇡
so
⇡ 1
ỹt = xt vt
1+ ⇡ 1+ ⇡
and hence inflation is
✓ ◆
⇡ 1 1 
⇡t = ỹt + xt =  xt + vt + xt = xt vt
1+ ⇡ 1+ ⇡ 1+ ⇡ 1+ ⇡

(c) To summarise

⇡ 1
ỹt = xt vt
1+ ⇡ 1+ ⇡
1 
⇡t = xt vt
1+ ⇡ 1+ ⇡

with interest rates then given by

rt = it = ⇢ + ⇡ ⇡t + vt
Monetary Economics: Problem Set #3 3

inflation

t = ỹt + xt AS

AD
xt 1
t = (ỹt + vt )

ỹt output gap

Figure 1: Static AS-AD model with demand shocks vt and supply shocks xt . Monetary policy sets
the slope of the AD curve. A more reactive policy makes a flatter AD curve.

(rt = it since expected inflation is zero). An adverse supply shock xt > 0 increases inflation
and decreases the output gap. This is a shift up of the aggregate supply relationship. A
contractionary monetary policy shock vt > 0 decreases inflation and decreases the output
gap. This is a shift in of the aggregate demand relationship. (see Figure 2).
Since inflation is increasing in xt , interest rates are too. What about the net e↵ect of vt ?
@it  ⇡ 1
=1 =
@vt 1+ ⇡ 1+ ⇡

which is positive (so long as ⇡ is, i.e., so long as the AD curve slopes down).
(d) To begin with, the variances of inflation and the output gap are
✓ ◆2 ✓ ◆2
⇡ 1
Var[ỹt ] = Var[xt ] + Var[vt ]
1+ ⇡ 1+ ⇡
✓ ◆2 ✓ ◆2
1 
Var[⇡t ] = Var[xt ] + Var[vt ]
1+ ⇡ 1+ ⇡
Adding these up and collecting terms, the loss function is
✓ ◆2
1
L= ( 2⇡ + 1)Var[xt ] + (1 + 2 )Var[vt ]
1+ ⇡

If there were no supply shocks, Var[xt ] = 0, then we would have the usual case from the
lectures where there is no tradeo↵ and the optimal policy would have ⇡ = +1. Here
Monetary Economics: Problem Set #3 4

inflation

(xt > 0)
AS

AD
(vt > 0)

ỹt ỹt output gap

Figure 2: An adverse supply shock xt > 0 shifts up of the AS curve, inflation rises and the output
gap falls. A contractionary monetary policy shock vt > 0 shifts in the AD curve, inflation and the
output gap both fall.

that means the AD curve would be completely flat. Then inflation would be constant at
zero and hence the variance of inflation would be zero too for any variance of the demand
shocks vt . However, if there are supply shocks then setting a high value of ⇡ exposes the
economy to large swings in the output gap and so when Var[xt ] > 0 there is a trade-o↵, as
we can see in the formula for the loss function.
The first order condition for the optimal ⇡ is
✓ ◆2
dL  2 2 1
= 2 ( ⇡ + 1)Var[xt ] + (1 +  )Var[vt ] + 2 ⇡ Var[xt ] = 0
d ⇡ (1 +  ⇡ )3 1+ ⇡

Cancelling the common factors and rearranging

2
(1 +  ⇡ ) ⇡ Var[xt ] = ( ⇡ + 1)Var[xt ] + (1 + 2 )Var[vt ]
2
Cancelling more common terms (and thankfully the terms in ⇡) we get

⇡ Var[xt ] = Var[xt ] + (1 + 2 )Var[vt ]

Hence
⇤ Var[vt ]
⇡ =  + (1 + 2 ) >>0
Var[xt ]
The optimal coefficient ⇤⇡ is increasing in  so that policy sets the slope of the AD curve
1/ ⇤⇡ to be flatter the steeper the slope of the AS curve. Intuitively, other things equal a
Monetary Economics: Problem Set #3 5

steeper aggregate supply curve makes for more inflation volatility at the expense of output
gap volatility and, since the preferences weigh each variance equally, policy counteracts
that by reducing inflation volatility. The extent of this adjustment depends on  and on
the relative size of the fundamental shocks Var[vt ]/Var[xt ].
⇤
(e) Yes, since ⇡ > 0. Here interest rates are given by
⇤
rt = it = ⇢ + ⇡ ⇡t + vt

Any positive coefficient will mean that nominal interest rates and hence real interest rates
rise with inflation. This is because with the IID shocks, inflation expectations are constant.
(Another way to put this is that any ⇤⇡ > 0 makes the aggregate demand curve slope down
here).

2. Interest rate versus money supply rules. Consider an economy again described by the
equilibrium conditions
1
ỹt = (it Et [⇡t+1 ] rtn ) + Et [ỹt+1 ]
⇡t = Et [⇡t+1 ] + ỹ

and now also a money demand equation of the form

mt pt = yt ⌘it , ⌘>0

where all variables are defined as usual. Both ytn and rtn evolve exogenously, independent of
monetary policy. The central bank seeks to minimize a loss function of the form

L = Var[⇡t ] + Var[ỹt ]

(a) Explain how the optimal monetary policy outcomes can be implemented by an interest
rate feedback rule. (10 points)
(b) Show that a constant money supply will generally not be optimal. (15 points)
(c) Derive a money supply rule that would implement the optimal monetary policy. (15 points)

Solutions:

(a) First observe that if we suppose an interest rate rule that does not depend on money mt
then this model can be solved without reference to the money demand condition mt pt =
yt ⌘it . All that condition will do is enable us to determine, residually, the quantity of
money consistent with the rest of the model.
Now consider an equilibrium outcome where ⇡t = 0 and ỹt = 0 for all t. Since the
unconditional variance of inflation and the output gap are both zero in such an equilibrium,
the loss is L = Var[⇡t ] + Var[ỹt ] = 0 and clearly the monetary policy authority cannot do
better than this. There are in general many interest rate rules that can implement this
outcome. We discussed one example in class, a simple feedback rule of the form

it = rtn + ⇡ ⇡t + y ỹt
Monetary Economics: Problem Set #3 6

If the coefficient on inflation ⇡ is sufficiently high (i.e., the rule is sufficiently “reactive”)
then this rule will lead to a unique equilibrium outcome and in that outcome we will
have ⇡t = 0 and ỹt = 0. More specifically, following the lecture notes, in this setting the
equilibrium dynamics are governed by
✓ ◆ ✓ ◆
ỹt Et [ỹt+1 ]
=A
⇡t Et [⇡t+1 ]
where ✓ ◆
1 ⇡ 1
A=⌦ , ⌦⌘
  + ( + y) + y+ ⇡

Now, when does A have both eigenvalues 1 , 2 inside the unit circle (thus ensuring a
unique outcome in the forward dynamics)? We have the following properties:
– determinant
✓ ◆
2 1 ⇡
det(A) = ⌦ det = = 1 2
  + ( + y) + y + ⇡

– trace
✓ ◆
1 ⇡ +  + ( + y)
tr(A) = ⌦tr = = 1 + 2
  + ( + y) + y+ ⇡

– polynomial at unity

y (1 ) + ( ⇡ 1)
p(1) = 1 tr(A) + det(A) = = (1 1 )(1 2)
+ y+ ⇡

Since the product 1 2 of the eigenvalues is positive, both eigenvalues must have the same
sign (either both positive or both negative). Since the sum 1 + 2 is positive, it then follows
that both eigenvalues must be positive. Since 0 < < 1 the product of eigenvalues is less
than one and so at least one of them is itself less than one. Therefore both eigenvalues are
less than one if and only if p(1) = (1 1 )(1 2 ) > 0, that is, if and only if

y (1 ) + ( ⇡ 1) > 0

Any interest sensitivity coefficient ⇡ that satisfies this (e.g., ⇡ > 1) will suffice to ensure
the unique equilibrium is ⇡t = 0 and ỹt = 0 for all t and thus will minimise the loss.
(b) Consider the equilibrium generated by the interest rate rule given in part (a) above with
⇡ such that this is the unique outcome. The money supply must then evolve according
to
mt = pt + yt ⌘it
or
mt = ⇡t + yt ⌘(it it 1 )
But in this equilibrium, ⇡t = 0 and ỹt = 0 so that the equilibrium nominal rate is just
it = rtn and equilibrium output is yt = ytn (since the output gap is zero). So money growth
will be
mt = ytn ⌘ rtn
Monetary Economics: Problem Set #3 7

The right hand side here is exogenous and generally varying over time. Notice that this
outcome does not depend on the parameters in the interest rate rule. Any interest rate
rule that implements the optimal policy will imply money growth of this form. In all these
situations, money growth would not generally be constant. Only in the very special case
that ytn and rtn are both constants would we have constant money growth.
(c) The basic idea here is to make find a money growth rule that clears the money market at
the same level of interest rates it = rtn as would prevail in the equilibrium described above.
Inverting the money market condition to write it in terms of the nominal rate
1
it = (mt pt ytn ỹt )
⌘
Plugging this into the dynamic IS curve gives
✓ ◆
1 1
ỹt = (mt pt ytn ỹt )) Et {⇡t+1 } rtn + Et {ỹt+1 }
⌘
Now consider the specific money supply rule

mt = pt + ytn + ỹtn ⌘(rtn + ⇡ ⇡t + y ỹt )

Plugging this into the dynamic IS curve gives

✓ ◆
1 1
ỹt = (pt + ytn + ỹtn ⌘(rtn + ⇡ ⇡t + y ỹt ) pt ytn ỹt )) Et {⇡t+1 } rtn + Et {ỹt+1 }
⌘
✓ ◆
1 1
= ( ⌘(rtn + ⇡ ⇡t + y ỹt ))) Et {⇡t+1 } rtn ) + Et {ỹt+1 }
⌘
✓ ◆
1 1
= ( ⌘(rtn + ⇡ ⇡t + y ỹt ))) Et {⇡t+1 } rtn + Et {ỹt+1 }
⌘
1
= ( ⇡ ⇡t + y ỹt Et {⇡t+1 }) + Et {ỹt+1 }

This is one equilibrium condition in ỹt and ⇡t . The other condition is given by the new
Keynesian Phillips curve. The equilibrium dynamics here satisfy the same system as in
part (a), namely ✓ ◆ ✓ ◆
ỹt Et [ỹt+1 ]
=A
⇡t Et [⇡t+1 ]
where ✓ ◆
1 ⇡ 1
A=⌦ , ⌦⌘
  + ( + y) + y+ ⇡

The coefficients ⇡ etc here now refer to the coefficients in the money supply rule above.
This money supply rule “mimics” the interest rate rule and, so long as

y (1 ) + ( ⇡ 1) > 0

we have a unique equilibrium with ⇡t = ỹt = 0 etc. In this equilibrium, money growth is
as in part (b), namely
mt = ytn ⌘ rtn
and so is not generally constant.
 Monetary Economics
Problem Set #4

Monetary Economics: Problem Set #4

Solutions

This problem set is marked out of 100 points. The weight given to each part is indicated below.
Please contact me asap if you have any questions.

1. Government purchases in the new Keynesian model. Consider a basic new Keynesian
model with the following (log-linearised) equilibrium conditions: for consumption, a dynamic
Euler equation
1
ct = (it Et [⇡t+1 ] ⇢) + Et [ct+1 ]
for labor supply
wt pt = ct + 'nt
the production function for firms
yt = nt
If prices were fully flexible, firms would set a constant markup over marginal cost. With sticky
prices, there is a new Keynesian Phillips curve in terms of the output gap

⇡t = Et [⇡t+1 ] + ỹt

The government purchases a fraction ⌧t of output each period with ⌧t varying exogenously.

(a) Derive a log-linear version of the goods market clearing equation of the form yt = ct + gt
where gt ⌘ log(1 ⌧t ). (5 points)
(b) Show that natural output is proportional to government purchases, ytn = gt , and give an
explicit formula for the coefficient . Does a fiscal expansion increase or decrease natural
output when prices are fully flexible? By how much? Explain. (10 points)

Now assume that monetary policy is set according to the simple interest rate rule

it = ⇢ + ⇡ ⇡t + y ỹt

and that government purchases gt follow an AR(1) process with persistence 0  ⇢g < 1.

(c) Show that the output gap ỹt satisfies a dynamic IS curve of the form
1
ỹt = (it Et [⇡t+1 ] rtn ) + Et [ỹt+1 ]

and derive a formula for the natural real rate rtn in terms of gt . (5 points)
(d) Use the method of undetermined coefficients to solve for the response of the key endogenous
variables —output, natural output, employment, inflation, interest rates— to an exogenous
increase in government purchases gt . Give economic intuition for your answers. (20 points)
Monetary Economics: Problem Set #4 2

(e) Explain how the response of output to government purchases depends on the monetary
policy coefficient ⇡ . Does a higher ⇡ increase or decrease the impact e↵ect of govern-
ment purchases on output? Similarly, explain how the response of output to government
purchases depends on the AR(1) persistence ⇢g . Does a more persistent process increase
or decrease the impact e↵ect of government purchases on output? Give economic intuition
for your answers. (10 points)

Solutions:

(a) Since the government purchases a fraction ⌧t of output, we have government purchases

G t = ⌧ t Yt

and the goods market clearing condition in raw levels is

Yt = C t + G t

so
(1 ⌧t )Yt = Ct
Taking logs of both sides
yt = ct + g t
where, as suggested, gt ⌘ log(1 ⌧t ) and as usual yt ⌘ log Yt , ct ⌘ log Ct .
(b) With fully flexible prices each period, a firm sets its price as a constant markup over
nominal marginal cost
"
Pt = Wt
" 1
or in logs
pt = µ + w t
where µ ⌘ log("/(" 1)) > 0 is the log markup. So the ‘natural real wage’ (for want of a
better term) is a constant
w t pt = µ
and so using the household’s labor supply condition and the production function for firms,
natural output ytn satisfies

µ = wt pt
= ct + 'nt
= (ytn gt ) + 'ytn
= ( + ')ytn gt

Solving for ytn then gives

1
ytn = µ+ gt (1)
+' +'
When prices are fully flexible, the multiplier is
@ytn
⌘ =
@gt +'
Monetary Economics: Problem Set #4 3

A fiscal expansion (an increase in gt ) increases the natural level of output. Why? Not for
Keynesian demand-side reasons, that’s for sure! The channel here is purely supply-side:
an increase in gt causes consumption ct to fall which is a negative ‘wealth e↵ect’ on labor
supply, i.e., households feel ‘poorer’ (have higher marginal utility of consumption) and so
work more at any given wage (an outward shift in the labor supply curve). That higher
labor supply translates into higher natural output.
(c) The output gap is defined by
ỹt ⌘ yt ytn
and so using the resource constraint from part (a) and the expression for natural output
from part (b) we have
1
ỹt = ct + gt ytn = ct + gt + µ gt
+' +'
Rearrange this to express consumption in terms of the output gap and government pur-
chases ✓ ◆
1 1 '
ct = ỹt µ+ 1 gt = ỹt µ gt
+' +' +' +'
We now want to plug this into the dynamic Euler equation given in the problem
1
ct = (it Et [⇡t+1 ] ⇢) + Et [ct+1 ]

Note that
h 1 ' i
Et [ct+1 ] = Et ỹt+1 µ gt+1
+' +'
1 '
= Et [ỹt+1 ] µ Et [gt+1 ]
+' +'
and since gt follows and AR(1) with persistence ⇢g we have

Et [gt+1 ] = ⇢g gt

so that
1 '
Et [ct+1 ] = Et [ỹt+1 ] µ ⇢g g t
+' +'
Plugging in for ct and Et [ct+1 ] in the Euler equation then gives
1 ' 1 1 '
ỹt µ gt = (it Et [⇡t+1 ] ⇢) + Et [ỹt+1 ] µ ⇢g gt
+' +' +' +'
Cancelling common terms and collecting things together
1 '
ỹt = (it Et [⇡t+1 ] ⇢) + Et [ỹt+1 ] + (1 ⇢g )gt (2)
+'
We can render this expression consistent with the dynamic IS curve given in the problem
if we define the natural real rate rtn appropriately. In particular, define
'
rtn = ⇢ + (1 ⇢g )gt ⌘ ⇢ + 'g gt (3)
+'
Monetary Economics: Problem Set #4 4

so that the natural real rate is increasing in government purchases gt . Then with this
definition of the natural real rate, from equation (2) we have
1 rtn ⇢
ỹt = (it Et [⇡t+1 ] ⇢) + Et [ỹt+1 ] +

or
1
ỹt = (it Et [⇡t+1 ] rtn ) + Et [ỹt+1 ]
as required.
(d) Given the policy rule
it = ⇢ + ⇡ ⇡t + y ỹt

the nominal rate less the natural rate is

it rtn = ⇡ ⇡t + y ỹt 'g g t

where 'g > 0 is the slope coefficient defined in (3) above. Plugging this into the dynamic
IS curve we have the reduced system of equations that needs to be solved
1
ỹt = ( ⇡ ⇡t + y ỹt 'g g t Et [⇡t+1 ]) + Et [ỹt+1 ] (4)
⇡t = Et [⇡t+1 ] + ỹt (5)

Now guess the linear functions

⇡t = '⇡g gt
ỹt = 'yg gt

so that

Et [⇡t+1 ] = '⇡g ⇢g gt
Et [ỹt+1 ] = 'yg ⇢g gt

for some as-yet unknown coefficients that we need to solve for. Plugging these into the
equations (4)-(5) and collecting terms gives

[ (1 ⇢g ) + y ]'yg +( ⇡ ⇢g )'⇡g = 'g (6)

(1 ⇢g )'⇡g = 'yg (7)

These are two equations to be solved for the two unknowns 'yg , '⇡g . Solving, we get

'⇡g = 'g (8)
(1 ⇢g )[ (1 ⇢g ) + y ] + ( ⇡ ⇢g )
(1 ⇢g )
'yg = 'g (9)
(1 ⇢g )[ (1 ⇢g ) + y ] + ( ⇡ ⇢g )

If we impose the Taylor Principle ⇡ > 1 (i.e., the standard sufficient condition for a
unique equilibrium, as discussed in class), then both of these coefficients are positive.
Under this parameter assumption, then, an increase in government purchases gt increases
both inflation ⇡t and the output gap ỹt . Since from equation (1) natural output ytn is
Monetary Economics: Problem Set #4 5

increasing in gt , output yt = ỹt + ytn is also increasing in government purchases. Since from
the production function yt = nt , labor nt is also increasing in gt . From the policy rule

it = ⇢ + ⇡ ⇡t + y ỹt

and since both inflation and the output gap increase, so too does the nominal interest rate.
(e) Since the denominators of (8) and (9) are both increasing in ⇡ , the response coefficients
'⇡g and 'yg are both decreasing in ⇡ . Moreover, since natural output does not depend
on ⇡ , the response of output itself is the same as the response of the output gap. So the
response of output is also decreasing in ⇡ . Intuitively, the more reactive the monetary
authority is to inflation, the smaller is the e↵ect of an increase in gt on the output gap
and hence the smaller is the rise in actual inflation. In this sense, the e↵ectiveness of fiscal
policy depends crucially on the monetary reaction.
The e↵ects of ⇢g are a little more complicated. In particular, ⇢g also enters through the
coefficient 'g on the natural real rate, as defined in (3). Plugging this is and tidying the
expression up gives
(1 ⇢g )(1 ⇢g )
'yg = (1 ) (10)
(1 ⇢g )(1 ⇢g ) + (1 ⇢g ) y + ( ⇡ ⇢g )
To see the main e↵ects of ⇢g intuitively, divide the numerator and denominator by (1
⇢g ) to write
1 ⇢g 1h  i
'yg = (1 ), ⌘ y + ( ⇡ ⇢ g ) >0
1 ⇢g + 1 ⇢g
where > 0 since ⇡ > 1 (and 1 > ⇢g ). Now first consider ⇢g = 0, a completely transitory
stimulus. Then we have
1
0 < 'yg = h i (1 )<1
⇢g =0 1
1+ y + ⇡

and so the multiplier, which is given by the e↵ect on output itself, rather than the output
gap, is greater than but less than one. Moreover, at the other extreme, consider ⇢g = 1,
a completely permanent stimulus (government purchases a random walk). Then we have
0
'yg = h i (1 )=0
⇢g =1 1 
0+ y + 1
( ⇡ 1)

in which case government purchases have zero e↵ect on the output gap and hence the e↵ect
on output is simply . In short, for ⇢g = 0 we have a multiplier less than one but larger
than while for ⇢g = 1 we have the lower multiplier . Hence we expect the multiplier
to be decreasing in ⇢g . The more persistent is the increase in government purchases, the
larger is the expected inflation response and hence (if the Taylor principle is satisfied), the
more vigorous is the monetary response.
For completeness, here is the calculation for general ⇢g . Dividing the numerator and
denominator in (10) by (1 ⇢g )(1 ⇢g ) we have

'yg = 1 ( ⇡ ⇢g )
(1 )
+ 1 ⇢g y +  (1 ⇢g )(1 ⇢g )
Monetary Economics: Problem Set #4 6

Hence the response coefficient 'yg is decreasing in ⇢g if and only if

1 ( ⇡ ⇢g )
D(⇢g ) ⌘ y +
1 ⇢g (1 ⇢g )(1 ⇢g )
is increasing in ⇢g . Di↵erentiating D(⇢g ) with respect to ⇢g gives, after some simplifications,
1 h  n oi
0
D (⇢g ) = y + ( ⇡ ⇢g )(1 + 2 ⇢g ) (1 ⇢g )(1 ⇢g )
(1 ⇢g )2 1 ⇢g
which is certainly positive so long as the term in braces {·} is. This term is quadratic in
⇢g and can be rearranged to give the condition

Q(⇢g ) > 0 ) D0 (⇢g ) > 0

where
Q(⇢g ) ⌘ ⇢2g 2 ⇡⇢ + (1 + ) ⇡ 1
Now observe that Q(0) = (1 + ) ⇡ 1 > 0 (since ⇡ > 1) but Q(⇢g ) is decreasing in ⇢g ,
so it may seem that for high enough ⇢g we could get Q(⇢g ) < 0 which would be a pain.
But in fact even at ⇢g = 1 we have Q(1) = (1 )( ⇡ 1) > 0 (again, since ⇡ > 1), so
luckily we know that Q(⇢g ) > 0 for all ⇢g 2 [0, 1]. In turn this implies D0 (⇢g ) > 0 for all
⇢g and so indeed, as conjectured, the response coefficient 'yg is decreasing in ⇢g .

2. Multipliers, the ZLB, and the duration of fiscal stimulus. Consider a new Keynesian
model with government purchases gt and shocks t to the interest rate facing households
1
ỹt = (it + t Et [⇡t+1 ] rtn ) + Et [ỹt+1 ]

⇡t = Et [⇡t+1 ] + ỹt
Monetary policy is set according to an interest rate rule but also faces a zero lower bound

it = max[0, ⇢ + ⇡ ⇡t + y ỹt ]

The natural real rate and natural output are given by

rtn = ⇢ (1 )Et [ gt+1 ], ytn = gt

The interest spread shock t can take on two values L , H with H = 0 and L > 0. The
economy starts in the L state. With probability ↵ it stays in the L state. With probability
1 ↵ it transitions to the H state. Once it enters the H (“normal”) state it stays there forever.
Suppose that gH = 0. We are interested in calculating economic outcomes as a function of fiscal
policy gL in the L (“crisis”) state.

(a) To begin with, suppose that gL = 0. Solve for the equilibrium values of inflation, the
output gap, and the nominal interest rate in the L state. (5 points)
(b) Now explain how an increase in government purchases to some gL > 0 would a↵ect in-
flation, expected inflation, output and the nominal and real interest rates in the L state.
Explain how your answers depend on the size of the interest spread L . How large is
the government purchases multiplier? Does this value depend on the the size of the fiscal
stimulus? (15 points)
Monetary Economics: Problem Set #4 7

Now consider the possibility that government purchases gL persist at an elevated level after the
crisis has abated. To be specific, imagine that there are three states, L, S, H. The economy
starts in the L state with L > 0. Suppose the initial interest spread L is sufficiently high that
the ZLB is binding in the L state. With probability ↵ the economy transitions to the S state.
In the S (“transitional”) state, the crisis is over S = H = 0. In the S state, with probability
the fiscal stimulus continues with government purchases gS = gL > 0. With probability 1
the fiscal stimulus ends with gS = gH = 0.

(c) Let ⇡S , ỹS , iS denote the equilibrium values of inflation, the output gap and the nominal
interest rate in the S state. Following the same logic as in part (a), solve for these
equilibrium values as a function of the size of the fiscal stimulus gL > 0. Explain how
your answers depend on whether the fiscal stimulus continues gS = gL , or not gS = gH .
(15 points)
(d) Using your results from part (c), now solve for the equilibrium values of inflation, the
output gap and the nominal interest rate in the initial L state given that with probability
the fiscal stimulus continues after the crisis has abated. How does the government
purchases multiplier compare to the one you found in part (b)? How does the multiplier
vary with the probability of the stimulus continuing? Explain. (15 points)

[Hint: this question is based on Woodford’s article “Simple Analytics of the Government Ex-
penditure Multiplier” American Economic Journal: Macroeconomics. 3(1): 1–35, especially
Section IV.B]
Solutions:

(a) From the new Keynesian Phillips curve in the low state we have

⇡L = ỹL
1 ↵
and the natural rate in the low state is

rLn = ⇢ + (1 )(1 ↵)ĝL

Plugging these into the IS curve in the low state we have

1 ↵
(1 ↵)ỹL = (rL iL ) + ỹL + (1 )(1 ↵)ĝL
(1 ↵ )

where rL ⌘ ⇢ L. So for a given iL the output gap is

ỹL = #r (rL iL ) + #g ĝL

with coefficients
(1 ↵ )
#r ⌘ >0
(1 ↵)(1 ↵ ) ↵
and
(1 ↵)(1 ↵ )
#g ⌘ (1 )>1 >0
(1 ↵)(1 ↵ ) ↵
Monetary Economics: Problem Set #4 8

But the nominal interest rate iL is endogenous, given by the policy rule
iL = max[0, ⇢ + ⇡ ⇡L + y ỹL ]
 ✓ ◆

= max 0, ⇢ + ⇡ + y ỹL
1 ↵
 ✓ ◆

= max 0, ⇢ + ⇡ + y (#r (rL iL ) + #g ĝL )
1 ↵
This is one nonlinear equation in one unknown, iL . Once we have solved for iL , we can
then recover ỹL and ⇡L from the expressions above.
Mathematically, we are solving a problem of the form
x = f (x), f (x) ⌘ max[0, b ax], x 0
given two constants a > 0 and b. If b > 0 then this equation has a unique solution x⇤ and
that solution lies on the positive branch, x⇤ = b/(1 + a) > 0. But if b < 0 then there is
a unique solution at x⇤ = 0. The size (and sign) of the intercept is determined by (i) the
size of the shock, L and (ii) the size of the fiscal stimulus ĝL . The ZLB will bind if either
the shock is large enough and/or the fiscal stimulus is too small.
(b) Specifically, suppose that ĝL = 0. Then the ZLB binds if
✓ ◆

⇢+ ⇡ + y #r r L < 0
1 ↵
or equivalently, whenever
✓✓ ◆ ◆ 1

rL < rL⇤ ⌘ ⇡ + y #r ⇢<0
1 ↵
Since rL = ⇢ L, this is also equivalent to
⇤
L > L =⇢ rL⇤ > 0
That is, the ZLB binds when the shock is large enough. Now if the ZLB binds (and
continuing to assume ĝL = 0) then iL = 0 and we have that the output gap is
ỹL = #r rL
with inflation

⇡L = #r r L
1 ↵
A small increase in government purchases ĝL would cause output to rise to
ỹL = #r rL + #g ĝL
with a corresponding increase in inflation (and expected inflation) so long as the ZLB
continues to bind. In this case, the government purchases multiplier is
dyL dỹL dyLn
= + = #g + >1 + >1
dĝL dĝL dĝL
If the increase in government purchases is large enough, however, the ZLB will cease to
bind and the multiplier will fall back to < 1. In short, output is a piecewise linear
function of ĝL with a slope that switches from above to below 1 at a critical “threshold”
fiscal stimulus.
Monetary Economics: Problem Set #4 9

(c) The approach here follows the same steps as in part (a) above except that the transition
probability ↵ is replaced with and the interest spread in the transitional state is S = 0.
In particular, from the new Keynesian Phillips curve in the transitional state we have

⇡S = ỹS
1
and the natural rate in the transitional state is

rSn = ⇢ + (1 )(1 )ĝS

Plugging these into the IS curve in the transitional state we then have
1 
(1 )ỹS = (⇢ iS ) + ỹS + (1 )(1 ↵)ĝS
(1 )
So for a given iS the output gap is

ỹS = #r, (⇢ iS ) + #g, ĝS

with coefficients
(1 )
#r, ⌘ >0
(1 )(1 ) 
and
(1 )(1 )
#g, ⌘ (1 )>1 >0
(1 )(1 ) 
(i.e., the same as in part (a) with replacing ↵). Again, the nominal interest rate is found
by solving

iS = max[0, ⇢ + ⇡ ⇡S + y ỹS ]
 ✓ ◆

= max 0, ⇢ + ⇡ + y ỹS
1
= max [0, ⇢ + (#r, (⇢ iS ) + #g, ĝS )]

where ⌘ ⇡ 1  + y > 0. Once we have solved for iS , we can then recover ỹS and ⇡S
from the expressions above. Since the crisis is over, the intercept term is strictly positive,

⇢+ (#r, ⇢ + #g, ĝS ) > 0

so the solution is on the positive branch where the ZLB is slack. Hence the solution for iS
is given by
#g,
iS = ⇢ + ĝS
1 + #r,
Plugging this into the expressions above for the output gap and inflation
#g,
ỹS = ĝS
1 + #r,

and
 #g,
⇡S = ĝS
1 1 + #r,
Monetary Economics: Problem Set #4 10

Hence, if in the transitional state the fiscal stimulus does not continue, i.e., if ĝS = ĝH = 0,
then iS = ⇢ and so in this case ỹS = 0 and ⇡S = 0. In short, if the fiscal stimulus is over,
then there is no di↵erence between the transitional S state and the full end-of-crisis H
state. But if the fiscal stimulus continues, i.e., if ĝS = ĝL > 0, then the interest rate is
greater than ⇢, there is a positive output gap ỹS > 0 and positive inflation ⇡S > 0 (and
positive expected inflation). In short, with an on-going active fiscal stimulus and away
from the ZLB, monetary policy responds by increasing interest rates.
(d) (Sketch) We can now roll back to the crisis L period to see how the possibility of the fiscal
stimulus continuing even after the crisis has abated a↵ects the equilibrium in this earlier
period. For example, we now have from the the new Keynesian Phillips curve in the L
state
⇡L = [↵⇡L + (1 ↵)⇡S ] + ỹL
so

⇡L = (1 ↵)⇡S + ỹL
1 ↵ 1 ↵
where ⇡S is given from part (c) above. The natural rate is

rLn = ⇢ (1 )(1 ↵)(ĝS ĝL )

which either simply equals ⇢ if the fiscal stimulus continues in the S state or is the same
as in part (a) if the fiscal stimulus shuts o↵ in the S state. Likewise, from the IS curve we
then have
1
ỹL = (iL + L ↵⇡L (1 ↵)⇡S rLn ) + ↵ỹL + (1 ↵)ỹS

where again ⇡S and ỹS are determined as in part (c) above. To solve the model, we then
need to determine the interest rate iL from

iL = max[0, ⇢ + ⇡ ⇡L + y ỹL ]

It should be clear that if the fiscal stimulus does not continue, i.e., if ĝS = ĝH = 0,
then everything reduces to being as in part (a) above. But if the fiscal stimulus does
continue, so that iS > 0, ỹS > 0 and ⇡S > 0 then following the discussion in Woodford
(pages 23–24) the size of the fiscal multiplier will be decreasing in the probability of the
transitional state recurring (i.e., in the expected duration of the “excess” period of fiscal
stimulus). Intuitively, this is because once inflation and the output gap increase in the
transitional phase, monetary policy is o↵ the ZLB and increases rates (via the interest rate
rule coefficients ⇡ , y ), i.e., monetary policy chokes o↵ some of the fiscal policy-induced
rise in demand. The more aggressive this monetary policy o↵setting response, the smaller
(or even negative) the fiscal policy multiplier.