Lecture Notes in Macroeconomics

John C. Driscoll
Brown University and NBER1

December 3, 2001

1
Department of Economics, Brown University, Box B, Providence RI 02912. Phone
(401) 863-1584, Fax (401) 863-1970, email:John Driscoll@[Link], web:http:\\
c
[Link]\ ∼jd. °Copyright John C. Driscoll, 1999, 2000, 2001. All rights
reserved. Do not reproduce without permission. Comments welcome. I especially
thank David Weil, on whose notes substantial parts of the chapters on Money and
Prices and Investment are based. Kyung Mook Lim and Wataru Miyanaga provided
detailed corrections to typographical errors. Several classes of Brown students have
provided suggestions and corrections. All remaining errors are mine.
ii
Contents

1 Money and Prices 1

1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Money . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 The History of Money . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 The Demand for Money . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 The Baumol-Tobin Model of Money Demand . . . . . . . 4
1.4 Money in Dynamic General Equilibrium . . . . . . . . . . . . . . 6
1.4.1 Discrete Time . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4.2 Continuous Time . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.3 Solving the Model . . . . . . . . . . . . . . . . . . . . . . 13
1.5 The optimum quantity of money . . . . . . . . . . . . . . . . . . 14
1.5.1 The Quantity Theory of Money . . . . . . . . . . . . . . . 14
1.6 Seigniorage, Hyperinflation and the Cost of Inflation . . . . . . . 16
1.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2 Nominal Rigidities and Economic Fluctuations 27

2.1 Old Keynesian Economics: The Neoclassical Synthesis . . . . . . 28
2.1.1 Open Economy . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1.2 Aggregate Supply . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Disequilibrium Economics . . . . . . . . . . . . . . . . . . . . . . 36
2.2.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2.2 The Walrasian Benchmark Case . . . . . . . . . . . . . . 37
2.2.3 Exogenously Fixed Price . . . . . . . . . . . . . . . . . . . 38
2.2.4 Exogenously Fixed Nominal Wage . . . . . . . . . . . . . 39
2.2.5 Both prices and wages inflexible . . . . . . . . . . . . . . 39
2.2.6 Analysis of this model . . . . . . . . . . . . . . . . . . . . 40
2.3 Imperfect Information Models . . . . . . . . . . . . . . . . . . . . 41
2.4 New Keynesian Models . . . . . . . . . . . . . . . . . . . . . . . . 43
2.4.1 Contracting Models . . . . . . . . . . . . . . . . . . . . . 43
2.4.2 Predetermined Wages . . . . . . . . . . . . . . . . . . . . 44
2.4.3 Fixed Wages . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.5 Imperfect Competition and New Keynesian Economics . . . . . . 50
2.5.1 Macroeconomic Effects of Imperfect Competition . . . . . 50

iii
iv CONTENTS

2.5.2 Imperfect competition and costs of changing prices . . . . 51

2.5.3 Dynamic Models . . . . . . . . . . . . . . . . . . . . . . . 56
2.6 Evidence and New Directions . . . . . . . . . . . . . . . . . . . . 57
2.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

3 Macroeconomic Policy 65
3.1 Rules v. Discretion . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.1.1 The Traditional Case For Rules . . . . . . . . . . . . . . . 66
3.2 The Modern Case For Rules: Time Consistency . . . . . . . . . . 68
3.2.1 Fischer’s Model of the Benevolent, Dissembling Government 68
3.2.2 Monetary Policy and Time Inconsistency . . . . . . . . . 72
3.2.3 Reputation . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3 The Lucas Critique . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.4 Monetarist Arithmetic: Links Between Monetary and Fiscal Policy 79
3.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4 Investment 87
4.1 The Classical Approach . . . . . . . . . . . . . . . . . . . . . . . 87
4.2 Adjustment Costs and Investment: q Theory . . . . . . . . . . . 88
4.2.1 The Housing Market: After Mankiw and Weil and Poterba 91
4.3 Credit Rationing . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.4 Investment and Financial Markets . . . . . . . . . . . . . . . . . 95
4.4.1 The Effects of Changing Cashflow . . . . . . . . . . . . . 97
4.4.2 The Modigliani-Miller Theorem . . . . . . . . . . . . . . . 98
4.5 Banking Issues: Bank Runs, Deposit Insurance and Moral Hazard 99
4.6 Investment Under Uncertainty and Irreversible Investment . . . . 103
4.6.1 Investment Under Uncertainty . . . . . . . . . . . . . . . 107
4.7 Problems: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

5 Unemployment and Coordination Failure 117

5.1 Efficiency wages, or why the real wage is too high . . . . . . . . . 117
5.1.1 Solow model . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.1.2 The Shapiro-Stiglitz shirking model . . . . . . . . . . . . 118
5.1.3 Other models of wage rigidity . . . . . . . . . . . . . . . . 120
5.2 Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.2.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.2.2 Steady State Equilibrium . . . . . . . . . . . . . . . . . . 122
5.3 Coordination Failure and Aggregate Demand Externalities . . . . 123
5.3.1 Model set-up . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.3.2 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.3.3 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.3.4 Propositions . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

Continuous-Time Dynamic Optimization 131

CONTENTS v

Stochastic Calculus 133

Introduction
Course Mechanics
• Requirements: Two exams, each 50% of grade, each covers half of material
in class. First exam: on Tuesday, March 12th. Second and final exam: on
Tuesday, April 30th.

• Problem sets: will be several, which will be handed in and corrected, but
not graded. Good way to learn macro, good practice for exams and core.

• On the reading list: It is very ambitious. We may well not cover every-
thing. That is fine, as not everything is essential. I may cut material as I
go along, and will try to give you fair warning when that happens.

• The lectures will very closely follow my lecture notes. There are two
other general textbooks available: Romer, which should be familiar and
Blanchard and Fischer. The latter is harder but covers more material.
The lecture notes combine the approaches of and adapt materials in both
books.

• References in the notes refer to articles given on the reading list. With
few exceptions, the articles are also summarized in Romer or Blanchard
and Fischer. It is thus not necessary to read all or even most of the ar-
ticles on the list. Since articles are the primary means through which
economists communicate, you should read at least one. Some of the ar-
ticles are in the two recommended volumes by Mankiw and Romer, New
Keynesian Economics, both of which will eventually be in the bookstore.
Just about all articles prior to 1989 are available via the internet at the
site [Link], provided one connects through a computer connected
to Brown’s network. I would ask that everyone not individually print out
every article, since that would take a lot of paper, energy and computing
power.

• Students considering macroeconomics as a field are strongly encouraged

to attend the Macroeconomics Workshop, on Wednesdays from 4:00-5:30
in Robinson 301.

Motivation
Consider the handout labeled “The First Measured Century.” It presents graphs
for the U.S. of the three most important macroeconomic statistics, output, un-
employment and inflation, since 1900. Essentially, Ec 207 tried to explain why
the graph of real GDP sloped upwards. It also tried to explain why there were
fluctuations around the trend, via real business cycle theory, but was much less
vi CONTENTS

successful. This course will explain the trend in and growth rates of inflation
and unemployment, and fluctuations in real GDP. It will also explain why these
variables move together- that is, unemployment tends to be low when output
growth is high, and inflation is often (but not always) low when output growth
is low.
[Omitted in Spring 2002: An important distinction that I have made implic-
itly above is the separation of variables into a trend component and a cyclical
component. The trend component can be thought of informally as the long-run
average behavior of the variable, and the cyclical component deviations from
that trend. For inflation and unemployment, the trend components appear to
be horizontal lines (with possible shifts in the level of the line for both over
time). When one assumes that a model like the Solow growth model explains
the long-run growth rate of output, but not the short run, one is already doing
such a division. There has been a debate in recent years over whether it is
appropriate to do such a division; some claim that variables like output, rather
than having a deterministic trend, as is claimed in the Solow model (where the
trend component, in log terms, is just proportional to time), instead have a
stochastic trend. Algebraically, the two cases are:

yt = α + βt + ²t (1)

for the deterministic trend case, and

yt = β + yt−1 + ²t (2)

in the stochastic trend case (a random walk with drift).1 yt = ln(GDP ) mea-
sured at time t. In the first case, βt is the trend component or GDP and ²t
is the deviation around the trend. Changes in ²t cause temporary variations
in GDP, but do not affect the long-run level of yt , which is only determined
by α + βt, trend growth. In contrast, in the second specification changes in ²t
permanently affect the level of yt .
In the stochastic-trend case, it may be more appropriate in some instances to
study the long-run and the short-run together. This was one of the motivations
of the RBC literature. For the purposes of this course, I am going to sidestep
this debate, partly because it requires some heavy-duty econometrics to fully
understand, but primarily because many macroeconomists have concluded that
even if output does have a stochastic trend, analyzes assuming it has a deter-
ministic trend will give many of the right answers. This is because computing
∆yt = yt − yt−1 gives the same answer in both cases, so that any finite-sample
time series with average growth rate of β can be represented by both processes.
For more information, see the first chapter of Blanchard and Fischer.]
We will cover the following topics in this course:

• Money and Prices: In Ec 207, although you may have occasionally referred
to variables denominated in dollars, the fact that transactions required a
1 This is a special case of what is known as a unit root process. See any time series textbook

for further discussion.

CONTENTS vii

medium of exchange wasn’t mentioned, and played no role in any of the

analyses you went through. This section will define what money is (which
turns out to be less obvious a question than one might immediately think),
describe theories of money demand, and describe the long-run behavior of
money and the price level.
• Nominal Rigidities and Economic Fluctuations. The previous section was
merely a prelude to this section, in a way. In the RBC section of 207,
you saw some explanations for why output and unemployment fluctuated
around their trend values (loosely speaking): variations in technology and
in tastes for leisure. In this section of the course you will see other ex-
planations. They all center around the notion that prices and wages may
be inflexible, and thus do not move rapidly enough to clear the markets
for goods or labor. This is an idea which dates back to the foundations of
macroeconomics, with the writings of Keynes. Over the years, in response
to problems fitting the model to empirical data and theoretical challenges,
people have made Keynes’ thinking more mathematically precise. Many
of the same conclusions remain. This section will essentially present these
models as they developed historically. Along the way, we will need to
think about how firms set prices and wages and about the macroeconomic
implications of imperfect competition.
• Macroeconomic Policy: Given an understanding of what causes economic
fluctuations, here we consider what policy can and should do about them.
We focus on whether policy should consist of adherence to (simple, but
possibly contingent) rules or should be permitted to vary at the policy-
maker’s discretion.
• Investment: Investment is the most volatile components of real GDP, and
is an important part to any serious theory of business cycles, as well as
growth. We will consider various theories of investment and also how
imperfections in financial markets may affect real economic outcomes
• Unemployment and Coordination Failure: We will conclude with a con-
sideration of several important kinds of macroeconomic models. We first
consider several reasons why the labor market fails to clear fully. We will
then think about models in which agents are searching for something- a
job, the best price, etc. These turn out to be important for determining
the average rate of unemployment. Next, we turn to models involving co-
ordination failure- that is, models in which all individuals would be better
off if they were allowed to coordinate among themselves. These models
are important for some theories of economic fluctuations.
viii CONTENTS
Chapter 1

Money and Prices

In Ec 207, there was scant reference to the fact that transactions needed a
medium of exchange to be carried out. The only references to money came
in the few cases where you were presented economic data denominated in some
currency. In this part of the course, we will see why it may have been acceptable
to ignore money, and look at the long-run relationship between money and
prices.

For some of this section, with an important exception, real output will be
exogenous with respect to money- that is, changes in the supply of money have
no effect on the level of real GDP (which is determined, for example, by the
neoclassical growth model). Later in the course, you will see models in which
changes in the nominal stock of money have real effects. Economists who believe
such models are sometimes referred to as Keynesians.

The models here obey what is known as the “classical dichotomy”- they will
have the property that real variables are determined by other real variables, and
not by nominal variables such as the money stock. Most, if not all, economists
believe that the classical dichotomy holds in the long run. Some also believe
this is true in the short run- such economists are known as “new classical”
economists; they are usually proponents of the Real Business Cycle view, which
you saw at the end of last semester. Note that the concepts of long-run and
short-run here are a little different from long-run and short-run for growth
models. There, the short run can be a few decades. Here, it’s a few years; this
is the more usual definition within macroeconomics.

We’ll begin with some definitions and history, continue with first a partial
equilibrium and then a general equilibrium model of the demand for money,
discuss the relation between money and prices, talk about how inflation can be
used as a source of revenue, and finally talk about what causes hyperinflations
and how to stop them.

1
2 CHAPTER 1. MONEY AND PRICES

1.1 Definitions
1.1.1 Prices
The price level is easier, so we’ll start with that. We want to measure the average
level of prices, i.e. the quantity which converts some measure of the quantity
of all goods and services into dollars. There are a variety of ways of doing this,
which essentially depend on whether you let the basket of goods you are using
be fixed over time or vary; for the purposes of this class, the distinction does
not matter.1 We’ll just call it P
Inflation is simply the percent change in the price level; negative inflation
is also called deflation. In discrete time, it is πt = PtP−P t−1
t−1
, or Ṗ in continuous
time.
We use the price level to deflate other nominal variables, such as the nominal
wage- thus if W is the nominal wage, W P is the real wage.
For rates, such as interest rates, we need to subtract the rate of inflation.
Thus if the real interest rate is r, and the nominal interest rate is i, then the real
interest rate r = i−π. Note that this is an approximation; the true relationship2
is that:

(1 + it )Pt
1 + rt = . (1.1)
Pt+1
If r is exogenous, this relationship is known as the Fisher equation, after
Irving Fisher. Since at any given time we know the nominal interest rate, but
do not know for certain what the rate of inflation will be for the future, we often
must distinguish between the ex ante real interest rate, it − Et πt+1 from the
ex post real interest rate, which is it − πt+1 , i.e. the actual real interest rate
observed after inflation has been observed. There is a large literature testing
whether the ex ante and ex post rates are equal.

1.1.2 Money
Defining money is harder. There are three classical qualifications for money:

• Medium of exchange
• Store of value
• Unit of account

The first simply means that money can be used to perform transactions.
This must be a fundamental reason why we hold it, and is a definite advantage
over a barter economy. Without money, in order to carry out trades we would
1 Although it does lie at the heart of the recent debate over whether CPI inflation mismea-

sures inflation. There is a good article by Shapiro and Wilcox in the 1997 NBER Macroeco-
nomics Annual on the subject
2 To see how the approximation is justified, take logs to get ln(1 + r ) = ln(1 + i ) + ln(P ) −
t t t
ln(Pt+1 ) and use the fact that for small x, ln(1 + x) ≈ x.
1.2. THE HISTORY OF MONEY 3

have to wait for a “double coincidence of wants” 3 - that is, we’d have to wait
until we found a person who would give us something we wanted in return for
something we have.
The second means that money can transfer purchasing power in the future-
it is a form of asset. It would be pretty dumb to use it solely as an asset, though,
because in general it pays less interest than other assets which are about as safe
(for example, treasury bills). Money is a dominated asset.
The third means that money is the unit in which we quote prices or in which
accounts are kept. Note that this doesn’t need to be the same as the transactions
medium- we could do our exchange with one kind of money and quote our prices
in another (true for countries with high inflation).
In principle, any asset which satisfies these criteria is money. In practice, it
is hard to measure.
The government computes a series of definitions of money which get pro-
gressively larger as they include more and more money-like objects. Currently,
three important definitions of money in the U.S. and their values (as of Decem-
ber 1998) are:
• Currency: $460 billion
• Monetary Base (Currency + Reserves): $515 billion
• M1 (Currency + checking accounts):$1.1 trillion
• M2 (M1+ savings accounts):$4.4 trillion
Remember that the technical definition of money here is different from the
popular definition, which equates money with the stock of wealth. The total
stock of wealth in the US is about three times GDP, or $24 trillion, much larger
than even the largest definition of money here.
We’ll ignore these distinctions, and just assume there is some aggregate M
which is paid no interest and used for transaction, and some other asset paid
nominal interest i which can’t be used for transactions.

1.2 The History of Money

The kinds of money in existence have changed over time. Originally, money
consisted of mostly precious metals, which were often made into coins. This
kind of money is known as commodity money, because money has an alternate
use (that is, the substance used as money is intrinsically valuable). This implies
that the stock of money is hard for the government to control, because it’s
determined by however much of the stuff is out there. A reduction in the money
supply could well occur by the sinking of a Spanish galleon.4
3 The phrase was coined by Jevons (1875), who also first presented the three qualifications

above. He added a fourth qualification, ‘Standard of Deferred Payment,’ but this is hard to
distinguish from the other three.
4 It’s been argued by some economic historians that the large influx of gold into Spain

during the 15th and 16th centuries led first to rapid economic growth, then high inflation.
4 CHAPTER 1. MONEY AND PRICES

Over time, people got tired of passing coins back and forth, so they starting
passing around pieces of paper which were promises to pay people in coins.
Holding a dollar bill once entitled you to march into a Federal Reserve bank and
demand a dollars worth of gold. Then, they forgot about the coins altogether,
but kept the pieces of paper. These could be passed around and exchanged
for other pieces of paper or for goods. This standard of money is known as
fiat money, because the money is valuable by fiat- in this case, because the
government says so. In a fiat money system, the government controls the stock
of money. In the U.S., this is done by an open market operation- the Federal
reserve exchanges bonds for money with the public.

1.3 The Demand for Money

We briefly mentioned the supply of money above. Particular details on how
money is supplied are important to some economic problems, but not the ones
we will study here. For a more full description of how this works in the U.S.,
consult Mankiw Chapter 18. Now let’s think about the demand for money. We’ll
start with a simple partial equilibrium model which is based on the transactions
role of money.

1.3.1 The Baumol-Tobin Model of Money Demand

Suppose that you want to spend some nominal amount P Y at a constant rate
over some interval of time. To buy stuff, let’s suppose that you have to have the
money in advance (this is known as a cash-in-advance constraint; Robert Clower
develops this idea in another set of models). You can hold your wealth in the
form of some asset or in the form of money. You have to go to the “bank” to
transfer wealth from the asset to money. It is costly for you to go to the bank,
and it is costly for you to hold money, so the question is how often do you go
to the bank, and how much money should you hold.
Suppose you went once during the period. Then, your average money hold-
ings would be: P Y /2. For N times per year, you would get P Y /2N (Figure
1.1)
What’s the opportunity cost of holding real balances? Since we assume
money doesn’t pay interest, the cost is equal to the interest rate (even the parts
of money that do pay interest generally pay much lower than other non-monetary
assets).
Which interest rate is it, real or nominal? Answer is nominal. To see this,
recall that assets may pay real return r, but that money declines in value with
inflation, so that it has real return −π. Hence the difference in real returns
between the two is just r − (−π) = r + π = i. The nominal interest rate is the
opportunity cost of holding cash.
Suppose the nominal cost of going to the bank is P F . This is due to the
time it takes to go to the bank.
The total cost of holding money is then: i P2NY + P F N
1.3. THE DEMAND FOR MONEY 5

We’d like to minimize this- to do so, find the first-order condition with
respect to N:
P Y −2
0 = −i N + PF (1.2)
2
which implies: r
iY
N= (1.3)
2F
PY
The amount withdrawn each time is N , so average nominal money holdings
are:
r
PY YF
M= =P (1.4)
2N 2i
A numerical example: for $720 worth of expenditures, a cost F of going to
the bank of $2, and a nominal interest rate of 5%, you should go the bank 3
times per month, withdraw $240 each time, and hold an average of $120 in your
wallet. Evidence suggests that people go much more often than is implied.
For other short-run models developed later in the class, we will want to see
how money holdings vary with income and the nominal interest rate. From the
expression above, we can derive:

• Interest elasticity = − 12
• Income elasticity =+ 12

The latter is too low empirically. However if F is proportional to income, as

it might be if it is an opportunity cost, then the elasticity will be close to 1.
This model links real balances with the price level, the nominal interest
rate and output. These links seem to be a general property of most monetary
models- the exact dependence is determined by the form of the model.
Other models of money demand:5 .

• Miller-Orr inventory model. This model was originally developed for

money demand by firms. The basic idea is that the firms will set trigger
points for their level of cash balances, at S and s. If because of receipts,
their cash balances rise until they hit the level S, the firms will go to the
bank and deposit S − r, where r is the return level, and is somewhere
between the trigger levels. Similarly, if cash balances fall to s, the firm
with withdraw funds until they are back up at the return level, r (Figure
1.2). One can show that this model produces an income elasticity equal
to 23 and an interest elasticity of − 13 . This is also known as a two-sided
(S,s) model, and it has figured in other areas of economics, including in-
ventories, price setting and portfolio management. We will see it again in
the next part of this course. It is generally the optimal policy in a model
when the state variable follows a random walk and there is a fixed cost of
adjustment.
5 See Walsh (MIT Press, 1999) for a good survey
6 CHAPTER 1. MONEY AND PRICES

• Clower Cash-in-Advance- a simpler version of the Baumol-Tobin model,

which just acknowledges that cash is needed in hand in order to buy goods-
thus the amount of cash one has is a constraint on consumption.

• Double-Coincidence of Wants Models- these model the decision to hold

money as a way of reducing search costs in an economy that is originally
a barter economy.

• Cash and Credit Models - these models incorporate the fact that some
goods can be bought with cash, and some with credit; thus the distribution
of goods is an important determinant of money demand.

• Overlapping Generations - these models use money as a way of passing

resources from one generation to another. The old can pass pieces of paper
along to the young in exchange for resources; the young will be willing to
accept those pieces of paper provided they in turn can pass them on to
other generations when they themselves are old. While favored by some
people, these don’t work very well when money is not the only asset.
However, they do explain the acceptance of fiat money quite well.

Many of these models can be subsumed into a more general model, the
money-in-the-utility-functionmodel. We will use this framework to integrate
money holdings into a general-equilibrium, growth framework.

1.4 Money in Dynamic General Equilibrium

Thus far we’ve seen an interesting partial-equilibrium version of the money
demand model. Now let’s see what happens when we introduce money into a
model of economic growth. We will use a model developed by Miguel Sidrauski,
who like Ramsey died at a tragically young age. The presentation below will
largely follow that developed by Blanchard and Fischer.
As noted above, the fact that money is a dominated asset makes it difficult
to model the demand for it; demand must come through its ability to facilitate
transactions. To get around this, we assume that money directly enters the
utility function; then people hold it because they like it.
This may seem a little crazy, but in fact Feenstra (JME, 1986) has shown that
many models of money demand can be written this way. The intuitive reason is
that money provides “liquidity” or “transactions” services which people value.
Putting money in the utility function is a way of giving it value without explicitly
modeling the way it provides value.6
I first present and solve the model in discrete time, and then solve it in
continuous time. The latter solution will introduce a solution method which is
very useful in dynamic models.
6 There are technical conditions on the way in which money provides ‘liquidity services’

which restrict the ability to do this. The Baumol-Tobin model under standard assumptions
satisfies the conditions.
1.4. MONEY IN DYNAMIC GENERAL EQUILIBRIUM 7

1.4.1 Discrete Time

Setup
For convenience, let’s assume there is a single agent and no population growth.
The agent derives (instantaneous) utility from two sources:

• Consumption, denoted Ct .
Mt
• Holding real balances, denoted Pt

With discount factor θ, the time-zero present value of felicities is:

X∞ µ ¶t
1 Mt
U (Ct , ) (1.5)
t=0
1+θ Pt

Each period, the consumer receives income7 Yt = F (Kt−1 ) and a lump-sum

transfer Xt . She also has money left over from last period, Mt−1 , whose current
real value is MPt−1
t
, and capital, Kt−1 (assume depreciation is zero). She must
choose to allocate these resources in three ways:

• As consumption, Ct .
Mt
• As new money holdings, with real value Pt .

• As capital, Kt .

Let rt−1 = F 0 (Kt−1 ), i.e. the marginal product of capital and the real
interest rate.
Hence we can write each period’s budget constraint as:

Mt Mt−1
Ct + Kt + = F (Kt−1 ) + Kt−1 + + Xt (1.6)
Pt Pt
The consumer maximizes utility subject to the above set of budget con-
straints.
Let λt denote the set of Lagrange multipliers for the time t flow constraint.
Assume that the transfer is provided by the government, which also supplies
nominal balances to meet demand. By supplying nominal balances (i.e. printing
money), the government creates revenue for itself. This revenue is known as
seigniorage. Assume that the revenue from seignorage is entirely refunded back
to the public in the form of the transfer (alternatively, you could assume that it
helps pay for some form of government expenditure, G). Then the government’s
budget constraint becomes:
Mt − Mt−1
= Xt . (1.7)
Pt
7 This is a shortcut for saying that the consumer receives labor income and capital income,

profits are zero, and there is no technical progress.

8 CHAPTER 1. MONEY AND PRICES

Solution: The Hard Way

Form the Lagrangian and differentiate with respect to the three choice variables
(Ct , Mt , Kt ) to obtain the following three sets of first-order conditions:
µ ¶t
1 Mt
UCt (Ct , ) − λt = 0 (1.8)
1+θ Pt
µ ¶t
1 1 Mt λt λt+1
UMt (Ct , )− + =0 (1.9)
Pt 1+θ Pt Pt Pt+1
−λt + λt+1 (1 + F 0 (Kt )) = 0. (1.10)

In addition, there are two transversality conditions for capital and money:

³ ´t (1.11)
1
limt → 0 1+θ λt Kt = 0
³ ´t (1.12)
limt → 0 1
1+θ λt M
Pt = 0.
t

Note that λt has an interpretation as the marginal utility of consumption.

For diminishing marginal utility of consumption and real balances, the bud-
get constraint will bind each period.
We now have all the ingredients to solve the model. In the end, we’ll be
interested in expressions for consumption growth, money growth, and output
growth. To derive these, let’s derive an intertemporal Euler equation for con-
sumption and an intratemporal one for consumption and real balances.
We can obtain the intertemporal condition by solving out for λt and λt+1 ,
and using the definition of rt :
1 + rt
UCt = UCt+1 (1.13)
1+θ
The intratemporal condition can similarly be obtained by solving out for λt
and λt+1 :
à Pt
!
Pt+1
UMt = 1− UCt . (1.14)
1 + rt
Also, let’s note that if we substitute the government’s budget constraint into
the consumer’s budget constraint, we obtain the following form of the constraint:

Ct + ∆Kt = F (Kt−1 ), (1.15)

or Ct + It = Yt , the national income accounting identity.
The consumption condition looks identical to the standard consumption
Euler equation. There is a key difference, however: the marginal utility of
consumption may be a function of real balances. Hence equation this may be
an intertemporal equation not just for consumption, but also for real balances.
1.4. MONEY IN DYNAMIC GENERAL EQUILIBRIUM 9

To explore the implications of this, let’s assume a particular form of the

felicity function:

(Ct mα
t)
1−σ
U (Ct , mt ) = , (1.16)
1−σ
where for convenience we’ve set mt = M Pt
t

This function has the advantage that if α = 0, we revert to the standard

CRRA utility function. Also note that in the general case where α 6= 0 that con-
sumption and real balances are nonseparable; this will turn out to be important
later.
By inserting these expressions into the first-order conditions, and using the
relation 1 + rt = (1+it )Pt
Pt+1 , we can rewrite them as follows:

µ ¶−σ µ ¶α(1−σ)
Ct 1 + rt mt+1
= (1.17)
Ct+1 1+θ mt
µ ¶
1
mt = α 1 + Ct (1.18)
it

Finally, taking logs, we can rewrite the two equations as follows:

rt − θ α(1 − σ)
∆ ln(Ct+1 ) = + ∆ ln(mt+1 ) (1.19)
σ σ
ln(mt ) = ln(α) − ln(it ) + ln(Ct ), (1.20)

where we’ve used the approximation that ln(1 + xt ) ≈ xt for small xt .

These two equations have nice economic interpretations. The first one is
the standard discrete-time expression for consumption growth from the Ramsey
model, with the addition of a term for growth of real balances. Note that in
a steady state where consumption and real balances are not growing (recall
that there is no population growth or technical progress), we have the standard
condition that rt = θ. This pins down the steady-state capital stock and thus
steady-state output. From (11), we can then get steady state consumption.
All three values are exactly the same as in the model without money. Hence,
in the long-run steady-state, this model obeys the classical dichotomy; no real
variable is affected by money growth (although, as we will see below, the level
of utility is). Out of steady-state, note that growth of real balances does affect
consumption growth, if α 6= 0. This arises from the fact that money is both a
store of value and provides utility through transactions services.
The second equation is simply a standard money demand equation relating
real balances to the nominal interest rate and expenditure (here, consumption).
The income elasticity is +1 and the interest elasticity is -1, both larger in ab-
solute value than the Baumol-Tobin model values.
The next section considers an easier way of solving this model.
10 CHAPTER 1. MONEY AND PRICES

Solution: The Easy Way

For an easier way of solving this model, let’s derive the intertemporal con-
sumption Euler equation and the intratemporal consumption-real balances Euler
equation from (xx)-(zz) using perturbation arguments.
First, let’s consider the experiment of, along the optimal path, reducing
consumption by one unit in period t, investing it in capital, and consuming
the proceeds next period. To first order, the utility change from reducing con-
sumption in t is:−UCt . The returns from investing in capital are 1 + rt . The
utility change from consuming the proceeds, discounted back to period t, are:
1+rt
1+θ UCt+1 . If we were on the optimal path to consumption, first-order changes
for small perturbations leave the discounted value of utility unchanged, so the
sum of the above changes is zero, or:
1 + rt
−UCt + UCt+1 = 0. (1.21)
1+θ
For the intertemporal condition, let’s consider the experiment of, again along
the optimal path, reducing consumption in period t by one unit, putting the
proceeds in the form of money, and consuming the resulting real amount next
period. To first order, the utility change from reducing consumption in t is
again:−UCt . One unit of consumption is Pt units of nominal money Mt , or 1
unit of real balances mt , yielding utility change of UMt . Next period, Pt dollars
have a real value of PPt+1
t 1
, yielding a (discounted) utility change of 1+θ Pt
Pt+1 UCt+1 .
These again sum to zero, yielding:
1 Pt
−UCt + UMt + UC = 0. (1.22)
1 + θ Pt+1 t+1
By substituting in the consumption Euler equation, we can proceed as in
the previous section.

1.4.2 Continuous Time

Assume that instantaneous utility per person is given by u(c, m), where m =
M
P N , i.e. real balances per person. We have real balances rather than nominal
balances because money is only desirable to the extent it can be used to purchase
real goods; the amount of real goods that a nominal amount of money M can
purchase is simple M P . As usual, people discount at the instantaneous rate θ,
meaning their lifetime utility is:
Z ∞
V = u(c, m)e−θt dt. (1.23)
0

As before, they will receive income in the form of (real) wages, w, and as
interest on their assets. They may hold assets in the form of either capital, K
in real terms, or as money, M ,
in nominal terms. The rate of return on capital is r, and money has no
nominal rate of return; we will derive its real rate of return below. We will also
1.4. MONEY IN DYNAMIC GENERAL EQUILIBRIUM 11

assume for now that they receive a real transfer X from the government; more
on this later.
Given this income, they can choose to consume it, accumulate capital or
accumulate real balances. In non-per-capita terms, the total budget constraint
is then:
Ṁ
C + K̇ + = wN + rK + X (1.24)
P
We can convert the third term on the left-hand-side into changes in real balances
¡ M˙ ¢
by noting that Ṁ M Ṗ
P = P + π P , where π = P , the rate of inflation. Hence:

µ ˙ ¶
M M
C + K̇ + = wN + rK − π +X (1.25)
P P

which illustrates the fact derived above for the Baumol-Tobin model that real
balances have a return equal to minus the rate of inflation. This may be written
in per-capita terms as:

c + k̇ + ṁ = w + rk + x − n(k + m) − πm (1.26)

where lower case letters denote real per capita terms (recall that m = PMN ).
Note that real balances enter the right-hand-side with a real return of −π; even
though their nominal return is zero, their real return is negative due to inflation.
To solve the optimization problem, it will be convenient to let a = k + m, where
a denotes total per person assets. Then we may write the budget constraint as:

ȧ = [w + x + (r − n)a] − [c + (r + π)m] (1.27)

r + π, which is the nominal interest rate, is the opportunity cost of holding
money.
Let’s also note that the growth in real balances per person, ṁ
m is by definition
equal to the
growth in nominal balances less growth in prices (i.e. inflation) and popu-
lation growth, or ṁ Ṁ Ṗ 8
m = M − P − n. If we let µ denote the growth in nominal
ṁ
balances, then: m = µ − π − n. µ is a policy variables which is set by the gov-
ernment. Thus if we can pin down the growth rate of real balances per person,
and are given a growth rate for nominal money and for population, we will know
the inflation rate.
Before we close the model, let’s note that we have enough information to
solve the consumer’s optimization problem. The solution is characterized by the
following first-order conditions from
the Hamiltonian:9
First for m and c
8 Since the growth rate of xy
= the growth rate of x - the growth rate of y.
9 You will have a lengthy description of the Hamiltonian in Harl Ryder’s math course. It
is best to see it as a continuous-time analog of the Lagrangian. See Blanchard and Fischer,
p. 39, for a brief economic introduction, and Appendix A of these notes.
12 CHAPTER 1. MONEY AND PRICES

uc (c, m) = λ (1.28)

um (c, m) = λ(r + π) (1.29)

Then from the asset accumulation equation:

λ̇ − θλ = −(r − n)λ. (1.30)

Note: I’ve done something a little sneaky here. When I did this, I wrote the
Hamiltonian as:

H = u(c, m)e−θt + λe−θt [w + x + (r − n)a − (c + (r + π)m)] (1.31)

That is, I replaced the usual Lagrange multiplier with one multiplied by
e−θt . This is
known as the present-value version of the Lagrange multiplier. This won’t
alter the results (remember, λ is a function of time, so multiplying it by another
function of time isn’t going to change its properties). It’s convenient to work
with because it allows us to collapse the whole Hamiltonian expression to a time
discount term multiplied by something which looks like a usual Lagrangian.
There are several things to note about this:

1. We can combine the first two first-order conditions to obtain: um (c, m) =

uc (c, m)(r +π), which can (usually) be solved for m = φ(c, r +π). In other
words, we have a money demand equation as a function of the nominal
interest rate and consumption. This is an illustration of Feenstra’s point
that we can get similar results to the model with money in the cash-in-
advance constraint.
2. The condition for the evolution of the state variable λ is the same as that
for the problem without money.

Now, let’s close the model. Let’s note that in the aggregate economy, r =
f 0 (k) and w = f (k) − kf 0 (k) (from CRS) as before. Also, let’s see where these
transfers come from.
Consumers demand the money, which must come from somewhere. The
money is issued by the government. We will assume that this issuance is done
through the transfer.10
What’s the revenue from printing money? Money is issued at rate Ṁ ; the
real value of this is just Ṁ
P . This means that transfers are such that X = P .
Ṁ

Note that this means that our original budget constraint is exactly the same
as it was in the Ramsey model; however, we’ll assume that consumers take the
government transfer as given when they decide how much money to accumulate.
10 Note in passing that by issuing money, the government is also creating revenue, which it

could use to finance spending. We will return to this issue in the chapter on macroeconomic
policy.
1.4. MONEY IN DYNAMIC GENERAL EQUILIBRIUM 13

1.4.3 Solving the Model

First, as usual differentiate the first-order condition for consumption with re-
spect to time, to obtain:

λ̇ = ucc (c, m)ċ + ucm (c, m)ṁ (1.32)

Combine this with the third first-order condition to obtain the following
equation of motion:

ċ (θ + n − f 0 (k))uc ucm m ṁ
= − . (1.33)
c ucc c ucc c m

Note that the first term is exactly the same as in the standard Ramsey growth
model.
From the budget constraint, the other equation of motion is:

k̇ = f (k) − c − nk (1.34)

In the steady-state, per-capita consumption, capital and real balances are not
growing (i.e.
ċ = ṁ = k̇ = 0). This implies that:

f 0 (k∗ ) = θ + n (1.35)
and:

c∗ = f (k∗ ) − nk (1.36)
The condition that real balances per person not grow in the steady state
(i.e. ṁ
m = 0) also implies that the rate
of inflation is pinned down by:

π =µ−n (1.37)
Let’s note several things about this:

1. The steady-state conditions for the capital stock and consumption are
exactly the same as those under the Ramsey model. Money does not
affect the steady-state values of the other variables. This implies that
money is superneutral in the steady state; that is, not only do nominal
balances have no real effects, but real balances have no real effects in
the long-run, either.11 In the long run, the model obeys the classical
dichotomy: nominal variables are determined by nominal variables, real
variables by real variables. Real balances do have an effect on the level of
utility, however.
11 There is some dispute over the meaning of the term ‘superneutral’
14 CHAPTER 1. MONEY AND PRICES

2. Out of steady state, note that if ucm = 0, i.e. utility is separable in

consumption and money, real balances are also superneutral, since the
equation for the evolution of consumption is the same as it was before.
If utility is not separable, then the accumulation of money has real ef-
fects along the transition path, because it affects the level of consumption
and of capital accumulation. However, as we saw above, money is a bad
store of value; thus, it is unlikely that money has large effects on capital
accumulation simply through its role as an alternate store of value, and
empirically the effects seem to be quite small.12 This is known as the
Tobin effect.

Thus we see it is possible to add money to a general equilibrium growth

model, where we specify all the agents and markets, and that it has no long-
run effects on capital accumulation. It may have effects along the transition
path, depending on whether ucm is nonzero. Later this semester you will see
an example of money in another general equilibrium model, in which nominal
balances may have real effects and the role of money in general will be somewhat
larger than it has been here.

1.5 The optimum quantity of money

The fact that money does affect utility but does not affect the steady-state
value of other real variables means that the government can maximize the utility
from holding money without having to worry about unfortunate real effects. In
particular, it should try to induce people to hold real balances until in steady-
state the marginal utility from holding additional real balances is zero.
Since in steady-state, changing the growth rate of nominal balances is equiv-
alent to changing the inflation rate, this is like picking the optimal inflation rate.
From first-order condition (10) (i.e. um (c, m) = λ(r + π)), we see that marginal
utility of money becomes zero if we have a rate of deflation equal to the interest
rate, or π = −r, which implies a rate of money growth (actually, shrinkage) of
−(θ + n). Note that by doing this, we have pushed the return of real balances
to be equal to the return on capital, or equivalently the nominal interest rate to
zero. This result was derived by Milton Friedman in an article on the reading list
some time before the Sidrauski article came out, and is known as the optimum
quantity of money result. In practice, though, nominal balances have shown
a distinct upward trend, and the Federal reserve has aimed for price stability,
not reduced prices. However, Japan has recently seen zero (and perhaps even
slightly negative) nominal interest rates.

1.5.1 The Quantity Theory of Money

Note that another characteristic of the steady state is that the growth rate of
nominal balances= the growth rate of prices (inflation) + the growth rate of
12 Recall the comparison of the various measures of money to the total stock of wealth.
1.5. THE OPTIMUM QUANTITY OF MONEY 15

output (population growth). This is a version of a more general proposition,

called the Quantity Theory of Money, which states that nominal balances are
proportional to nominal income, or M V = P Y , where V is a constant of pro-
portionality known as velocity. V has this name because it tells us how much
a given quantity of money circulates through the income for a given stock of
transactions (nominal income). Note that this proposition implies that

Ṁ V̇ Ṗ Ẏ
+ = + (1.38)
M V P Y
which yields the result of the first sentence, assuming that VV̇ = 0.
Where does this result come from, and what is its relation to theories of the
demand for money? Suppose we think of money demand as M P = L(i, Y ), and
M
assume a unit income elasticity so that we may write it as P = L(i)Y . Then
we see that if we define V = 1/L(i), we have the quantity theory.
Looked at this way, velocity can change for two reasons:

• The nominal interest rate changes

• The form of the money demand function changes

Velocity changes due to the first reason are known as endogenous changes
in velocity, since they are changing due to the nominal interest rate, which is
something specified within the model. The second kind of velocity changes are
exogenous- they are disturbances which are not specified in the model. They
include things like changes in the demand for money due to the introduction
of new monetary assets, or due to the introduction of ATM machines. The
quantity theory’s assumption is that movements in velocity due to both of these
reasons are small and on average about zero. It’s not clear that this is true
in practice- velocity appears to have a permanent upward trend. Nevertheless,
over very long periods of time, it looks like the quantity theory approximately
holds (see Friedman and Schwartz, A Monetary History of the United States,
for a full description). This theory is one of the oldest theories in economics still
in use, older even than Adam Smith’s “Invisible Hand.” One can find references
to the quantity theory in the writings of David Hume in the 1750s, and possibly
even earlier in the 18th century. The modern version has been popularized
by Milton Friedman, and the quantity theory shows up in many (although, in
fact not all) monetary models. This makes sense - money should certainly be
proportional to the number of transactions, and the number of transactions is
probably proportional to income. Let’s use this theory to look at the behavior
of prices in a number of different cases. Output growth is zero in both, but in
one, money is growing a 5% and in the other, money is growing at 10%. It is
useful in such models to make a plot with time on the x axis and the log of
prices on the y axis, so that things growing at a constant rate are straight lines
13
. If money grows at rate g, and velocity and output are not growing, then
13 Why? well, Ṁ = g implies that M = M egt (differentiate the latter to check), so that
M 0
log(M ) = log(M0 ) + gt
16 CHAPTER 1. MONEY AND PRICES

prices will also grow at rate g. Hence the first economy will have prices which
are a straight line with slope 5, and the second a straight line with slope 10
(Figure 1.3).
Now suppose velocity increases suddenly, for some reason. What happens
to prices? Well, we can use the quantity equation to write:P = MV Y , or
logP = logM +logV −logY . So if velocity jumps up, prices will immediately
jump up, too (Figure 1.4).
Finally, suppose that money is growing at 5%, but the rate of growth sud-
denly increases to 10%. What happens? Well, we know from the first part, that
we have an increased slope. However, in this case, velocity will also change.
Why? Because increased money growth will lead to increased inflation, which
will lower desired real balances through a higher nominal interest rate and raise
velocity. Thus prices will experience a one-time jump. We’ll see this kind of
analysis a lot more formally and in depth in the next section.

1.6 Seigniorage, Hyperinflation and the Cost of

Inflation
We’ve seen how money can be integrated into a growth model, and a little bit
of the long-run dynamics of money and prices. Now, let’s look at the short-run
dynamics. Let’s assume a particular form of the money demand function, and
look at things in discrete-time (this will be a discrete-time version of a model
by Sargent and Wallace (Econometrica, 1973)).

lnMt − lnPt = −α(it ) + lnYt (1.39)

Now, recall from the Fisher equation that it = rt +Et πt+1 , where Et denotes
expectations taken with respect to the information set at time t and that πt+1 =
Pt+1 −Pt
Pt . Recall that for x near 1, ln(x) = x − 1. Hence we may rewrite πt+1 as
ln(Pt+1 ) − ln(Pt ), so that Et πt+1 is approximately Et [ln(Pt+1 )] − ln(Pt ). Given
this, we can rewrite the money demand equation as (now using lower case letters
to indicate logs):

mt − pt = −α(rt + Et pt+1 − pt ) + yt (1.40)

Let’s ignore the real interest and output terms, and normalize them to zero,
since they’re exogenous (we could alternatively write them as a constant and
not change the subsequent results). Then we have that:

mt − pt = −α(Et pt+1 − pt ) (1.41)

This says that if we knew the expected future price level and the current
stock of money, that would tell us where the current price level has to be. The
big question is then how to get the expected future price level. We will get it
by assuming rational expectations, that is that people form expectations of the
price level using the model. How do we apply this?
1.6. SEIGNIORAGE, HYPERINFLATION AND THE COST OF INFLATION17

Well, first let’s solve the above equation for pt :

α 1
pt = Et pt+1 + mt (1.42)
1+α 1+α
Now, let’s lead it one period, to obtain:
α 1
pt+1 = Et+1 pt+2 + mt+1 (1.43)
1+α 1+α
and let’s take expectations:
α 1
Et pt+1 = Et (Et+1 pt+2 ) + Et mt+1 . (1.44)
1+α 1+α
Now, using the law of iterated expectations, Et (Et+1 pt+2 ) = Et pt+2 , since
your best guess today of your best guess tomorrow of the price the day after
tomorrow is just your best guess today of the price the day after tomorrow.14
Let’s substitute this into our original equation for pt , to obtain:
µ ¶2 µ ¶
α 1 α
pt = Et pt+2 + mt + Et mt+1 (1.45)
1+α 1+α 1+α
repeating this operation, we obtain:

µ ¶n à µ ¶2 !
α 1 α α
pt = lim Et pt+n + mt + Et mt+1 + Et mt+2 + . . .
n→∞ 1+α 1+α 1+α 1+α
(1.46)
If we assume that the expected path of the price level is nonexplosive, so that
the first term goes to zero,15 then we’re left with only the second parenthetical
term.
This implies that the current price level depends on not only the current but
also the entire expected future path of the money stock. To go further, I would
either need to give you:

• Explicit values for the expected future path of money

• A stochastic process for money, from which you could then derive the
expected future path
of the money stock.

We can use this framework to analyze what happens when there are an-
ticipated and unanticipated changes in monetary policy (i.e. in the way M
moves).
For example, suppose it is announced that M is supposed to jump at some
point in the future.
14 Think about this.
15 Technically, we are ruling out bubbles in the price level.
18 CHAPTER 1. MONEY AND PRICES

What should happen to the price level?

Well, it must be the case that the price level jumps at the time of the
announcement, because the expected future value of money has changed. The
price level can’t jump at the time of the change.
Why? Well, you’ve seen the math, which has that implication. The intuition
is that the price level is expected to be higher than it was in the future. If there
were a jump at the time of the change, we would then expect that holding money
at the time of the change would involve a capital loss.
But we can’t have an expected capital loss like this; just before the time of
the change, people would be off their money demand curves. Hence prices must
jump now, then gradually increase, to insure that there is no expected discrete
capital loss.

Seigniorage
When we discussed the Sidrauski model, we noticed that in the act of issuing
money, the government was in fact also collecting revenue. This revenue is
known as seigniorage, from the French seigneur, or lord16 . It has been an
important source of revenue for some countries.
The problem with seigniorage, the revenue from printing money, is that
printing money increases inflation, which reduces the amount of revenue the
government can collect. If the government wishes to continue to pay its bills by
printing money, it has to print it at a faster and faster rate.
This can be seen by looking at the definition of seigniorage: Ṁ Ṁ M
P = M P The
first part is the growth in the nominal money stock; increasing money growth
increases the revenue from seigniorage.
The second part is real balances, which decreases with increased money
growth because increased inflation reduces money demand. To talk about this
decrease sensibly, we need a specification for the demand for money.
Let’s do this by using the Cagan money demand function, which specifies
that:

M
= Y eα(−r−π) (1.47)
P
If we substitute this into the expression for seigniorage, and note that the
rate of inflation is just equal to the growth rate of the money stock in steady
state, we can write the expression for seigniorage as just:

Y πeα(−r−π) (1.48)

Choosing the rate of inflation which maximizes this has the first order con-
dition:
Y (1 − απ)eα(−r−π) = 0 (1.49)
16 After the Norman conquest, lords were allowed to use their own coins in paying workers

on their manors. The Oxford English Dictionary also admits “seignorage” as a valid spelling.
1.6. SEIGNIORAGE, HYPERINFLATION AND THE COST OF INFLATION19

Thus seigniorage is zero when π = 0, is maximized when π = α1 and then

decreases thereafter, although it never goes to zero. Seigniorage has a Laffer
curve, like any other kind of tax.
Note that this tax is distortionary like many other kinds of taxes, so we
should have a mix of distortionary taxes.

Hyperinflation
The principle problem with seigniorage is that governments tend to overdo it
when they use it; that is, they tend to print increasingly large amounts of
money, leading to large amounts of inflation. Why? Well, often governments
which resort to seigniorage to pay their debts have no other source of finance-
in particular, they may be unable or unwilling to raise taxes.
Furthermore, the government may not understand the economics, or may
hope that they can consistently fool people by printing money at a faster rate
than people expect prices to rise.
The usual outcome in such cases is hyperinflation, defined as a rate of infla-
tion exceeding 50 percent per month. This is important because inflation rates
at this level or above tend to be very costly; we’ll say more about the costs
of inflation in a little bit, but you might imagine if inflation gets very high,
people spend a lot of time trying to hold as little money as possible, and the
distribution of relative prices becomes somewhat confused.
One could model hyperinflations by specifying an expectations mechanism
in which people don’t realize what’s happening immediately- in other words,
with non-rational expectations. Blanchard and Fischer, pp. 195-201, have such
a model. I won’t derive it here, but essentially what one can show is that
if people form inflationary expectations based on past rates of inflation, the
government can always stay one step ahead of them by increasing the rate of
money growth, and thus getting a little more revenue. In the limit, both the
rate of money growth and inflation approach infinity.
Fortunately, stopping hyperinflations is something that economists think
they actually know how to do. The root cause of hyperinflations is usually
the inability to raise revenues by other means to pay off government debts.
Therefore putting the fiscal house in order by reducing spending or increasing
taxes is an important step. Since many taxes are usually collected in nominal
terms,
hyperinflation reduces the value of taxes collected, so stopping the hyperin-
flation will in and of itself help this (this is known as the Olivera-Tanzi effect).
Simultaneous with the fiscal reform, the government should stop printing money.
Note that if it does so cold turkey, the resulting drop in expected inflation would
result in increased money demand immediately; to satisfy this, we have to have
a one-time increase in the level of the money stock, followed by the reduction
in money growth (Figures 1.5(a), 1.5(b)).
The problem with implementing this, of course, is that after printing money
in large quantities for a long-period of time, announcing that there will be a
one-time only increase in the stock of money followed by zero money growth
20 CHAPTER 1. MONEY AND PRICES

may not be credible. There is a large literature on how to establish credibility

in monetary policy, in general, which I won’t go into here. Essentially, that
literature argues that credibility is built up from one’s past reputation. In this
case, that would seem to imply that the old central bankers be replaced, since
their reputation would work against them.

Costs of Inflation
Let me conclude with a brief discussion of the costs of inflation.
I have already talked a little about this in the context of hyperinflation- it
should be intuitively pretty clear there that inflation is quite costly. But what
about the costs of inflation when the level of inflation is moderate?
This is a subject which isn’t well understood- basically, people seem to find
inflation much more costly than economists can find reason for. It’s traditional
to divide the costs of inflation into two parts: The costs of anticipated inflation,
and the costs of unanticipated inflation.
Anticipated Inflation:

• Shoeleather costs: people have to go to the bank more frequently (and

hence wear out their shoes).

• Menu costs: costs of changing prices; L.L. Bean has to print catalogs more
often if inflation is higher.

• Liquidity effect: High inflation means high nominal interest rates, which
means that payments made in nominal terms on debt (such as a mortgage)
will be high. In real terms, this won’t matter, but if for some reasons you
are limited in nominal terms by the amount you can borrow, high inflation
will limit your borrowing.

Unanticipated Inflation:
The costs here are a little more obvious

• The biggest effect is redistribution from creditors to debtors. For example,

consider thirty-year mortgages written in the 1960s; the average (nominal)
interest rate on mortgages in 1965 was 5.38 percent. Over the next thirty
years, inflation averaged 6.5 percent.
This led to a negative real interest rate, which was good for homeowners
but bad for banks, which had their comeuppance in the mid 1980s. This
fact has sometimes led people mistakenly to think that inflation is always
good for debtors but bad for creditors. For example, during the late 19th
century, when prices were falling, farmers, who were debtors, agitated for
allowing silver as well as gold to be used as money. This would have
produced inflation, which the farmers reasoned would have allowed them
to pay back their loans much more easily. But if this were expected,
nothing would change. Economic historian Hugh Rockoff The Wizard of
1.7. PROBLEMS 21

Oz is in fact a monetary allegory which advocates just what the farmers

wanted.17
The solution to such problems is simply to index things. It’s not clear
why nominal assets aren’t indexed- some governments (such as that of the
U.K.) do issue them; the U.S. just started last year. It would be really
helpful for economists, because then we’d have a measure of both the real
interest rate and expected inflation.
• Many taxes are in fact on nominal quantities. We get what’s called
“bracket creep” when peoples’ income rises due to inflation and they are
pushed into higher tax brackets. This has been eliminated for labor in-
come from the U.S. tax code, but capital gains taxes are still not indexed,
for example.

Inflation can actually be beneficial by increasing the flexibility of wages and

prices. If nominal wages can’t be lowered for some reason, inflation can lower
real wages.

1.7 Problems
1. Consider an economy with a money-demand function given by the Baumol-
Tobin model. Y and M are constant. Assume now that the real interest
rate has been constant at 2%, and that it now jumps to 8%, and will
remain at that level. How will the price level in the new steady-state
compare to the price level in the old steady state?
2. (Old final question) Suppose that per capita money demand is given by
the standard Baumol-Tobin model:
r
M FY
= (1.50)
P 2i

Where Y is real income per capita, M is nominal money balances per

capita, P is the price level, i is the nominal interest rate, and F is the cost
of going to the bank. Inflation is equal to zero and the real interest rate
is constant. Suppose that total real income in this economy is growing
at a rate of 10% per year. For each of the two cases below, calculate the
rate at which the money supply must be growing, and describe what is
happening to velocity over time.

(a) Income per capita is constant, and the population is rising at 10%
per year.
(b) Population is constant, and income per capita is rising at 10% per
year.
17 The movie obscures the allegory by, among other things, changing Dorothy’s magical

slippers from silver to ruby.

22 CHAPTER 1. MONEY AND PRICES

3. Consider an economy with a money demand function:

M
= Y e−i (1.51)
P
Assume that the real interest rate is 5%, Y is growing at 3% and money
is growing at a rate of 20% (and that people expect it to keep growing at
this rate), All of a sudden, the growth rate of money falls to zero (and
people expect it to remain at zero). What will happen to the price level?
To the rate of inflation? Draw a picture of the log of the price level.
4. (Old midterm question) Consider the problem of maximizing seigniorage
in a growing economy. Money demand is given by:

M
= Y e−i (1.52)
P
The real rate of interest is zero. Output grows exogenously at rate g.
Find the rate of inflation which maximizes steady-state seigniorage as a
fraction of output. Explain why the seigniorage-maximizing rate of output
depends on the growth rate of output.
5. Consider an infinitely-lived individual who receives real wage of 100 per
period. The price level at the beginning of her life is 1, and she buys a
house of value 500 the moment she is born. She pays for the house by
taking out a mortgage for its full price. She never repays any of the any
of the nominal principal of her loan; instead, she pays interest at rate i for
perpetuity. The real interest rate is r and the rate of inflation is π.
What money she does not spend on mortgage payments she consumes.
She is unable to borrow for purposes other than buying a house. Her
utility from non-housing consumption is:
Z ∞
V = e−θt cdt (1.53)
0

where her time discount rate is θ. Note that the instantaneous utility
function is assumed to be linear.

(a) Find the path of her real mortgage payments

(b) Calculate the path of her real consumption
(c) Why does inflation make this person worse off? Assume r = 5%,
θ = 10% and
π = 10%. Calculate the reduction in her real wage she would accept
in order to live in a world
with zero inflation.

6. Analyze the relation between inflation and real seigniorage in the case
where money demand is described by the Baumol-Tobin model and the
real interest rate is zero
1.7. PROBLEMS 23

7. A person lives for two periods. His utility function is U = ln(c1 ) + ln(C2 ).
In the first period, he does not earn any income, while in the second period
he earns a wage with a real value of 100. In the first period, he can borrow
at a nominal interest rate of 10% per period. It is known that the actual
rate of inflation will be 0% with probability .5 and 20% with probability
.5.
What would be the value to this person, expressed in terms of real income
in the second period, of being able to borrow at a guaranteed real interest
rate of 0%.
8. Assume the money demand function is (in logs):

mt − pt = −α(Et pt+1 − pt ) (1.54)

and that expected inflation is set equal to last period’s inflation:
(Et pt+1 − pt ) = (pt − pt−1 ) (1.55)

Solve for the price level in period t as a function of the level of the money
stock in current and past periods. Why is it the case that, holding money
in period t constant, a higher level of money in period t − 1 leads to a
lower price level in period t.
9. (Old final exam question) Assume that money demand in each period is
given by:
mt − pt = γ − α(Et pt+1 − pt ) (1.56)
where m and p are the logs of the money supply and the price level. It is
known with certainty from time zero that the pattern of the money supply
will be as follows: for periods 0, . . . , s m = 1. For periods s + 1, . . . , ∞
m = 2. Solve for the price level in period s.
10. Suppose that the demand for money is described by the following familiar
model:
mt − pt = −α(Et pt+1 − pt ) (1.57)
Suppose that at time T − 1, the stock of money is M̄1 , and is expected to
remain at that level for the indefinite future. Now, at time T , the stock
of money is raised to level M̄2 , and it is expected to remain there for the
indefinite future. Derive the price level for periods T −1, T, T +1, T +2, . . ..
11. Consider an economy in which money has been growing at rate µ for some
time and is expected to continue to grow at that rate, so that:
mt = µ + mt−1 + ²t (1.58)

where mt = ln(Mt ) and ²t is iid with zero mean. Money demand is

described by the familiar function:
Mt Pt+1 −Pt
= Yt e−α(rt +Et Pt ) . (1.59)
Pt
24 CHAPTER 1. MONEY AND PRICES

You may normalize any irrelevant exogenous variables as you see fit. As-
sume the classical dichotomy holds. Also, recall that ln(1 + x) ≈ x for
small x. You need not completely simplify your results.
(a) Solve for the price level at time T .

Now suppose that a monetary reform is announced at time T . If the

reform is successful, the money stock will be:

mt = mt−1 + ²t (1.60)
from time T + 1 onwards. The reform has probability θ of success. If
it is unsuccessful, the money supply process remains unchanged from
the previous equation, and it is believed that the reform will not be
tried again.
(b) What is the new price level at time T after the announcement has
been made?
(c) Work out the new price level at time T + 1, first assuming that the
reform succeeds and then assuming that it fails. Provide intuition
for your results.

Now suppose it is believed that if the reform fails, it will be tried once
and only once more. The new reform has a probability of success
φ < θ.
(d) Work out the price level at time T after the announcement of the
reform package has been made. Compare your answer to the one
obtained in part (b) and explain.
(e) Work out the price level at time T + 1 first assuming the reform has
been successful and then assuming it has been unsuccessful. Compare
your answer to the one obtained in part (c) and explain.
12. Consider an economy where money demand has the Cagan form:
M
P = Y e−αi .
Assume the Fisher equation relating nominal and real interest rates holds.
Suppose there are two types of beings in the economy: economists, who
form expectations rationally, and normal people, who expect prices to grow
at rate γ forever. Normal people comprise fraction β of the population.
(a) Solve for the price level and the inflation rate, assuming the Classical
dichotomy. You may also make any convenient normalizations you
wish.
(b) Solve for the price level and inflation rate under the assumption that
Alan Greenspan has in fact been implementing a policy of growing
the money stock at γ percent per year. What happens to prices and
inflation as β gets close to one? What happens as it gets close to
zero?
1.7. PROBLEMS 25

(c) Now suppose that Fed Chairman Alan Greenspan is replaced by his
evil Mirror Universe counterpart. The Evil Greenspan is only in-
terested in maximizing the amount of seignorage collected by the
government. What growth rate of the money stock maximizes the
amount of seignorage?
(d) Assume the growth rate of the money stock you derived in the pre-
vious part is implemented. At some point, the normal people realize
that a change in regime has occurred, and adjust their inflationary
expectations to be equal to the growth rate of the money stock you
derived in the previous part. Describe the path of the price level and
inflation from a time before this realization to a time after it.

13. Consider the following variant of the Sargent and Wallace/Brock model
of forward determination of the price level:

mt − pt = −α(rt + (Et (pt+1 ) − pt )) + yt (1.61)

yt = −βrt (1.62)
yt = d(pt − Et−1 pt ), (1.63)

where all notation is standard and all variables (except for rt ) are in
natural logs.

(a) Provide economic interpretations for these equations.

(b) Suppose the money stock has been growing for since time t = 0 at
rate γ, and is expected to grow at that rate forever. Define m0 = m̄.
Solve for all endogenous variables at time t. (Hint: For this part and
the next two, it isn’t necessary to use recursive substitution to solve
the expectational difference equation. Use intuition to get a much
simpler solution.)
(c) Suppose the assumptions of the previous question hold, but unex-
pectedly at time T the money stock jumps by an amount ². This
jump is expected never to happen again. Solve for all endogenous
variables.
(d) Now assume instead that at T , the subsequent growth rate of the
money stock is expected to increase to γ + ² and remain permanently
at that rate. Solve for all endogenous variables.

14. Consider a version of the Sidrauski money-in-the-utility-function model in

which nominal government debt is used to carry out transactions instead
of money. There are two stores of value: capital, which has real return
r, and nominal debt, which has nominal return z. Let π denote the rate
of inflation and assume that z < r + π. The government also purchases
a constant amount g per capita of goods and services; this spending does
26 CHAPTER 1. MONEY AND PRICES

not contribute to utility. The spending and debt service is financed by

per-capita lump-sum taxes in the amount τ and issuance of new nominal
debt. Assume agents have discount rate θ, and that there is no population
growth or technical progress. Let k denote the per-capita capital stock and
b the per-capita stock of nominal debt. Production is given by y = f (k).

(a) Write down the consumer’s optimization problem.

(b) Write down the equation of motion for nominal debt.
(c) Solve the optimization problem to give the joint behavior of con-
sumption, capital and debt.
(d) Under what conditions does nominal debt have no effect on consump-
tion or capital accumulation?
(e) Suppose there is a permanent rise in per-capita taxes, τ . Describe the
transition path of and steady-state effects on capital, consumption
and debt.
(f) Suppose there is a permanent rise in per-capita government pur-
chases, g. Describe the transition path of and steady-state effects on
capital, consumption and debt.
(g) What happens if z = r + π? If z > r + π?

15. Consider the following model of money and growth (due to Tobin):
There are two assets: money and capital. Population grows at rate n.
There is no technical progress. The government prints nominal money at
rate µ. Seignorage is remitted back to the public as a lump-sum transfer.
Denote the per-capita production function by f (k). Depreciation occurs at
rate δ. The representative agent saves a constant fraction s of her income.
Suppose that money demand per capita is proportional to the product of
consumption and the nominal interest rate, so that PMN = φ(r + π)c. All
notation is standard.

(a) Write down the equation of motion for the capital stock for this
particular economy.
(b) Derive the steady-state growth rates and levels of output and capital
stock.
(c) Do any of your answers in part (b) depend on the growth rate or
level of the money stock in the steady state?
(d) Compare results with the results from the Sidrauski money-in-the-
utility-function growth model. Try to explain any differences between
the two models.
Chapter 2

Nominal Rigidities and

Economic Fluctuations

In Ec207 and so far in Ec208, we have only seen models in which the classical
dichotomy holds- that is, real variables are determined by other real variables,
and nominal variables by nominal and real variables, but nominal variables don’t
determine real variables.1
In Economics 207, you saw that completely real models had significant diffi-
culties explaining the joint behavior of real macroeconomic variables in the short
run. In this section of the course, we will pursue an alternative set of models,
sometimes called Keynesian economics, which involve interactions between real
and nominal variables. By allowing for some form of nominal price or wage
rigidity, these models will imply that changes in the nominal stock of money,
and more broadly changes in aggregate demand, will have effects on output,
unemployment and other real variables. Hence the potential list of candidates
for the causes of economic fluctuations becomes much larger.
We will approach this topic chronologically. In the first section, we will
begin by looking at the version of the model generally agreed upon by the mid
1960s. This version, referred to as the “neoclassical synthesis,” is still taught
in undergraduate classes today. It involved simple aggregate models and a
large number of ad hoc assumptions, but provided a fairly useful framework for
describing the short-run dynamics of money and prices.
The lack of microeconomic foundations for this model was troubling to many.
In the late 1960s and early 1970s, several economists attempted to remedy this
deficiency by taking a small-scale Walrasian GE model and adding fixed prices
and wages to it. The resulting models, known as disequilibrium economics,
shared many predictions with the original Keynesian models. They are the
subject of section 2.
1 An exception is the behavior of the Sidrauski model along the transition path. The effects

of money on capital accumulation in that model are not empirically large enough to explain
the short-run effects of money on output.

27
28CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

In the third section, we present an attempt to generate Keynesian results

in a model in which expectations are considered more carefully than in the
previous model. This model, by Robert Lucas, has the troubling implication
that anticipated monetary policy has no real effects.
In the fourth section, we show that models with rational expectations need
not imply impotence of monetary policy. We first present models with long-
term nominal contracts. After a brief discussion of the macroeconomic effects of
imperfect competition, we go on to discuss models in which prices are inflexible
because it is costly to change prices. We study these “menu cost” models first
in a partial equilibrium framework, then in a static GE framework, and finally
in a dynamic GE framework. These models, written in the 1980s and 1990s, are
often lumped together until the title “New Keynesian Economics.”
We will conclude by looking at some empirical studies of these models, and
new directions for this literature. We shall see that although the forms of these
models have changed over the years, many of the implications have not.

2.1 Old Keynesian Economics: The Neoclassical

Synthesis
The standard model of the economy in the short-run which is still presented
in undergraduate macroeconomics textbooks (e.g. Mankiw) is the Aggregate
Demand/Aggregate Supply model. The Aggregate Demand side was developed
by John Maynard Keynes and John Hicks in the 30s and 40s. The Aggregate
Supply side, although referred to by Keynes, did not receive its full development
until the 1950s and early 1960s. The resulting model is also known as the
“neoclassical synthesis.”
In general terms, the model assumes there are two relationships between
output and the price level: a downward sloping relationship, aggregate demand,
and an upward sloping one, aggregate supply. Aggregate demand is usually
associated with the joint demand for goods and services of consumers, firms and
the government2 , while aggregate supply arises from the pricing and production
decisions of firms. The intersection of the two curves determines the price level
and output. Fluctuations in either aggregate demand or aggregate supply can
lead to short-run variations in output and prices. It is assumed that in the
long-run, aggregate supply is vertical, and variations in aggregate demand only
lead to variations in prices (Figure 2.1).
For the moment, make the extreme assumption that prices are completely
inflexible in the short run, so that aggregate demand disturbances affect output
completely. Under this assumption, we shall derive the components of aggregate
demand, and look at the short-run effects of various shocks to output.
The detailed version of the aggregate demand model is known as the IS/LM
2 Despite its name, aggregate demand is not exactly analogous to the standard single-market

demand curve, since in the absence of money there is no reason for demand for all goods to
depend on the price level.
2.1. OLD KEYNESIAN ECONOMICS: THE NEOCLASSICAL SYNTHESIS29

model (so named by Hicks). This model specifies equilibrium behavior in the
markets for goods, money and bonds.
Recall that from the national income accounting identities, Income=Output=Expenditure.3
In a closed economy, this implies that Y = C + I + G, where Y is real GDP,
as usual, C is consumption, I is investment and G is government purchases of
goods and services. Next, make the following behavioral assumptions about
each of the components of expenditure:

• C = C(Y − T ). Consumption is a positive function of current disposable

income. It is assumed that 0 < C 0 < 1. This function, the Keynesian
consumption function, is clearly not consistent with the pure LCH/PIH.
Interest rates and expected future income should appear. However, it is
consistent with LCH/PIH if a large fraction of the population is liquidity
constrained.

• I = I(r). r, the real interest rate, is the opportunity cost associated with
investing. Thus the dependence is negative. We shall see later that this
may be a valid simplification under some circumstances.

• G and T , the fiscal policy variables, are assumed exogenous and fixed.

Given this, one can write the national income accounting identity as: Y =
C(Y − T ) + I(r) + G. Note that as usual, one can rewrite this as: Y − C(Y −
T ) − G = I(r), or National Saving=Investment. This relationship is known as
the IS curve, for investment=saving. Provided aggregate supply is horizontal,
this equation also represents goods market equilibrium.
Equilibrium in the money market requires that money supply equals money
demand. We can write this condition as: M P = L(i, Y ), where i is the nominal
interest rate. This condition is uncontroversial, since one can derive the money
demand function from more microeconomic models such as the Baumol-Tobin
model. It is known as the LM curve, from demand for liquidity=supply of
money.
Since prices are assumed fixed, this implies that expected inflation is zero.
Through the Fisher equation, this implies that ex ante real rates are equal to
the nominal interest rate, so that r = i. The results would not be modified
qualitatively if one assumed a constant expected rate of inflation.4
We need not specify equilibrium conditions in the market for bonds (which
comprise the alternative to holding money), because by Walras’ Law, that mar-
ket clears provided the other two markets clear.
Thus our two equilibrium conditions are:

Y = C(Y − T ) + I(r) + G (2.1)

3 Seean undergraduate text, such as Mankiw, chapter 2.
4 There is an old literature on policy which distinguishes between the short-term interest
rate which is the appropriate rate as an alternative for holding money, and a long-term rate
which is appropriate for investment. This is practically important for formulating policy, but
not germane to the point here.
30CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

M
= L(r, Y ) (2.2)
P
The exogenous variables are G, T , M and P , and the endogenous variables
are Y and r.
We can plot these relationships in a graph with the two endogenous
variables on the axes (Figure 2.2)
This yields the usual predictions that expansionary fiscal policy (i.e. G up
or T down) will increase output and raise real interest rates (crowding out).
Expansionary monetary policy will increase output and lower real interest rates
(Figure 2.3). We can state these results more formally by totally differentiating
the two equations, to yield:

dY = C 0 (dY − dT ) + I 0 dr + dG (2.3)

µ ¶
M
d = Lr dr + LY dy, (2.4)
P
where derivatives are indicated by primes or subscripts, as appropriate. We
can solve out for dY and dr, either by direct substitution, or by using Cramer’s
rule. Either way, solving out for dY obtains:
µ ¶
0 I0 M
dY = −(ΓC )dT + ΓdG + Γ d (2.5)
Lr P

where

Lr
Γ= (2.6)
Lr (1 − C 0 ) + I 0 LY
LY
If the LM curve is nearly horizontal, so that Lr is nearly zero, then approx-
1
imately Γ = 1−C 0 , and

−C 0 1
dY = dT + dG (2.7)
1 − C0 1 − C0
1 0
The term 1−C 0 is known as the Keynesian multiplier. Note that since C ,

the marginal propensity to consume out of disposable income, is less then one,
this number is greater than one. Thus changes in government spending have a
more than one-for one effect on output.
This works through an expenditure mechanism. The initial increase in
spending, dG represents income to someone. Some fraction of this income,
C 0 dG is spent. This additional spending is also income to someone; these peo-
ple spend some fraction C 0 of it, so that their increased spending is (C 0 )2 dG.
The process continues, so that the total increase in expenditure (and therefore
output) is:
dG + C 0 dG + (C 0 )2 dG + (C 0 )3 dG + . . . (2.8)
2.1. OLD KEYNESIAN ECONOMICS: THE NEOCLASSICAL SYNTHESIS31

1
which simplifies to: 1−C 0 dG, as above.

Note that the assumption LLYr ≈ 0 implies that the effects of monetary policy
on output are small. This conclusion was disputed by a group of economists
who believed that most short-run fluctuations in output were due to shocks
to the money supply. Because of this view, they were known as Monetarists;
their leader was Milton Friedman (who, with Anna Schwartz, wrote an 800
page monetary history of the United States referred to in the previous chap-
ter). Although at the time, their views were seen as a significant challenge to
Keynesian views, from today’s perspective the controversy can simply be cate-
gorized as disagreement over the magnitudes of derivatives within the Keynesian
model. Thus, Monetarists believe C 0 to be small, and Keynesians believed it
to be large. Monetarists believed Lr to be small, an implication one would get
from the quantity theory of money, while Keynesians believe it to be large. To-
day, the Monetarist perspective has been largely accepted, and money is seen
as an important component of models of economic fluctuations.
Finally, note that the IS and LM curves can jointly be solved to produce a
relationship between output and prices. Invert the LM curve, to obtain the real
interest rate as a function of output and real balances, and then plug it into the
IS curve. The result is that Y = g( M P ). Note that if this function is log-linear,
we return to the quantity theory, since that can be rewritten as Y = V ( M P ).
We could also have obtained the same result graphically, be observing that an
increase in the price level is identical to a contraction in the money stock, and
will reduce the level of output (Figure 2.4).

2.1.1 Open Economy

This model may easily be extended to the open economy. In this case, the
national income accounting identity tell us to rewrite the IS curve as:

Y = C(Y − T ) + I(r) + G + N X(²), (2.9)

where N X=net exports, exports-imports, and ² is the real exchange rate. If the
nominal exchange rate, defined as foreign currency units per domestic unit, is
labeled e, and if asterisks denote foreign variables, then
² = PeP∗ .5 For fixed domestic and foreign price levels, increases in the nominal
exchange rate e lead to real appreciations, since it now requires fewer domestic
goods to obtain a given foreign good.
The dependence of N X on ² is negative, since appreciations should lead to
an increase in imports and a decline in exports, and vice-versa. Note that we
can manipulate this relationship to yield S − I = N X, which in other words
states that the current account=- the capital account, or that trade deficits
must be paid for by other countries accumulating claims on domestic assets. In
5 One can show through dimensional analysis that this yields foreign goods per domestic

good. The nominal exchange rate is often defined as the reciprocal of the definition given here
(e.g. the British pound is $1.50).
32CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

principle, this also implies that S and I need not move together- in other words,
a country with a relatively low S can have a high I.
Also assume that there is perfect capital mobility, so that by arbitrage i =
i∗ − E[ eė ]. One can derive this by noting that if one invests $1 in an asset
denominated in domestic currency, one will get $1 + i next period. One should
get the same expected return by investing abroad. For each dollar, one gets e1
units of foreign currency. Per unit of foreign currency, one will get 1 + i∗ next
period. Converting this back into U.S. currency, at rate e2 , one’s total return
∗
is (1+ie2 )e1 . Taking logs, and using the approximation that log(1 + x) = x for
small x, we have that: i = i∗ − e2e−e2
1
, which implies the result given above.
For now, let’s look at the completely static case, and assume that E[ eė ] = 0.
This implies that the nominal interest rate is completely determined by the
foreign nominal interest rate, a small open-economy assumption. This case
was first analyzed by Mundell and Fleming.6 Then our two relationships are
(assuming fixed prices, again):

eP
Y = C(Y − T ) + I(r) + G + N X( ) (2.10)
P∗

M
= L(i∗ , Y ) (2.11)
P
Here, the exogenous variables are G,T ,M , P , P ∗ and i∗ and the endogenous
Y and e. Plot the two relationships with Y on the horizontal and e on the
vertical axes. The IS relationship is a downward-sloping relationship and the
LM relationship is a vertical line (Figure 2.5).
We can consider the effects of policies and shocks in this model under two
cases: one in which the nominal exchange rate is allowed to vary freely, and one
in which it is held fixed. Nominal exchange rates are fixed through government
intervention in currency markets. If the government wishes to cause an appre-
ciation of its currency, it sells foreign currency from its reserves to buy back its
own currency. If it wishes to cause a depreciation, it sells its own currency to
buy foreign currency.7
In the first case, fiscal policy or other shocks to the IS curve simply leads to
a change in the nominal exchange rate, but has no effect on the level of output.
Intuitively, this occurs because the increase in nominal interest rates which an
increase in government spending would normally cause leads to an appreciation,
which reduces the level of net exports until the increase in aggregate demand is
negated. Monetary policy has effects on output, as usual.
In the case of fixed exchange rates, the opposite is true. There cannot
be independent monetary policy, because the central bank must use monetary
policy to keep the nominal exchange rate at the same level as before. Any
6 Mundell won the 1999 Nobel Prize for this and other contributions.
7 Note the fundamental asymmetry here: since the government prints its own currency,
it will never run out of it. But governments can and often do run out of foreign currency
reserves. This last fact has led to several recent ‘speculative attacks’ on fixed exchange rate
regimes in Southeast Asia.
2.1. OLD KEYNESIAN ECONOMICS: THE NEOCLASSICAL SYNTHESIS33

increase in the money supply must be completely counteracted. Conversely,

increase in fiscal policy are very effective, because the central bank must then
increase monetary policy in order to bring the exchange rate back to its previous
level. Now, let’s turn to the dynamic case, where expected appreciations are
no longer zero. Invert the LM curve, and substitute in for the nominal interest
rate to write:

ė M
E[ ] = i∗ − i( , Y ) (2.12)
e P
Let’s also assume that output adjusts sluggishly, so that

eP
Ẏ = φ[A(Y, )−Y] (2.13)
P∗
∂A
where A denotes absorption, the level of domestic demand, and ∂Y < 1.8
This gives us a dynamic system in two variables. We can solve for the
two steady-state conditions (namely, Ẏ = 0 and ė = 0) and plot them as
follows:((Figure 2.6).
Using this, we can then determine the dynamics of the system, and find that,
as usual, it is saddle-path stable, with a unique stable arm (Figure 2.7).
Consider the following policy experiment: There is an increase in the money
supply. This shifts the ė = 0 schedule to the right. The new stable arm and new
system are as in the diagram below (Figure 2.8). Output adjusts sluggishly, but
the exchange rate may move instantaneously. Thus, it falls, and then slow rises
over time to its new, lower value.
Intuitively, what has happened is that the increase in the money supply leads
to a lower domestic nominal interest rate. In order for uncovered interest parity,
the condition that rates of return must be equalized, to hold, it must be the
case that people expect an appreciation. Thus the exchange rate must initially
drop by more than it eventually will to cause this appreciation.
This result is known as “overshooting,” and was first derived by Dornbusch
(JPE, 1974). It is a more general feature of rational-expectations models, that
variables must initially move more than they will in the end.
The foregoing was a somewhat simplistic introduction to the subject of in-
ternational finance. For a much more detailed presentation, see Obstfeld and
Rogoff, Foundations of International Macroeconomics.

2.1.2 Aggregate Supply

The details of aggregate demand are fairly uncontroversial. There is signifi-
cantly more controversy regarding aggregate supply. Keynes initially focused
on the case where (short-run) aggregate supply is horizontal, which implies that
movements in aggregate demand have no effect on prices in the short run. This
is somewhat unrealistic, and so it was quickly assumed that aggregate supply
was upward-sloping. While we will see the theoretical reasons for this view
8 One could get very similar results in what follows by assuming that prices adjust sluggishly.
34CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

(which in most cases were not made explicit until much later), it gained con-
siderable empirical support in the 1950s by the discovery of British economist
A.W. Phillips of a seeming long-run negative relationship between inflation and
unemployment. Thus, one could simply write down π = f (u).9
It was later found that the relationship worked better if one wrote πt =
πe + f (U ), where π e is expected inflation. In practice, it was assumed that πte =
A(L)πt−1 , or that expected inflation was a distributed lag of past inflation.10
This expectations formation mechanism is known as adaptive expectations, since
expectations of inflation evolve over time in response to past inflation.
People told several stories which were consistent with this relationship:

1. The sticky wage story: Suppose workers had to set their wage some time
in advance. Suppose they have a target real wage ω. They will then set
their nominal wage W = ωP e . The actual ex post real wage paid after
Pe
P is realized will then be W
P =ω P .
If we assume thate firms are on their labor demand curves, then LD =
LD ( W D P
P ) = L (ω P ). This implies that if prices are un expectedly high,
the actual real wage paid will be low, labor demand will be high, and
therefore output will be high.
2. Money illusion or worker misperception story: Suppose workers are igno-
rant of the price level, but firms are not. Then while labor demand will
be based on the actual real wage, labor supply will only be based on the
expected real wage, so that LD = LD ( W S S W P
P ), but L = L ( P P e ). Here, if
e
P > P , labor supply will increase, the amount of labor hired will go up,
and output will go up. Workers are in effect tricked into working more
because they confuse nominal wage increases with real wage increases.
3. Sticky prices: Assume some firms are able to adjust prices completely,
but some firms cannot change their prices, but must set them in advanced
based on expectations of the price level. Then clearly the AS curve will
slope upwards, where the slope depends on how many firms are able to
change prices.

All three stories imply that unexpected inflation will be associated with a
higher level of output. The first two stories (the first of which dates back to
Keynes himself) have the implication that the real wage is countercyclical (that
is, covaries negatively with the deviations of output from trend). This was found
by Tarshis and Dunlop to be untrue in the 1930s shortly after the publication of
the General Theory, and is still untrue today. Keynesians have tended to rely
more on the last theory as a result.11
9 By Okun’s law, which inversely relates output growth and changes in unemployment, one

can translate this relationship into a relationship between output and prices.
10 A(L) = A + A L + A L2 + . . ., where Lx = x
0 1 2 t t−1 , so that A(L)πt−1 = A0 πt−1 +
A1 πt−2 + A2 πt−3 + . . ..
11 Although more recently, Rotemberg and Woodford have argued that the markup of price

to marginal cost may vary countercyclically, which would allow for nominal wage inflexibility
and for real wage acyclicality.
2.1. OLD KEYNESIAN ECONOMICS: THE NEOCLASSICAL SYNTHESIS35

Even given the presence of several theories and strong empirical support,
several economists were uncertain about the formulation of aggregate supply
given above. Milton Friedman, in his 1968 AEA presidential address, expressed
doubts that the formulation of inflationary expectations as a distributed lag
was valid. In particular, since the formulation was purely backwards-looking,
it implied that the Federal Reserve could consistently apply expansionary mon-
etary policy to increase output, because P − P e could be permanently made
greater than zero. Friedman instead hypothesized that people had forward
looking-expectations, and that they would response to changes in Federal Re-
serve policy. Hence, one could expect the Phillips Curve relationship to break
down.
As the graph below (Figure 2.9) shows, it did shortly after his statement.
This failure of the Phillips curve in fact discredited Keynesian economics in
many peoples’ view.
In fact, it need not have done any such thing. It has turned out that if one
replaces inflationary expectations with a more sophisticated measure, and if one
allows for certain discrete shifts in the aggregate supply curve corresponding to
events such as the raising of oil prices by OPEC, that one again achieve a stable
aggregate supply curve. Thus empirically, it seems that aggregate supply curves
of the following form:
π = π e + f (u) + ² (2.14)
where πe is determined rationally, that is from the economic model into
which this Phillips curve is embedded, and ² is an aggregate supply shock, are
quite successful.
It is not clear that even if the aggregate supply curve were stable, that
one should expect to see the pattern of inflation and unemployment seen in
the data by Phillips. Recall that although theory suggests that the behavior
of the economy is determined by aggregate demand and aggregate supply, in
practice we only observe the intersection of aggregate demand and aggregate
supply. Suppose in principle there are shocks to aggregate supply and aggregate
demand.
Consider the following three cases:
1. Most shocks to the economy are to AD: Then the resulting equilibrium
prices and wages will trace out the aggregate supply curve, as Phillips saw
(Figure 2.10).
2. Most shock are to AS. Then the equilibria will trace out the AD curve, a
downward sloping relationship (Figure 2.11).
3. Shocks are equally distributed. Then we get no relationship between prices
and output (Figure 2.12).
Phillips, who just looked at a simple regression involving wage inflation and
unemployment, was lucky to have obtained the results he did. Over that period
in British history, it must have been the case that most shocks were aggregate
demand shocks, and therefore the results traced out the aggregate supply curve.
36CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

In general, this would not be true, and the observed pattern need not follow
either of the two curves. In general, the problem of estimating a structural
relationship from data generating by equilibrium outcomes by several structural
relationships is known as the identification problem. A solution to this problem
is to find a variable which shifts one of the relationships, but not the other. This
variable, known as an instrument, permits estimation of the relationships.
This solution was not known as of the early 1970s, and it appeared that
Keynesian economics had serious empirical problems. In part because of this,
economists also focused attention on three significant problems with Keynesian
theory:

1. Keynesian models were very ad hoc; they posited aggregate relationships

not derived from microeconomic theory.

2. Their treatment of expectations was quite primitive, and seemed to involve

large elements of irrationality

3. The models also lacked an explicit treatment of dynamics

Thus, while over the next twenty years some economists would invent Real
Business Cycle Theory and write articles with titles such as “The Death of
Keynesian Economics” or “After Keynesian Economics,” others would try to
remedy each of these three deficiencies. We turn to these, in chronological
order, in the next sections.

2.2 Disequilibrium Economics

The first group of economists who attempted to reform Keynesian economics
tried to give it microeconomic foundations. They reasoned that since Keynesian
models obtained most of their results from the assumption of inflexible wages
and/or prices, that one might be able to obtain similar results by imposing
inflexible wages and prices on an otherwise Walrasian model. This turned out
to be true. The resulting models were referred to as Disequilibrium models,
because they had the implication that markets did not always clear. The main
developers of these models were Robert Barro and Herschel Grossman. These
models later became very popular in France, where Edmond Malinvaud and
Jean-Pascal Benassy also worked on them.

2.2.1 Setup
The economy consists of a representative agent, a representative firm, and the
government. Assume the representative agent has the following utility function,
defined over leisure, consumption and real monetary balances:

M β
U = C α( ) − Lγ . (2.15)
P
2.2. DISEQUILIBRIUM ECONOMICS 37

L is labor supply, so that −Lγ represents the disutility of working (which

could alternatively be represented by the utility of leisure). The time endowment
is L̄.
Her budget constraint in real terms is:

M M0 W
C+ = + L+Π (2.16)
P P P
where M 0 is initial nominal money holdings and Π is profit. The production
function is Y = Lη . The firm maximizes profit, Π = Y − WP L.
The government is completely passive, and simply supplies money to equal
demand; there will be no seigniorage in equilibrium.
Finally, note that the national income accounting identity for this economy,
without investment or government spending, is Y = C.

2.2.2 The Walrasian Benchmark Case

Suppose that prices and wages are purely flexible. The first order conditions for
the consumer’s maximization problem are:

M β
αC α−1 ( ) =λ (2.17)
P
M β−1
βC α ( ) =λ (2.18)
P
W
γLγ−1 = λ−ν (2.19)
P
where λ and ν are the Lagrange multipliers associated with the budget con-
straint and the time endowment constraint respectively. The time endowment
constraint will not bind, so that ν = 0.
The first two first-order conditions can be combined to yield:

M β
= C (2.20)
P α
We can use the definition of profits to rewrite the budget constraint as:

M M0
C+ = +Y (2.21)
P P
Using the optimality condition for M
P derived above, and the national income
accounting identity, we can then solve for aggregate demand:

αM
Y = (2.22)
β P
We can also use the condition that labor demand equal labor supply to solve
for the level of output. Labor demand comes from profit maximization, which
implies:
38CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

1 W η−1
1
L=( ) (2.23)
η P
For labor supply, one can combine the first order conditions for C and L,
and using the production function and the fact that Y = C, after some tedious
manipulations one can solve for the level of LS as:

γ 1 W 1
L=( ) η(α+β−1)+1−γ ( ) −1+γ−η(α+β−1) (2.24)
α1−β β β P
Equating the two, one can see that the equilibrium level of L will be a
function of the parameters η, α, β and γ, and not of M . Hence equilibrium
supply is constant. Money is neutral, as expected.
One can graph the usual supply and demand curves for the labor and goods
markets as follows (Figure 2.13).

2.2.3 Exogenously Fixed Price

Let us now consider the case where the price level is exogenously fixed at P = P̄ .
There are in principle two cases to consider:

1. P̄ is above the goods market clearing level

2. P̄ is below the goods market clearing level

I shall only consider the first case, for reasons that will become apparent
later in the course.12 Note that in this case, the desired supply of goods is
greater than the level of demand. Let us make the general assumption that
when supply and demand are not equal, the quantity transacted is equal to the
minimum of supply or demand. Then, the firm is rationed on the goods market.
A key result of these models of disequilibrium is that the presence of rationing
in one market may imply alterations in behavior in other markets. In particular,
in this case since the firm is rationed on the goods market, it will never be willing
to hire more labor than is necessary to produce the level of demand implied by
P = P̄ . Thus, labor demand will be vertical at that point. The new labor
demand curve is an example of what has been termed an effective demand
curve, that is a demand curve derived under the presence of a constraint in
another market. This contrasts with the unconstrained, or notional demand
curve (Figure 2.14).
The consumer is not rationed, and therefore her demand remains Y D = α M
β P̄ .
Output is demand determined, as derived above, so for the fixed price level
money is now nonneutral. The aggregate supply curve is horizontal at P = P̄ .
12 Briefly, in this case price is greater than marginal cost, which is a natural outcome of

imperfectly competitive models.

2.2. DISEQUILIBRIUM ECONOMICS 39

2.2.4 Exogenously Fixed Nominal Wage

In this case, the market for goods clears, but the market for labor does not
clear. We again need to in principle consider two cases:
W̄
1. W̄ is such that P is above the labor market clearing level
W̄
2. W̄ is such that P is below the labor market clearing level.

I shall only consider the first case, because it implies that since LD < LS
that there is involuntary unemployment.
In this case, the consumer is rationed on the labor market, since she is not
able to supply as much labor as she would like at the given level of the real wage.
Her constraint on the labor market implies that her demand for goods might be
affected. In particular, the constraint in her maximization problem that L ≤ L̄
is replaced by the constraint that L ≤ LD ( W̄
P ), which will be binding.
The consumer’s three first-order conditions remain the same, but now ν 6= 0.
However, we can see that this fact does not affect our solution for C in terms of
M
P . Hence we may solve for the same expression for aggregate demand as before,
so that effective demand for goods and notional demand for goods coincide
(Figure 2.15). This is a special case, and arises solely from the assumption that
consumption and leisure are separable in utility.
Hence aggregate demand is still:

αM
YD = (2.25)
β P
We may determine aggregate supply from the production function and the
fact that firms are on their labor demand curves, so that
µ ¶ η−1
η

S 1 W̄
Y = . (2.26)
η P
This is upward sloping in P . Money is again non-neutral. In terms of the
graphical analysis, an increase in M leads to a shift outwards in the demand
curve. Were the price level to increase (as would happen in the Walrasian
benchmark case), this would imply a lower real wage, and therefore a greater
degree of labor demanded and therefore employed. So output also increases.

2.2.5 Both prices and wages inflexible

The logical next step is to consider cases in which both the nominal wage and
the price level are fixed. There are four cases to be considered:

• Both the wage and the price level are above the market-clearing level. This
has been called the Keynesian Unemployment case, because it combines
the case of unemployment derived above with the case of real effects of
aggregated demand disturbances.
40CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

• The wage is above, but the price level is below the market-clearing level.
Here, consumers are rationed on the goods market. In principle, their
rationing implies that effective labor supply is different from notional. In
practice, that isn’t the case, for the same reason effective and actual goods
demand coincided above. This is known as the Classical Unemployment
case. Since labor supply exceeds labor demand, there is unemployment,
but changes in aggregate demand are neutral, since firms are not rationed
and are on their Walrasian supply curve.

• The wage is below, but the price level is above the market-clearing level.
This implies that the firm is constrained in both goods and labor markets.
This cannot happen in a static model, so this case is degenerate.

• Both the wage and the price level are below the market-clearing level. In
this case, the consumer is rationed in the goods market, and the producer
in the labor market. This is known as the Repressed Inflation case, since
prices are ‘too low’. One can show that in general, increases in aggregate
demand, because they increase the degree of rationing faced by the con-
sumer, lead to shifts inward in labor supply as the consumer takes more
leisure.

2.2.6 Analysis of this model

These models provide some microeconomic foundations for the effects of nominal
money and government spending on output. Some of these results are Keyne-
sian, some are similar to classical models, and some are neither. The versions
presented above are extremely simplified; the models of Barro and Grossman
(1976) and Malinvaud (1977) extend these models by adding in more than one
kind of consumer, and making the models more dynamic.
Despite the successes in providing microeconomic foundations to Keynesian
models, these models were ultimately abandoned. In particular, there were three
problems with them:

• The results of the model depend on the rationing rules involved. In models
which abandon the representative consumer assumption, changes in the
types of consumers rationed greatly change the end results. Also, changes
from the short-side rationing rule employed here, where it was assumed
that the lesser of supply and demand was the quantity traded, lead to
large changes in results.

• The model is still static. This can be remedied.

• The model assumes prices and wages are sticky, without explaining why.
In this sense, it is still ad hoc.
We shall see later that the reasons why prices and wages are sticky may not
make the effects of money and government spending on output any different
from what is implied in these models, and that those models also implicitly
2.3. IMPERFECT INFORMATION MODELS 41

involve rationing. Before we come to these models, we will first look at another
class of models which were developed at the same time as the disequilibrium
models, but took a different route to fixing old Keynesian models.13

2.3 Imperfect Information Models

These models assume that prices and wages are flexible, but allow mispercep-
tions about changes in relative prices versus changes in the price level to affect
real economic decisions. Their innovation over the Keynesian models is in their
careful treatment of expectations. Recall that the older Keynesian models para-
meterized expectations in an ad hoc fashion. The insight of Robert Lucas, who
wrote the first of the imperfect information models, was to assume that people
optimize over expectations as well as over other things in the models. This
implies that agents’ expectations should be identical to mathematical expecta-
tions, in particular the mathematical expectations one would get from solving
out the model.14
The model consists of a large number of competitive markets. As an analogy,
they may be thought of as farmers living on different islands. There are two
shocks in this economy: one to the aggregate stock of money, and one a sector-
specific price shock. The farmers do not observe these shocks separately; rather,
they only observe the individual price of their good. Thus, the aggregate price
level is unobserved. Given the observation of a change in the price of their good,
the farmers would like to change their output only in the case where there has
been a change in the relative demand for their good, that is in the sector-specific
price shock case. If there has been a change in the aggregate stock of money,
they do not want to change output.
Assume that the supply curve for market i is:

yit = b(pit − Et pt ) (2.27)

where the Et denotes the expectation taken conditional on the information set
at time t, which is assumed to include all variables through time t − 1 and pit .
This supply curve is rather ad hoc, but it has the justification that the firm
cares about its relative real price.
Aggregate demand is as usual assumed from the quantity theory to be:

yt = mt − pt , (2.28)

where log velocity has been normalized to zero.

Now, we need to think about Et pt and how it is computed.
The consumer knows the past series of pt . From this, one can compute a
ˆ ≡ Et−1 pt and σˆ2 . We will assume that
sample expected value and variance, Ep P
13 I have two recent papers which look at the relationship between these models and New

Keynesian models, and on the conditions needed to derive microfoundations for AD and AS
separately.
14 This idea had first been though of by John Muth, in the early 1960s, but it was not until

the 1970s that it was widely applied.

42CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

all disturbances will be normally distributed, so that the firms’ prior is that pt
is distributed normally with the above mean and variance.
The firm also knows that since there are only two shocks to the economy,
the price of his or her own good pit = pt + zit , where zit is the unobserved
idiosyncratic shock. This shock averages to zero over the cross-section, and is
also assumed to have zero time-series mean. From past values of this shock, one
can compute a sample variance, σˆz2 .
Given the observed pit , the consumer has to decide how much of the observed
change in that variable comes from a price-level change, and how much comes
from a relative real price change. This essentially reduces to the problem of
finding the mean of the posterior distribution of pt given the prior distribution
assumed above. One can show (as in Casella and Berger, Ex. 7.2.10) that:
ˆ
Et pt = θpit + (1 − θ)Ep (2.29)

where
σˆp2
θ= (2.30)
σˆp2 + σˆz2
In words, the expected change in price is partly attributed to a change in
the price level, and partly due to a change in the relative real price.
Note that this implies that the supply curve is:
ˆ
yit = b(1 − θ)(pit − Ep) (2.31)

If the variance of the price level (i.e. the variance of monetary shocks) were
extremely high relative to the variance of idiosyncratic shocks, this means that
θ would be close to one, and output would not vary. In other words, since most
shocks were aggregate demand shocks, there would be no reason to supply more
output, because it’s very likely that any given observe change in prices is due to
changes in the money stock and not changes in relative demand for the good.
The opposite happens if there is a high variance to the idiosyncratic shocks.
If we aggregate over consumers by summing over the i (which, given the log-
linearity assumed here, means that we are using the geometric mean of output,
rather than the average), we get that
ˆ
yt = β(pt − Ep) (2.32)

where β = b(1 − θ).

ˆ and Pt , we get:
Combining this with aggregate demand, and solving for Ep
β
yt = (mt − Et mt ) (2.33)
1+β
This has a very striking implication: only unanticipated shocks to the money
supply have an effect on output. Any shocks which were anticipated will be
correctly perceived to have resulted in changes to the price level, and no one
will change his or her supply of goods.
Additional implications:
2.4. NEW KEYNESIAN MODELS 43

• Minimizing the variance of output implies setting the variance of monetary

shocks equal to zero- in other words, following a strict money
supply rule.
³ 2 ´
σz
• The slope of the Phillips curve, β = b σ2 +σ 2 , will be inversely related
p z
to the variance of aggregate demand shocks. This seems to be true in the
data.

• Shocks are not persistent.

Both despite and because of the model’s striking implications, there are a
number of problems with it. They include:

1. The lack of motivation for the supply curve (which is now known as a
Lucas supply curve). This was later derived by Robert Barro from an
intertemporal substitution argument. It arises rather more naturally in
other contexts.

2. The lack of persistence of shocks and dynamics. This can be readily fixed
by adding capital markets, inventories or through other means.

3. The dependence of the results on the inability to observe the price level.
In fact, observations on the price level are published weekly. Furthermore,
if it really were such a key variable, presumably there would be a large
incentive to gather information on it.

This last problem has turned out to be a fatal flaw, and it is why this model
is no longer regarded as the key reason for why changes in nominal money may
have real effects. However, the idea that the economy is noisy and people must
do signal extraction to get the information they need is a powerful one worth
pursuing.
Reactions to this model took two forms:
One group concluded that in fact, anticipated money doesn’t matter, and
went off to write Real Business Cycle models. They generally assumed that any
model with rational expectations would have this implication.
Another group concluded that in fact, it could, and went off to write models
which had rational expectations but also had real effects of money on output.
We turn to them next.

2.4 New Keynesian Models

2.4.1 Contracting Models
The next set of models to be developed also assumed that prices or wages were
exogenously sticky, as did the disequilibrium models. These models, however,
incorporate dynamics and expectations more thoroughly. In principle, there are
44CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

two ways in which prices and wage may be set: they may be set in a time-
dependent manner, in which prices or wages are changed at fixed time intervals,
or in a state-dependent manner, in which they are changed depending on the
value of some state-variable. We will consider only time-dependent changes
here, not state-dependent ones (which we will return to later).
We will first consider a model in which the level of the price or wage is
determined in advance, but not constrained to be equal over all periods the
contract is in place. We will next consider cases in which the level of the wage
is fixed over all periods of the contract. We will consider models in which the
wage is set rather than the price, to conform with the original statement of the
models and to provide a contrast with the textbook treatments.

2.4.2 Predetermined Wages

Assume the production function Yt = L1−γ t . The profit-maximizing condition
that the marginal product of labor equal the real wage for this production
function is then:
Wt
(1 − γ)L−γ
t = (2.34)
Pt
After taking logs, this implies:
1 1
lt = ln(1 − γ) − (wt − pt ) (2.35)
γ γ
1
Define α = γ ln(1 − γ) and β = γ1 . Then we can show that

β−1
yt = α − (β − 1)(wt − pt ) (2.36)
β
Below, even though we will have fixed wages, we will assume that firms are
always on their labor demand curves. In the context of the previous models,
this implies that the real wage is assumed to be above the market-clearing level.
Assume that aggregate demand follows the usual quantity-theoretic formu-
lation:

yt = mt − pt + vt (2.37)
Finally, we will need to specify how wages are set. Let’s consider two cases:
one in which wages are set one period in advance, one in which they are set two
periods in advance.

One-period-ahead wages
In this case, we shall assume that wt = Et−1 pt , so that the expected (log) real
wage is set to zero (a convenient normalization). Note that this implies that we
can rewrite output as:
β−1
yt = α − (β − 1)(Et−1 pt − pt ) (2.38)
β
2.4. NEW KEYNESIAN MODELS 45

But this is exactly the same as the Lucas Supply curve, derived above. Again,
unexpected inflation will be associated with higher output than usual, but only
for one period. Only unanticipated money will matter, and then only for one
period.
This is a nice result, because it confirms the results of the Lucas model,
but in a somewhat more natural setting, where the appropriate supply curve
is derived from more basic assumptions. It still has the unattractive aspect
of implying that only unanticipated money matters (which has at best weak
support in the data). We shall now review the implications of assuming that
wages must be set two periods in advance.

Two-period-ahead wages
Now assume that wages must be set two periods in advance; i.e. wages are set
at time t for periods t + 1 and t + 2. We will also assume that not all wages
are set at the same time: half of the wages will be set every other period. This
implies that the wage at period t is equal to:
1
wt = (Et−1 pt + Et−2 pt ), (2.39)
2
because half of the economy’s wages were set at time t − 1 and half at time
t − 2.
By substituting this into the supply expression for output and equating it
to aggregate demand, one can show that the system reduces to:
1−β β−1 1
pt = α+ (Et−1 pt + Et−2 pt ) + (mt + vt ) (2.40)
β2 2β β
This is an expectational difference equation. As we did in the section on
money and prices, let’s try to solve it by recursive substitution- that is, by
taking expectations as of times t − 1 and t − 2, and plugging the results back
into the initial equation. Solving for the two expectations yields:

1−β β−1 1
Et−1 pt = 2
α+ (Et−1 pt + Et−2 pt ) + Et−1 (mt + vt ) (2.41)
β 2β β
and
1−β β−1 1
Et−2 pt = 2
α+ (Et−2 pt ) + Et−2 (mt + vt ) (2.42)
β β β
where the second result has used the law of iterated expectations. We can
see from this that the second expression can be solved for Et−2 pt , and then
plugged into the first to solve for Et−1 pt . The results are:
1−β
Et−2 pt = α + Et−2 (mt + vt ) (2.43)
β
and:
46CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

1−β β−1 2
Et−1 pt = α+ Et−2 (mt + vt ) + Et−1 (mt + vt ) (2.44)
β 1+β 1+β

Substituting, this implies:

1−β β−1 β−1 1

pt = α+ Et−2 (mt +vt )+ Et−1 (mt +vt )+ (mt +vt ) (2.45)
β 1+β β(1 + β) β

This expression gives the price level in terms of levels and expectations of
exogenous variables. It is as far as we may go in solving for it, without specifying
stochastic processes for mt and vt .
We can plug this back into aggregate demand, and show that output is the
following:

β−1 β−1 β
yt = (α+(mt +vt )− Et−2 (mt +vt )− Et−1 (mt +vt )). (2.46)
β 1+β β(1 + β)

Assume for simplicity that for all s, Et−s vt = 0, so that velocity is unpre-
dictable.
We can then rewrite the answer as:

(β − 1)α (β − 1) β−1 β−1

yt = + (mt − Et−1 mt ) + (Et−1 mt − Et−2 mt ) + vt
β β 1+β β
(2.47)
Note that the first term here is like that of the Lucas model, again. But
now we have the second term, which states that a fraction β−1 1+β of changes in m
which become known at time t − 1 pass into output. In other words, expected
money matters, and is passed into output.
Clearly, this arises from the fact that not all wages are set each period. Peo-
ple setting wages later have an opportunity to respond to shocks. In particular,
people setting wages at period t − 1 can respond to changes in the money stock
which have been revealed between t − 2 and t − 1. But people who have set
wages in period t − 2 cannot, giving the monetary authority a chance to act on
the basis of predetermined variables.
Note that this fraction is not equal to 12 , as one might expect simply from the
fact that half of the wages are set each period. This is because β is a measure
of how wages respond to labor supply. If β increases, real wages respond less
to changes in labor supply, and the effects of money are greater, because firms
will respond less to the change in money.
Another clear implication of this result is that monetary policy may now
be used to offset shocks and stabilize output. This result thus helped estab-
lish the proposition that monetary policy need not be irrelevant under rational
expectations.
2.4. NEW KEYNESIAN MODELS 47

2.4.3 Fixed Wages

The next model we will consider assumes that wages are not only set in advance
for two periods, but that they are in fact fixed (i.e. they must be the same over
two periods). Here, the wage will be set at time t for periods t and t + 1. The
model is due to John Taylor.
As before, assume that aggregate demand is determined by:

yt = mt − pt (2.48)

(where velocity is eliminated for simplicity).

Also assume that:
1
pt = (wt + wt−1 ) (2.49)
2
This may be justified by looking at it as a markup equation. If production
is CRS, marginal cost will be the wage. The wage at t is the average of the
two contracts in effect at t; that set at t, and that set at t − 1. The preceding
equation then simply says that the markup of price over marginal cost is unity.
Assume wages are set in the following manner:
1 a
wt = (wt−1 + Et wt+1 ) + (yt + Et yt+1 ) (2.50)
2 2
The first term may be thought of as reflecting a desire by workers to have
their wages be equal to those of the other workers in the economy. Over the
two periods that this worker’s contract is in effect, the other workers will be
receiving wt−1 and Et wt+1 respectively.
The second term reflects the fact that excess demand puts upward pressure
on nominal wages 15
One can first substitute in the expression for prices into aggregate demand,
and then substitute the aggregate demand expression into the wage-setting equa-
tion. This will yield the following expectational difference equation in wages and
the money stock:
2−a a
wt = (wt−1 + Et wt+1 ) + (mt + Et mt+1 ) (2.51)
2(2 + a) 2+a
2−a
Define A = 2(2+a) . Then the expression for wt is:

1 − 2A
wt = A(wt−1 + Et wt+1 ) + (mt + Et mt+1 ) (2.52)
2
Recursive substitution will not be helpful in solving this problem (try it
and see- the fact that this is a second-order difference equation makes it more
difficult). We must thus turn to other methods. There are two commonly used
methods:
15 We can think of this coming from a labor supply effect, where higher output requires more

labor supply through the production function. To induce people to supply more labor, wages
must be raised.
48CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

• Undetermined Coefficients In this method, one makes an educated guess at

the form of the solution. One then plugs the solution into the difference
equation, and tries to match up the coefficients of one’s guess with the
coefficients implied by the substitution.

• Lag Operators, or Factorization In this method, we use lag operators to

rewrite the solution as a set of lag polynomials in the levels of the endoge-
nous and exogenous variables, then solve out for the endogenous variables.

Here, we will use the second solution method. For an example of the first,
since Romer, section 6.8.
Note that we can write wt−1 = Lwt , and LEt wt+1 = Et wt = wt , which
implies that Et wt+1 = L−1 wt . (We could define another operator, B, which
lags the information set over which the condition expectation is taken as well
as the variable. This is not needed to solve this problem, however).
Given this, and letting I denote the identity operator, we can factor the
expression above as follows:

1 − 2A
(I − AL − AL−1 )wt = (I + L−1 )mt (2.53)
2
The first parenthetical term is simply a polynomial in L, which we may factor
into:

A 1 − 2A
(I − λL−1 )(I − λL) wt = (I + L−1 )mt (2.54)
λ 2
where

1 + (1 − 4A2 ).5
λ= , (2.55)
2A
which is less than one. In order to solve this, we would like to get rid of the
Et wt+1 term. We see that we may do this by using the fact that:

1
= I + λL−1 + (λL−1 )2 + (λL−1 )3 + . . . (2.56)
I − λL−1
and multiplying through on both sides. If we do so, we will see that we get the
following expression:

λ 1 − 2A ¡ ¢
wt = λwt−1 + mt + (1 + λ)(Et mt+1 + λEt mt + 2 + λ2 Et mt+3 + . . .)
A 2
(2.57)
In words, the wage depends on the lagged wage on the expected future path
of the money stock.
We could plug this into the expression for the price level and then into
aggregate demand to find output. I will do so, but first I want to make a
simplifying assumption: namely, that mt follows a random walk. Thus, let
2.4. NEW KEYNESIAN MODELS 49

mt = mt−1 + xt , where xt is serially uncorrelated. Then, Et mt+i = mt . If we

do this, and also set the supply shock ut equal to zero, we get:

1+λ
yt = λyt−1 + xt (2.58)
2

(where the algebra may be obtained by looking at Romer, pp. 290-291).

This model has some strong implications. In particular, it implies that since
λ is a positive number, shocks to aggregate demand (here, shocks to the money
supply) will have long-lasting effects, even after all contracts set before the
shock have expired.
To see this, note that the effects in period zero of a unit shock to aggregate
demand will be 1+λ 1+λ
2 . In period one, they will be λ 2 . In period two,λ
2 1+λ
2 ,
and so on. (Note that technically, this may be obtained by multiplying through
1
both sides by 1−λL and solving for y. The resulting expression is known as the
Moving Average Representation or MAR, for y, and the tracing out of an effect
of a unit shock is known as the impulse response function (Figure 2.16). You
will see both of these concepts in time series econometrics).
This occurs because λ, which remember is ultimately a function of a, is
a measure of the inertia in nominal wages. It parameterizes to what degree
nominal wages will adjust to shocks, and to what degree they will simply remain
at whatever level they were last period. Thus, multi-period contracts with
staggering lead to long-lasting effects of money on output, even under rational
expectations. As a final note, I could have derived the same results assume price
contracts rather than wage-contracts; see Romer or Blanchard and Fischer for
examples.
These models, although in some ways similar to the disequilibrium models,
add to them dynamics and rational expectations. They do use a simplistic
version of aggregate demand, which may lead to some of the effects of the
Barro-Grossman model to not be present (they essentially do not consider one
of the three regimes). They also share Barro and Grossman’s shortcoming, in
that they assume wages inflexible without explaining why. Fischer has argued
that as a positive fact, we do see wages set in the way of his or Taylor’s model.
Robert Barro and others have argued that wages may seem sticky because they
represent payments in a long-term contract between firms and workers, who
want stable wages.16 Several papers have explored the implications of indexing
in these models.
Because of this last deficiency, however, researchers in the 1980s and 1990s
have moved to models in which prices are inflexible because there are small
costs of changing prices. They have modeled this in the context of imperfect
competition. It is this we now turn to.

16 We will return to this issue in the chapter on Unemployment and Coordination Failure.
50CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

2.5 Imperfect Competition and New Keynesian

Economics
The next set of models looked at circumstances in which firms set prices, but
occasionally chose not to change prices because of costs of doing so. Since
we need to think about cases where firms set prices, rather than take them
as given, we must first consider the macroeconomic implications of imperfect
competition. I will first do this in the context of a model without costs of
changing prices, to show that imperfect competition per se can yield Keynesian
effects of government spending on output.

2.5.1 Macroeconomic Effects of Imperfect Competition

The following model is after one by Mankiw (1986). It has the same simple
flavor as the Disequilibrium model presented above.

Consumers
Suppose utility for consumers is:

U = αlogC + (1 − α)logL, (2.59)

where L now denotes leisure, not labor. Suppose L0 is the consumer’s time
endowment.
Let leisure be the numeraire good, so that the wage is unity. Then we can
write total income as (L0 − L) + Π − T . Here, Π represents profits by firms and
T is a lump-sum tax.
If we let P be the relative price of the consumption good, we may write the
budget constraint as:
P C = L0 − L + Π − T (2.60)
Cobb-Douglas utility implies that the solution to the consumer’s maximiza-
tion problem is:
P C = α(L0 + Π − T ) (2.61)
We may think of α as the MPC.

Government
The government imposes taxes T , and uses it to purchase goods, G, or to hire
government workers, LG .17
Total expenditure on the consumption good in this economy is then Y =
P C + G, or substituting, is equal to:

Y = α(L0 + Π − T ) + G (2.62)
17 We need the latter condition, given the absence of money and the static nature of the

model, to allow G − T to be nonzero.

2.5. IMPERFECT COMPETITION AND NEW KEYNESIAN ECONOMICS51

Firms
Assume that N firms produce the single consumption good. Total expenditure
is taken as given. If we let Q denote output, then Q = YP .
Assume that firms have an IRS production function. They must pay a fixed
cost F in labor, and after that, face a constant marginal cost of c units of labor
per good. Total cost is then:

T C = F + cQi (2.63)

Now assume that the market is imperfectly competitive. The firms play
some game to jointly determine the markup of price over marginal cost. Denote
this markup by µ, so that
P −c
µ= (2.64)
P
Note that if the firms are playing a Cournot game, µ = N1 . If they are
playing Bertrand, i.e. getting close to the perfectly competitive level, µ = 0. If
there is perfect collusion, µ = 1. Given this, we can rewrite Q = 1−µ c Y , and
profits as:
Π = µY − N F (2.65)

Equilibrium and Comparative Statics

Our two equilibrium conditions are:

Y = α(L0 + Π − T ) + G (2.66)

Π = µY − N F (2.67)
In words, expenditure depends on profits, and profits on expenditure. From
this, one can show that dY −α dY
dT = 1−αµ , and, more importantly, that dG = 1−αµ ,
1

which is clearly greater than one for nonzero µ. Thus we have a Keynesian
multiplier.
This result comes about for the same reasons the Keynesian multiplier ex-
isted in the old Keynesian model. Namely, if G increases, this leads directly to
higher expenditure, which, given that profits are increasing in expenditure due
to the imperfect competition, leads to higher profits and higher income, which
leads to higher expenditure, and so on. With perfect competition, µ = Π = 018

2.5.2 Imperfect competition and costs of changing prices

In this section, we will look at a models of imperfect competition with costs to
changing prices. We will start with the simple case of a single firm in a static,
one-firm model. We will then proceed to a static, multi-firm GE model, and
conclude with a dynamic GE model.
18 Although since there are fixed costs and increasing returns, it’s not clear that an equilib-

rium exists in the perfectly competitive case.

52CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

Static, One Firm

The static model presented here is due to Mankiw (QJE, 1985). Suppose there
is a single firm, a monopoly. Suppose its nominal cost function is C = kqM ,
where q is the number of units produced, and M is the nominal stock of money
(if the nominal wage is proportional to the nominal stock of money, we could
get this from a CRS production function).
Suppose the inverse demand curve is P = f (q)M ; this arises naturally by,
for example, assuming a quantity-theoretic version of aggregate demand where
q=VM P , which would then imply that P = q M .
V

Since the firm is a monopoly, it produces where M R = M C. M C = kM ,

and M R = (qf 0 + f (q))M . In other words, both M R and M C are linear in
M , so the level of production is independent of M . In the absence of costs of
changing prices, increases in M will just shift up the demand, marginal cost
and marginal revenue curves proportionately, which will mean that the nominal
price charged will be higher. Note that, as usual, the price charged (which we
will denote pm ) is greater than the competitive price, and the quantity is less
than the competitive quantity (Figure 2.17).
C P
To simplify things, let c = M and p = M . We will measure welfare as the
sum of consumer surplus and producer surplus (we haven’t specified utility, so
just assume a constant marginal utility of wealth). The welfare maximizing
point is at the perfectly competitive point.
Now suppose that the firm has costs of price adjust equal to z. Suppose
further that it needs to set its nominal price one period ahead, based on expec-
tations of what the future money stock is going to be.e It will set its nominal
price to be pm M e . The actual price is then p0 = pm M M . The firm can choose
to pay z to reset the actual price to pm after the realization of the monetary
shock.
Suppose that money is unexpectedly low. Then the actual price will be
higher than the monopoly price. Profits will decline, as will the sum of producer
and consumer surplus. In terms of the diagram below (Figure 2.18, same as in
Mankiw article), the reduction in profits is B − A and the reduction in surplus
is B + C, which is greater than B − A.
Note that the firm will only choose to change its price if B − A > z. But
the socially optimal condition for changing price is B + C > z. Hence, if
B + C > z > B − A, there will be cases in which it is in fact socially optimal for
the firm to change its price, but not privately optimal to do so. This is because
there is an external benefit of C + A to changing prices.
A good measure of the externality may be the ratio of the social benefit
to the private benefit, B+C
B−A . Let’s try to figure out how big this is. Note
that B − A is equal to the lost profit, or Π(p0 ) − Π(pm ) Recall that pm is the
profit-maximizing price for the monopolist. Hence, at this price, Π0 (pM ) = 0. A
second order Taylor series expansion of profits around this point yields, Π(p0 ) ≈
Π(pm ) +Π0 (pm )(p0 − pm ) + 12 Π00 (pm )(p0 −pm )2 , which would imply that the loss
in profits Π(p0 ) − Π(pm ) ≈ 12 Π00 (pm )(p0 − pm )2 , given the optimality condition
above. This loss is only second order in the difference between the actual and
2.5. IMPERFECT COMPETITION AND NEW KEYNESIAN ECONOMICS53

ideal price p0 − pm .
In contrast, the numerator is first-order in p0 − pm , since social welfare is
maximized at marginal cost k. Hence the social benefit may far exceed the
private benefit. Thus, very small menu costs may lead to potentially large
welfare effects (Figure 2.19). One can show that expansions in aggregate demand
may lead to welfare increases, given sticky prices, because output is now closer
to its socially optimal level.
Another way to derive this model is not to have costs of changing prices
per se, but to assume that firms do not always optimize due to small costs
of optimization. This approach was taken by Akerlof and Yellen, also on the
reading list.

Static Multi-firm
The next model expands the previous model to the case of imperfect competition
with many firms under general equilibrium. It is a version of the model of
Blanchard and Kiyotaki, and is adapted from the versions presented in Romer
and in Blanchard and Fischer.
We will now assume that there are N producer-consumers. Each producer
produces a differentiated good. The production function for each producer-
consumer is Yi = Li .
We will assume that each consumer has a utility function
µ ¶g à Mi
!1−g
Ci P 1 γ
Ui = − L (2.68)
g 1−g γ i
where g is a constant, Mi is individual money holdings and Ci is the following
index over all the consumption goods:

XN η−1
1 η
Ci = N 1−η ( Cjiη ) η−1 (2.69)
j=1

and the price index is defined over the prices of the individual goods to be:
N
1 X 1−η 1−η1
P =( Pi ) (2.70)
N i=1

This utility function is known as a Dixit-Stiglitz utility function. The basic

idea is that the goods are differentiated products and are potentially close sub-
stitutes to one another. Hence firms will have some monopoly power over their
goods, but not total, since if firms try to raise prices too high consumers will
simply substitute towards other goods.
You may see this again in Micro.19
The consumer’s nominal budget constraint is:
19 Although perhaps in a different form- the monopolistic competition model often arises in

a model of spatial dispersion.

54CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

X
Pj Cji + Mi = Πi + W Li + M̄i (2.71)
j

where Πi is profits and M̄i is individual initial money holdings.

One can work out that the demand for each good by consumer i is:
Pj −η gIi
Cji = ( ) ( ) (2.72)
P NP
where Ii = πi + wLi + M̄i . This comes from the fact that utility is Cobb-
Douglas in C and M , so that total nominal consumption is a constant fraction
g of income Ii .
It follows that Ci = g IPi
and that M Ii
P = (1 − g) P .
i

This implies that we can rewrite utility as:

Ii 1
Ui = − Lγi (2.73)
P γ
Thus, utility is linear in wealth.20
Now note that we can obtain the total level produced for an individual good
by summing over the demands for all consumers. This implies that:
µ ¶−η
Pj g X Ii
Yj = (2.74)
P N i P
P
If we use the facts that IPi = PPi Yi + M i
and that Y = i Pi
P Yi and define
P P IiP
M̄ = i M̄i , then we can show that i P = Y + M̄ P and that
µ ¶
g M̄
Y = (2.75)
1−g P
Thus, we have a standard aggregate demand equation.
P
We can use our expression for total real wealth, i IPi to substitute into the
demand curves for each good. This will be:
µ ¶−η
Pi Y
Yi = (2.76)
P N
Substitute this back into the utility function, to obtain the following:
µ ¶µ ¶−η
Pi W Pi Y W 1
Ui = − + Li − Li γ (2.77)
P P P N P γ
Now the producer-consumer has two choice variables: Pi and Li . Let’s solve
the model by looking at the first-order condition for each of these.
First, for Pi :
20 Because the model is static, the distinction between income (a flow variable) and wealth

(a stock variable) is blurred.

2.5. IMPERFECT COMPETITION AND NEW KEYNESIAN ECONOMICS55

Pi η W
= (2.78)
P η−1 P
Note that the constant in front of wages is greater than one. Thus, we have
a markup µ of price over marginal cost of greater than one, as expected.
The second condition is:
W
= Lγ−1
i (2.79)
P
To see the full import of the last equation, let’s note that the symmetry of
the model implies that, in equilibrium, all firms will produce the same output
and sell it at the same price. Thus, Pi = P and Yi = Yj for all i and j. This is
an important modeling strategy, and is commonly used. We can then combine
the two first-order conditions to solve for Y and P :
Y η − 1 γ−1
1
=( ) (2.80)
N η
and:

g M̄
P = (2.81)
1−g Y
Let’s note that so far, without any costs of price adjustment, the price level is
proportional to the money stock (since Y is a constant, as solved for above), and
that therefore money is neutral. One can also show that the competitive solution
simply involves maximizing L − γ1 Lγ , which yields a level of output per firm of
one. Therefore, the level of output produced under the monopolistic competition
case is smaller than the equilibrium level of output, as expected. The level of
output small as the elasticity of substitution across goods, η, diminishes, because
the monopoly power of each of the firms increases.
Let us now consider imposing small costs of changing prices. Consider the
effects of a shock to the stock of money. What we notice is that since the firm
is very close to its optimum, but society is far away from its optimum, we will
have the same first-order loss in social welfare of not adjusting versus a second-
order loss for the firm of not adjusting. Thus, the same result will hold as in
the Mankiw menu-cost article.
However, there is an additional factor here which is solely from the multi-
firm aspect. Let’s note that the demand for each good is a function both of
the relative price of the good and of the level of aggregate demand. Let’s think
about a decrease in M .. From the perspective of an individual firm, lowering
its price would raise the demand for its good directly through the relative price
effect. But, given all other prices, lowering the price of its good would also
have a slight effect on the price level. Lowering the price level slightly would
raise the level of aggregate demand, and raise the demand for all goods. The
firm does not take this effect into account in making its decisions. This second
effect on aggregate demand is called the aggregate demand externality. It has
the following implications:
56CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

Suppose that there is a small cost of changing prices, so that it is not in-
dividually optimal for a firm to change prices in response to a change in the
money stock. It may be optimal for all firms to do so, though, due to this ag-
gregate demand externality. We may therefore arrive at a situation of multiple
equilibria, with the following two equilibria:

• All firms adjust prices. They are all better off, even net of the menu cost,
due to the aggregate demand externality

• No firm changes its price. All firms are worse off than they would be under
the first equilibrium, but no individual firm has an incentive to change its
price.

Because all firms are better off under the first equilibrium, that equilibrium
is Pareto superior to the second equilibrium. If all firms were able to coordinate
on an equilibrium, they would choose the first one. They may not get there
because of the coordination problem. This multiplicity of Pareto-rankable equi-
libria is known as coordination failure, and is a more general feature of other
macroeconomic models we will return to later in the course.

2.5.3 Dynamic Models

Let’s think about dynamic versions of these models. In logs, we can write
individual demand functions, which is what these monopolistically competitive
firms care about, as:

yi (t) = m(t) − p(t) − η(pi (t) − p(t)) (2.82)

For simplicity, let η = 1, so that we may write the right-hand-side of the
above as m(t) − pi (t).
Assume the quantity theory R is true, so that y(t) = m(t) − p(t), and let the
symmetric price index p(t) = pi (t)di.
Suppose that firms must pay a fixed menu cost z to change their nominal
price pi (t).
Then, as in the Miller-Orr model of money demand, we will assume the
optimal policy for firms with regard to pricing is an (s, S) policy.
For now, we will assume that the nominal money stock M (t) is monotonically
increasing. Hence the (s, S) policy is a one-sided policy; firms will change their
nominal price when m(t) − pi (t) = S, and will change it so that m(t) −pi (t) = s,
so that the price changes by S − s. Firms in the interior of the interval will not
change their price, so that m(t) − pi (t) will simply increase by the amount that
m increases (Figure 2.20).
Finally, assume that the initially distribution of relative prices is such that
m(t) − pi (t) is distributed uniformly over the interval (s, S].
Now consider a small change in nominal money of size ∆m << (S−s). Given
the menu cost, only firms within ∆m of the top of the distribution will want
∆m
to change prices. From the uniformity assumption, there are S−s of such firms.
2.6. EVIDENCE AND NEW DIRECTIONS 57

Each of these firms changes its price by an amount S −s. Thus, the total change
∆m
in prices is S−s (S − s) = ∆m. Therefore, ∆p = ∆m, so ∆y = ∆m − ∆p = 0.
Money is neutral at the aggregate level.
Neutrality comes at the aggregate level because the distribution of prices
is unchanged, as illustrated by the picture below. All firms individual change
output in response to the monetary shock, but the net aggregate change in
quantities is zero (Figure 2.21). This result breaks down if the initial distribution
is not uniform (although one can show that under some conditions non-uniform
distributions will eventually converge to uniform ones). It also breaks down if
the money shocks are so large that all firms become concentrated at s.
This result, by Caplin and Spulber (1989) is not meant to be realistic, but is
just meant to show that adding dynamics to menu cost models may significantly
alter the traditional results.
Caplin and Leahy (1991) subsequently showed that the results of the model
are reversed if one makes different assumptions about the money supply process.
In particular, assume that m follows a Brownian motion, so that ∆m is equally
likely to be positive as negative. Then the optimal policy under a menu cost is
to follow a two-sided (s, S) rule. Assume that the firm’s objective function is
symmetric in the state variable m(t) − pi (t), so that the s, S rule is symmetric:
the firm will adjust its price so that the state variable equals 0 if m(t)−pi (t) = S
or m(t)−pi (t) = −S. Now assume that the initial distribution of firms is uniform
over an interval of length S somewhere within the interval (−S, S).
In this case, changes in nominal money will not cause any firms to change
their prices as long as the distribution of firms does not hit S or −S. The
distribution of firms will be unchanging. Since prices are inflexible and money
is changing, money is nonneutral. If there is a series of positive or negative
money shocks, the distribution of firms may hit the upper or lower bound, and
the results will revert to the Caplin and Spulber model (Figure 2.22).

2.6 Evidence and New Directions

The most important empirical results they models confirm is the long-standing
body of results which argues that aggregate demand disturbances affect output.
These are too numerous to go into here, but include Friedman and Schwarz’s
Monetary History, referred to above, and much of empirical monetary eco-
nomics21 (surprisingly much less work has been done on the effects of gov-
ernment spending on output). The new theoretical results presented here to a
large degree only reconfirm older results which are already known.
There are some new implications of these models, however. The first to be
tested was simply the idea that the slope of the short-run Phillips curve should
be related to the average level of inflation as well as the variance of inflation. If
inflation is high, firms are changing prices more often, and it is easier to respond
to aggregate demand disturbances by changing prices rather than by changing
21 See the syllabus of my second-year course, Economics 213, for a summary of more recent

work.
58CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

output. Ball, Mankiw and Romer have tested this, and found it to be so. Recall
that Lucas’s imperfect information model implied that the Phillips curve became
steeper. This is also a prediction of the Ball, Mankiw and Romer model. Thus,
Ball Mankiw and Romer have a model which makes the same prediction as the
Lucas model for one relationship, and postulates another relationship (between
inflation and the slope of the Phillips curve) which is true in the data but not
predicted by the Lucas model (Figure 2.23).
This is a very active and large area of research. Among the topics within it:

• The evaluation of monetary policy rules is done within dynamic versions

of these models.

• The real effects of imperfect competition. Julio Rotemberg and Michael

Woodford have written a variety of RBC-like models with imperfect com-
petition, and explored other effects on output different from the ones pre-
sented here

• Dynamic GE models with Keynesian features. Many RBC proponents and

Keynesians are starting to write down dynamic models which have fixed
prices or menu costs. Authors include Patrick Kehoe, Martin Eichenbaum
and Miles Kimball, to name a very few.

• Aggregation issues. Ricardo Caballero, Eduardo Engel and John Leahy

have written papers which think about the interactions between firms
more carefully, and the relationships between microeconomic behavior and
macroeconomic behavior. This is done even for more general Keynesian
models.

2.7 Problems
1. Suppose the economy is described by the following
aggregate demand and supply curves:

yt = mt − pt + vt (2.83)
yt = α(pt − Et−1 pt ) + ²t (2.84)

where vt and ²t are iid disturbances. Suppose that private agents are mis-
informed, and believe that the economy is described by the same aggregate
demand curve as above, but the following aggregate supply curve:

yt = β(pt − Et−1 pt ) + ²t (2.85)

where β 6= α
2.7. PROBLEMS 59

(a) Will this ignorance lead to any real effects of anticipated money under
any of the following monetary policies? (Where relevant, assume
the central bank knows the true aggregate supply curve, but people
believe the other aggregate supply curve).

mt = m̄ + ut (2.86)
mt = ρmt−1 + ut (2.87)
mt = −cyt−1 + ut (2.88)

Here, ut is a random error, and 0 < ρ < 1.

(b) Now suppose agents make a different mistake. They think the supply
function is:

yt = α(pt − Et−2 pt ) + ²t (2.89)

Answer the same question as in (a)

(c) Give a brief intuitive explanation of the differences in your answers
between parts (a) and (b).

2. Suppose we are in an economy in which wages are set in even periods

for each of the following two periods. Thus in period zero, wages are set
for periods 1 and 2; in period 2, for periods 3 and 4, and so on. Wages
are predetermined, not fixed, and thus may differ across the two periods.
Wages are set so that the expected real wage is constant. Expectations
are rational. The production function is: Yt = Lαt , where α < 1. Assume
that aggregate demand is well represented by the quantity theory. There
are no shocks to aggregate supply.

(a) Will anticipated money have real effects in odd periods? In even
periods?
(b) For how many periods will a shock to unanticipated money have real
effects if the shock occurs in an odd period? In an even period?
(c) Suppose velocity is distributed iid N (0, σv2 ). Are real wages acyclical,
countercyclical or procyclical?

3. Consider an economy with a representative firm with a production function

of Yt = At Lβt , where β < 1.
The firm faces a competitive market for its output, Yt and for labor Lt .
The quantity theory of money, Mt Vt = Pt Yt holds. Vt is constant at V̄ ,
and Mt = M̄ for t = T − 1, T − 2, .... Today is time t = T . Labor supply
is given by LSt = ( Wt θ
Pt ) , θ > 0. Your answers to each of the following may
be qualitative.
60CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

(a) Suppose that the wage and price level are perfectly flexible. What
is the effect on output, labor and the real wage if the money stock
doubles from time T − 1 to T (i.e. MT = 2M̄ )?
What is the effect on output, labor and the real wage if the produc-
tivity parameter A doubles from time T −1 to T ? For each of the two
shocks, state whether the real wage and employment are procyclical,
countercyclical or neither.
(b) Now suppose the nominal wage is perfectly flexible, but the price
level Pt = P̄ , which is greater than the level which clears the goods
market. What is the effect on output, labor and the real wage if the
money stock doubles from time T − 1 to T ?
What is the effect on output, labor and the real wage if the produc-
tivity parameter A doubles from time T − 1 to T ? For each case,
assume the price level remains above the market-clearing level after
the shock.
For each of the two shocks, state whether the real wage and labor
are procyclical, countercyclical or neither.
(c) Now suppose the nominal price is perfectly flexible, but the nominal
wage is fixed at a level Wt = W̄ , greater than the level which clears
the labor market. What is the effect on output, labor and the real
wage if the money stock doubles from time T − 1 to T ?
What is the effect on output, labor and the real wage if the produc-
tivity parameter A doubles from time T − 1 to T ?
For each case, assume the wage remains above the market-clearing
level after the shock.
Does it matter whether the doublings are anticipated or unantici-
pated?
For each of the two shocks, state whether the real wage and labor are
procyclical, countercyclical or neither. Account for any differences in
your answer to this part and your answer to the previous part.

4. Consider a firm with the following loss function:

Π(pi , p, m) = −γ(pi − p − α(m − p))2 , (2.90)

where Pi is the firm’s nominal price, p is a price index, m is the money

supply and γ and α are positive constants. All variables are in logs.
Suppose that initially Pi = p = m = 0.

(a) Suppose the firm incurs a fixed cost β each time it alters its nominal
price. Characterize the optimal policy in the face of a once and for all
change in m given that the price index remains constant and that the
firm discounts future losses at rate r. How does this policy depend
on the curvature of the loss function (γ) and the discount rate (r)?
2.7. PROBLEMS 61

(b) Suppose that the economy is made up of many firms just like the one
aboveR and that the log price index is just the average of log prices
(p = pi di). What policy would these firms follow in the face of a
once and for all change in the money stock M if they could agree
to follow identical strategies? Account for any differences in your
answer from the answer given in part (a). Interpret the role of α.

5. Suppose the economy consists of a representative agent with the following

utility function:
¡ ¢1−α
U = Cα M P . Labor is supplied exogenously at level L̄. Output is sup-
plied by a perfectly competitive representative firm, which has production
function Y = Lη . Any profits are remitted to the representative agent.
There is a government, which purchases amount G of goods and services.
This amount is financed by a lump-sum tax, of level T and seigniorage.
Money is the only asset in the economy; initial money holdings are given
by M 0 . Money is supplied exogenously by the government.
Assume both prices and wages are flexible.

(a) Write down the consumer’s and the firm’s intertemporal maximiza-
tion problems.
(b) Solve for expressions for aggregate demand and supply (that is, out-
put as a function of the price level) and for the equilibrium price and
wage levels as a function of exogenous variables.
(c) Describe the effects of increases in initial money holdings M 0 on all
the endogenous variables in this economy.
(d) Describe the effects of increases in government purchases G on all
the endogenous variables in this economy.
Now assume that the nominal wage is permanently fixed at W = W̄ .
In solving the problems below, you may make further convenient
assumptions about the level of W̄ , with justification.
(e) Solve for expressions for aggregate demand and supply (that is, out-
put as a function of the price level).
(f) Solve for the level of unemployment, if applicable.
(g) Describe the effects of increases in initial money holdings M 0 on all
the endogenous variables in this economy.
(h) Would your answers to the previous parts qualitatively change if the
price level were
fixed instead of the nominal wage level?

6. Consider the following dynamic, perfect-foresight version of the IS-LM

model:
62CHAPTER 2. NOMINAL RIGIDITIES AND ECONOMIC FLUCTUATIONS

Ṙ
r =R− (2.91)
ρ
r = αy − βm (2.92)
ẏ = γ(d − y) (2.93)
d = λy − θR + g, (2.94)

where R is the long-term interest rate, r is the short-term interest rate, y

is output, m is real money balances, d is demand, g is fiscal policy. All
parameters are positive. In addition, λ < 1.

(a) Give interpretations for these equations.

(b) Write the model using two variables and two laws of motion. Identify
the state (non-jumping) variable and the costate (jumping) variable.
(c) Draw the phase diagram, including the steady-state conditions, the
implied dynamics, and the saddle-point stable path.
(d) Describe the effects of an immediate permanent decrease in g. What
happens to short-term and long-term interest rates?

7. Consider the following version of the IS-LM model which incorporates the
stock market (after Blanchard, AER, 1981):

q̇ π
+ =r (2.95)
q q
π = α0 + α1 y (2.96)
r = cy − βm (2.97)
ẏ = γ(d − y) (2.98)
d = λy + θq + g, (2.99)

where q is the market value of firms, π is profits, r is the short-term interest

rate, y is output, m is real money balances, d is demand and g is fiscal
policy. All parameters are positive. Assume λ < 1 and that, near the
steady-state value of q (i.e. where both q̇ and ẏ equal zero), cq − α1 < 0.
The first and second equations respectively define an arbitrage condition
between the value of the firm and the real interest rate, and define profits.

(a) Write the model using two variables and two laws of motion. Identify
the state (non-jumping) variable and the costate (jumping) variable.
(b) Draw the phase diagram, including the steady-state conditions, the
implied dynamics, and the saddle-point stable path.
(c) Describe the effects of an immediate small decrease in g on y, q and
r.
2.7. PROBLEMS 63

(d) Describe the effects of an immediate small decrease in m on y, q and

r.
(e) Describe the effects of an announced future small decrease in g on y,
q and r.