Вы находитесь на странице: 1из 65

Contents

Lecture Notes for Macro 2 2001 (first year PhD


course in Stockholm)
Paul Soderlind1

June 2001 (some typos corrected later)

Money Demand (and some Supply)


1.1 Money Supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Overview of Money Demand . . . . . . . . . . . . . . . . . . . . . .
1.3 Money Demand: A General Equilibrium Model with Money in the
Utility Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4 The Mechanics of Money Supply . . . . . . . . . . . . . . . . . . .
The Price of Money
2.1 UIP, Fisher Equation, and the Expectations Hypothesis of the Yield
Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 The Price Level as an Asset Price: Cagans (1956) Model with Rational Expectations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 A Simple Model of Exchange Rate Determination . . . . . . . . . . .

A Derivations of the Pricing Relations


A.1 A Real Bond . . . . . . . . . . . . . . . .
A.2 A Nominal Bond . . . . . . . . . . . . . .
A.3 A Nominal Foreign Bond . . . . . . . . . .
A.4 Real Effects of Money? . . . . . . . . . . .
A.5 Empirical Evidence on the Pricing Relations
3

University of St. Gallen and CEPR. Address: s/bf-HSG, Rosenbergstrasse 52, CH-9000 St.
Gallen, Switzerland. E-mail: Paul.Soderlind@unisg.ch. Document name: MacAll.TeX.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Money and Sticky Prices: A First Look


3.1 Basic Models of the Effects of Monetary Policy Surprises . . . . . . .
3.2 Money and Wage Contracts in an Optimizing Model of the Business
Cycle, by Benassy . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Money and the Business Cycle, by Cooley and Hansen . . . . . . .

4
4
4
7
16
24
24
25
31
38
40
40
41
41
42
45
45
46
56
1

3.4

X Sticky

Wages or Sticky Prices? . . . . . . . . . . . . . . . . . . . .

61

Money in Models of Monopolistic Competition


4.1 Monopolistic Competition . . . . . . . . . . . . . . . . . . . . . . .

63
63

Money and Price Setting


5.1 Dynamic Models of Sticky Prices . . . . . . . . . . . . . . . . . . .
5.2 Aggregation of One-Sided Ss Rule: A Counter-Example to 1M
1Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

69
69

A Summary of Solution Method for Linear RE Models


A.1 Summary . . . . . . . . . . . . . . . . . . . . . . .
A.2 Special Case: Scalar Second Order Equation . . . . .
A.3 An Alternative for the Scalar Second Order Equation:
tion Method . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . .
. . . . . . . . .
The Factoriza. . . . . . . . .

88
88
91
97

Empirical Measures of the Effect of Money on Output


7.1 Some Stylized Facts about Money, Prices, and Exchange Rates
7.2 Early Studies of the Effect of Money on Output . . . . . . . .
7.3 Early Monetarist Studies of the Effect of Money on Output . .
7.4 Unanticipated or Anticipated Money . . . . . . . . . . . . .
7.5 VAR Studies . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.6 Structural Models of Monetary Policy . . . . . . . . . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

124
124
125
125
126
126

84

.
.
.
.
.
.

81
81
82

85

A Derivations of the Aggregate Demand Equation

Money and Prices in RBC Models . . . . . . . . . . . . . . .


Money and Monopolistic Competition . . . . . . . . . . . . .
Sticky Prices . . . . . . . . . . . . . . . . . . . . . . . . . .
Monetary Policy . . . . . . . . . . . . . . . . . . . . . . . . .
Empirical Measures of the Effect of Money on Output . . . . .
The Transmission Mechanism from Monetary Policy to Output

80

B Calvos Model: An Alternative Derivation


Monetary Policy
6.1 The IS-LM Model . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 The Barro-Gordon Model . . . . . . . . . . . . . . . . . . . . . . . .
6.3 Recent Models for Studying Monetary Policy . . . . . . . . . . . . .

0.3
0.4
0.5
0.6
0.7
0.8

104

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

106
106
107
108
115
116
119

Reading List
123
0.1 Money Supply and Demand . . . . . . . . . . . . . . . . . . . . . . 123
0.2 Price Level and Nominal Assets . . . . . . . . . . . . . . . . . . . . 123
2

are used in many different models, for instance as the LM curve is IS-LM models. Mt in
(1.1) is often a money aggregate like M1 or M3. In most of the models on this course, we
will assume that the central bank have control over this aggregate.

Money Demand (and some Supply)

Main references: Romer (1996) (Romer), Blanchard and Fischer (1989) (BF), Obstfeldt
and Rogoff (1996) (OR), and Walsh (1998).

1.1

Money Supply

References: Burda and Wyplosz (1997) 9, OR 8.7.6 and Appendix 8B, and Mishkin
(1997).
The really short version: the central bank can control either some monetary aggregate
or an interest rate or the exchange rate. How they do that is typically not very important
for most macroeconomic questions. Still, this is discussed in Section 1.4. Why they do it,
that is, the monetary policy, is much more importantand something we will discuss at
length later.

1.2

Overview of Money Demand

1.2.1

Money in Macroeconomics

Reference: Goldfeld and Sichel (1990).


Applied money demand equations often take the form
ln

Mt1
Pt
Mt
= b0 + b1 ln Yt + b2 i t + b3 ln
+ b4 ln
+ ut .
Pt
Pt1
Pt1

(1.2)

(1.1)

References: Romer 5.2, BF 4.5, OR 8.3, Burda and Wyplosz (1997) 8.


The standard money demand equation
Mt
= constant + ln Yt i t
Pt

Applied money demand equations

where Mt is nominal money holdings, Pt the price level, Yt some measure of economic
activity, and i t the net nominal interest rate (like 0.07). The inclusion of Mt1 /Pt1 and
Pt /Pt1 is thought to capture partial adjustment effects due to adjustment costs of either
nominal (b4 6 = 0) or real money balances (b4 = 0).
For instance, the estimate for Germany (69:1-85:4) reported by Goldfeld and Sichel
(1990) is {b1 , b2 , b3 , b4 } = {0.3, 0.5, 0.7, 0.7}. (They interest rate used in their estimation is in percentages, that is, like 7 instead of the 0.07 used here, so I have scaled their
b2 = 0.005 by 100.)

Traditional money demand equations

ln

Money Demand and Monetary Policy

There are many different models for why money is used. The common feature of these
models is that they all generate something pretty close to (1.1). But why is this broader
money aggregate related to the monetary base, which the central bank may control? Short
answer: the central bank creates a demand for narrow money by forcing banks to hold it
(reserve requirements) and by prohibiting private substitutes to narrow money (banks are
not allowed to print bills).
The idea behind central bank interventions is to affect the money supply. However,
most central banks use short interest rates as their operating target. In effect, the central
bank has monopoly over supply over narrow money which allows it to set the short interest
rate, since short debt is a very close substitute to cash. In terms of (1.1), the central bank
may set i t , which for a given output and price level determines the money supply as a
residual.
1.2.4

Roles of money: medium of exchange, unit of account, and storage of value (often dominated by other assets).
Money is macro model is typically identified with currency which gives no interest.
The liquidity service of money ( medium of exchange) is emphasized, rather than store of
value or unit of account.
1.2.2

1.2.3

In general, this type of equation worked fine until 1975, overpredicted money demand
during the late 1970s, and underpredicted money demand in the early 1980s. Financial
innovations? (1.2) has been refined in various ways. Various disaggregated money measures have been tried, a wealth of different interest rates and alternative costs have been
used, the income variable has been disaggregated, and fairly free adjustment models have
been tried (error correction models). Single equation estimation of (1.2) presumes that
this is a true demand function, with monetary authorities setting the interest rate, and with
the other right hand side variables being predetermined.

we introduce leisure or credit goods.


Shopping-time models typically have a utility function is terms of consumption and
leisure

X
s U (Ct , 1 lt n t ) ,
(1.5)

1.2.5 Different Ways to Introduce Money in Macro Models

1.3

Reference: OR 8.3 and Walsh (1998) 2.3 and 3.3.


The money in the utility function (MIU) model just postulates that real money balances
enter the utility function, so the consumers optimization problem is
max

{Ct ,Mt }t=0

X
t=0



Mt
t u Ct ,
.
Pt

(1.3)

s=0

where lt is hours worked, and n t hours spent on shopping (supposed to give disutility).
The latter is typically modelled as some function which is increasing in consumption and
decreasing in cash holdings

Money Demand: A General Equilibrium Model with Money in


the Utility Function

Reference: BF 4.5; OR 8.3; Walsh (1998) 2.3; and Lucas (2000)


1.3.1

Model Setup

The consumers optimization problem is

One motivation for having the real balances in the utility function is that
 having cash may
shopping
save time in transactions. The correct utility function would then be u Ct , L L t
,

max

{Ct ,Mt }t=0

shopping

is a decreasing function of Mt /Pt .


where L t
Cash-in-advance constraint (CIA) means that cash is needed to buy (some) goods, for
instance, consumption goods
Pt Ct Mt1 ,
(1.4)

X
t=0



Mt
t u Ct ,
Pt

(1.6)

subject to the real budget constraint


K t+1 +

Mt
Mt1
= (1 + rt ) K t +
+ wt Ct Tt ,
Pt
Pt

(1.7)

where Mt1 was brought over from the end of period t 1. Without uncertainty, this
restriction must hold with equality since cash pays no interest: no one would accumulate
more cash than strictly needed for consumption purposes since there are better investment
opportunities. In stochastic economies, this may no longer be true.
The simple CIA constraint implies that money demand equation does not include
the nominal interest rate. If the utility function depends on consumption only, then all
rates of inflation gives the same steady state utility. This stands in sharp contrast to the
MIU model, where the optimal rate of inflation is minus one times the real interest rate
(to get zero nominal interest rate). However, this is not longer true if the cash-in-advance
constraint applies only to a subset of the arguments in the utility function. For instance, if

where rt is the (net) real interest rate (from investing in t 1 and receiving the return in
t), and wt the real wage rate. Labor supply is normalized to one. The consumer rents his
capital stock to competitive firms in each period. Tt denotes lump sum taxes.
Production is given by a production function with constant returns to scale

Yt = F (K t , L t ) ,

(1.8)

and there is perfect competition in the product and factor markets. The firms rent capital
and labor from the households and make zero profits.
I assume perfect foresight in order to simply the algebra somewhat. It is straightforward to derive the same results in a stochastic setting, at least if we assume that variances

and covariances do not depend on the level of the other variables.

Under perfect foresight, (1.12) can then be written






it
Mt
Mt
/u C Ct ,
,
= u M/P Ct ,
1 + it
Pt
Pt

1.3.2 Optimal Consumption and Money Holdings


Use (1.7) in (1.6) to get the unconstrained problem for the consumer

max
t u (1 + rt ) K t +
{ K t+1 ,Mt }
t=0 t=0

Mt1
Mt Mt
.
+ wt Tt K t+1
,
Pt
Pt Pt

The first order condition for K t+1 is






Mt
Mt+1
u C Ct ,
= (1 + rt+1 ) u C Ct+1 ,
,
Pt
Pt+1

(1.9)

(1.10)

which is the traditional Euler equation for real bonds (with uncertainty we need to take
the expected value of the right hand side, conditional on the information in t). It would
also hold for any other financial asset.
The first order condition for Mt is






Mt
Mt
Mt+1
Pt
u C Ct ,
= u M/P Ct ,
+ u C Ct+1 ,
.
(1.11)
Pt
Pt
Pt+1 Pt+1
If money would not enter the utility function, then this is a special case of (1.10) since the
real gross return on money is Pt /Pt+1 . It is not obvious, however, that we get an interior
solution to money holdings unless money gives direct utility.
The left hand side of (1.11) is the marginal utility lost because some resources are
taken from time t consumption, and the right hand side is the marginal utility gained by
having more cash today and the extra consumption this allows tomorrow (cash provides
utility and is also a form of saving, whose purchasing power depends on the inflation).
Substitute for u C (Ct+1 , Mt+1 /Pt+1 ) from (1.10) in (1.11) and rearrange to get





1
Mt
Pt
Mt
1
= u M/P Ct ,
.
(1.12)
u C Ct ,
Pt
1 + rt+1 Pt+1
Pt
The Fisher equation is
1 + i t = Et (1 + rt+1 )

Pt+1
,
Pt

(1.14)

which highlights that the nominal interest rate is the relative price of the money services
we get by holding money one period instead of consuming it. Note that (1.14) is a relation between real money balances, the nominal interest rate, and an activity level (here
consumption), which is very similar to the LM equation.
Example 1 (Explicit money demand equation from Cobb-Douglas/CRRA.) Let the utility
function be
"   #1


1
Mt 1
Mt
,
=
Ct
u Ct ,
Pt
1
Pt
in which case (1.14) can be written
Mt
1 1 + it
= Ct
,
Pt

it
which is decreasing in i t and increasing in Ct . This is quite similar to the standard
money demand equation (1.1). Take logs and make a first-order Taylor expansion of
ln [(1 + i t ) /i t ] around i ss
ln

1
Mt
= constant + ln Ct
it .
Pt
i ss (1 + i ss )

Compared with the money demand equation (1.1), ln Yt is replaced by ln Ct and =


1/ [i ss (1 + i ss )]. If i ss = 5%, then 20, which appears to be very high compared to
empirical estimates.
Example 2 (Explicit money demand equation from Lucas (2000). Let the utility function
be
" 


1 #1
1
Ct
Mt
=
Ct 1 + B
,
u Ct ,
Pt
1
Mt /Pt

(1.13)

where the convention is that the nominal interest rate is dated t since it is known as of t.
8

1.3.4

in which case (1.14) can be written


it
=
1 + it

2

Two definitions:

Ct
B, or
Mt /Pt
1/2

Mt
it
,
= Ct B 1/2
Pt
1 + it

Neutrality of money: the real equilibrium is independent of the money stock.


Superneutrality of money: the real equilibrium is independent of the money growth
rate.

which can be written (approximately) on the same form as (1.1)


ln

Mt
it
1
.
constant + ln Ct
Pt
2 i ss (1 + i ss )

This gives a value of the in the money demand equation (1.1) which is only half of that
in Example 1.
1.3.3

Steady State

Let the money growth rate be , so Mt /Mt1 = 1 + . A steady state implies


that inflation, consumption, and real money balances are constant: Pt /Pt1 = 1 + ,
Ct /Ct1 = 1, and (Mt /Pt ) / (Mt1 /Pt1 ) = 1. This implies that = .
The first order condition for K , (1.10), can then be simplified as 1 + rss = 1/.
Combining with (1.16) gives that steady state capital stock must solve

General Equilibrium with MIU

We now add a few equations to close the MIU model in a closed economy. The government budget is assumed to be in balance in every period (not restrictive since Ricardian
equivalence holds in this model)
Ms Ms1
Ts =
,
Ps

(1.15)

so the seigniorage (right hand side) is distributed as lump sum transfers (negative taxes).
Note that this is taken as given by each individual agent.
Competitive factor markets, constant returns to scale, and a fixed labor equal to one
(recall that is was not part of the utility function) give
F (K t , 1)
, and
K
F (K t , 1)
wt = F (K t , 1) K t
.
K
rt =

(1.16)

(This follows from that w = F/ L and r = F/ K and that constant returns to scale
implies F = L F/ L + K F/ K .)
In general, the price level is determined jointly with the rest of the dynamic equilibrium. In special cases, as with log utility and complete depreciation of capital (as in the
model of Benassy (1995)) there is a closed form solution. However, in most cases, the
equilibrium must be computed with numerical methods.
10

FK (K ss , 1) = rss
= 1/ 1,

(1.17)

which depends only on the technology and the real discount rate, not on the money stock
or growth. In steady state, the capital stock is constant so Css = F (K ss , 1), so consumption in steady state is uniquely determined by the real side of the economy: in the steady
state of this model, money is neutral and superneutral. Note that this is not true for the
dynamics around the steady state unless marginal utility of consumption is independent
of the real money balances (see Walsh (1998) 2.3 for a textbook treatment).
With a value for steady state consumption, we can solve for the steady state real money
balances, Mss /Pss , by combining the first order




Mss
Mss
i ss
= u M/P Css ,
/u C Css ,
,
(1.18)
1 + i ss
Pss
Pss
and the Fisher equation
1 + i = (1 + rss ) Pt+1 /Pt
= (1 + rss ) (1 + ) .

(1.19)

Using (1.19) in (1.18) gives an equation in Mss /Pss and known model parameters.

11

1.3.5 The Welfare Cost of Inflation


The welfare cost of inflation is typically analyzed for the steady state, since we can then
make use of the superneutrality of money. The growth rate of money, and therefore the
inflation rate and nominal interest rate, can then be changed without affecting the real
equilibrium.
The welfare loss from a higher nominal interest rate is often measured as the extra
consumption needed in order to achieve the same utility as in the case with lower interest
rate. The approach is typically to find the money demand function which expresses real
money balances as a function of the consumption level and the nominal interest rate M
P =
f (i, C), and calculate utility as the value of the period utility function u [C, f (i, C)].


If C = 1 in steady state, a certain interest rate i 0 gives the utility u 1, f i 0 , 1 . The
welfare loss from another nominal interest rate, i 1 , is the value of C which solves
h

i
h

i
u 1, f i 0 , 1 = u C, f i 1 , C .
(1.20)

Fraction of consumption, %

Money in utility function: cost of i>3%


1.5

CobbDouglas, =0.99
Lucas, B=0.0018

0.5

0
3

5
6
7
8
Nominal interest rate, %

10

Figure 1.1: Utility loss, in terms of consumption, of inflation in two MIU models.
To get the same utility as with C = 1 and i = 3% consumption must be

This value of C is the compensation that the consumers need to be as well off with the
interest rate i 1 as with i 0 . Note that C 1 can be interpreted as the percentage change in
consumption needed to compensate for the higher nominal interest rate.
Example 3 (Welfare loss with Cobb-Douglas/CRRA.) From Example 1, the utility at the
nominal interest rate i is
" 
 #1
1 1 + i 1
1
C C
.
u [C, f (i, C)] =
1

i
Consider a steady state where C = 1 and suppose that inflation is zero, so i = r . For
instance, to get the same utility as at C = 1 and i = 3%, u [1, f (0.03, 1)], then consumption must be


1.03/ (1 + i) 1
= C.
0.03/i

1 + B 1/2
1 + B 1/2

i
1+i

1/2


0.03 1/2
1.03

= C.

Figure 1.1 illustrates the result for B = 0.0018 (Lucas point estimate).
Also a cash-in-advance model (see, for instance, Cooley and Hansen (1989)) can generate welfare costs of inflation (more precise: of a non-zero nominal interest rate) if the
cash-in-advance restriction applies only to a subset of the arguments in the utility function. A positive inflation acts like a tax on those goods that must be paid in cash, and
thereby creates a distortion.
Friedmans Rule for Optimal Money Supply
Reference: Romer 9.8, BF 4.5.
Friedman suggested a money rate growth which would set the nominal interest rate to
zero and thereby saturate money demand. The idea is that bills are (virtually) costless to
print and it has a (utility) value for agents, so why not give them as much as they would
possibly would like to have?

Figure 1.1 illustrates the result for = 0.99.


Example 4 (Welfare loss from Lucas (1994).) From Example 2 we get


1/2 !1 1
1
it
1/2

u [C, f (i, C)] =


C 1+ B
.
1
1 + it
12

13

We know that the steady states of the real variables is unaffected by the inflation rate
(see above), so if we concentrate attention to the steady state, then (1.14) tells us that
consumers are satiated with real money balances if u M/P = 0, that is, if i = 0.
By the Fisher equation (1.13) this means that the monetary policy should set (1 + rt ) Pt+1 /Pt =
1, so the rate of deflation should equal the real interest rate. In this way, holding cash gives
the same return as a real bond, so savers will be happy to keep large real money balances
and to get the utility out of it.
In steady state, inflation equals the money growth rate, so a deflation requires a shrinking money supply, which means that seigniorage is negativesee (1.15). In this setting,
this is compensated by lump-sum taxes, which highlights the assumption that the government revenues from the inflation tax is either wasted or can be raised in other, nondistortive, ways. If, instead, a certain revenue must be raised and the alternative taxes are
distortive, then it may no longer be optimal with a zero inflation rate. See Walsh (1998).
If the utility function is separable in consumption and real money balances, then this
result hold in general, not just in steady state.
The Welfare Cost of Inflation - Other Arguments
Reference: Fischer (1996), Romer 9.8, Driffil, Mizon, and Ulph (1990), and Walsh (1998)
4.5-4.6.
1. Inflation raises the effective capital income tax (subsidy), since the nominal return
(loss) is taxed (part of which is just compensation for inflation). The real net of tax
return is
r net = (1 ) i
= (1 ) r ,

(1.21)

where the Fisher relation gives i = r + E and we assume that = E . This


distorts the savings decision. Some calculation for the US (Feldstein, NBER, 1996)
suggest that this effect is large (twice as large as the effect on government revenues).
Counter-argument: a lower inflation and therefore lower government revenues from
capital income taxation is likely to bring higher tax rates.
2. Costs of price adjustments and indexation.
14

3. Some empirical evidence that really high inflation is bad for growth. It is (both
theoretically and empirically) unclear if zero inflation is better for growth than 5%
inflation.
4. Seigniorage is low for most OECD countries (less than one percent of GDP, see OR
8.2).
5. Low inflation means that it will be hard to drive down the real interest really low to
stimulate output. (The nominal interest rates cannot be negative since the nominal
return on cash is zero.).
6. Variable inflation may lead to large inflation surprises which redistribute wealth,
increases uncertainty (affects savings in which way?), and increases the information
costs.
1.3.6

The Relation to Traditional Macro Models

Equation (1.14) is a money demand equation, which in many cases can be approximated
by
ln Mt ln Pt = 1 ln Ct 2 i t ,
(1.22)
which is a traditional LM equation.
When the utility function is separable in consumption and real money balances, then
the optimality condition for consumption (1.10) can often be approximated by
ln Ct = ln (1 + rt+1 ) + ln Ct+1 ,

(1.23)

where consumption growth is related to the real interest rate. From the Fisher equation,
we can replace ln (1 + rt+1 ) by i t Et (ln Pt+1 ln Pt ). This is clearly reminiscent of an
IS equation.
1.3.7

The Price Level

The price level is determined simultaneously with all other variables, and there is typically
no closed form solution.
In the special case where the utility function in (1.6) is separable, so the Euler equation
for consumption (1.10) is unaffected by real money balances, and where money supply
15

is exogenous is might be possible to arrive at an analytical expression for the price level.
In this case, we can solve for the real equilibrium (consumption, real interest rates, etc)
without any reference to money supply. The price level can then be found by solving
(1.11) and information about money supply. This is an example of a classical dichotomy.

a. Interest rule
supply

Example 5 (Solving for the price level.) Use the approximate Fisher equation, i t =Et ln Pt+1
ln Pt + rt+1 , in the approximate money demand equation in Example 1

b. Money rule
supply

1
ln Mt ln Pt = a + ln Ct
(Et ln Pt+1 ln Pt + rt+1 ) ,
i ss (1 + i ss )
and rewrite as
ln Pt

i ss (1 + i ss ) + 1
1
1
= a ln Ct + ln Mt +
rt+1 +
Et ln Pt+1 ,
i ss (1 + i ss )
i ss (1 + i ss )
i ss (1 + i ss )

c. Mixture

supply

which is a forward looking difference equations for ln Pt in terms of the exogenous


variables ln Ct , ln Mt , and rt+1 .
M

1.4

The Mechanics of Money Supply

References: Burda and Wyplosz (1997) 9, OR 8.7.6 and Appendix 8B, and Mishkin
(1997).
The short version: the central bank can control either some monetary aggregate or an
interest rate or the exchange rate. This section is about how they do that, even if this is
not particularly important for most macroeconomic issues. Why they do it, that is, the
monetary policy, is much more importantand something we will return to later.
1.4.1

Operating Procedures of the Central Bank

Suppose demand for the monetary base is decreasing in the nominal interest rate. Suppose
the central bank does no interventions at all. Shifts in the demand curve for money will
then lead to movements in the nominal interest rate. Alternatively, suppose the central
bank announces a discount rate where any bank can lend/borrow unlimited amounts. This
will fix the interest rate and any shifts in the demand curve leads to movements in the
monetary base (as the banks are free to borrow reserves and currency at the fixed rate).
Finally, it is possible to strike a compromise between these two extremes letting banks
16

Figure 1.2: Partial equilibrium on money market


lend at increasing interest rates. This effectively creates an upward sloping supply curve
for the monetary base. This is illustrated in Figure 1.2.
1.4.2

Money Supply and Budget Accounting

Money supply has a direct effect on government finances. Consider the consolidated government sector (here interpreted as treasury plus central bank). The real budget identity
is
Mt1
Bt
Mt
Bt1
Gt +
= Tt +
+
,
(1.24)
(1 + i t1 ) +
Pt
Pt
Pt
Pt
where G t and Tt are real government expenditures revenues, respectively, Bt is nominal
debt, i t the nominal interest rate, Mt is the monetary base (the central bank liabilities),
and Pt the price level.

17

1.4.4

Assume the Fisher equation holds, so the nominal interest rate is


1 + i t1 = Et1 (1 + rt )

Pt
,
Pt1

(1.25)

where rt is the real interest rate. The convention is that the nominal interest rate is dated
t 1 since it is known as of t 1. To simplify, assume rt is known in t 1. We then get
from (1.25) that the real debt in t is
Bt
Bt1 Pt1
Pt
Mt Mt1
= G t Tt + (1 + rt )
Et1

.
Pt
Pt1 Pt
Pt1
Pt

(1.26)

Consider the case where real government expenditures and tax revenues are unaffected
by monetary policy (money is neutral), and where the central bank increases money supply, Mt > Mt1 . This drives down the real value of government debt, Bt /Pt in two
ways. First, the real revenues from money creation (printing), called seigniorage, is
(Mt Mt1 )/Pt . Second, the money supply increase will probably increase the price
level. If this increase is unanticipated, then actual inflation exceeds expected inflation,
Pt1 /Pt Et1 (Pt /Pt1 ) < 1, so the real value of government debt brought over from t 1
decreases.
1.4.3 Money Aggregates and the Balance Sheet of the Central Bank
The liabilities of the central bank are currency (Cu) plus banks reserves (Re) deposited
in the central bank, the assets are the foreign exchange reserve, the holding of domestic
bonds, and perhaps gold
Balance Sheet of Central Bank
Liabilities
Assets
Domestic bonds Currency (Cu)
Foreign currency Reserve deposits (Re)
Foreign bonds
Net worth
Gold

18

Reserve Requirements and Deposits

Money stock, M, is currency, Cu, plus deposits, D, (also called inside money since it is
generated inside the private banking system)
M = Cu + D.

(1.27)

Suppose that private banks (because of reserve requirements or prudence) hold the
fraction r of deposits, D, in reserves, Re. This means that an increase in reserves, 1Re,
allows the bank to increase deposits with the reserve multiplier, 1/r ,
1D =

1Re
.
r

(1.28)

These new deposits may be lent to someone (an thereby bring in profits to the bank,
assuming the lending rate is above the deposit rate). Note that if r goes to zero, then the
central bank cannot control the creation of new deposits by affecting the availability of
reserves.
Reserve requirements mean that a private bank must hold a fraction (usually a few
percentage) of the (checkable) deposits in either cash (in the vault) or as reserves with
the central bank. This fraction is often specified as an average over some period (two
weeks in the US). Suppose a bank needs to get more reserves (maybe depositors withdrew
money during the preceding week). It can then either sell some assets, borrow from other
banks (federal funds market in the US), or borrow from the central bank (at the Feds
discount window in the US). Note that borrowing from the central bank is effectively
a decrease (as long as the loan lasts) in the reserve requirement. Therefore something
has to be done to make the reserve requirements bite in spite of the possibility to borrow
from the central bank. This is typically a penalty rate for these loans, or some kind of
administrative rationing of loans.
It is the fact that r < 1 that makes banks different from other financial institutions.
To see why, suppose r = 1. Then all deposits would have to be kept as reserves and
couldnt be used for lending. Consequently, any lending has to be done from the banks
own capital. In this sense, the bank is not an intermediary any more and cannot create
money.
The money stock deposits discussed above can be interpreted/measured in different
ways. The most common monetary aggregates are: M1 (currency, travellers checks,
19

checkable deposits); M2 (M1 plus small denomination time deposits, savings deposits,
money market funds, repos, Eurodollars); M3 (M2 plus large denomination time deposits,
and money market funds held by institutions)
1.4.5 Interventions and the Money Multiplier
The central bank can typically not control the reserves directly, only the sum of reserves
and currency (the private sector can always convert the currency into reserves and vice
versa). This sum is called the monetary base (B) (also called high-powered money, M0,
central bank money, or outside money since it is generated outside the private banking
system), is
B = Cu + Re.
(1.29)
The monetary base can be increased by, for instance, an open market operation where
the central bank buys government bonds and pays by either cash (increases Cu) or cheque
(increases Re). The same effect is achieved by a foreign exchange intervention; the central
bank buys a foreign asset and pays by either cash or cheque denominated in the domestic
currency.
Suppose private agents wants to hold the fraction c of the money stock in currency.
We can then write (1.29) and (1.27) as
B = cM + r D

(1.30)

M = cM + D.
The money multiplier, M divided by B, can then be written
M
M
=
B
Cu + Re
D/ (1 c)
=
Dc/ (1 c) + r D
1
=
.
c + r (1 c)

1
B.
c + r (1 c)

Example 6 (US data 1994, Burda and Wyplosz (1997) 9) For the US 1994 c = 0.29,
M1/M0 = 2.83, which by (1.32) should imply that r = 0.089. The reserve requirements
on demand deposits was (at least in 1995) virtually zero for small demand deposits..
Explanations: voluntary reserves (prudence or transaction purposes) and leakages
(deposits in nonbank financial institutions).
Example 7 (The Great Depression.) Private sector decisions can lead to important
changes in both c and r . During the Great Depression, B was not changed much (counter
to the conventional wisdom about a contractionary monetary policy), while savers withdrew deposits from the banks and the banks increased the voluntary reserves (both c and
r increased substantially), with the result that M decreased some 30%. Fear of banks
going bust?
Example 8 (Money creation.) The central bank makes an open market purchase of
bonds. This increases the monetary base by 1B. Recall Re = r D and Cu = cM,
so according to (1.30)


c
+ r D.
B=
1c
The increase in the monetary base will therefore be split into increases in reserves and
currency according to

(1.31)

r (1 c)
c + r (1 c)
c
1Cu = 1B where =
.
c + r (1 c)

(1.32)

For concreteness, assume this could happen if the central bank buys the bonds from the
(consolidated) private bank, which in turn buys some of them from savers. The extra
reserves allow the (consolidated) private bank to take extra deposits. This can be done by

20

21

1Re = 1B where =

The relation between money stock and monetary base is therefore


M=

c is often thought of as being determined by bank/households, so the central bank can


affect money supply by either influencing r (reserve requirements) or by changing the
monetary base (interventions). The money multiplier is decreasing in fraction of currency,
c, since it acts like a leakage from the private banking system. At c = 0, that is, when
there is no currency, then the money multiplier is at a maximum and coincides with the
reserve multiplier.

Burda, M., and C. Wyplosz, 1997, Macroeconomics - A European Text, Oxford University
Press, 2nd edn.

lending money to a customer by crediting his deposit account. This extra deposit is

1B
r
1c
1B.
=
c + r (1 c)

1D =

Cooley, T. F., and G. D. Hansen, 1989, The Inflation Tax in a Real Business Cycle
Model, The American Economic Review, 79, 733748.
Driffil, J., G. E. Mizon, and A. Ulph, 1990, Costs of Inflation, in Benjamin M. Friedman,
and Frank H. Hahn (ed.), Handbook of Monetary Economics . , vol. 2, North-Holland.

Clearly,
1M = 1D + 1Cu
1
=
1B,
c + r (1 c)

Fischer, S., 1996, Why Are Central Banks Pursuing Long-Run Price Stability, in
Achieving Price Stability, pp. 734. Federal Reserve Bank of Kansas City.

as expected and all desired ratios are fulfilled (check this).

Goldfeld, S. M., and D. E. Sichel, 1990, The Demand for Money, in Benjamin M. Friedmand, and Frank H. Hahn (ed.), Handbook of Monetary Economics, vol. 1, . chap. 8,
North-Holland.

1.4.6 More on Interventions


Reference: OR
A sterilized foreign exchange intervention is when the effects on the money supply of
a foreign exchange intervention is nullified by an open market operation; the central bank
buys foreign assets, pays with cash, but sell domestic assets (government bonds) to get
the cash back.
An intervention on the forward market is very similar to a sterilized intervention.
Suppose the bank enters a forward contract to buy domestic currency tomorrow and to
sell foreign currency at the same time. This decreases the supply of domestic currency
tomorrow, that is, of domestic bonds, while increasing the supply of foreign bonds - just
like in a sterilized intervention.
It seems as if sterilized interventions can have effects on exchange rates and interest
rates, but we are not sure why. Portfolio-balance effect (changing supply changes the risk
premium, but what about Ricardian equivalence?) or signalling?

Lucas, R. E., 2000, Inflation and Welfare, Econometrica, 68, 247274.


Mishkin, F. S., 1997, The Economics of Money, Banking, and Financial Markets,
Addison-Wesley, Reading, Massachusetts, 5th edn.
Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT
Press.
Romer, D., 1996, Advanced Macroeconomics, McGraw-Hill.
Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

Bibliography
Benassy, J.-P., 1995, Money and Wage Contracts in an Optimizing Model of the Business
Cycle, Journal of Monetary Economics, 35, 303315.
Blanchard, O. J., and S. Fischer, 1989, Lectures on Macroeconomics, MIT Press.
22

23

The Price of Money

Main references: Romer (1996) (Romer), Blanchard and Fischer (1989) (BF), Obstfeldt
and Rogoff (1996) (OR), and Walsh (1998).

2.1

UIP, Fisher Equation, and the Expectations Hypothesis of the


Yield Curve

i t,t+s =

1
(Et i t + Et i t+1 + + Et i t+s1 ) + ti ,
s

(2.3)

where ti is a risk premium (term premium). The expectations hypothesis of interest rates
(yield curve) assumes that the risk premium is zero or constant.
Derivations are given in the Appendix.

Reference: OR 8.7.1-8.7.3
2.1.1

(The sign of the risk premium is just a matter of definition. Here ts > 0 means that
an investment in foreign bonds require a positive risk premium.) Uncovered interest rate
parity, UIP, assumes that the risk premium is zero or constant. Both (2.1) and (2.2) have
similar forms for multi-period investments.
Finally, let i t,t+s be the continuously compounded per period nominal interest rate on
a discount bond traded at t and maturing at t + s. The relation between long interest rates
and expected future short interest rates is

Pricing Relations for Nominal Returns

This section gives three very important relations, which we will use in the subsequent
analysis. These relations can be stated in several different forms, but here we use the
log-linear form which fits nicely into the linear models used in most of this class.
Let i t be a continuously compounded (per period) nominal interest rate on a discount
bond (no coupons) traded at t and maturing at t + 1, and let t+1 = ln(Pt+1 /Pt ) be
the corresponding inflation rate. The relation between nominal interest rates, real interest
rates and expected inflation is
i t = Et t+1 + rtr + t ,

(2.1)

where rtr is a real interest rate and t a risk premium (inflation risk premium). The Fisher
equation assumes that the risk premium is zero or constant, and sometimes also that the
real interest rate is constant.
The relation between domestic interest rates, foreign interest rates (indicated by a
star ), and expected exchange rate depreciation is
i t i t = Et ln (St+1 /St ) ts ,

(2.2)

where St is the exchange rate (units of domestic currency per unit of foreign currency,
for instance, 8 SEK per USD), and ts is a risk premium (exchange rate risk premium).
24

2.2

The Price Level as an Asset Price: Cagans (1956) Model with


Rational Expectations

Reference: BF 4.7 and 5.1, Romer 9.7, OR 8.2, Walsh (1998) 4.3.
This classic model was first used to discuss hyperinflation. In this case prices are
driven almost entirely by the dynamics of money supply (Phillips effects and other influences of real variables are of minor importance). Many hyperinflation episodes originate
in the need to generate government revenues, why we will take a look at seigniorage. This
model also allows us to discuss the asset pricing aspect of the price level, and how to
solve such models.
The model is an approximation to the general equilibrium model with money in the
utility function, where the real side of the economy (output, consumption, real interest
rate) is kept constant.
2.2.1

Determination of the Price Level under Rational Expectations

Suppose the money demand function (LM curve) is


ln Mt ln Pt = ln Yt i t , with > 0.

(2.4)

25

Prices are assumed to be completely flexible. Assume that income and the real interest
rate are constant, so by the Fisher equation
i t = Et (ln Pt+1 ln Pt ) + constant,

Substituting for Et ln Pt+1 in (2.7) gives




ln Pt = (1 ) ln Mt + Et (1 ) ln Mt+1 + Et+1 ln Pt+2 ,
|
{z
}

(2.5)

ln Pt+1

= (1 ) ln Mt + (1 ) Et ln Mt+1 + 2 Et ln Pt+2 ,

and the money demand equation can be normalized as


ln Mt ln Pt = (Et ln Pt+1 ln Pt ) .

(2.6)

These assumptions could either be motivated by that this is a steady state situation (general
equilibrium with MIU) or that we want to look at hyperinflation, where movements in the
real interest rate and output are of trivial importance compared with the movements in the
money stock.

This says that the price of money equals a dividend, ln Mt , and a discounted capital
gain. To see that ln Mt is like a dividend, suppose utility is U (C, M/P) = u 1 (C) +
ln M/P, then U M = 1/M. We could therefore think of ln Mt as an approximation of
the marginal utility of money.

(2.9)

where we use the law of iterated expectations. By continuing the substitution we end up
with
K
X
ln Pt = (1 )
s Et ln Mt+s + K +1 Et ln Pt+K +1 .
(2.10)
s=0

If lim K K +1 Et ln Pt+K +1 = 0, then we can write


ln Pt = (1 )

Remark 9 (The price of money) Note that if one unit of the good costs Pt units of money,
then one unit of money costs Ft = 1/Pt units of goods. We can then rewrite (2.6) as
ln Ft = ln Mt + (Et ln Ft+1 ln Ft ) .

(2.8)

s Et ln Mt+s ,

(2.11)

s=0

where ln Pt is the bubble-free (or fundamental) solution. Note that money is neutral in the sense that changing the level of money supply (Mt ) by a factor in each
period increases the price level by the same factor. This follows from the fact that
P
s
(1 )
s=0 = 1.
Example 11 (Log money supply is random walk plus drift.) Suppose the growth rate of
money is plus a (serially uncorrelated) shock, or

Rewrite (2.6) as

ln Mt+1 = + ln Mt + t+1 .

ln Pt = (1 ) ln Mt + Et ln Pt+1 , with = / (1 + ) < 1.

(2.7)

The price level today depend on the money supply, but also on the expected price level
tomorrow. If the price level tomorrow is expected to be very high, then currency will
be worth little tomorrow (the value of money in terms of goods is 1/Pt ). Like any other
asset, the value of money will then decrease already today, which means that Pt increases.
Remark 10 (Law of iterated expectations.) We must have

Then, the log price level in (2.11) is


ln Pt = (1 )

s (ln Mt + s) .

s=0

Since

s=0

s s

= / (1 ) , the rational expectation equilibrium price level is


2

ln Pt = ln Mt +

Et Et+1 ln Pt+2 = Et ln Pt+2 ,

Mt

or ln =
,
1
Pt
1

which we can write as (since = / (1 + ))

since the information set in t is a subset of the information set in t + 1. (Senility is not
allowed.)
26

ln Pt = ln Mt + or ln

Mt
= ,
Pt
27

so the log real balances are a decreasing function of the growth rate of money. In this
special case, real money balances are not affected by the shocks. (See Romer 391-394 for
a diagrammatic description and a discussion of the case with price inertia.)
Example 12 (Log money supply is random walk plus drift, continued.) From Example
11, inflation is equal to money growth

Since < 1, the expected value Et bt+s explodes.


Ruling out bubbles, that is, using the solution (2.11), therefore amounts to finding
the unique stable solution of an unstable difference equation, that is, exploiting the saddle point property. Equations (2.12)-(2.14) shows that any other solution explodes (in
expectation).
2.2.3

t = 1 ln Pt = 1 ln Mt = + t .
A higher growth rate of money supply drives up the expected inflation and therefore the
nominal interest rate (the real interest rate is assumed to be constant) which decreases
demand for real money balances. With a given money supply Mt , the price level must
increase to keep the money market in equilibrium.
Example 13 (Money supply and the nominal interest rate.) Consider a money demand
equation ln Mt ln Pt = ln Yt i t , and assume that ln Yt is constant. What is the
effect on the nominal interest rate, i t , of a shock to money supply? When money supply
is a random walk, then the effect is zero, since ln Pt increases as much as ln Mt . If
ln Mt = ln Mt1 + t with || < 1, then ln Pt = [(1 )/(1 )] ln Mt so ln Pt reacts
less than ln Mt to shock. In this case, i t must decrease in response to a positive shock
to money growth. To get a positive effect on i t , the shock to money supply must be more
persistent than a random walk, for instance, by letting 1 ln Mt be an AR(1).
2.2.2 Bubbles and Saddle Point Properties
Equation (2.11) is not the unique solution, only the unique fundamental solution. Let
us call it ln Pt , and postulate that any solution can be written as a sum of the fundamental
solution and a bubble bt ,
ln Pt = ln Pt + bt .
(2.12)

Seigniorage

Seigniorage can be an important source of funds for the government. It has historically
been very important during specific episodes, often in conjunction with wars. (Very historically, it was the fee the authorities asked for the service of minting your silver or gold.
To increase the demand for this service, it was often forbidden to mint your own coins or
to use foreign coins or even domestic coins older than a certain number of years.) The
need for seigniorage is reputed to be the main cause of most hyperinflations.
Real revenues from money creation, seigniorage, is
Mt Mt1
P
t

Mt
Mt1
=
1
.
Pt
Mt

Seignioraget =

(2.15)

This is the real revenues the government get by printing more money. We can think of
Mt /Pt as the tax base and 1 Mt1 /Mt as the tax rate. In fact, seigniorage is often
called inflation tax. The tax rate is essentially the money growth rate, which is strongly
correlated with inflation and the nominal interest rate (the Cagan model, they are the
same). We know from the money demand equation that real balances are decreasing in
the nominal interest rate (for a given output level), so increasing the money growth rate
therefore increases the tax rate and decreases the tax basethe result is often a Laffer
curve. Mt in (2.15) should be interpreted as the monetary base. Seigniorage is fairly
unimportant for most OECD countries; it is typically less than 1% of GDP.

Use this in (2.7)




ln Pt + bt = (1 ) ln Mt + Et ln Pt+1
+ bt+1 ,

Example 14 From Example 11 we have


(2.13)
Mt = Mt1 exp( + t ), and Pt = Mt exp () .

and use (2.11) to obtain the requirement that any bubble must satisfy
bt = Et bt+1 Et bt+s = s bt .

(2.14)
28

29

2.2.5

Use this in (2.15) to get


Seignioraget =

Mt1 exp( + t ) Mt1


Mt1 exp( + t ) exp ()

Empirical Illustrations

Burda and Wyplosz (1997) Fig 8.9, 8.12, Box 8.5, Table 16.3, and Boxes 16.4-5; Walsh
Fig 4.3.

= exp() exp [(1 + ) t ] .


This is always increasing in the money supply surprises t , but will typically show a
Laffer curve with respect to .
Example 15 (Tax smoothing and seigniorage.) Suppose the distortionary effects of taxes
are convex functions of the tax rates. The optimal way to finance government expenditures
is then to keep tax rates more or less constant over time. Temporary changes in government consumption should be met by lending/borrowing. Seigniorage is a tax, so this
theory suggests that seigniorage should be relatively constant. In fact, however, seigniorage seems to much more correlated with government consumption than other taxes. War
financing is a particularly clear case. See, for instance, Walsh (1998) 4.
2.2.4

Causality or Only an Equilibrium Condition?

In Cagans model, money supply is exogenous, so (2.11) shows how the expectations for
money supply determine the price level.
This would no longer be true if allowed output to change and to depend on prices
(something we will see later in the course), or if we assumed that money supply was
a function of output and inflation (something we will also see later in the course). In
this case, we could still combine the money demand equation and the Fisher equation to
express the price level in terms of a discounted sum of expected money supply and output,
but it would only be an equilibrium condition.
To demonstrate the last point, suppose both the real interest rate, rt , and log output,
ln Yt , vary. Then (2.6) should be
ln Mt ln Yt + rt ln Pt = Et (ln Pt+1 Pt ).

(2.16)

Therefore, if we replace ln Mt in (2.11) with ln Mt ln Yt + rt , we have a solution of


(2.16). However, this cannot really be interpreted as the cause of the price level until we
have specified how money supply, output, and the real interest rate are determinedin
particular, if they depend on the price level.
30

2.3

A Simple Model of Exchange Rate Determination

Reference: OR 8.2.7, 8.4.1-4, Burda and Wyplosz (1997) 18-21, Isard (1995).
2.3.1

UIP and the Exchange Rate Equation

The traditional monetary model of exchange rates starts out from a money demand equation, which is combined with an UIP condition.
The UIP (uncovered interest rate parity) condition is
Et 1 ln St+1 = i t i t ,

(2.17)

where i t and i t are the (per period) domestic and foreign currency interest rates, respectively, and St is the number of units of domestic currency per unit of foreign currency
(example: 8 SEK per USD). The condition says that the expected returns (measured in
a common currency) of lending in the domestic currency or in the foreign currency are
equal. This typically requires full capital mobility and risk neutrality.
The money demand equation is
ln Mt ln Pt = ln Yt i t , or
1
i t = ( ln Yt ln Mt + ln Pt )

(2.18)

There is a similar demand equation in the foreign country, so the interest rate differential
can be written
i t i t =




1
ln Yt ln Yt ln Mt ln Mt + ln Pt ln Pt .

(2.19)

The real exchange rate is defined as the relative price of foreign goods
Qt =

St Pt
8 kronor per dollar 1 dollar per US hamburger
=
.
Pt
10 kronor per Swedish hamburger

(2.20)

31

You may note that the purchasing power parity (PPP) issue is about whether Q t is constant
or not. The flexible-price monetary model of exchange rates would set Q t constant. We
will not impose that, so the equation for the exchange rate that we will arrive at can only
by regarded as an equilibrium condition.
Use (2.20) to substitute for ln Pt ln Pt = ln Q t +ln St in (2.19), and then combine
with the UIP condition (2.17) to get
Et 1 ln St+1 =



 1
1
ln Yt ln Yt ln Mt ln Mt ln Q t + ln St .

(2.21)

Collect the fundamental driving variable into




vt = ln Yt ln Yt + ln Mt ln Mt + ln Q t ,

(2.22)

and rewrite (2.21)as


vt
1
+ ln St or

ln St = vt + Et 1 ln St+1 .

Et 1 ln St+1 =

(2.23)

Note how the exchange rate is like any other asset: the current price is a function of some
dividend, vt , plus a discounted capital gain, Et 1 ln St+1 . Note that (2.23) is just an
equilibrium conditionit cannot be given a causal interpretations without making further
assumptions.
2.3.2

A Fixed Exchange Rate (St fixed, Mt variable)

Suppose the central bank pursues a policy of a unilateral fixed exchange rate, and that it
manages to make this policy credible. For instance, let ln St =Et 1 ln St+1 = 0 in (2.23),
and note that this requires vt = 0. From (2.22), we see that this means that

increases money demand according to (2.18); the central bank increases money supply
by either an open market operation (buying bonds, selling domestic currency) or an intervention on the foreign exchange market (buying foreign exchange, selling domestic
currency). This restores money market equilibrium at an unchanged interest rate. By UIP
(2.17), this is compatible with a fixed exchange rate.
Instead of a unilateral peg, suppose the home country and one foreign country decide
to fix their bilateral exchange rate (for instance, St = 0). According to (2.24), this puts a
requirement on the relative money supply, Mt /Mt . The level of money supply (nominal
anchor) can be set to meet some other objective (the exchange rate with a third currency,
the price level, or for stabilization purposes). (This is often called the N-1 problem.)
Example 16 (Bretton Woods.) The Bretton Woods system was originally based on the
USD being pegged to gold, and the other currencies being pegged to the USD. The fixed
exchange rate forced other countries to behave according to (2.24). As a consequence, the
other countries more or less adopted the US inflation rate, see (2.11). The gold peg meant
that foreign central banks (not private agents) could buy gold per 35 USD per ounce.
It was expected to discipline the US since too fast US money creation would lead other
central banks to convert dollars into gold. However, this did not work as expected during
the 1960s when the US money growth rate increased. Several countries, like Germany
and Japan, had strong anti-inflationary preferences but abstained from converting dollars
into gold, possibly because of the strong political dependence of the US. At the end of the
1960s/beginning of the 1970s a series of speculative attacks toppled the Bretton Woods
regime.

If the money stock is the monetary policy instrument, then this equation shows how the
central bank must act to keep the exchange rate fixed. In this sense, the central bank has
no control over the money stock in a fixed exchange rate regime. By fixing the exchange
rate (or any other financial price) the country looses its monetary policy independence.
The mechanism is the following: suppose we get a temporary shock to ln Yt . This

Example 17 (EMS.) Until the mid 1980s EMS was characterized by a series of small
realignments, but thereafter it was very much a system for fixed exchange rates where
most European currencies were effectively pegged to the DM. The boom in Germany after
the unification lead to inflationary pressure, while most other European countries were in
a fairly deep recession with almost deflationary tendencies. Bundesbank wanted to keep
monetary policy tight, which forced other countries to follow (see(2.24)). In the end, the
political pressure for looser monetary policy (for instance, in the UK) undermined the
credibility of the system and a series of speculative attacks forced a number or currencies
to abandon the peg. Why? Central banks should in most cases be able to buy back the
monetary base and thereby restore the exchange rate. However, by (2.18) this would

32

33



ln Mt = ln Mt + ln Yt ln Yt + ln Pt ln Pt .

(2.24)

(at unchanged prices and output) lead to very high interest rates, which may disrupt the
economy (especially the banking sector which typically borrows short and lends long).
The Swedish central bank was willing to accept extreme interest rates during the first
attack on the krona in September 1992, but not during the second attack two months later.
2.3.3 A Floating Exchange Rate
Suppose instead that the central bank lets the exchange rate move. In the extreme case,
Mt is fixed, but that is clearly not necessary for the exchange rate to move. Rearrange
(2.23) to express ln St as a function of vt and Et St+1 on the left hand side
ln St = vt + Et ln St+1 ln St
= (1 ) vt + Et ln St+1 , with = / (1 + ) < 1

(2.25)

that as becomes large, ln St will be very similar to a martingale (for instance, a random
walk).
(2.26) gives a key role to the expectations formation. An unanticipated increase in
the fundamental will cause a large increase in ln S1 , while an anticipated increase in the
fundamental causes the exchange rate to increase already at the date of announcement. A
simple example illustrates that.
Example 19 (Anticipated versus unanticipated shocks.) Suppose
v0 =
0
vi = 1 for i 1.
If the change in vi is completely unanticipated, then
ln S0 = 0
if E0 vi = 0 for i 1.
ln S1 = 1

because > 0. The stable solution (ruling out bubbles) is then


ln St = (1 )

s Et vt+s .

(2.26)

In contrast, if the change in vi is anticipated in t = 0, then

s=0

This expresses the current exchange rate in terms of the expected values of future dividends. This is the asset pricing view of exchange rate determination. Since we have
not said anything about how vt is determined, this is only an equilibrium condition. In
particular, there is plenty of evidence that the real exchange rate (which is in vt ) is affected
by the nominal exchange rate, at least in the short to medium run.
Example 18 (AR(1) fundamental.) If vt+1 = vt + t+1 with t+1 iid, then Et vt+s =
s vt , so (2.26) becomes
ln St = vt (1 )

P
s
ln S0 = 0 + (1 )
s=1 =
if E0 vi = 1 for i 1.
ln S1 =
1
Is often claimed that the exchange rate depreciation, 1 ln St , and the interest rate
differential, i t i t , are correlated. However, this depends on which kind of shock that
hits the economy. To get a positive correlation between 1 ln St and i t i t , the shock
must create a change in the exchange rate which is expected to be followed by changes in
the future in the same direction.
Example 20 (A mean reverting shock to the fundamental.) When vt is iid, then (2.26)
becomes
ln St = (1 ) vt ,

()s

s=0

1
= vt
,
1

so

where the last term is the effect on 1 ln St+1 of a shock to vt+1 . This effect is increasing
in , since a larger means that the shock has a more long lasting effect on vt+s .

1 ln St = (1 ) (vt vt1 ) , and


Et 1 ln St+1 = (1 ) Et (vt+1 vt ) = (1 ) vt .

We have, once again, nominal neutrality in the sense that increasing Mt in all periods
by the factor increases the exchange rate with the same factor. Also note from (2.25)
34

35

Since i t i t =Et 1 ln St+1 we get

The depreciation is then

Cov(1 ln St , i t i t ) = (1 )2 Var (vt ) ,

1 ln St = u t + (1 ) 0 t (u t1 + (1 ) 0 t1 )

which is negative. It is straightforward to show that we get Cov(1 ln St , i t i t ) < 0


for any stationary AR(1) specification of vt . In contrast, most empirical estimates of this
covariance are positive.

= [1 + (1 ) 0 ]t (1 ) 0 t1 ,
since u t u t1 = t . Et t+1 = 0, so the expected depreciation must be
Et 1 ln St+1 = (1 ) 0 t

Example 21 (Permanent shock to the level.) Consider the case when there are permanent
shocks to the level of money supply

Combine to get Cov(1 ln St , i t i t ) as


Cov(1 ln St , Et 1 ln St+1 ) = E{[1 + (1 ) 0 ]t (1 ) 0 t1 }{ (1 ) 0 t }

vt+1 = vt + t+1 ,

= [1 + (1 ) 0 ] (1 ) 0 Var (t )
so (2.26) is
ln St = vt ,
1 ln St = vt vt1 = t ,
Et ln St+1 = vt , so Et 1 ln St+1 = i t i t = 0.
This means that Cov(1 ln St , i t i t ) = 0, which is still too low compared with most
empirical estimates.
Example 22 (A more than permanent shock.) Let the fundamental be a sum of a random
walk and the shock to the random walk

which can have either sign. For 0 = 0 it is zero, since this is the random walk case
discussed above. For 0 > 0 it is always negative, since t > 0 shifts the permanent level
of vt up, but also gives a temporary positive blip in t, so 1 ln St > 0 but Et 1 ln St+1 < 0.
For 0 < 0, we can get a positive covariance, since t > 0 gives a temporary negative
blip, that is, the upward permanent shift in vt is not realized until t + 1. For instance, for
0 = 1, the covariance is (1 )Var(t ), which is positive. The basic mechanism is
that the expected effect of the shock on the fundamental grows over time.
2.3.4

The Correlation Between Real and Nominal Exchange Rates

A stylized fact is that the real, Q t , and the nominal, St , exchange rates are strongly positively correlated. If we assumed that St does not affect Q t , then the only explanation for
Cov(Q t , St ) > 0 in this setting is that shocks to the real exchange rate is the main driving
force behind the nominal exchange rate. As seen from (2.22) and (2.23), the nominal exchange rate is an increasing function of the real exchange rate. Monetary shocks should
be small. There is plenty of empirical evidence against this view. The basic mechanism
would then be that monetary shocks drive St which in turn drive Q t , because nominal
prices Pt /Pt tend to be sticky.

vt = u t + 0 t , where u t = u t1 + t and t is iid.


This means that
Et vt+s = Et u t+s + 0 Et t+s
(
u t + 0 t if s = 0
=
.
ut
if s 1
By (2.26) the exchange rate is
ln St = u t + (1 ) 0 t

36

37

2.3.5 Capital Controls and Risk Premia


The preceding discussion shows that the country can chose either fixed exchange rate or
monetary policy independence. It cannot have both, unless there is some way to break the
UIP condition (which here serves as the equilibrium condition for the capital market). One
possibility of doing that is to let the central bank affect risk premia by changing the relative
supplies of different assets. For instance, the portfolio-balance approach emphasizes risk
premia and discusses how sterilized interventions change the risk characteristics of the
private-sector portfolio and thereby asset prices. Another possibility is to impose capital
controls, which make it costly to move capital.
2.3.6

Other Candidate Exchange Rate Equations

There are several other candidate exchange rate equations. Monetary sticky-price models
relaxes the assumption of instantaneous price flexibility, which often adds inflation terms
to vt and makes the real exchange rate endogenous. The Dornbusch model (see OR 9.2)
is one example.
Structural exchange rate equations, which try to explain exchange rates in terms of
macro variables often fail to improve upon a simple random walkat least for relatively
short horizons.
2.3.7


i
1 = Et exp rt+1
+ qt+1 .

(A.2)

Assume that q and r j are conditionally normally distributed


"
# "
# " q #
" q #
" # "
#!
qt+1
Et qt+1
t+1
t+1
0
qq q j
=
+
with

N
,
.
j
j
j
j
rt+1
Et rt+1
t+1
t+1
0
q j j j
(A.3)
Example 23 (CRRA utility.) The real discount factor, q, would be normally distributed,
for instance, if ln Ct+1 and ln(Mt+1 /Pt+1 ) are normally distributed and the utility func
1
tion is isoelastic, so U (Ct , Mt /Pt ) = Ct (Mt /Pt )1
/ (1 ). In this case,
qt+1 = ln[u C (Ct+1 , Mt+1 /Pt+1 ) /UC (Ct , Mt /Pt )]
Ct+1 (Mt+1 /Pt+1 )(1)(1 )

= ln

Ct

(Mt /Pt )(1)(1 )

= ln (Ct+1 /Ct ) + (1 ) (1 ) ln(Mt+1 Pt /Pt+1 Mt ).


Remark 24 If x N (Ex, Var(x)), then

Empirical Illustrations

Burda and Wyplosz (1997) Figs. 8.9, 13.6, and 19.1; OR Figs. 9.1; Isard (1995) Figs.
3.2, 4.2, and 11.1.

Note that the log real discount factor must be decreasing in Ct+1 (and ln Ct+1 ), since the
utility function is concave.
j
j
Let rt+1 = ln Rt+1 be the log gross return, and rewrite the first order condition as

Derivations of the Pricing Relations


j

Consider an agent who chooses consumption and asset holdings optimally. Let Rt+1 be
the one-period gross return from investing in asset j in t. The first order condition is
i
1 = Et Rt+1
exp (qt+1 ) ,

(A.1)

where qt+1 is the log real discount factor between t and t + 1. For instance, if the utility
s u(C
function is time separable, Et 6s=0
t+s , Mt+s /Pt+s ), then qt+1 = ln[u C (C t+1 , Mt+1 /Pt+1 ) /UC (C t , Mt /Pt )].
38

E exp (x) = exp [Ex + Var (x) /2] .


The distribution could be interpreted as a conditional or an unconditional distribution.
Take logs of (A.2), and apply the rule for Eexp (x) when x is normally distributed


j
j
(A.4)
0 = Et rt+1 + Et qt+1 + Vart rt+1 + qt+1 /2.
This can be written as


1
j
j
Et rt+1 = Et qt+1 Vart rt+1 + qt+1
2


 1

1
j
j
= Et qt+1 Vart rt+1 Vart (qt+1 ) Covt rt+1 , qt+1
2
2

(A.5)

39

The variance terms are due to the non-linear transformation (recall Jensens inequality)
and typically not very interesting. The covariance is more important, since it captures risk
aversion.

A.1

A Real Bond
j

A real bond has a known real interest rate (it is safe), rt+1 = rtr . In this case (A.5)
simplifies to
1
rtr = Et qt+1 Vart (qt+1 ) .
(A.6)
2
We can use this equation to rewrite (A.5) as




1
j
j
j
Et rt+1 rtr = Vart rt+1 Covt rt+1 , qt+1
2
j
= t ,

(A.7)
(A.8)

which shows how the expected return in excess of the safe return depends on a Jensens
inequality term and the covariance of the return with the stochastic discount factor. A
negative covariance means that the asset tends to have an unexpectedly low return when
marginal utility is unexpectedly high. This is like a negative insurance, so investors will
j
require a positive risk premium, t .

A.2

A Nominal Bond
j

A nominal bond has an uncertain real return, rt+1 = i t t+1 , since inflation is uncertain.
In this case (A.7) is
1
i t Et t+1 rtr = Vart (t+1 ) Covt (t+1 , qt+1 ) ,
2

A.3 A Nominal Foreign Bond


A nominal foreign bond has also an uncertain real return, i t t+1 + ln (St+1 /St ), since
both inflation and the exchange rate depreciation are uncertain. To see that this is the
real return, note that giving up one unit of goods today gives Pt units of domestic cur
rency, and therefore Pt /St units of foreign currency. In t + 1, this gives exp i t Pt /St

units of foreign currency, or exp i t Pt St+1 /St units of domestic currency, and therefore

exp i t (Pt /Pt+1 ) (St+1 /St ) units of goods.
In this case (A.7) is
i t Et t+1 + Et ln (St+1 /St ) rtr = t ,

(A.10)

where the variance and covariance terms are collectively denoted t . We can use (A.9) to
substitute for Et t+1 + rtr to get
i t i t + Et ln (St+1 /St ) = ts ,

(A.11)

which is the same as (2.2). It is straightforward to show that








1
ts = Vart ln (St+1 /St ) Covt t+1 , ln (St+1 /St ) Covt ln (St+1 /St ) , qt+1 .
2
(A.12)
If the last covariance term is negative, then the exchange rate tend to unexpectedly appreciate (the real return, measured in domestic goods, of the foreign bond is low), when
marginal utility is unexpectedly high. Investors will then require a positive exchange rate
premium.

A.4

Real Effects of Money?

(A.9)

which has the same form as (2.1). Note that i t is known, so only t+1 enter the variance
and covariance terms. If the covariance is negative, then inflation tend to be unexpectedly
high (the real return on the nominal bond is unexpectedly low), when marginal utility is
unexpectedly high. Investors will then require a positive inflation risk premium. The sign
of the inflation risk premium can clearly be different in different economies. It can also
change over time if the conditional covariance does.

The derivation of the pricing relations does not rule out real effects of money. In fact, nothing has been said about how the conditional distribution (A.3) is determined: it could be
the case that money supply changes affect output and therefore the optimal consumption
decision. Alternatively, it could be the case that monetary policy cannot affect the average
level of output, but it may have an effect on the volatilities. In this case, monetary policy
affects the risk premia for financial assets, which could affect the consumption/savings
trade-off.

40

41

A.5

Empirical Evidence on the Pricing Relations

realignment eventually came. In many studies, the sample ended before the realignment:
the peso problem.) Evidence from survey data suggest that this might be the case.

Reference: Soderlind and Svensson (1997)


A.5.2
A.5.1

Fisher Equation

UIP

Several of these relations have been studied by running ex post regressions. For instance, to test the UIP, we could add the innovation in the log exchange rate, u t+1 =
ln St+1 Et ln St+1 , to both sides of (2.2) to get
ln St+1 ln St = i t i t + u t+1 ,

The Fisher equation is typically tested in much the same way as UIP: inflation is related
to the nominal interest rate and the null hypothesis is that the sum of the real interest rate
and the inflation risk premium is a constant. Most empirical evidence suggests that this is
not true, in particularly not for short maturities.

(A.13)
A.5.3

provided UIP holds, that is, if ts = 0. We would test this relation by running the regression

ln St+1 ln St = a + b i t i t + t+1 ,
(A.14)
and test the null hypothesis that a = 0 and b = 1. Under the null hypothesis that UIP
holds, this regression should give consistent estimates of a and b, since the innovation
u t+1 must (by definition) be uncorrelated with the regressors, or for that matter, everything
else in period t or earlier.
This testing approach is a special case of the more general implication of UIP: ln St+1
ln St i t + i t should be unforecastable, and therefore uncorrelated with all information
in period t.
The test of the null hypothesis is, of course, a joint test of rational expectations (that
i t i t equals Et ln St+1 ln St rather than some other expectation of the exchange rate
depreciation) and of no risk premia. If a 6 = 0 but b = 1, then this might (under RE)
be interpreted as constant risk premia, which in (A.11)-(A.12) corresponds to constant
second moments.
The typical result from a large number of studies is that a 6= 0 but b 6 = 1 (often
b < 0). This could be due to time-varying risk premia (which requires time-varying
variances and covariances in the theory for risk premia presented above). An alternative
explanation is that the sample is not long enough for ex post data to produce all the
jumps (devaluations) and other features that seems to be part of market expectations of
exchange rates. (The Mexican peso is the classic case, where the interest rate differential
to USD interest rates was positive for many yearsand it took a very long time before the

42

Expectations Hypothesis of Interest Rates

The expectations hypothesis of interest rates is also tested in a similar way: future long
interest rates are related to current long interest rates. Most evidence suggest that the
expectations hypothesis of interest rates works fairly well for very short maturities, but
perhaps less well for maturities of 6 months up to a couple of years. The empirical evidence for really long maturities is very mixed.
A.5.4

Empirical Illustrations

McCallum (1996) Fig 9.1; Soderlind and Svensson (1997) Fig 5; Soderlind (1998) Fig 2.

Bibliography
Blanchard, O. J., and S. Fischer, 1989, Lectures on Macroeconomics, MIT Press.
Burda, M., and C. Wyplosz, 1997, Macroeconomics - A European Text, Oxford University
Press, 2nd edn.
Isard, P., 1995, Exchange Rate Economics, Cambridge University Press.
McCallum, B. T., 1996, International Monetary Economics, Oxford University Press,
Oxford.
Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT
Press.

43

Romer, D., 1996, Advanced Macroeconomics, McGraw-Hill.


Soderlind, P., 1998, Nominal Interest Rates as Indicators of Inflation Expectations,
Scandinavian Journal of Economics, 100, 457472.
Soderlind, P., and L. E. O. Svensson, 1997, New Techniques to Extract Market Expectations from Financial Instruments, Journal of Monetary Economics, 40, 383420.

Money and Sticky Prices: A First Look

Main references: Romer (1996) (Romer), Blanchard and Fischer (1989) (BF), Obstfeldt
and Rogoff (1996) (OR), and Walsh (1998).

Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

3.1

Basic Models of the Effects of Monetary Policy Surprises

3.1.1

One-Period Wage Contracts

References: BF 8.2, Romer 6.8, Walsh 5.3.


The firm has a Cobb-Douglas production function so log output is
ln Yt = ln Z t + ln K t + (1 ) ln h t ,

(3.1)

where Z t is the productivity level, K t capital stock, and h t employment.


On a competitive labor market, the log nominal wage would be
ln wt = ln Pt + ln (1 ) + ln Yt ln h t .

(3.2)

Instead, nominal wages are here set, one period in advance, equal to the expected value
of the market clearing wage (in logs, to simplify)
ln wt = Et1 ln Pt + ln (1 ) + Et1 ln Yt Et1 ln h t .

(3.3)

In period t, shocks are realized and firms employ labor until the nominal marginal
product of labor equals the nominal wage, that is, until (3.2) holds, but where is ln wt set
in t 1. This implicitly assumes that households have a flat labor supply curve.
Use (3.3) in (3.2) to get
ln h t Et1 ln h t = ln Pt Et1 ln Pt + ln Yt Et1 ln Yt .

(3.4)

If K t = (1 ) K t1 + It , so K t is known in t 1, then the innovation in log output is


ln Yt Et1 ln Yt = (ln Z t Et1 ln Z t ) + (1 ) (ln h t Et1 ln h t ) .
44

(3.5)
45

Now, use (3.4) to substitute for the labor input innovation in (3.5). After rearrangement
we get


1
1
ln Yt Et1 ln Yt = (ln Z t Et1 ln Z t ) +
(ln Pt Et1 ln Pt ) . (3.6)

This model is a general equilibrium model with money in the utility function, but
sticky nominal wages.
The key equations are:
Utility function : E0

In this model monetary policy surprises can have affect on output by causing a price
surprise, while anticipated monetary policy cannot. For instance, if monetary policy could
react to innovations in the productivity level, then it might be possible to stabilize output.
3.1.2

1
( pi Ei p) ,




Mt
t ln Ct + ln
+ V h h t
Pt

Mt
wt
Mt1
+ K t+1 =
h t + r t K t + t
Pt
Pt
Pt
Yt
Yt
wt
1
= (1 ) , rt = .
Production function : Yt = Z t K t h t

Pt
ht
Kt

Real budget constraint : Ct +

Capital accumulation : complete depreciation, Yt = Ct + K t+1 .

Reference: BF 356-361, Romer 6.1-6.4, .


The Lucas model is another way to get one-period effects of monetary supply shocks.
The basic mechanism is that firm i has the supply curve
(3.7)

where pi is the firms price, and Ei p the expectation about the general price level based
on the information set of firm i. It is assumed that firm i observes only pi and yi and uses
these to infer the general price level. The firm cannot distinguish between real shocks
(to the relative price, pi p) and nominal shocks. It therefore reacts, to some extent,
to all observed movements in pi : we get real effects of monetary policy as long as the
policy surprise lasts - the macro implication is very similar to the sticky wage model (3.6).
However, the coefficient in front of the innovation in prices here depends on the volatility
of prices, rather than on the production function parameter.
A major criticism against the Lucas model is that the misperception story is weak: the
consumer price index is published with a very short lag.

3.2

t=0

Lucass Model of the Phillips Curve

yi =

(3.8)
(3.9)
(3.10)
(3.11)

This is the same model as in Long and Plosser (1983), but with money in the utility
function. The notation is standard, except that the nominal value in the beginning of t
of money is t Mt1 , where t may be different from unity. Benassy refers to t as a
multiplicative monetary shock. One possible interpretation is a stochastic tax/subsidy
on cash holdings, where the government prints (t 1)Mt1 new money at the beginning
of period t and distributes it as interest rate payments on cash holdings. The exact
details might not be too important. The essential feature is that there is a stochastic money
supply.
3.2.2

Flexible Prices

The Lagrangian is
E0





wt
Mt1
Mt
K t+1 + h t + rt K t + t
t ln Ct + ln Mt /Pt + V h h t + t Ct
.
Pt
Pt
Pt
t=0
(3.12)

Money and Wage Contracts in an Optimizing Model of the Business Cycle, by Benassy

3.2.1

Baseline Model

Reference: Benassy (1995), Walsh 5.3.

46

47

which shows that a stable solution (with limT ()T Et K T +1 /C T = 0) must satisfy

The first order conditions are


1
t = 0
Ct

wt
h t : V 0 h h t + t
=0
Pt

Ct :

K t+1 : t + Et t+1rt+1 = 0
1
t+1 t+1

t + Et
= 0.
Mt :
Mt
Pt
Pt+1

(3.14)
(3.15)
(3.16)

(3.18)

that is, the Long-Plosser solution

To determine consumption, use (3.13) in (3.15)


1
rt+1
= Et
,
Ct
Ct+1
Yt+1
1
= Et
/*rt+1 = Yt+1 /K t+1 */
Ct
Ct+1 K t+1
(Ct+1 + K t+2 )
K t+1
= Et
/* K t+1 , use Yt+1 = Ct+1 + K t+2 */
Ct
Ct+1
K t+1
K t+2
= + Et
.
Ct
Ct+1

K t+1
,
=
1
Ct
Yt C t
=
Ct
Yt
=
1,
Ct

(3.13)

Ct = (1 ) Yt
K t+1 = Yt .

(3.19)
(3.20)

To determine labor supply, use (3.13) in (3.14)



1 wt
,
V 0 h h t =
Ct Pt
(3.17)

Note that we have used som aggregate relations (equilibrium rental rate and aggregate
resource constraint), which makes a lot of sense once we realize that tthe behaviour of a
representative agent must be compatible with equilibrium.
Solve (3.17) recursively forward


K t+1
K t+3
= + Et + Et+1
Ct
Ct+2
K t+3
2
2
= + () + () Et
Ct+2
T
X
K T +1
,
=
()s + ()T Et
CT
s=0

48

(3.21)

and then note that wt /Pt = (1 ) Yt / h t , which together with (3.19) gives

1
h t V 0 h h t =
,
1

(3.22)

which is an implicit function for h t . Clearly, labor supply is a constant, which we denote by h (ruling out any odd V function which would make the left hand side of (3.22)
constant in spite of changes in h t ). Note that this implies that the dynamics of log output
is
ln Yt+1 = ln Z t+1 + ln K t+1 + (1 ) ln h
= ln Z t+1 + (ln + ln + ln Yt ) + (1 ) ln h. /*K t+1 = Yt */

(3.23)
(3.24)

Solving this equation recursively backwards shows that only shocks to Z t drive output:
monetary shocks do not matter for real variables. The reason is that the marginal utility
of consumption/leisure does not depend on the real money balances: the utility function
is separable. Therefore, money is neutral both in the steady state (as in most models with
money in the utility function) and along the adjustment path to steady state.
Since labor supply is constant, it is clear from the production function (3.10) that the
real wage and output have a correlation of one.
49

Finally, to determine the price level, use (3.13) in (3.16)

t+1
1
+ Et
=
Mt
Pt+1 Ct+1
Pt Ct
Mt+1
Mt
+ Et
=
/* Mt , Mt+1 = t Mt */,
Pt+1 Ct+1
Pt Ct

condition with respect to Bt+1 is

(3.25)

(3.26)

Equation (3.26) shows that, as long as money supply is exogenous, prices and consumption are negatively correlated. The reason is that money has no effect on real variables in this model with flexible prices. Formally, we have
Cov (ln Pt , ln Ct ) = Cov (ln Mt ln Ct , ln Ct ) /*from (3.26)*/
= Cov (ln Mt , ln Ct ) Var (ln Ct ) ,

(3.27)

which is negative since Cov(ln Mt , ln Ct ) = 0 as long as Mt is uncorrelated with Z t (and


hence output and consumption). The intuition for this result is that ln Pt is affected by
both nominal and real shocks, while ln Ct is affected by real shocks only. It is clear from
(3.26) that real shocks must drive prices and consumption in different directions, since
the money stock is unaffected by these shocks.

1
1
1
Et
= Et Ct /Ct+1 Et
,
1 + rtr Pt+1 /Pt
Pt+1 /Pt

t +

1 + rtr

Et t+1

1
= Et t+1 /t = Et Ct /Ct+1 .
= 0, or
1 + rtr

3.2.4

Discussion of the Money Demand Equation

The money demand equation (3.26) looks a bit odd, since it does not include the nominal
interest rate. This section shows that this comes from the dividends payed to holders
of cash. To see this, add a lump sum transfer to the budget constraint (3.9), and assume
that money supply evolves as, Mt+1 = t+1 Mt . If t+1 6 = t+1 , then the central bank
balances its budget by changing the lump sum transfer.
The first order condition for money holdings is then still characterized by (3.13) and
(3.16). The first line in (3.25) still holds, but the second does not. To see what we get
instead, suppose t+1 and t+1 are known in t. In this case, by multiplying the first line
in (3.25) by Pt Ct we have

Pt Ct
Pt Ct
+ t+1 Et
= 1.
Mt
Pt+1 Ct+1

(3.31)

By using the nominal interest rate from (3.29), (3.31) can be written

(3.28)

(3.30)

which looks pretty much like the nominal interest rate in (3.29). The difference is, of
course, the risk inflation premium, that is, the conditional covariance of the numerator
and denominator in the last term of (3.29). (Recall that E x y = E x E y + Cov (x, y).)
This is a Fisher equation plus an inflation risk premium.

3.2.3 Asset Pricing


 r
r
r
Add a real bond to the budget constraint: add 1 + rt1
Bt to the revenues and Bt+1
to
r
the expenditures. Since the real interest rate rt is known in t, the first order condition with
r
respect to Bt+1
is

(3.29)

We can also note that if we multiply (3.28) by Et Pt /Pt+1 , then we get

where the second line uses the fact that money demand, Mt+1 , equals money supply,
t+1 Mt . Solving recursively forward shows that a stable solution must be
Mt

.
=
Pt Ct
1

t
t+1
1
t+1 /t
Ct /Ct+1
+ (1 + i t ) Et
= 0, or
= Et
= Et
.
Pt
Pt+1
1 + it
Pt+1 /Pt
Pt+1 /Pt

t+1
Pt Ct
+
= 1.
Mt
1 + it

(3.32)

We could also add a nominal bond to the budget constraint: add (1 + i t1 ) Bt /Pt
to the revenues and Bt+1 /Pt to the expenditures. Since i t is known in t, the first order

This shows that when t+1 is proportional to 1 + i t , then we get a quantity equation. This
is satisfied when Mt+1 = t+1 M, since by combining (3.26) and (3.29) we also see that
1 + i t = Mt+1 /(Mt ) = t+1 /. It can be noted that the nominal interest rate must be
higher than the direct dividends on cash to compensate for the fact that cash has a direct

50

51

effect on utility. Otherwise no one would like to buy nominal bonds.


In most monetary macro models, the dividend on money is not proportional to the
nominal interest rate. To demonstrate this, consider the very simple case where (i) cash
has do dividends (t+1 = 1 in all periods); (ii) consumption and the real interest rates
are constant (more generally, they follow an exogenous process since they are unaffected
by money supply); (iii) the Fisher equation holds; (iv) and money supply follows some
stochastic process. This is Cagans model, where we easily can generate movements in
the nominal interest rates by making money supply something other than a random walk.
Even if the nominal interest rate cancels from the money demand equation, it is not
obvious it should simplify to a quantity equation. It is straightforward to show that this
happens only when the utility function has a Cobb-Douglas form or logarithmic form in
terms of consumption and real money balances.
3.2.5

Nominal Wage Contracts

Assume now that nominal wages for t are set in t 1, with the agreement that households
will supply any labor actually demanded by firms in t (compare with the right to manage
in the wage literature). The log nominal wage is set equal to the log expected (as of t 1)
nominal value of the marginal product of labor in t in the case of no stickiness. This
assumption is, of course, ad hoc but not too unreasonableand very convenient.
The household still optimizes (3.12), but it now takes h t as given. Since the utility
function is separable, the first order conditions for Ct , K t+1 , and Mt are unchanged.
To find the wage, note that from the production function we have wt = Pt (1 ) Yt / h t ,
which we rewrite as
Ct 1
/*Yt = Ct / (1 ) from (3.19)*/
1 h t
1
(1 ) (1 ) 1
= Mt
. /*Pt Ct = Mt
from (3.26)*/
(1 ) h t

Use (3.34) in (3.33) to solve for h t


ln h t = ln Mt Et1 ln Mt + constant.

(3.35)

Combining this with (3.24), but replacing h with h t+1 gives


ln Yt+1 = ln Z t+1 + ln Yt + (1 ) (ln Mt+1 Et ln Mt+1 ) + constant,

(3.36)

which shows that money supply surprises affect current output. It also affects future
output via capital accumulation.
Since labor supply is now positively correlated with output, real wages are no longer
perfectly correlated with output.
Since consumption and output are positively correlated with money supply, the correlation between prices and output does not have to be negative, see (3.27). Note that ln Mt
equals ln Pt + ln Ct plus a constant. Since real shocks have no effect on ln Mt they must
move ln Ct and ln Pt in opposite directions. In contrast, nominal shocks move both ln Mt
and ln Ct in the same directions, but ln Ct moves less (see (3.36)), so ln Pt also moves in
the same direction. The covariance of ln Ct and ln Pt is an average of these two forces and depends therefore on the relative volatility of real and nominal shocks.
Example 25 (Correlation of output and real wage in a special case.) Suppose = 0 (or
that K t is very stable so it doesnt matter for the business cycle movements). By using the
definitions in (3.10), Yt = Z t h t and wt /Pt = Yt / h t = Z t , we then get


wt
= Cov (ln Z t + ln h t , ln Z t )
Cov ln Yt , ln
Pt
= Var (ln Z t ) ,

wt = Pt (1 )

(3.33)

since (3.35) shows that ln h t depends only on money supply shocks, which are assumed to

Since h t = h with flexible prices, the log nominal wage set in t 1 is


ln wt = Et1 ln Mt + constant.

(3.34)

52

53

labour supply in t exactly offsets the increased propensity to consume so investment is


unaffected by the shock to At .
It is also possible to think of direct shock to labour supply (add a stochastic parameter
in the V function) or to money demand (let be stochastic). Shock to labour supply will
certainly affect output and prices, but shocks to money demand will (under flexible prices)
probably not affect the real side of the economy since the utility function is separable.

be uncorrelated with the productivity level. The correlation is therefore






Cov ln Yt , ln wPtt
wt


Corr ln Yt , ln
=
Pt
Std ln wt Std (ln Y )
t

Pt

Var (ln Z t )
1/2

Std (ln Z t ) Var (ln Z t ) + Var (ln Mt Et1 ln Mt )
1
=
1/2
1 + Var (ln Mt Et1 ln Mt ) /Var (ln Z t )

3.2.7

which decreases towards zero as the volatility of the money supply shocks increase, and
increases towards one as the volatility of the Solow residual increases.
3.2.6 Extensions of the Model: Demand Shocks
This model has two shocks: to productivity and to money supply. It is easy to add demand
shocks, which could useful for discussing monetary policy. The easiest way to do that is
to change the utility function (3.8) by multiplying ln Ct with a random taste parameter,
At . In this case, the first order condition for optimal consumption (3.13) becomes
At /Ct = t .

(3.37)

This changes a number of things. In particular, (3.17) becomes


At+1 K t+2
At K t+1
= Et At+1 + Et
.
Ct
Ct+1

(3.38)

Solving recursively forward, and assuming that limT ()T Et A T K T +1 /C T = 0)


gives
T
X
K t+1
=
(3.39)
()s Et At+1+s /At .
Ct

Extensions of the Model: Monetary Policy

So far, monetary policy has been described as some type of random process for money
supply. In reality, monetary policy is typically pursued with a purpose: to stabilize inflation or perhaps output, or even to maximize welfare.
With flexible prices, monetary policy can clearly not affect the real variables like
output and consumption. It can, however, affect prices. We see from (3.26) that any
change in Mt makes Pt change to keep Mt /Pt unchanged (recall that Ct is unaffected).
This shows that the central bank can affect prices, and even control them if it can set Mt
after the productivity shock is realized, but that this does not affect utility (none of the
arguments in the utility functionCt , Mt /Pt , and h t is affected).
This is no longer true when there is nominal stickiness, provided monetary policy can
surprise private agents. Note from (3.36) that monetary policy surprise have real effects.
If money supply can be set after the productivity (or demand) shock is observed, then
monetary policy can clearly stabilize output. Too see this, decompose log productivity in
z
t + 1 into its expectation in t and the shock: ln Z t+1 = Et ln Z t+1 + t+1
and do the same
m
for money supply ln Mt+1 = Et ln Mt+1 + t+1 . We can then write (3.36) as
z
m
ln Yt+1 = Et ln Z t+1 + t+1
+ ln Yt + (1 ) t+1
.

(3.40)

Suppose central bank has the policy rule

s=0

This shows how the taste parameter affects the trade-off between consuming today and
investing in capital for future consumption. For instance, if the current taste parameter is
higher than expected future taste parameters, At > Et At+1+s , then K t+1 /Ct is lower than
otherwise. This means that consumption is higher.
Will a temporary shock to At affect output? Yes, it is likely to affect Yt if labour
supply increases (it typically will)and yes, it will also affect Yt+s unless the increased
54

z
m
t+1
= t+1
/(1 ),

(3.41)

ln Yt+1 = Et ln Z t+1 + ln Yt .

(3.42)

then log output is

In this case, output can be perfectly predicted one period ahead (but not two). This certainly affects real variables and utility.
55

3.3

Money and the Business Cycle, by Cooley and Hansen

case letters denote values for a representative household, whereas upper case letters denote aggregates.)

Reference: Cooley and Hansen (1995).


Utility function : E0

3.3.1 Stylized Facts

Corr(xt2 , ln GDPt )
0.63
0.41
0.08
0.03
0.01
0.72

Corr(xt , ln GDPt )
1
0.33/ 0.12
0.37
0.40
0.34
0.52

Corr(xt+2 , ln GDPt )
0.63
0.12
0.33
0.44
0.44
0.17

Some facts:
1. Log money and log velocity (ln V = ln Y + ln P ln M) are procyclical.
(a) Log money peaks before log output (Corr(ln M1t1 ,ln Yt ) > Corr(ln M1t ,ln Yt )
> Corr(ln M1t+2 ,ln Yt )).
(b) Log velocity lags output.
2. The nominal interest rate (short) and inflation rate are also procyclical, but peaks
after output.
3. The log price level is counter cyclical.
3.3.2

t [a ln c1t + (1 a) ln c2t h t ]

t=0

Hodrick-Prescott filtered data (data minus a moving average: cycles longer than 8 years
are virtually eliminated).

Variable (xt ):
ln Yt
ln M1/1 ln M1
ln Velocity
Nominal interest rate:
Inflation:
ln Price level:

Real budget constraint : c1t + c2t + xt +

m t+1
wt
mt
Tt
=
h t + rt kt +
+ .
Pt
Pt
Pt
Pt

Cash-in-advance constraint : Pt c1t = m t + Tt


Production function : Yt = e z t K t Ht1 .
Capital accumulation : kt+1 = (1 ) kt + xt .
Government budget constraint : Tt = 1Mt+1 .
Money supply : 1 ln Mt+1 = 0.491 ln Mt + t+1 , ln t+1 N , known at t.


Log productivity : z t+1 = 0.95z t + t+1 , t+1 N 0, 4.9 105
(Note: it should be Tt /Pt in the real budget constraint; there is a typo in the book.) The
notation is: capital stock (K ), money stock (M), price level (P), wage rate (W ), hours
worked (H ), output (Y ), investment (X ), and productivity (z). Note that the notation
differs somewhat from the model in Benassy (1995): the money stock held at the end of
period t is denoted Mt+1 (Mt in Benassy).
Private consumption consists of a cash good, c1t , and a credit good, c2t . One
interpretation of the trading sequence within a time period t is the following.
1. In the beginning of the period, the household carries over m t from t 1, and gets
Tt is cash transfers from the government. Households also own all physical capital
(kt ). Firms hold no cash or physical capital. The government finances the transfers
by printing new money.
2. Firms rent capital and labor (the rent and wages are paid somewhat later in the
period), and produce goods.

Inflation Tax Model

This is a fairly standard real business cycle model, with some additional features. A
stochastic money supply interacts with a cash-in-advance transaction technology to create some real effects of money supply shocks. The key equations are listed below. (Lower

3. The household buys the cash good with the available cash, where the cash-inadvance restriction Pt c1t m t + Tt must hold. (The log-normal distribution of
the money supply shock t means that the money stock can never decrease, which
is enough to ensure that the CIA constraint always binds: positive nominal interest
rate with probability one.) Firms now hold m t + Tt in cash.

56

57

4. The household receives nominal factor payments wt h t + Pt rt kt from the firms (exhausts all profits), and buys credit goods (Pt c2t ) and investment goods (Pt xt ). The
firms now hold no cash; households own the physical capital kt+1 = (1 ) kt + xt ,
and the cash m t+1 = wt h t + Pt rt kt Pt c2t Pt xt .
5. In equilibrium, the money stock held by the households (m t+1 ) must equal money
supply by the central bank (m t + Tt = Mt+1 ).

this is not reflected by the headers: they should read (x(+5), ..., x, ..., x(5).)
Output is virtually neutral with respect to money supply shocks, even if the composition of aggregate demand is affected by the inflation tax. Intuition: 1 ln Mt expected
inflationexpensive to hold money so households substitute credit goods (and saving,
that is future consumption) for cash goods.
The following table summarizes some properties of data and simulations (simulation
results in parentheses):

Calibration
The parameters in the production function, depreciation, Solow residual, and time preference are chosen as in standard RBC models. The money supply process (for M1) is
estimated with least squares. The a parameter is estimated from the first-order condition
C1t + C2t
Pt (C1t + C2t )
Pt Ct
1 1
=
=
= +
*interest rate,
C1t
Pt C1t
mt

(3.43)

where the paper uses the portion of M1 held by households as a proxy for m t (this differs
from how they estimate the AR(1) for money supply, where they use all of M1). Identifying a from the intercept, they get a = 0.85. (If they had identified from the slope
instead, then they would have got = 0.9.)
To sum up, they use = 0.4, = 0.019, = 0.989, = 2.53, and a = 0.84.
Solving the Model
The inflation tax means that the competitive solution will not coincide with the social
plannerss solution. The solution algorithm is therefore a based on the concept of recursive
competitive equilibrium. Solving a quadratic approximation (in logs) of the model results
in a set of linear decision rules in terms of the state of the economy. Productivity is
stationary (|| < 1), but the money supply is not, so prices will also be non-stationary. It
is therefore very convenient to detrend all nominal variables by dividing by Mt before
the solution algorithm is applied.

Variable (xt ):
ln Yt
ln M1/1 ln M1
ln Velocity
Nominal interest rate:
Inflation:
ln Price level:

Corr(xt2 , ln G D Pt )
0.63 (0.44)
0.41 (?)
0.08 (0.48)
0.03 (0.01)
0.01 (-0.06)
0.72 (-0.06)

The model is simulated 100 times to generate artificial samples of 150 quarters, and the
simulated data is subsequently filtered with the Hodrick-Prescott filter. (Note: Tables 7.37.5 have the x(s) columns in the reverse order compared with Tables 7.1-7.2, even if
58

Corr(xt+2 , ln G D Pt )
0.63 (0.44)
0.12 (?)
0.33 (0.35)
0.44 (0.01)
0.44 (0.04)
0.17 (-0.16)

1. Correlation money growth and real variables very small (except for the correlation
between money and the cash good which is very negative. In data, all correlations
between real variables and money growth are small.
2. Correlation output and interest rate/inflation is completely wrong.
3. Correlation output and velocity is much too strong.
4. For impulse response function, see Cooley Fig 7.6.
3.3.3

A Model with Nominal Wage Stickiness

The wage contract is based on the one-period ahead expectation of the marginal product
of labor. The first order condition for profit maximization is
wt = (1 ) Pt e z t

Results

Corr(xt , ln G D Pt )
1 (1)
0.33/ 0.12 (?/0)
0.37 (0.95)
0.40 (-0.01)
0.34 (-0.14)
0.52 (-0.22)

Kt
Ht

ln wt = ln (1 ) + ln Pt + z t + (ln K t ln Ht ) .

(3.44)
(3.45)

It is assumed (ad hoc) that ln wt is set equal to the expectation of the right hand side of
(3.45), conditional on the information in t 1. Note that K t is in the information set at
59

t 1, while Pt , Ht , and z t are not. The deterministic steady state of the economy with
the this type of wage contracts is the same as in the economy without wage contracts
(simplifies a lot).
The nominal wage is fixed in t 1, and the price level is observed in t. Money
supply shocks may therefore affect the real wage by affecting the price level. Workers are
assumed to supply inelastically at the going real wage (firms are on their labor demand
schedules). A positive money supply shock will decrease the real wage and therefore
increase labor demand and output. As usual, this effect lasts as long as some prices
remain fixed: here it is one period since we have one-period labor contracts. Consumers
(which own both the firms and the labor resources and therefore get all output) choose
to consume only a fraction of the temporary income increase, so most of output increase
spills over to investment (saving).
Results:
The following table summarizes some properties of data and simulations (with the
simulation results in the parentheses).

Variable (xt ):
ln Yt
ln M1/1 ln M1
ln Velocity
Nominal interest rate:
Inflation:
ln Price level:

Corr(xt2 , ln G D Pt )
0.63 (0.20)
0.41 (?)
0.08 (0.17)
0.03 (-0.05)
0.01 (-0.12)
0.72 (-0.27)

Corr(xt , ln G D Pt )
1 (1)
0.33/ 0.12 (?/0.61)
0.37 (0.98)
0.40 (0.48)
0.34 (0.56)
0.52 (-0.01)

Corr(xt+2 , ln G D Pt )
0.63 (0.20)
0.12 (?)
0.33 (0.12)
0.44 (0.01)
0.44 (0.02)
0.17 (-0.06)

1. Much stronger correlation of 1 ln Mt and real variables than in inflation tax model
(at odds with data).
2. Better fit of the contemporaneous correlation of nominal interest rate/inflation and
ln G D Pt , but worsens the correlation of the price level and ln Yt . But the fit for
leads and lags still poor.
3. For impulse response function, see Cooley Fig 7.7.

60

3.4

Sticky Wages or Sticky Prices?

Reference: Romer 5.4, Mankiws comment to Rotemberg (1987).


The distinction between wages and prices is often blurred in models of nominal stickiness. It is still interesting to take a quick look at the different possibilities.
3.4.1

Sticky Wages and Flexible Prices

The combination of sticky wages, flexible prices, and an additional assumption about that
the firms are on their labor demand curve means that workers have a completely elastic
labor supply at the given real wage (see Keynes, Benassy (1995), Cooley and Hansen
(1995)): unemployment/overemployment. Firms hire labor until FL (L) = W/P holds,
so demand shocks lead to counter cyclical real wages (a demand shock increases the price
level and therefore decreases the real wage which makes it profitable to hire more workers;
both P and L increases so FL (L) decreases if F (L) is a concave production function).
However, this implication can be overturned by assuming that the markup of prices
over marginal costs, , is counter-cyclical. Since nominal marginal costs is W/FL (L),
we have FL (L) / = W/P. Consider a fixed nominal wage and an increase in demand.
If the markup is constant, then we have the previous case. However, if is lower in
booms (higher demand elasticities in booms because shopping around or more difficult to
sustain collusion in booms?), then prices need not increase and the real wage is constant.
3.4.2

Flexible Wages, Sticky Prices with Monopolistic Competition

In this case, prices are fixed, but MC > P so firms are willing to supply demand (up
to where MC = P, assume this never happens), so we here assume a completely elastic
supply of goods. Labor demand is then found by inverting the production function L d =
F 1 (Y ). Workers are on their labor supply curve (no unemployment/overemployment),
so the real wage is given by the condition that the labor market clears L d = L s , or
F 1 (Y ) = L (W/P). A demand shock increases Y and L which requires that W/P
increases: demand shocks lead to pro-cyclical real wages.
However, the implication of no unemployment can easily be overturned by adding
real labor market frictions where there is equilibrium unemployment (efficiency wages,
insider-outsiders). A common specification is that there is a real wage function W/P =

61

w (L) which together with labor demand determine a real wage rate, where workers would
be willing to supply more labor.

Bibliography
Benassy, J.-P., 1995, Money and Wage Contracts in an Optimizing Model of the Business
Cycle, Journal of Monetary Economics, 35, 303315.
Blanchard, O. J., and S. Fischer, 1989, Lectures on Macroeconomics, MIT Press.
Cooley, T. F., and G. D. Hansen, 1995, Money and the Business Cycle, in Thomas F.
Cooley (ed.), Frontiers of Business Cycle Research, Princeton University Press, Princeton, New Jersey.
Long, J. B., and C. I. Plosser, 1983, Real Business Cycles, Journal of Political Economy,
91, 3969.
Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT
Press.

Money in Models of Monopolistic Competition

4.1

Monopolistic Competition

Reference: Walsh (1998) 5.3 (See also Romer (1996) 6.6, Blanchard and Fischer (1989)
8.1, and Obstfeldt and Rogoff (1996) 10.1)
Background: Monopolistic competition do not, in itself, lead to a role for monetary
policynominal stickiness does. Monopolistic competition is only a starting point for
discussing price setting behaviour.
4.1.1

Producers of the Final Good

The final good (the good that enters the utility function of agents) is produced by competitive firms. The production function is a CES function (with constant returns to scale) of
a continuum of intermediate goods Y (i), indexed by i [0, 1]
1

Romer, D., 1996, Advanced Macroeconomics, McGraw-Hill.

Yt =

!1
q

, 0 < q < 1.

Y (i) di
q

(4.1)

Rotemberg, J. J., 1987, New Keynesian Microfoundations, in Stanley Fischer (ed.),


NBER Macroeconomics Annual . pp. 69104, NBER.
Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

Example 26 (CES production function with two inputs.) Consider the CES production
function
q
q 1/q
.
y = x1 + x2
We get a linear production function, y = x1 + x2 , if q = 1; a Leontief production function,
1/2 1/2
, y = min (x1 , x2 ), if q = ; and a Cobb-Douglas production function, y = x1 x2 ,
if q = 0.
The price of Y (i) is P (i). Profits of a firm in the final good sector are
= PY

P (i) Y (i) di

Z
=P
0

62

!1
q

Y (i) di
q

P (i) Y (i) di.

(4.2)

63

The first order condition for profit maximization with respect to input i is that the
(nominal) marginal product equals the price
1
P
q

! 1 1
q

Y (i) di
q

qY (i)q1 = P (i)

PY

q( q1 1)

Y (i)

q1

= P (i) /*Y =
q

P (i)
P

Y (i) di*/ or

1
q1

Y.

P (i)
Y (i)
= P (i)
.
(4.8)
Y (i)
Y (i)
Example 27 (Cobb Douglas production function.) Suppose the production function is
Cobb-Douglas
Y (i) = Z K (i)a L (i)b ,
P (i) + Y (i)

Y (i) =

The real unit cost will depend on the productivity level (negatively), and the factor prices
(positively, being homogenous of degree one in factor prices). The first order condition,
with respect to Y (i), for profit maximization is that (nominal) marginal revenues equal
(nominal) marginal costs

(4.3)

Since all producers of the final good have the same production functions and their mass
is one, this is also the (aggregate) demand curve for intermediate good i.
Using the demand equations in the expression for profits, (4.2), we get

which has decreasing returns to scale if 1/ = a + b < 1. The real cost function of
producing Y (i) is
1

= PY

P (i)

P (i)
P
(i)
P

q
q1

Divide by PY to get

=1
PY

1P

1
q1

Y di.

di

(4.4)

(4.5)

Zero profits (there is perfect competition on the market for the final good) means that the
output price, P, must make the right hand side of (4.5) zero, so
"Z
P=

P (i) q1 di

 (i) Y (i)k = Z a+b r a+b


{z
|

w

b
a+b

Y (i) a+b ,

P
}

(i)

where is a constant ( = (a/b)b/(a+b) + (a/b)a/(a+b) ). Marginal cost is increasing


in output if there is decreasing returns to scale, 1/ = a + b < 1.
Invert the demand function (4.3)
P (i) = P

# q1

Y (i)
Y

q1

(4.9)

(4.6)

Use (4.7) and (4.9) in (4.8) to get

This expression defines the price of the final good, which we can think of as a CPI or an
aggregate price level.

P (i) + Y (i) (q 1) P

Y (i)
Y

q1

1
= P (i) Y (i)1 .
Y (i)

(4.10)

Simplify and use (4.3) or (4.9) to substitute for Yi


4.1.2

Producers of the Intermediate Goods


q P (i) = P (i)

Each of the intermediate goods is produced by a monopolist. The profits of firm i are
(i) = P (i) Y (i) P (i) Y (i) , > 1,

(4.7)

P (i)
=
P

"

P (i)
P

#1

1
q1

  q1
q1
(1)(q1)
q
 (i) q Y q .
q

or

(4.11)

where  (i) is the real unit cost function, 1/ is the returns to scale in the production
function (that is, the production function is assumed to be homogenous of degree 1/).

This is the profit maximizing choice of price for firm i. Note that the firm takes the
aggregate price level (the prices of other firms) as given when setting its price.

64

65

4.1.3 Price Setting Behaviour


Firms set their prices taking other prices, P, and aggregate output, Y , as given. Changes
in the aggregate price level are clearly matched one for one. Changes in aggregate output
acts like a demand shifter for firm i, see (4.3). Suppose that Y increases, but that real
unit cost, (i), is unchanged. In this case, firm i will raise its price (the exponent of Y
is positive since 0 < q < 1 and > 1). For instance, if = 3/2 and q = 1/2 then
(4.11) becomes P (i) /P = 31/2  (i)1/2 Y 1/4 , which is increasing in Y . The intuition is
that, with decreasing returns to scale, it is too expensive for the firm to meet an increase
in demand by just increasing output. Instead, both the price and output will be increased.
This feature will be important when we analyze price setting in an environment with some
kind of costs associated with changing nominal prices.
4.1.4 General Equilibrium and Growth
This model of the production side of the economy is easily combined with a model for
household optimization. First, both consumption and accumulation of physical capital
is done in terms of the final good. Second, firms producing intermediate goods make
profits, which must be distributed to the owners (households). Third, households must
be able to buy and sell shares in the firms: we need a market for equity. Fourth, the
standard expressions for the rental rate of capital and wage rate no longer apply since
factor payments do not exhaust the value added.
This model has decreasing returns to scale, which we probably should think of as
a short-term to medium-term feature. There are two ways to fix the model in order to
fit stylized facts of long run data. The mass of intermediate goods could be allowed to
expand over time, as in many models of endogenous growth. Alternatively, we could
change the production function for intermediate firms by adding a semi-fixed production
factor which can be accumulated over time (for instance, some type of physical capital).
There could then be short-run decreasing returns to scale, but long-run constant returns to
scale.
4.1.5 Equilibrium Output with Flexible Prices
Suppose all firms face the same real unit cost,  (i) = . This would, for instance, be the
case if all firms have the same production functions, the same productivity levels and face
66

the same factor prices. Since all intermediate goods enter the production function of the
final good symmetrically, all prices will be equal in equilibrium; the relative price on the
right hand side of (4.11) will be one. This means that we can solve for aggregate output,
Y , in terms of the real unit cost, 
Y =

q 

1
1

 1 ,

(4.12)

which is decreasing in the real unit cost.


It can also be noted that it is increasing in q (assuming we can keep  (i) more or less
unchanged): a larger value of q means that the intermediate goods are closer substitutes
in the production function for the final good (4.1) and the demand curve (4.3) becomes
more elastic. With less monopoly power, output is higher. The flip side of this is that
marginal costs are lower than prices in the intermediate goods industry.
It is clear from (4.12) that money supply can only affect output through real unit
cost, , only. This means that unless monetary policy can affect the supply of labour or
capital, then it cannot affect output. This is the same result as in MIU models with perfect
competition: market imperfections do not imply a role for monetary policy. (Recall, for
instance, that money supply in MIU models is neutral in steady state, and neutral also
along an adjustment path if the utility function is separable in goods and real money
supply.)
4.1.6

Looking Ahead: Nominal Stickiness

In order to generate a really important role for monetary policy, we need some type of
nominal stickiness. One straightforward way is to add nominal wage stickiness as in
Benassy (1995). In this case, (4.12) still holds, and effect of a money supply surprise
comes from decreasing the real unit cost (by decreasing the real wage). Alternatively, we
could add some type of stickiness in prices of intermediate goods. In this case, the profit
maximizing price is no longer (4.11), so (4.12) does no longer hold. We will study such a
case later in the course.
It was earlier argued that intermediate firms will try to raise the price in a boom,
provided the real unit cost does not decrease at the same time. However, (4.12) shows
that the only way we can get a boom is by a decrease in the real unit cost, which seems to
undermine the importance of the argument. It indeed does in an equilibrium with flexible
67

prices, but it need not do so in an equilibrium with (some) sticky prices. For instance,
when (some) prices are sticky and quantities are demand determined (for some reason,
firms post prices in advance and agree to supply whatever is demanded at the price), then
those firm that do not have sticky prices will increase them in booms.

Bibliography

Main references: Romer (1996) (Romer), Blanchard and Fischer (1989) (BF), Obstfeldt
and Rogoff (1996) (OR), and Walsh (1998)

Benassy, J.-P., 1995, Money and Wage Contracts in an Optimizing Model of the Business
Cycle, Journal of Monetary Economics, 35, 303315.
Blanchard, O. J., and S. Fischer, 1989, Lectures on Macroeconomics, MIT Press.
Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT
Press.
Romer, D., 1996, Advanced Macroeconomics, McGraw-Hill.
Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

5.1

Money and Price Setting

Dynamic Models of Sticky Prices

References: BF. 8.2, Romer 6.7, Rotemberg (1987).


This section deals with the effect of price rigidities in dynamic models. Prices are set
in advance and firms are assumed to supply whatever demand happens to be (which is reasonable only as long as demand shocks do not force marginal costs above the price). This
clearly assumes that firms can expand production, for instance, by hiring more labour, so
there must be a fairly elastic factor supply. If factor supply is not particularly elastic, then
marginal costs will increase rapidly so the assumption that marginal cost is always below
the price becomes implausible.
Aggregate demand shocks (or money supply) will usually have real effects when
prices adjust slowly. This is certainly the case when prices are changed with prespecified intervals (time-dependent rules), and the main issue is instead how long the effects
last. It is typically also the case when prices are changed when the old prices are too far
from the frictionless optimum (state dependent rules).
In general, we would like to find a reasonable model which can explain both why
average prices seem to adjust gradually to monetary expansions and why price changes
of individual firms appear to be lumpy. This is hard.
5.1.1

Quadratic Costs of Price-Adjustment

Reference: Rotemberg (1982a), Rotemberg (1982b), and Walsh 5.5.


Firm i is a monopolist on its market and sets the log price, pit , to maximize the value
of the firm: the expected discounted sum of profits. If there were no costs of adjusting
this price, then the price would be equal to some value, pit , which we call the flex price
optimum.
68

69

With costs of adjusting the price we formulate the maximization problem in two steps.
First, find the flex price optimum, pit . Second, minimize the loss from not being at pit
and from incurring adjustment costs. For the moment, we will take the time series process
of pit as given and focus on the second part of the maximization problem. To make any
progress, we also approximate the objective function in the second step by a quadratic
function
min Et

{ pit+s }s=0

min

pit+s pit+s

2

i
+ c ( pit+s pit+s1 )2 or

(5.1)

s=0

{ pit+s }
s=0

pit pit

2

+ c ( pit pit1 )2 + Et pit+1 pit+1

2

o
+ cEt ( pit+1 pit )2 + ... .

The first order condition with respect to pit is


pit pit + c ( pit pit1 ) cEt ( pit+1 pit ) = 0 or

1
Et 1pit+1 +
p pit = 1pit .
c it

(5.2)
(5.3)
pit

There is no lumpiness in individual price changes. Since both deviations from the
and
prices changes are much more costly when they are large (the loss function is quadratic),
the optimal policy will be to converge to pit by taking many small steps rather than a
few large. In a symmetric equilibrium pit = pt and pit = pt . It can also be noted
that situations with a high surprise inflation will lead to a higher pit pit , so the price
adjustment is then faster.
The smooth individual price changes carry over to the average prices, since all firms
are similar. Let pit = pt and pit = pt be the common prices and write (5.3) as
1pt = Et 1pt+1 +


1
pt pt .
c

(5.4)

Special Case: No Adjustment Cost (c = 0)


If c = 0, then (5.2) shows that pit = pit , so the firm will always set its actual price equal
to the unrestricted optimal price (quite obvious since the price is then unrestricted).
5.1.2

In contrast to the model with quadratic adjustment costs, this model has lumpy individual price changesbut it gives the same evolution of the average price level (shown
below).
Prices are changed at exogenous random points in time only. The fraction q of the
firms are allowed to set a new price in a period, and the fraction 1 q must keep their old
price. As in the model with quadratic adjustment costs, the model is set up in two steps.
First, find the flex price optimum, pit . Second, minimize the loss from not being at pit ,
taking into account the firm only gets a time to change its price at random occasions.

Let pit+s
be the flex price optimum of firm i in period t + s, and pit the actual price
set in t (and perhaps still valid in future periods if no opportunity to change the price has
occurred). The frequency of price changes is, unrealistically, assumed to be constant.
Suppose firm i gets to change its price in t. The minimization problem in t is to choose
the price, pit , to minimize the expected discounted loss of being away from the optimal
price. The loss function is once again approximated by a quadratic function

Calvos Model

Reference: Rotemberg (1987) and Walsh 5.5.


70

L t = Et

s p t+s pit+s

2

(5.5)

s=0

where p t+s denotes actual price in t + s, which might have been set in the same or some
earlier period.
This loss function is potentially problematic since p t+s has a mixture (both discrete
and continuous parts to it) distribution, which can make it complicated to calculate the

expected value (note that p t+s and pit+s


are likely to be correlated in a complicated way).
However, we can simplify the problem by noting that all we care about are the terms that
involve p t+s = pit . First, note that the next time the firm will get to change its price is
in the random time t + 1 , the second time in t + 2 , and so forth. The probability that it
does not get to change the price in t + 1 is 1 q, that it does not get to change in either
t + 1 or in t + 2 is (1 q)2 , and so forth. This means that Pr( p t+s = pit ) = (1 q)s .
R
)2 =

2
Second, note that Et ( p t+s pit+s
 ( p t+s pit+s ) d F( p t+s , pit+s ), where F is
R
the joint distribution function and  be understood as a Stieltjes integral over , which
}. We can split up  into two subsets: one
is the space of values of the pair { p t+s , pit+s
where p t+s = pit and the rest. Clearly, only the first subset matters for the first order

condition with respect to pit . This subset has the probability (1 q)s and pit+s
has some
distribution (only the first moment will be important).
71

We therefore rewrite the loss function (5.5) as


Lt =

s (1 q)s Et pit pit+s

2

+ terms not involving pit ,

(5.6)

s=0

where pit is the price set in t (and still in effect in t + s with probability (1 q)s ). (An
alternative, and more careful, derivation is given in Appendix B.)
The first order condition with respect to pit is

s (1 q)s pit Et pit+s


= 0.

(5.7)

pit = pt




1
1q
1
1q
pt
pt1 = [1 (1 q)] pt + (1 q) Et
pt+1
pt , or
q
q
q
q
(5.13)

q

(5.14)
1pt = Et 1pt+1 +
[1 (1 q)] pt pt .
1q
This is of the same form as (5.4), but where the coefficients have different interpretations.
Special Case: Flexible Prices (q = 1)

s=0

Since

s=0

If q = 1, so all firms get to change their prices in every period, then (5.10) becomes

(1 q)s = 1/[1 (1 q)], we get

pit = [1 (1 q)]

pit = pit ,

s (1 q)s Et pit+s

so prices are set equal to the flex price optimal.

s=0

= [1 (1 q)] pit + [1 (1 q)] (1 q)

(5.15)

(5.8)

s (1 q)s Et pit+1+s
. (5.9)

s=0

This can be rewritten as a forward looking difference equation in pit by noting that the
last term in (5.9) equals (1 q)Et pit+1 , see (5.8). (This is the price they would set in
t + 1 if they get to change the price then.) This gives
pit = [1 (1 q)] pit + (1 q) Et pit+1 .

(5.10)

If all firms are similar, then pit and pit are the same for all firms. The aggregate
price level is the weighted average of those who get to change the price and those who
did not. (This can be seen as an approximation to a true CPI. For instance, if the utility
function/production function for the final good is CES, then the exact CPI is also a CES
function.) The latter have, on average, a price equal to the price level in t 1 (since the
draws are independent over time). The average price level is therefore
pt = q pit + (1 q) pt1 , so
1q
1
pt1 (if q > 0).
pit = pt
q
q

Average Time between Price Changes


Price changes follows a Poisson process, so time to a price change has an exponential
distribution.(The exponential distribution is q exp (q ) for > 0, with E = 1/q and
Var( ) = 1/q 2 .)
5.1.3

Other Ways to Model Price Stickiness

It is straightforward (but a bit tedious) to model multi-period price contracts. It could,


for instance, be a case where half the firms set prices for two periods in every odd time
period, and the other half in every even period (staggered price setting).
The Fischer model allows the contract to stipulate different (but predetermined) prices
for the different subperiods of the contract. The Taylor model is similar, but has the same
price for all subperiods (fixed prices), which gives more price inertia.

(5.11)

5.1.4

(5.12)

What is the unrestricted optimal price, pit , which plays such an important role in the
previous model? A typical formulation is that it represents a monopolists price in a flexprice equilibrium. That price is typically an increasing function of aggregate demand and

Use this to substitute for pit and pit+1 in (5.10) and let all firms have the same flex price,
72

Price Setting in a Model with Monopolistic Competition

73

a decreasing function of the productivity level. In logs, we write


pt = pt + yt + t ,

(5.16)

where t is interpreted as the negative of a productivity shock (negative supply shock).


Note that > 0. It is typically increasing in slope of the marginal cost curve (the degree
of decreasing returns to scale) and decreasing in the elasticity of substitution between
goods in consumer preferences. In most models, we need an upward sloping marginal
cost curve to get > 0, which could be motivated by some fixed factors of production.
If these fixed factors are not completely fixed, but can be accumulated over time, then
the problem becomes more complicated (dynamic) and (5.16) can only be interpreted as
an approximation that might be valid for short to medium run horizons (a business cycle,
say).
Using (5.16) in (5.4) gives
1
(yt + t ) .
c

(5.17)

q
[1 (1 q)] (yt + t ) .
1q

(5.18)

1pt = Et 1pt+1 +

As in any Phillips curve, it appears as if inflation is a real phenomenon! This is quite


the opposite to the Cagan model, where it is assumed that both output and the real interest
rate are constant. This suggests that this model of price setting is certainly not suitable for
understanding a permanent change in the money supply trend. It is not plausible that the
model parameters, for instance q and c, would remain unchanged in such a case.
5.1.5

Closing the Model: The Demand Side

We need a model for yt to illustrate how this price setting works. Suppose we have
macro model with money in the utility function/cash in advance, optimizing households,
imperfect competition with (at least temporarily) decreasing returns to scale, and costs of
changing nominal prices. We also add the assumption that firms with sticky prices agree
to sell any quantity demanded at the prevailing price.
In this setting, (5.19) describes how firms set prices. The rest of the model would be
something like the following. Money demand is
m t pt = yt i t .

(5.21)

Similarly, using (5.16), in (5.13) gives

1pt = Et 1pt+1 +
We write both these equations as

(5.19)

which can be thought of as an expectations-augmented Phillips curve. It is in an sense


similar to the Keynesian AS curve, which has positive relation between output and the
price level.
Recursion forward gives

= (1 + rt+1 )Et Ct+1 , which is approximately

ct = Et ct+1 + rt+1 or rt+1 = (Et ct+1 ct ) ,

1pt = Et 1pt+1 + (yt + t ) ,

1pt =

The Euler equation for consumption is Ct


equal to the following (in logs)

s Et (yt+s + t+s ) ,

(5.20)

s=0

provided lims s+1 Et 1pt+s = 0. Note that Et yt+s has a large effect on inflation is is
high (strong decreasing returns to scale and/or strong market power), and Et (yt+s + t+s )
has a large effect if is high (small c in (5.17) or large q in (5.19)).

74

(5.22)

where rt+1 is the real interest rate. If assume that consumption is approximately proportional to output, then (5.22) can be written
rt+1 = (Et yt+1 yt ) + constant.

(5.23)

i t = Et ( pt+1 pt ) + rt+1 .

(5.24)

The Fisher equation is

Finally, we have to add an assumption about what the central bank uses as its instrument:
m t or i t . This is the whole model. We could clearly add more features, for instance, a
labour supply decision (must be elastic to let the firms expand output), but we disregard
that in order to keep the model as simple as possible.
Consider the case when the central bank sets m t . When money is interest rate elastic,

75

6 = 0, then (5.19) and (5.21) are no longer sufficient to close the model, since the
nominal interest rate remains to be determined. From the Fisher equation (5.24) we see
that determining the nominal interest rate requires both inflation (already captured by
(5.19)) and the real interest rate. We therefore need a model of the real interest rate, for
instance, (5.23), to close the model. In contrast, when = 0, then we do not need a
model of the real interest rate in order to solve for the equilibrium price and output. Of
course, we can always plug in the equilibrium process for yt in (5.23) to calculate the
implied real interest rate, but we do not have to.
Now, consider the case when the central bank sets i t . In this case, we do not need the
money demand equation (5.21) for determining the price and output. Instead, we combine
(5.19) with (5.23) and (5.24) to give us a model in terms of the price, output, real interest
rate, and the nominal interest rate (set by central bank). Of course, we can always plug in
the equilibrium values in (5.21) to calculate money demand, but we do not have to.
5.1.6

sticky price model. Substitute for yt in (5.19) by using (5.25)


1pt = Et 1pt+1 + (m t pt ) + t
pt1 + pt (1 + + ) Et pt+1 = (m t + t ) .

This is a second-order expectational difference equation, which can be solved with a


variety of methods. The perhaps most straightforward one is to specify a time-series
process for the exogenous driving process, and transform the system to a vector first-order
system and then use a decomposition of the resulting matrix to decouple the variables in
those that are predetermined in t (typically the exogenous variables and values determined
in previous periods like the capital stock and lagged variables) and those that can jump
in t in response to changes in expectations about future values (typically asset prices and
anything else that depend on expected future values).
A trivial step is to note that (5.26) can be rewritten
1
1 + +

Et pt+1 = pt1 + pt
(m t + t ) .

Example: Calvo Model in a Very Simple Macro Model

For simplicity, assume that the quantity equation holds. In logs we have
m t = pt + yt .

(5.26)

(5.27)

Suppose t = 0 and that m t is an AR(1)


(5.25)

This can be taken to represent aggregate demand. Aggregate supply is represented by


the price setting rule, and it is assumed that firms supply whatever the market demands
at the going price: output is demand determined. In traditional monetarist models, the
quantity equation is aggregate demand, without much discussion of where it comes from.
In a Keynesian model, the quantity equation would be an approximation to the Keynesian
AD curve (the combination of the IS and LM curves which traces out the relation between
output and prices). Both these interpretations assume a negative relation between the price
level and output. In some modern dynamic general equilibrium models, the quantity
equation can be shown to be the money demand equation (see, for instance, Benassy
(1995)).
We now use this very simple model of demand to illustrate some properties of the

76

m t = m t1 + mt .

(5.28)

We can then write the model on state space form as


0
0
m t+1
mt
mt+1

0
1
pt
= 0
pt1 + 0
.
1 1++
Et pt+1
p
0
t

(5.29)

Some impulse response functions (dynamic simulations obtained from setting mt = 1


in t = 0 but zero in all other periods) are shown in Figure 5.1. In Figure 5.1.a, price adjustment is fairly slow (many prices are fixed in spite of an increase in nominal demand),
so a monetary shock leads to a relatively large effect on output: money is far from neutral.
In Figure 5.1.b, price adjustment is much faster (the rate at which an occasion to change
the price arrives is much higher), so the monetary shock has almost no effect on output:
money is almost neutral. In Figure 5.1.c also has fat price adjustment, but now because
is high (quickly decreasing returns to scale or strong monopoly power), which makes it
too costly for firms to keep their old prices.
77

on which type of policy experiments which are meaningful to analyze with the help of
this model: we should probably only use this model for policy changes which keeps the
average inflation rate unchanged. In many applications, the Phillips equation is assumed
to refer to detrended output (as a measure of the business cycle). The main reason is
that the Phillips effect is typically only relevant for as long as the production function
has decreasing returns to scale, see the discussion of (5.16). Since detrended output per
definition has a zero mean the kind of experiments that changes Eyt must be ruled out.

b. Frequent price adjustments, q=0.99


1

a. Baseline model
1

0.5

0.5
money
price
output

0
2

2
4
period

0
8

2
4
period

5.1.8
c. Prices sensitive to demand, =2
1

Relation between output and unemployment: Okuns law, 1 ln Yt = 31u t .

Calvo model, response to money supply shock


Parameter values (base line):
=0.95, =3/7, q=0.875, =0.96

Phillips curve unstable, in particular if not accounting for Et t+1 . (Burda and
Wyplosz (1997) Figs. 12.4 and 12.6)

0.5

OR Box 10.2 (substantial evidence of imperfect competition from micro data of 50


US industries)

0
2

Empirical Illustration

2
4
period

OR Box 10.1 (12-18 months between price changes on selected items in US mail
catalogues)

Figure 5.1: Impulse responses in the Calvo model

Rotemberg (1982b) finds that for Germany the average time between price changes
was 4 quarters. It must therefore be the case that 1/q = 4, so q = 1/4. He reports
even lower numbers for the US.

5.1.7 The Calvo Model and the Natural Rate Hypothesis


Reference: Walsh 5.5.
The natural rate hypothesis states that the mean of output cannot be affected by any
monetary policy. Suppose the central bank can change the inflation rate by changing its
policy instrument. Take the unconditional expectation of the Calvo model (5.19) and use
iterated expectations and Et = 0 to get
Eyt =

E1pt E1pt+1
.

(5.30)

If = 1 ( < 1), and inflation is a stationary series so E1pt = E1pt+1 , then this
means that inflation cannot (can) affect average output. Irrespective of whether = 1 or
not, a drifting inflation rate (E1pt 6=E1pt+1 ) can certainly affect average output.
This should probably be regarded as an artifact of the Calvo model. It puts restrictions
78

Roberts (1995) estimates




1pt Et 1pt+1 = co + yt +c1 1 log real oil pricet +c2 1 log real oil pricet1 +t ,
where 1pt is measured by the percentage increase December to December in CPI,
yt GDP detrended with a deterministic trend (time and time squared), and Et 1pt+1
is approximated by inflation expectation surveys. The sample is annual and covers
1949-1990. The results are 0.3. If we use this together with q = 1/4 (from
Rotemberg (1982b)) and = 1 in (5.18), we get 3.6, so a one percent increase
in aggregate demand drives up the desired relative price of a monopolist with 3.6
percent.

79

5.2

Aggregation of One-Sided Ss Rule: A Counter-Example to 1M


1Y

[s, S] with pdf


1
.
(5.35)
Ss
What is the effect on real balances of an infinitesimal increase in money stock dm.
For all firms which have p p [s, S), that is, below the point which triggers a price
change, all firms are simply shifted up in parallel. For all firms which are moved to the
trigger point ( p p = S), we see an immediate adjustment of the price to, so p p = s.
These two movements completely offsets each other, so the average p p is unchanged.
If supply depends on real balances, then output is unaffected by the change in money
supply.
Critical assumptions and discussion. The initial distribution is uniform and shocks
are small. The general contribution of the model is to highlight the importance of the distribution of price/frictionless price. One could envision a situation where a large fraction
of firms are close to adjusting there price upwards. I small (but not infinitesimal) change
in money supply can then trigger a disproportionate jump in the average price level, so
that a monetary expansion leads to a recession.

f p p =

Reference: BF. 408-414, Romer 273-276.


Basic point: individual price stickiness does not necessarily imply aggregate price
stickiness.
If there where no adjustment costs the optimal nominal price of a firm i is
pit = m t .

(5.31)

m t m t1t ,

(5.32)

Assume that m t is non-decreasing, so

which makes the frictionless optimum pit drift upwards. Suppose the actual price set by
the firm is pit , and that there is a menu cost for changing the price. It can be shown that a
one-sided Ss rule is optimal
(
pit1t if pit1t > pit S, S > 0
pit =
,
(5.33)
pit s else, s < 0.
where S and s depend on the drift and the menu cost.
This rule implies that the nominal price is unchanged as long as the frictionless optimum p is not much larger that the actual price. At a price change, the actual price is
set somewhat higher than the frictionless price pit = pit s, s < 0, in expectation of
increases in pi (m t drifts upwards all the time).
Consider an example. We start with p0 = p0 , and then pt increases until it reaches
p0 + S in period . At that time the actual price is changed to
p = p s
= p0 + S s /*price is changed when p = p0 + S*/
> p0 + S /*since s < 0*/

(5.34)

A
A.1

Summary of Solution Method for Linear RE Models


Summary

The model is

"

x1t+1
Et x2t+1

"
=A

x1t
x2t

"
+

t+1
0

#
,

(A.1)

where x1t is an n 1 1 vector of predetermined variables, x2t is an n 2 vector of forward


looking variables, and t is a white noise process. All dynamics of the exogenous processes have been placed in x1t . A necessary condition for a saddle path equilibrium is that
A has as many unstable roots (inside unit circle) as there are elements in x1t .
Decompose A as
A = Z T Z 1 ,
(A.2)

The deviation p p moves within the band [s, S].


In general, this type of setup implies a uniform distribution of p p over the band

where T is (at least) upper block diagonal. Note that we require Z to be invertible. In
some cases we could let T be a diagonal matrix with eigenvalues along the principal
diagonal and with the corresponding eigenvectors in the columns of Z (if the eigenvectors

80

81

are linearly independent).


This decomposition should be reordered so that the blocks corresponding to the stable
eigenvalues (in or on the unit circle) comes first. Partition conformably with the stable
and unstable roots
#
"
#
"
T T
Z k Z k
T =
and Z =
.
(A.3)
0 T
Z Z
The solution can then be shown to be
1
x1t+1 = Z k T Z k
x1t + t+1

x2t =

A.2

We assume that a and b are such that |1 | < 1 but |2 | > 1.


Example 28 a = 1.1 and b = 0.3 gives 1 0.23 and 2 1.323.
In this case, (A.4) gives
xt = 1 xt1 ,

since Z k = 2 /b and T = 1 .
It is straightforward to show that (A.11) satisfies (A.6). From (A.11), xt = 1 xt1 ,
so xt+1 = 1 xt = 21 xt1 , so (A.6) becomes

(A.4)

1
Z Z k
x1t .

21 a1 b,

which should be zero. Substituting for a and b from (A.10) immediately shows that this
is the case.
We now generalize the scalar difference equation to

Reference: Romer 6.8, BF Appendix to chap 5, Hamilton (1994) 2.3.


Consider the homogenous scalar second order difference equation

Et xt+1 axt bxt1 = Et z t ,

xt+1 axt bxt1 = 0.

(A.6)

We can write the difference equation as a system of first order equations as in (A.1)
(with t = 0 for all t)
"
# "
#"
#
xt
0 1
xt1
=
.
(A.7)
xt+1
b a
xt
The matrix can be decomposed in terms of eigenvalues and eigenvectors. It can be shown
that the decomposition is

where

0 1
b a

"
=

2 /b 1 /b
1
1
"

1
2

#"

"
=

1 0
0 2

1
2a
1
2a

#"

2 /b 1 /b
1
1

(A.12)

(A.5)

Special Case: Scalar Second Order Equation

"

(A.11)

1
2
2 a
1
2
2 a

+ 4b
+ 4b

#1
,

xt = 1 xt1



1 X 1 s
Et z t+s .
2
2

(A.14)

s=0

We now demonstrate that this solution indeed satisfies (A.13). First, lead (A.14) ones
period, take expectations as of time t and then use this to substitute for Et xt+1 in (A.13).
"
#


1 X 1 s
Et z t = 1 xt
Et z t+s+1 axt bxt1 .
(A.15)
2
2
s=0

Second, note that

(A.9)





X
1 X 1 s
1 s
Et z t+s+1 =
Et z t+s Et z t ,
2
2
2
s=0

which we use in (A.15)


"

and where the eigenvalues satisfy


a = 1 + 2 and b = 1 2 .

where z t is some exogenous process.


It can then be shown that the solution is

(A.8)

#
,

(A.13)

(A.10)

Et z t = 1 xt

(A.16)

s=0

#


X
1 s
Et z t+s + Et z t axt bxt1 .
2

(A.17)

s=0

82

83

Third, subtract Et z t from both sides and use (A.14) to substitute for xt in (A.17)
0 = (1 a) 1 xt1 (1 a)

1
2


X
s=0

1
2

s
Et z t+s


X
s=0

1
2

s

Factor the polynomial within parenthesis as


1 aL bL2 = (1 1 L) (1 2 L)

Et z t+s bxt1 .

(A.18)
From (A.10) we know that (1 a) /2 = 1, so the terms involving Et z t+s cancel.
Since (1 a) 1 = b also the terms involving xt1 cancel (bit we already knew that
from the homogenous equation).
Example 29 (z t is AR(1).)

z t+1

xt
Et xt+1

Suppose z t+1

0

= 0 0
1 b

xt

= 1 (1 + 2 ) L + 1 2 L , so

(A.22)

a = 1 + 2 and b = 1 2 , with
a 1p 2
1,2 =
a + 4b.
2 2

(A.23)

L1 (1 1 L) (1 2 L) Et xt = Et z t


(1 1 L) L1 2 Et xt = /*multiply in L1 */


1
1
(1 1 L) 1 L1 Et xt = Et z t /* divide both sides with 2 */
2
2

and we could employ the spectral decomposition to solve this problem. Alternatively, note
that the AR(1) for z t gives Et z t+s = s z t , so (A.14) can be written

(A.24)

Assume |1 | < 1 and |2 | 1 (must be shown for each model). Use (A.22) in (A.19)

= z t + t+1 . The state space form is then

0
zt
t+1

1 xt1 + 0
,
a

(A.21)
2

(A.25)

Apply the operator



1

1
1
zt .
xt = 1 xt1
2 1 /2

1 1
L
2

1
=1+



1 1
1 1 2
L +
L
+ ...
2
2

(A.26)

on both sides of (A.25)

A.3

An Alternative for the Scalar Second Order Equation: The Fac-

!


1 1
1 1 2
L +
L
+ ... Et z t
2
2
s

X
1
1
xt = 1 xt1
Et z t+s
2
2

(1 1 L) Et xt =

torization Method
Reference: Romer 6.8, BF Appendix to chap 5.
The scalar difference equation is

Et xt+1 axt bxt1 = Et z t




1 aL bL2 Et xt = Et z t

1
2

1+

(A.27)

s=0

(A.19)
(A.20)

where z t is some exogenous process. The lag operator (L) affects the dating of the variable, but not of the expectations operator: L1 Et xt =Et xt+1 , Et xt = xt , LEt xt =Et xt1 =
xt1 .

Calvos Model: An Alternative Derivation

The actual price p t+s will then be one of pit , ..., pit+s , depending on the realization of
some random variable. In particular, let us write
p t+s =

s
X

I (vt+s ) pit+ ,

(B.1)

=0

where vt+s is a random variable (note that there is a different random variable for each
t + s) and I an indicator function. For a given realization of vt+s , one of the functions
84

85

I0 , ..., Is is unity and all other are zero. For instance, if I2 is unity, then the actual price in
t + s (assuming s 2) equals the price set in t + 2. We can then write (5.5) as
L t = Et

X
s=0

s
X

!2

I (vt+s ) pit+ pit+s

Hamilton, J. D., 1994, Time Series Analysis, Princeton University Press, Princeton.


2

= pit
I0 (vt+1 ) pit + I1 (vt+1 ) pit+1 pit+1
+

2
2

Et I0 (vt+2 ) pit + I1 (vt+2 ) pit+1 + I2 (vt+2 ) pit+ pit+2 + ...

Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT


Press.
Roberts, J. M., 1995, New Keynasian Economics and the Phillips Curve, Journal of
Money, Credit, and Banking, 27, 975984.
(B.3)

=0

s=0

Romer, D., 1996, Advanced Macroeconomics, McGraw-Hill.





= pit pit + Et I0 (vt+1 ) pit + I1 (vt+1 ) pit+1 pit+1


I0 (vt+1 ) +


2

Et I0 (vt+2 ) pit + I1 (vt+2 ) pit+1 + I2 (vt+2 ) pit+ pit+2 I0 (vt+2 ) + ...

Rotemberg, J. J., 1982a, Monopolistic Price Adjustment and Aggregate Output, Review
of Economic Studies, 49, 517531.

Since I (vt+s ) I0 (vt+s ) = 0 for 6 = 0


Et

s I0 (vt+s ) pit pit+s


I0 (vt+s ) = 0.

Rotemberg, J. J., 1982b, Sticky Prices in the United States, Journal of Political Economy, 60, 11871211.
(B.4)

s=0

Now use the facts that Et I0 (vt+s ) =Et I0 (vt+s ) I0 (vt+s ) = Pr (I0 (vt+s ) = 1), pit is

known in t, pit+s
and I0 (vt+s ) are independent to get
Et

Burda, M., and C. Wyplosz, 1997, Macroeconomics - A European Text, Oxford University
Press, 2nd edn.

(B.2)

=0
2
pit + Et

The first order condition with respect to pit is then


!

s
X
X
s

0 = Et

I (vt+s ) pit+ pit+s I0 (vt+s )

Blanchard, O. J., and S. Fischer, 1989, Lectures on Macroeconomics, MIT Press.

s Pr (I0 (vt+s ) = 1) pit Et pit+s


= 0.

Rotemberg, J. J., 1987, New Keynesian Microfoundations, in Stanley Fischer (ed.),


NBER Macroeconomics Annual . pp. 69104, NBER.
Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

(B.5)

s=0

Finally, recall that Pr (I0 (vt+s ) = 1) = (1 q)s , which gives


Et

s (1 q)s pit Et pit+s


= 0.

(B.6)

s=0

Bibliography
Benassy, J.-P., 1995, Money and Wage Contracts in an Optimizing Model of the Business
Cycle, Journal of Monetary Economics, 35, 303315.

86

87

Poole (1970), Mishkin (1997) 23) Suppose the goal of monetary policy is to stabilize
output. The central bank must set its instrument (either m t or i t ) before the shocks have
been observed. Which instrument should it choose? If i t is kept fixed, then

Monetary Policy

Main references: Romer (1996) (Romer), Blanchard and Fischer (1989) (BF), Obstfeldt
and Rogoff (1996) (OR), and Walsh (1998).

6.1

The IS-LM Model

yt + yt
,

(6.1)

where yt is a real (demand) shock. The LM curve (in logs) is


m t pt = yt i t + mt i t =

since the money demand shocks are not allowed to spill over to output, and the interest
rate is not allowed to cushion real shocks. If m t is kept fixed, then
dyt
1
dyt
1
=
< 0 and
=
< 1, (m t fixed)
dmt
/ +
d yt
1 + /

Reference: Romer 5, BF 10.4, and King (1993).


The IS curve (in logs) is
yt = i t + yt i t =

dyt
dyt
= 0 and
= 1, (i t fixed)
dmt
d yt

yt + mt m t + pt
,

(6.2)

where mt is a money demand shock. Consider fixed prices, which amounts to assuming a
perfect elastic aggregate supply schedule: income is demand driven, which is the opposite
to RBC models where income is essentially supply driven. Increasing m t lowers the
interest rate, which increases output. An outward shift in the IS curve because of an
increase in yt , increases both output and the nominal interest rate.
The most important problem with this model is that there are no supply-side effects,
that is, prices are fixed. As a logical consequence, the IS curve is written in terms of
the nominal interest rate, which differs from the real interest rate by a constant only.
At a minimum, this model need to be amended with a model for prices (and thus price
expectations), and also a term Et 1pt+1 in the IS curve to let demand depend on the ex
ante real interest rate.
The IS-LM framework has, in spite of these problems, been used extensively to discuss many important monetary policy issues. The following examples summarize two of
them.
Example 30 (Monetary Policy: Interest Rate Targeting or Money Targeting? BF. 11.2,
88

since money demand shocks now increase the nominal interest rate and thereby decreases
output, but the real shocks are cushioned by the increase in interest rates. Pooles conclusion was that interest rate targeting is preferred if most shocks are money demand shocks,
while money stock targeting is better if most shocks are real. This is illustrated in Figure
6.1.
Example 31 (The Mundell-Flemming Model and choice of exchange rate regime, Reference: OR 9.4, Romer 5.3, and BF 10.4) Add a real exchange rate term to the IS curve
(6.1)

yt = i t + st + pt pt + yt

yt + yt + st + pt pt
,
it =

and let asset market equilibrium be given by the UIP condition


i t = i t + E1st+1 .
Assumptions: fixed prices, foreign and domestic goods are imperfect substitutes, foreign
and domestic bonds are perfect substitutes. Assume also that E1st+1 = 0 so i t = i t (this
does, of course, allow st to changeand makes a lot of sense if all shocks are permanent).
If m t is fixed, so the exchange rate is floating (set m t = 0, for simplicity), then the LM

equation gives i t = ( yt + mt ) / or yt = i t mt / (since i t = i t ) so
dyt
1
dyt
= < 0 and
= 0 (m t fixed, st floating).
dmt

d yt
89

to accommodate the extra money demand to keep the exchange rate fixed (that is, the
output shock is not allowed to increase the nominal interest rate). A fixed exchange rate
(or a currency union) means that the country abandons the possibility to use monetary
policy to buffer country specific real shocks (a common real shock among the participating
countries can be buffered), but all money demand shocks are buffered. The extent of
country-specific shocks is a main determinant behind optimum currency areas (the other
is the degree of factor mobility). The conclusion from this analysis is that a floating
exchange rate is better at stabilizing output if real shocks dominate, while a fixed exchange
rate is better if money demand shocks dominate.

a. Interest rate targeting


Real shock

Money demand shock

IS

LM

b. Money stock targeting


Money demand shock

Real shock

6.2

The Barro-Gordon Model

6.2.1

The Basic Model

References: Walsh 8, OR 9.5, BF 11.2 and 11.4, and Romer 9.4 and 9.5.
Use the LM curve (6.2) in the IS curve (6.1) to derive the aggregate demand curve
ytd =
y

yt .
(m t pt mt ) +
+
+

(6.3)

For simplicity, merge mt + / ( + ) yt into a composite demand shock, td ,

Figure 6.1: Pooles analysis of different monetary policy instruments in an IS-LM model.
The real shock is a positive aggregate demand shock, and the money demand shock is a
positive shock to money demand.
A money demand shock has a negative effect on output (similar to a closed economy
model), while a real shock has not (different from a closed economy model). yt cannot
increase unless m t , i t or mt does. If they do not, then any real shock must simply spill
over into an exchange rate appreciation. If the exchange rate is fixed, say st = 0, then the

IS equation gives i t = yt + yt / or yt = i t + yt (since i t = i t ) so
dyt
dyt
= 0 and
= 1 (st fixed).
dmt
d yt

ytd =

(m t pt ) + td .
+

(6.4)

This is a very common formulation of aggregate demand; it shows up in Lucas model


of the Phillips curve, and also in several monetary models with monopolistic competition
(see, for instance, BF 8.1). Note, however, that if the IS curve depended on the ex ante
real interest rate instead of the nominal interest rate, then a term Et 1pt+1 /( + )
is added to (6.4).
We now also introduce an aggregate supply side inspired by Lucas version of the
Phillips curve or by a model with predetermined prices (or long nominal contracts)


e
yts = b pt pt|t1
+ ts ,
(6.5)


e
= b t t|t1
+ ts /* pt1 */
(6.6)

All shocks to the LM curve must be accommodated by corresponding changes in m t to


keep st fixed. Any real shocks feed right through, since the money stock is expanded

e
where pt|t1
is the log price level in t which private agents expect based on the informa-

90

91

e
e
e
tion in t 1, and t|t1
is the corresponding expected inflation rate, t|t1
= pt|t1
pt1 .
e
Let expectations be rational, so t|t1 in (6.6) is the mathematical expectation
e
t|t1

= Et1 t .

(6.7)

To simplify the algebra we note that the central bank can always generate any inflation it
wants by manipulating the money supply, m t . We therefore treat inflation t as the policy
instrument (the required m t can be backed out from the equilibrium).
The loss function of the central bank is
Lt =

t2

+ (yt y ) ,
2

(6.8)

so the central bank want to stabilize inflation around its natural level (normalized to zero),
but output around y , which may be different from the natural level (once again normalized
to zero). The target level for output, y , is typically positiveperhaps the natural level of
output (zero) is not compatible with full employment (due to labour market imperfections)
or because the natural level of output is affected by product market imperfections. Using
monetary policy to solve such imperfections is probably not the best idea; in this model,
it will not even work.
The central bank sets the monetary policy instrument after observing the shock, ts .
(This is different from the two examples given at the beginning of this note, where policy
had to be set before the shocks were realized.) In practice, monetary policy can react
quickly, although perhaps not completely without a lag. However, the main point in this
analysis is that the monetary policy can react more quickly than the private sector (price
and wage setters). This is probably a realistic assumption. This opens a channel for
monetary policy to have effect.
6.2.2

that the policy rule is on the form


t = + ts + td ,

(6.9)

where we have to find the values of , , and . The publics expectations must be
e
t|t1
= Et1 t

= ,

(6.10)

provided the shocks are unpredictable. Note that is not determined yet. The idea is that
whatever value of that the central bank would happen to choose, the public knows it and
will adjust their expectations accordingly. This means that the central bank can influence
the publics expectations and that it makes use of this in the optimization problem.
Using the supply function (6.6) and (6.9)-(6.10) in the loss function (6.8), and taking
expectations as of t 1 gives the optimization problem
Et1 L t = Et1 + ts + td

2



2
+ Et1 b + ts + td + ts y . (6.11)

The first order condition with respect to gives


= 0.

(6.12)

The first order condition with respect to is


2dd + 2b2 dd = 0 or = 0,

(6.13)

provided the shocks are unpredictable and also uncorrelated, Et1 td ts = 0. Finally, the
first order condition with respect to is then
2ss + 2b (b + 1) ss = 0 or

Monetary Policy with Commitment

In the commitment case, the central bank chooses a policy rule in t 1 and precommits
to it. It will therefore choose a rule which minimizes Et1 L t . Since the model is linearquadratic, we can assume that the policy rule is linear. Since only innovations can affect
output we can safely restrict attention to policy rules in terms of a constant (there is no
dynamics in the model) and the shocks. We therefore assume (correctly, it can be shown)

92

b
.
1 + b2

(6.14)

The policy rule (6.9) is therefore


t = ts ,

(6.15)

93

with given by (6.14). Output is then


yt = (b + 1) ts .

(6.16)

If the central bank targets inflation only, = 0, then = 0, which by (6.15) and (6.16)
means that inflation is completely stable and that output shocks are not cushioned. Conversely, if the central bank targets output only, , then = 1/b (apply lHopitals
rule) so output is now completely stable, but inflation varies.
More generally, note that

All parameters are positive. A positive shock to td increases both output and price proportionally, so a decrease in m t can stabilize the effects completely. This can also be seen
directly from (6.4). In contrast, a positive shock to ts increases output but decreases the
price. Since the effect of m t on output and price has the same sign, the central bank cannot use monetary supply to stabilize both when the economy is hit by a supply shock. If
it opts for increasing m t , then this may stabilize the price but destabilizes output further,
and vice versa.
6.2.3


2

b
2b2
1
Var(t ) =
=

=
(6.17)
3 > 0 and
s
2
Var(t )

1 + b
1 + b2

2
1

b2
b2

+
1
= 2
Var(y
)
=
+
1)
=
(b
t
3 < 0.
Var(ts )

1 + b2
1 + b2
(6.18)
As expected, the variance of is therefore increasing in . Conversely, the variance of
output decreasing in .
Example 32 When b = 1, then t = /(1+)ts and yt = 1/(1+)ts so Var(t )/Var(yt ) =
2 , which is clearly increasing in .
The policy rule implies that average inflation is zero, = 0. There is no point in
creating a non-zero average inflation, since anticipated inflation does not affect output.
The policy rule also implies that demand shocks should always be completely offset:
they do not enter either inflation (6.15) or output (6.16). The reason is that demand shocks
push prices and output in the same direction, so there is no trade-off between price and
output stability. Only supply shocks, which push inflation and output in different directions, gives a trade-off.
e
To see this, let us simplify by setting price expectations in (6.5), pt|t1
, to zero and
also revert to considering m t as the policy instrument (there is a one-to-one relation to the
inflation rate). We can then solve the system (6.4) and (6.5) for output and price as
#
"
#"
#
"
# "
yt
b
b ( + )

td
=
mt +
[b ( + ) + ]
pt

+
( + )
ts

94

Monetary Policy without Commitment (Discretionary)

One problem with the commitment equilibrium is that the policy rule announced in t 1
may no longer be the optimal rule in t. At that time, inflation expectations can be treated
as given (for instance, inflation expectations might enter the model because they represent nominal contracts written in t 1). The central bank could have an incentive to
exploit this: the policy rule is then not time consistent. If the central bank cannot commit to a policy rule, then the time inconsistent rule is not credible, and the commitment
equilibrium falls apart.
We now assume that the central bank cannot commit to a rule. Instead, we look
for a policy that is optimal in t (after the shocks have been observed), when t|t1 is
taken as given. If this happens to be the same decision rule as above, then there is no
time inconsistency problemotherwise there is. With discretionary monetary policy, the
choice of inflation minimizes

2
e
t2 + bt bt|t1
+ ts y .
(6.19)
There is no expectations operator, since the central bank makes its decision after the
shocks are realized, and it does not precommit (before the shock) to follow any particular
decision rule.
The first order condition with respect to t is
e
t = t b2 + b2 t|t1
bts + b y ,

(6.20)

with (two times the) marginal cost of inflation on the left hand side and (two times the)
marginal benefits on the right hand side. The public knows that (6.20) will determine
how the central bank acts. They therefore form their expectations in t 1 by rationally
95

using all available information. Taking mathematical expectations of (6.20) based on the
information available in t 1 and rearranging gives that expectations formed in t 1 must
be
e
t|t1
= b y .
(6.21)

The high inflation between mid 1960s and early 1980s could possibly be due to the
lack of commitment technology combined with more ambitious employment goals. An
alternative explanation is that the policy makers believed in a long run trade-off between
unemployment and inflation.

Combine this with (6.20) to get


b
t = b y
s
1 + b2 t
= b y + ts

6.2.4

Empirical Illustration

Walsh Fig 8.5 (relation between central bank independence and average inflation).
(6.22)

This rule has the same response to the output shock as the commitment rule, but a higher
average inflation (if both and y are positive). The first of these results means that
the variances are the same as in the commitment equilibrium. The reason is that there
is no persistence in this model. In a model with more dynamics this will no longer be
truein that case we can intuitively think of the natural output level, here normalized to
zero, as time varying. This makes the difference between commitment and discretionary
equilibrium more complicated.
The second of the results, the higher average inflation, is due to the incentive to deviate
from the commitment ruleand that the public incorporates that when forming inflation
e
expectations. To understand the incentives to inflate consider (6.20) when t|t1
= ts =
0. If the central bank then sets t = 0 (so there is no policy surprise), then the marginal
cost of inflation (left hand side) is zero, but the marginal benefit (right hand side) is b y .
If both and y are positive, then there is an incentive to inflate. Private agents will realize
e
this and form their expectations accordingly. The equilibrium is where Et t = t|t1
and
marginal cost and benefits are equal.
It is often argued that making the central bank more independent of the government
is quite similar to a lower , that is, to a lower relative weight on output. From (6.22) we
see that this should lower the average inflation rate. At the same time, it should lower the
variability of inflation, but increase the variability of output, see (6.17)-(6.18).
It is still unclear if the inflation bias is important. There are many other cases where
the logic of the discretionary equilibrium seems unappealing, for instance, in capital income taxation (why is not all capital confiscated every year?). It might be the case that
society has managed to set up institutions and informal rules which create some kind of
commitment technology.
96

6.3

Recent Models for Studying Monetary Policy

This section gives an introduction to more recent models of monetary policy. Such models
typically combine a forward looking Phillips curve, for instance, from a Calvo model,
with an aggregate demand equation derived from an optimizing consumers intertemporal
consumption/savings decision, and some kind of policy rule or objective function for the
central bank.
6.3.1

A Simple Model

Price are set as in the Calvo model. In this model, a fraction q of the firms are allowed
to set a new price in a period, and the fraction 1 q must keep their old price. When
allowed to change the price, the firms chooses a price to minimize a discounted sum of
the squared deviations of the actual price and the flex price. (See Rotemberg (1987) and
MacPri.TeX for details.) We also assume that the flex price is determined as in model
of monopolistic competition, pit = pt + yt + t , where measures how much price
setters wants to increase the relative price when demand increases ( is high when the
substitution elasticities between goods is low and when the marginal cost curve is steep).
The supply side of the economy can then be summarized by the Phillips curve
t = Et t+1 + (yt + t ) ,

(6.23)

where is increasing in the fraction q.


The aggregate demand curve is derived in Section A from an Euler condition for
optimal consumption choice with taste shocks, combined with the assumption that con-

97

sumption equals output. It is


Et yt+1 = yt +

1
(i t Et t+1 ) + yt ,

(6.24)

where yt is a negative shock to current (time t) demand.


The central bank sets short interest rate, i t . This can have effect on output since prices
are sticky, so the nominal interest rate affects the real interest rate. This, in turn, affects
demand, and thus inflation through the Phillips effect. Suppose the reaction function,
also called simple policy rule, of the central bank is a Taylor rule
i t = t + yt .

t+1 = t + t+1
(6.26)

0
y
0
1

0
0
1
0

0
t+1

0
yt+1

0 it + 0

1
0

(6.25)

This is a sub-optimal commitment policy. It is a commitment rule since the policy setter
will stick to this rule, even if it would be optimal to deviate from it in certain states. The
optimal commitment rule, however, would not restrict the decision rule to be a function
of yt and t only.
Note that there is no money demand function in this model. The reason is that monetary policy is specified in terms of the interest rate, so the money stock becomes demand
determined (the money supply curve is flat at the chosen nominal interest rate). Of course,
in order for the central bank to control anything of importance, there must be a demand
for money. The money demand function could be added to the model, but its only role is
to determine the money stock.
Suppose the shocks in (6.23) and (6.24) follow

yt+1 = y yt + yt+1 .

We can write (6.23)-(6.26) as

1 0 0 0
t+1

0 1 0 0 yt+1

0 0 0 E

t t+1
0 0 1 1
Et yt+1

with

it =

0 0

t
yt
t
yt

(6.27)

This system is in state space form and could be summarized as


"
#
"
#
x1t+1
x1t

t + t+1 , and
A0
=A
+ Bi
Et x2t+1
x2t
"
#
x1t
.
i t = F
x2t

(6.28)

(6.29)
(6.30)

where x1t is a vector of predetermined variables (here t and yt , which happens to


be exogenous, but also endogenous variables can be predetermined) and x2t a vector of
forward looking variables (here t and yt ). Premultiply (6.29) with A 1
0 to get
"
#
"
#
x1t+1
x1t
=A
+ Bi t + t+1 , where
(6.31)
Et x2t+1
x2t
 

1
1
10
A = A 1
(6.32)
0 A, B = A0 B, and Cov (t ) = A0 Cov t A0 .
By using the policy rule (6.30) in (6.31) we get
"
#
"
#
x1t+1
x1t
= (A B F)
+ t+1 .
Et x2t+1
x2t

98

t
yt
t
yt

(6.33)

99

This system of expectational difference equations (with stable and unstable roots) can
be solved in several different ways. For instance, a decomposition of A B F in terms of
eigenvalues and eigenvectors will work if the latter are linearly independent. Otherwise,
other techniques must be used (see, for instance, Soderlind (1999)). A necessary condition
for a unique saddle path equilibrium is that A B F has as many stable roots (inside the
unit circle) as there are predetermined variables (that is, elements in x1t ).
To solve the model numerically, parameter values are needed. The following values
have been used in most of Figures 6.2-6.4 (exceptions are indicated)

y
y i
0.99 2.25 2/7 2 0.5 0.5 0.5 1.5 0.5 0

Suppose the central banks loss function is

s L t+s , where

(6.34)

s=0

L t+s = t+s

2

+ y yt+s y

2

+ i i t+s i

y
i

2
4
period

c. Large output coefficient

Optimal Monetary Policy

Et

b. Large inflation coefficient


4

2

(6.35)

A particularly straightforward way to proceed is to optimize (6.34), by restricting the


policy rule to be of the simple form discussed above, (6.25). Optimization then proceeds
100

2
4
period

Persistent price shock: simple policy rule

The choice of implies relatively little price stickiness. The choice of means that a 1%
increase in aggregate demand leads to a desired increase of the relative price of 2/7%. The
choice of the relative risk aversion implies an elasticity of intertemporal substitution of
1/2. The and are those advocated by Taylor. The loss function parameters (see next
section) means that inflation is twice as important as output, and that the policy maker
does not care about fluctuations in the nominal interest rate.
The first subfigure in Figure 6.2 illustrates how the model with the policy rule (6.25)
works. An inflation shock in period t = 0 increases inflation. The policy maker reacts by
raising the nominal interest even more in order to increase the real interest rate. This, in
turn, has a negative effect on output and therefore on inflation via the Phillips curve. The
central bank creates a recession to bring down inflation. The other subfigures illustrates
what happens if the coefficients in the reaction function (6.25) are changed.
6.3.2

a. Baseline model
4

2
0
2
2

2
4
period

Figure 6.2: Impulse responses to price shock; simple policy rule


as follows: guess the coefficients and , solve the model, use the time series representation of the model to calculate the loss function value. Then try other coefficients and
, and see if they give a lower loss function value. Continue until the best coefficients
have been found.
The unrestricted optimal commitment policy and the optimal discretionary policy rule
are a bit harder to find. Methods for doing that are discussed in, among other places,
Soderlind (1999).
Figure 6.3 compares the equilibria under the simple policy rule, unrestricted optimal
commitment rule, and optimal discretionary rule, when it is assumed that = y = 0.
It is clear that the optimal commitment rule achieves a much more stable inflation and
output, in spite of a less vigorous increase in the nominal interest rate. This is achieved by
credibly promising to keep interest rates high in the future (and even raise further), which
gives expectations of lower future output and therefore future inflation. This, in turn,
101

a. Simple policy rule

b. Commitment policy

a. Simple policy rule

b. Commitment policy

y
i

2
4
period

c. Discretionary policy

2
4
period

2
0

2
4
period

2
4
period

2
4
period

2
4
period

Persistent demand shocks

y
i

c. Discretionary policy

Persistent price shocks

Figure 6.3: Impulse responses to price shock: simple rule, optimal commitment policy,
and discretionary policy

Figure 6.4: Impulse responses to positive demand shock: simple rule, optimal commitment policy, and discretionary policy

gives lower inflation and output today. The discretionary equilibrium is fairly similar to
the simple rule in this model. Note that there is no constant inflation bias when target
levels are at their natural levels (zero) as they are in these figures. The discretionary rule is
still different from the commitment rule (they are, after all, outcomes of different games).
The intuition is that there is a time-varying bias since the conditional expectations of
output and inflation in the next periods (their conditional natural rates) typically differ
from the target rates (here zero).
Figure 6.4 makes the same type of comparison, but for a positive demand shock, yt .
In this case, both optimal rules kill the demand shock, which is seen almost directly
from (6.24): any shock yt could be met by increasing i t by yt . In this way output is
unaffected by the shock, and there will then be no effect on inflation either, since the only
way the demand shock can affect inflation is via output (see (6.23)). This is very similar

to the static model discussed above: the demand shock drives both prices and output in
the same direction and should, if possible, neutralized. Of course, the result hinges on
the assumption that the policy maker is not averse to movements in the nominal interest
rate, that is, i = 0 in (6.35). (It can be shown that this case can be approximated in the
simple policy rule (6.25) by setting the coefficients very high.) Many studies indicate that
central banks are unwilling to let the nominal interest rate vary much. This is sometimes
interpreted as a concern for the banking sector, and sometimes as due to uncertainty about
the state of the economy and/or the effect of policy changes on output/inflation. In any
case, i > 0 is often necessary in order to make this type of model fit the observed
variability in nominal interest rates.

102

103

Derivations of the Aggregate Demand Equation

King, R. G., 1993, Will the New Keynesian Macroeconomics Resurrect the IS-LM
Model?, Journal of Economic Perspectives, 7, 6782.

The period utility function is


U (Ct ) =

At
1
C
,
1 t

(A.1)

Mishkin, F. S., 1997, The Economics of Money, Banking, and Financial Markets,
Addison-Wesley, Reading, Massachusetts, 5th edn.

where At is a taste shift parameter. The Euler equation for optimal consumption is


U (Ct+1 )
U (Ct )
= Et
Q t+1 ,
(A.2)
Ct
Ct+1

Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT


Press.

where Q t+1 is the gross real return.


The marginal utility of Ct is

Rotemberg, J. J., 1987, New Keynesian Microfoundations, in Stanley Fischer (ed.),


NBER Macroeconomics Annual . pp. 69104, NBER.

U (Ct )

= At C t ,
Ct

(A.3)

so the optimality condition can be written




At+1 Ct+1
1 = Et Q t+1
At
Ct
= Et exp (ln Q t+1 + 1 ln At+1 ln Ct+1 + ln Ct ) .

Romer, D., 1996, Advanced Macroeconomics, McGraw-Hill.

Soderlind, P., 1999, Solution and Estimation of RE Macromodels with Optimal Policy,
European Economic Review, 43, 813823.
Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

(A.4)

Assume that ln Q t+1 , ln At+1 , and ln Ct+1 are jointly normally distributed. (Recall Eexp (x) =
exp (Ex + Var (x) /2) is x is normally distributed.) Take logs of (A.4) and rewrite it as
0 = ln + Et ln Q t+1 + Et 1 ln At+1 Et ln Ct+1 + ln Ct
+ Vart (ln Q t+1 + ln At+1 ln Ct+1 ) /2, or
1
1
Et ln Ct+1 = ln Ct + Et ln Q t+1 + Et z t+1 ,

(A.5)

where Et z t+1 = ln +Et 1 ln At+1 +Vart (.). The most important part of Et z t+1 is Et 1 ln At+1 .
If ln At+1 = ln At + u t+1 , then Et 1 ln At+1 = ( 1) ln At , so the AR(1) formulation
carries over to the expected change, but the sign is reversed if > 0.

Bibliography
Blanchard, O. J., and S. Fischer, 1989, Lectures on Macroeconomics, MIT Press.

104

105

9. The monetary contraction in many countries after the 1929 crash in the US can
probably be regarded as exogenous shifts. Countries which left the gold standard
early had weaker recessions. (OR Fig 9.10)

Empirical Measures of the Effect of Money on Output

Reference: Romer 5.6, Walsh (1998) 1, Mishkin (1997) 25, Isard (1995), Meese (1990),
and Obstfeldt and Rogoff (1996) 9.1

7.1

Some Stylized Facts about Money, Prices, and Exchange Rates

1. The correlation between long run inflation and money growth is almost one across
countries (BW Fig 8.9a).
2. The correlation between short run inflation and money growth is more uncertain
(BW Fig 10.10).
3. There is no clear long run correlation between inflation and the growth of real output
or between money growth and the growth of real output.
4. Money stock innovations and output innovations are correlated. Money stock changes
seems to lead output changes. Most of this correlation is between output and inside
money (created within the banking system, for instance, deposits).
5. PPP does not hold, except possibly for long horizons. Realignments seem to have
long lasting (although perhaps not permanent) effects on the real exchange rate.
(BW Fig 8.9b)
6. Countries with weak current accounts, and rapidly expanding money supply often
experience depreciation of the exchange rate.
7. Real exchange rates are much more volatile under flexible exchange rate regimes
than under fixed exchange rates. Real and nominal exchange rates are very strongly
correlated (this is evidence of monetary non-neutrality only if we can prove existence of important nominal shocks). (Isard Fig 3.2 and 4.2)

10. Most contracts are written in nominal terms, and prices are typically changed fairly
seldomeven at the fairly high inflation rates of the late 1970s. This seems to
change as we move into very high inflation rates. The way it changes is by indexation or dollarization.
11. Sharp exogenous monetary contractions (or a sudden and unexpected increase in
the short interest rates by the central bank) seems to have an effect on output and
employment which may last for years. (Walsh Fig 1.3)

7.2

Early Studies of the Effect of Money on Output

Early Keynesians (until the 1960s, say) thought that money has little effect on output.
There were several reasons for this. First, nominal interest rates (on high grade bonds)
were very low during the Great Depression, but that did not appear to boost output. Second, investment and consumption regressions showed very little effect of nominal interest
rates.
The idea that the Great Depression was a period loose monetary policy was heavily
challenged by Friedman and Schwartz (1963a). They showed that the decline in money
supply was the largest ever in US history. (A possible counter argument to this is that the
monetary base changed fairly little, and that most of the change in the money aggregates
were due to the fact that the public chose to hold more cash and less deposits, and that
banks chose to hold more reserves.)
Another weak spot of the early Keynesian interpretation of the Great Depression is
that it is based on low nominal interest rates. Prices were falling, so the real interest rates
were actually very high, perhaps as high as 10%, which could be interpreted as a very
tight monetary policy.

8. The log exchange rate behaves almost like a random walk (that is, ln St ln St1 +
u t ). Equations for predicting ln St ln St1 typically have R 2 < 0.1.
106

107

7.3

Early Monetarist Studies of the Effect of Money on Output

7.3.2

7.3.1 Friedman&Schwartz and St. Louis Equations


Friedman and Schwartz (1963a) and Friedman and Schwartz (1963b) study 100 years
of US data and find that money aggregates lead output, in the sense that all recessions
were preceded by declining money growth rates. However, the response to money growth
changes showed long and variable lags. They also argue that many of the money supply changes can be regarded as exogenous with respect to output (changes within the
monetary sector).
The St. Louis equation (see, for instance, Andersen and Jordan (1968)) is a regression of output (growth) on current and lagged money (growth) and perhaps some other
variables. This can be thought of as a formalization of the approach of Friedman and
Schwartz, although no attention is paid to whether the money supply changes are exogenous or not. In its simplest form it is
yt =

s m ts + t ,

(7.1)

s=0

On the Interpretation of Cov(yt , m t ) I: Reverse Causality

The problem with a causal interpretation of the correlation between money and output,
or of the St. Louis equation, is that most of the correlation between money and output is
between output and inside money (deposits which is money created within the private
banking system, as opposed to the monetary base which is outside money).
It is possible that banks extend more loans (which generates deposits) as the business
conditions are about to pick up. The correlation between money and output is then due to
reverse causality as discussed in King and Plosser (1984). The idea is that money may not
have any effect on output, but output affects money aggregates positively, which explains
the positive correlation of output and money. For this to make sense, it must be the case
that the central bank cannot, or does not want to, control the broad monetary aggregates.
Example 33 (Regressing output on money, two-way causality.) Suppose the structural
model of output and money is
yt = m t + u yt
m t = yt + u mt ,

and the s coefficients are sometimes interpreted as the effect of the money stock, m ts ,
on output, yt . (Initially, St. Louis equations were used to explain nominal output, but here
the focus is on real output.)
The St. Louis equation is not a structural model; it is a reduced form. Monetarists
would perhaps argue that the transmission mechanism from money to output is very complex, and that it makes sense to use (7.1) since it can potentially summarize the effects.
Keynesians would perhaps be more structural in the sense that their view of the transmission mechanism is quite clear: money supply affects the interest rate (the LM equation), which in turn affects output (IS equation).
Any causal interpretation of the correlation between money and output, or of the St.
Louis equation relies on the assumption that most movements in the money stock are due
exogenous forces and not due to changes in output. These exogenous forces could be
policy shocks or shifts in money demand which are not caused by output.

108

where the shocks are assumed to be uncorrelated. Output and money are here allowed to
depend on each other (in the same period). The reduced form is
"
#
"
#"
#
1
1
u yt
yt
=
.
1 1
u mt
mt
Suppose we run a very simple St. Louis equation
yt = m t + t ,

109

The least squares (LS) estimate of is (in probability limit, plim)

7.3.3

Cov (m t , yt )
Var (m t )

Cov u yt + u mt , u yt + u mt

=
Var u yt + u mt

Var u yt + Var (u mt )

= 2
Var u yt + Var (u mt )

Var u yt /Var (u mt ) +

= 2
.
Var u yt /Var (u mt ) + 1

On the Interpretation of Cov(yt , m t ) II: Common Driving Force

Instead of assuming that output affects money stock directly, it may be that there is a third
variable which drives both output and money. This will also make the interpretation of
money-output correlations very murky, since it gives an omitted-variables bias.

plim L S =

Example 37 (Regressing output on money, unobservable driving force.) Suppose the


structural model of output and money is
m t = z t + u mt
yt = m t + z t + u yt

Example 34 (Exogenous money supply.) If m t is exogenous, = 0, then plim L S =


in Example 33 and the regression coefficient captures the effect of money shocks on output.
This is the interpretation in Friedman and Schwartz (1963a). We get the same result if

Var u yt /Var(u mt ) 0, that is, when most of the movements in output (and money) is
caused by the exogenous money shocks, so money is once again essentially exogenous.

Example 35 (Output shocks dominate.) Conversely, Var u yt /Var(u mt ) gives
plim L S 1/ in Example 33 (use lHopitals rule to show this), so the regression
coefficient merely reflects how money (and output) are both driven by output shocks (reverse causality).
The reverse causality story can be turned on its head, however. Suppose money do
have effect on output, but that the central bank tries to stabilize output, that is, output has
a negative effect on money supply. In this case, the correlation of money and output will
underestimate the effect of money on output.
Example 36 (Countercyclical policy.) From the reduced form in Example 33, we see that
output follows

1
u yt +
u mt .
yt =
1
1
Suppose > 0, then a negative makes the effect of an output shock small. This gives
plim L S < , so the St. Louis equation has a negative bias in the direct effect of money
on output.

= ( + ) z t + u yt + u mt
where the shocks are uncorrelated with each other and with z t . Money affects output if
6 = 0, but output does not affect money. However, both money and output are driven by
the common factor z t . The LS estimate of in the St. Louis equation, yt = m t + t , is
then
Cov (m t , yt )
Var (m t )

Cov z t + u mt , ( + ) z t + u yt + u mt
=
Var (z t + u mt )
( + ) Var (z t ) + Var (u mt )
=
Var (z t ) + Var (u mt )
( + ) Var (z t ) /Var (u mt ) +
=
.
Var (z t ) /Var (u mt ) + 1

plim L S =

This equals only if z t is (effectively) a constant, Var(z t ) /Var(u mt ) = 0, or if output is


not affected by z t , = 0, so there is no common driving force.
Example 38 (Unobservable driving force and reverse causality.) Suppose = 0 in
Example 37, so money has no effect on output (nor has output any effect on money). In
this case, the estimate of in the St. Louis equation, yt = m t + t , becomes
plim L S =

Var (z t ) /Var (u mt )
,
Var (z t ) /Var (u mt ) + 1

which has the same sign as . If we interpret yt as output in the next period, which
thus depends on todays z t , then we have the case of King and Plosser (1984). Their
110

111

mechanism is that z t signals high future productivity, which leads to more purchases of
production factors today (needs to be accumulated in advance). This is a version of the
traditional reverse causality, with the twist that future output affects todays money
demand.

Run a regression m t = At + t , where we get from the reduced form that


plim =
=

Example 39 (Countercyclical policy.) Suppose Var(u yt ) =Var(u mt ) = 0 in Example 37


and that the central bank sets = . This achieves complete stabilization of output,
and plim L S = 0 in the St. Louis equation. In this case, the successful monetary policy
makes it look as if monetary policy cannot affect output.
One way of getting around the problem with the common driving force is to estimate
an extended St. Louis equation, where output is related to both money and a vector of
other variables capturing the driving force, xt ,
yt =

X
s=0

s m ts +

s xts + t .

(7.2)

s=0

Of course, this approach assumes that we can observe the relevant variables and that they
are exogenous. The problem with a direct reverse causality (direct effect of output on
money, possibly with leads/lags) remains in this equation, however.

Cov(At , m t )
Var (At )

Cov(At , 1
u yt +

1
1

At +

1
1 mt )

Var (At )

1
=
,
1
since At is assumed to be uncorrelated with u yt and mt . Form a fitted value of m t as
m t = At , or
1
m t =
At ,
1
and use this instead of m t in the St. Louis regression, yt = m t + t . This gives a new
estimate

Cov m t , yt

plim iv =
Var m t


1
Cov 1
At , u yt + At + mt


=
1
Var 1
At
= ,

7.3.4

St. Louis Equation with Instrumental Variables

Suppose we could get exogenous indicators of the exogenous movements in monetary


policy. As a first step, regress m t on the instruments and construct a series of fitted
policy, m t . In a second step, use m t instead of m t in a regression like (7.1). This is the
IV/2SLS method, which is consistent as long as the instruments are not driven by output.
The trick in the IV/2SLS method is to discard all variation in m t which is not driven
by the instruments. This side-steps all movements in m t which are due to reverse causality,
that is, due to the output shock. The regression will then look at how output moves in
response to the exogenous changes in money.
Example 40 (IV and the case of two-way causality.) Consider the model in Example 33,
and suppose we now that u mt = At +mt , where At are observable shifts in money supply.

112

where we once again use the fact that At is uncorrelated with u yt and mt . This is the
correct value.
Romer and Romer (1990) went through the Feds minutes to identify (exogenous)
policy shifts. They found six such policy shifts during the post war period. In each case,
short interest rates increased, while money aggregates and output decreased. They then
estimated a St. Louis equation with both least squares and instrumental variables, where
dummies for these episodes were used as instruments for money. They found that the
instrumental variables method gave larger effects of money on output, as expected if the
central bank uses money in a systematic way to stabilize output.
The analysis of Romer and Romer (1990) is very much like a formalization of what
Friedman and Schwartz (1963b) did: pick out a number of monetary contractions which
look exogenous and study what happens to output after that. Overall, the evidence from
the historical episodes in these and other studies point in the direction that money probably
113

have an effect on output. This conclusion is strengthened if we look at the reaction of


output of large exchange rate realignments (another type of monetary policy).
Example 41 (Summary of Romer and Romer (1990) and Romer 232-236.) Main issue:
to find instruments for exogenous shifts in monetary policy. The problem with possibly
endogenous money is circumvented by focusing on episodes of exogenous (according to
Romer and Romer) shifts in monetary policy: September 1955, December 1968, April
1974, August 1978, and October 1979. According to the authors, these monetary contractions were brought about by a desire to take down inflation, with little concern about
the effects on output. To investigate if there really were monetary contractions, univariate
forecasting equations for log money stock where estimated
1m t =

24
X

s 1m ts + u t

s=1

on monthly postwar data. Dynamic forecast for each period of 36 months after the 5 break
dates indicate that the money stock were a lot lower (in data) than predicted: there seem
to have been contractions. Also, the interest rates show fairly consistent increases over
the same 36 months, so any effect on output could be consistent with either a monetarist or
a Keynesian model. To see if theses contractions mattered for output, a St. Louis equation
was estimated
24
24
X
X
1yt = a + bt +
ci 1yts +
di 1m ts ,
s=1

7.3.5

Dynamic Effects and Causality

Suppose the structural model consists of a reaction function where money is predetermined
p
p
X
X
s m ts + u mt
(7.3)
s yts +
mt =
s=1

s=1

and a somewhat extended St. Louis equation


yt = 0 m t +

p
X
s=1

s m ts +

p
X

s yts + u yt .

(7.4)

s=1

The shocks are assumed to be uncorrelated.


This is a fully recursive system of simultaneous equations, so least squares on (7.3)
and (7.4) separately is consistent. The reason is that the simultaneity problem is assumed
away: money is not contemporaneously affected by the output shock. It is also assumed
that the lagged money and output capture all relevant common driving forces, so there is
no omitted-variables bias.
However, if we care about more than just the impact effect of money supply shocks,
then money is endogenous in the economic sense since we have the following chain of
effects: u mt m t yt m t+1 yt+1 ... We therefore need to estimate both
equations in order to trace out the effect of a monetary policy shock on output. This is the
VAR approach discussed below.

s=0

with OLS and instrumental variables (IV). The instruments were a constant, a trend,
1yts , 1m ts , and {P Sts }36
s=0 , where P St = 1 if t is one of the five policy shift dates,
and zero otherwise. If m t is affected (negatively - to stabilize output) by Yt , OLS on the
output equation gives a downward bias in di . The IV approach should give consistent estimates if the P St are not affected by output (the policy shifts were exogenous with respect
to output, driven by concern about inflation, and inflation does not affect output per se),
and if P St affects output via money only (probably reasonable). The result is that dI V is
larger than dO L S .

7.4

Unanticipated or Anticipated Money

Reference: Romer 6.4, Mishkin (1983) 6, and Barro (1977).


Both the new monetarist (Lucas model for the Phillips curve ) and new Keynesians
(micro based models of nominal rigidities) emphasize the distinction between anticipated and unanticipated policy changes. The theoretical predictions, based on rational
expectations, are that anticipated policy changes should have no or only small effects on
output, while unanticipated changes could have large effects.
For simplicity, assume that the structural model for output is
yt =

p
X

s (m ts Ets1 m ts ) + u yt ,

(7.5)

s=0

114

115

so output depends on money stock surprises and an output shock, u yt .


The money supply rule is assumed to depend on past information and a policy shock;
money supply is predetermined in relation to output
m t = 0 z t1 + u mt Et1 m t = 0 z t1 .

(7.6)

The output and policy shocks are uncorrelated. The vector z t1 typically involves lagged
output, money supply, and some other variables. As in (7.3) and (7.4), the econometric
simultaneity problem is assumed away by making money supply predetermined in relation
to the output shock.
Use the policy rule (7.6) in the output equation (7.5)
yt =

k
X
s=0

n
 X
s m ts 0 z ts1 +
0 z ts1 + u yt ,

(7.7)

s=0

where = 0 if only unanticipated money matters. Suppose the system (7.6) and (7.7)
is estimated jointly and rational expectations is imposed (restricting to be the same in
both equations). It is then straightforward to test if = 0. Mishkins results indicate that
anticipated policy does matter.
Under the maintained hypothesis that only unanticipated money matters, only the s
coefficients are needed in order to study the effect on money on output. This is different
from (7.3) and (7.4), where we needed the whole system. The reason is that in (7.5)-(7.6)
the feedback from yt to m t+1 (due to an initial money supply shock u mt ) has no effect on
yt+1 .

7.5
7.5.1

VAR Studies
VAR Models and Simultaneous Equations Systems

The VAR approach is an attempt to capture the main time series properties of money
and output (and sometimes other variables) at the same time as enough restrictions are
imposed to identify exogenous policy shifts. It is basically an attempt to estimate a system
of simultaneous structural equations.
For instance, (7.3) and (7.4) is a typical VAR model. In order to illustrate how VAR
models are handled, we look at that simple bivariate system of output and money once
116

again. The two (structural, it is assumed) equations can be written


"
#
"
#
"
# "
#
mt
m t1
m t p
u mt
B0
= B1
+ ... + B p
+
,
yt
yt1
yt p
u yt

(7.8)

B0 , ..., B p are 2 2 matrices.


The parameters (B0 , ..., B p , and the covariance matrix of the shocks) cannot be estimated without imposing some identifying restrictions. In fact, all data can tell us is the
parameters of the reduced form
"
#
"
#
"
#
mt
m t1
m t p
= A1
+ ... + A p
+ t ,
(7.9)
yt
yt1
yt p
The problem is that there may be many structural forms that generate the same reduced
form.
By (7.8), the reduced form can be written as
"
#
"
#
"
#
"
#
mt
m t1
m t p
u mt
1
1
1
= B0 B1
+ ... + B0 B p
+ B0
,
(7.10)
yt
yt1
yt p
u yt
so
"
Ai =

B01 Bi

for i = 1, ..., p and Cov (t ) =

B01 Cov

u mt
u yt

#!

0
B01 .

(7.11)

In the structural form (7.8), there are (1 + p) 4 coefficients in B0 , ..., B p and 3 parameters
in covariance matrix of the structural shocks. In the reduced form (7.10), there are p4
coefficients in A1 , ..., A p and 3 parameters in covariance matrix of the reduced form
shocks. We therefore need to impose at least 4 restrictions to calculate the parameters
in the structural form.
The following is a common set of restrictions. First, the structural shocks are uncorrelated (one restriction, the off-diagonal element in the covariance matrix is zero).
This means that the structural shocks have an economic interpretation as money or output
shocks. Second, the diagonal elements of B0 are unity (two restrictions); alternatively we
could assume that the variances of the structural shocks are unity. Third, B0 is triangular. Often (see, for instance, Sims (1980))), the assumption has been that money is not

117

contemporaneously affected by output, so B0 has the form


"
#
1
0
B0 =
.
0 1

(7.12)

Using (7.12) in (7.8) gives us a system on the same form as (7.3) and (7.4). As
discussed before, these equations can be estimated with LS as they stand. Alternatively,
the reduced form (7.9) can be estimated with LS. Imposing the assumptions allows us to
solve for the structural parameters (typically a matter of solving non-linear equations, but
can be simplified by using some straightforward matrix decompositions like the Cholesky
decomposition).
This illustrates that the VAR does not allow us to escape the tricky question of endogenous money, or more generally, monetary policy. The simultaneous equations bias may
therefore affect the VAR estimateif the identifying assumptions we make are wrong.
Including lagged money and output in both equations is an attempt to control for common
movements in money and movements which are driven common factors (to get away from
a potential omitted-variables bias).
Once we have an estimate of the structural form (7.8), we may trace out the effect on,
for instance, output of shocks to monetary policy (impulse response function). We may
also calculate how large a fraction of the variance of the forecast error of output that is
explained by the monetary policy shocks (variance decomposition).
7.5.2 Typical Results from VAR Studies of Money and Output
The typical result for the US on quarterly data for the last 30 years or so is that output has
a humped-shaped response to u mt which may last for several years. However, monetary
policy shocks seems to have accounted for a fairly small fraction of the variance of output,
in particular after 1982. This does not, of course, mean that systematic monetary policy
has not been important or that monetary policy shocks cannot be important. For the latter,
the impulse response function is much more informative than the variance decomposition.
It is unfortunate that the VAR analysis typically says very little about the importance of
the systematic (feedback) part of policy, which by many is believed to be very important
(partly in opposition to the idea that only unanticipated money matters).
Some authors (for instance, Bernanke and Blinder (1992)) argue that M1 is not a good
proxy for the monetary policy instrument, and that a short nominal interest rate should
118

be used instead. First, there is evidence (see, for instance, Sims (1980)) that if output,
money, and the nominal interest rate are included in a VAR, then money can no longer help
predict output, but that interest rates help predict both output and money. (This suggests
that much of the movements in monetary aggregates are endogenous, not exogenous as
often assumed in macro models of monetary policy.) Second, the impulse responses to
money aggregates like US M1 is often weird: a positive innovation in M1 gives a decline
in output and an increase in nominal interest rates (see, for instance, Eichenbaum (1992)).
In contrast, a positive shock to the federal funds rate gives decline in output, which seems
much more reasonable. Third, many analysts argue that most central banks discuss and
set monetary policy in terms of a short interest rate.
Often, the dynamic response to a policy shock (stricter policy) is that the price level
increases instead of decreases: the price puzzle. This is often the case when a short interest
rate is taken to be the policy instrument. The reason is probably that the VAR contains too
little information compared to what the policy makers use. Including a commodity price
solves the puzzle. The interpretation is that the central bank reacts to commodity price
increases (by raising the interest rate/decreasing money supply) since they signal future
inflation. Suppose the monetary policy is unable to kill the inflation impulse completely.
Unless you control for the commodity prices, it will appear as if tighter monetary policy
leads to higher inflation. (See Walsh Fig 1.4)
Critique against VAR models: too little forward looking information included (asset
prices?), policy shocks differ wildly between different studies, relatively little information
about the effect of systematic policy, what if policy changes? (Lucas critique).

7.6

Structural Models of Monetary Policy

Reference: Fuhrer and Moore (1995), Fuhrer (1997), Soderlind (2000), and Clarida, Gal,
and Gertler (1998).
Another line of research is to estimate a structural economic model directly (although
the requirement for the label structural may have shifted over time). Of course, there
is a long tradition of estimating AS-AD (IS-LM) models with adaptive expectations. Recently, this have gained new popularity, but we have also seen a number of papers estimating similar models while allowing for rational expectations. Many central banks still use
fairly large macro models, even if there is a tendency towards smaller models (in order to
119

handle rational expectations, among other things).


Consider the simple model
t = Et t+1 + yt + t
1
Et yt+1 = yt + (i t Et t+1 ) + yt

i t = t + yt + it .

(7.13)
(7.14)
(7.15)

The first equation comes from a Calvo model of price setting for a firm under monopolistic
competition; the second from the intertemporal optimality condition for a consumers (plus
the assumption that consumption equals output); and the third is Taylors policy rule,
which describes how the central banks sets the short nominal interest rate.
It is clear that this is a complicated model to estimate: there is plenty of simultaneity,
and expectations about future values are needed.
This model can be estimated by maximum likelihood by specifying the distribution
of the shocks ( t , yt , and it ). The MLE is found by iteration on the following steps
(i) guess a vector of parameters (, , , , , and the variances of the three shocks);
(ii) solve for the equilibrium and times series process (of t , yt , and i t ); (iii) evaluate the
likelihood function; (iv) improve the guess of the parameter vector. This gives a set of
estimates where rational (model consistent) expectations have been imposed. A possible
alternative is to use survey information as proxies for some or all expectations.
Another approach is taken by Clarida, Gal, and Gertler (1998) who want to estimate
the monetary policy rule (reaction function) for several countries. They specify a policy
rule for the short interest rate, i t , of the form
i t = Et t+12 + Et yt + i t1 + u t ,

(7.16)

where Et yt is the best estimate of the current output gap and where t+12 is the inflation
over the next twelve months. This formulation is can be seen as an approximation of the
optimal policy in many models where there is some price stickiness and where the central
bank wants to stabilize inflation and output, but also avoiding excessive movements in the
short interest rate. The equation is estimated with an instrumental variables method. To
see why that makes sense, not that we can rewrite (7.16) as
i t = t+12 + yt + i t1 + u t + t+n ,

where t+n is a linear combination of the surprise in inflation, t+12 Et t+12 , and in the
output gap, yt Et yt . The new residual in (7.17), u t +t+n , is clearly not uncorrelated with
the regressors, so we need to apply an instrumental variables method. All information at
t or earlier should be valid instruments.
Clearly, this approach allows us to understand the monetary policy rule only. To see
how a policy change, for instance, a shock u t to (7.16) affects output and inflation, we
have to estimate the rest of the model.

Bibliography
Andersen, L. C., and J. L. Jordan, 1968, Monetary and Fiscal Actions: A Test of Their
Relative Importance in Economic Stabilization, Federal Reserve Bank of St. Louis
Review, 50, 1124.
Barro, R. J., 1977, Unanticipated Money Growth and Unemployment in the United
States, American Economic Review, 67, 101115.
Bernanke, B. S., and A. S. Blinder, 1992, The Federal Funds Rate and the Channels of
Monetary Transmission, American Economic Review, 82, 901921.
Clarida, R., J. Gal, and M. Gertler, 1998, Monetary Policy Rules in Practice: Some
International Evidence, European Economic Review, 42, 10331067.
Eichenbaum, M., 1992, Comment on Interpreting the Macroeconomic Time Series
Facts: The Effects of Monetary Policy by Christopher Sims, European Economic
Review, 36, 10011011.
Friedman, M., and A. J. Schwartz, 1963a, A Monetary History of the United States, 18671960, Princeton University Press, Princeton.
Friedman, M., and A. J. Schwartz, 1963b, Money and Business Cycles, Review of Economics and Statistics, 45, 3264.
Fuhrer, J., and G. Moore, 1995, Inflation Persistence, Quarterly Journal of Economics,
110, 127159.

(7.17)
120

121

Fuhrer, J. C., 1997, Inflation/Output Variance Trade-Offs and Optimal Monetary Policy,
Journal of Money, Credit, and Banking, 29, 214234.
Isard, P., 1995, Exchange Rate Economics, Cambridge University Press.

King, R. G., and C. Plosser, 1984, Money, Credit and Prices in a Real Business Cycle
Model, American Economic Review, 74, 363380.
Meese, R. A., 1990, Currency Fluctuations in the Post-Bretton Woods Era, Journal of
Economic Perspectives, 4, 117133.

Reading List

Main references: Romer (1996) (Romer), Blanchard and Fischer (1989) (BF), and Obstfeldt and Rogoff (1996) (OR). For an introduction to many of the issues, see Burda
and Wyplosz (1997) (BW). Walsh (1998) (Walsh) is a more advanced text, and is recommended for further study.
Papers marked with ( ) are required reading.

Mishkin, F. S., 1983, A Rational Expectations Approach to Macroeconometrics, NBER.


Mishkin, F. S., 1997, The Economics of Money, Banking, and Financial Markets,
Addison-Wesley, Reading, Massachusetts, 5th edn.

0.1

Money Supply and Demand

1. Lecture notes
Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT
Press.
Romer, C. D., and D. H. Romer, 1990, New Evidence on the Monetary Transmission
Mechanism, Brookings Papers on Economic Activity, 1, 149213.
Sims, C. A., 1980, Macroeconomics and Reality, Econometrica, 48, 148.

2. OR 8.1, 8.3.1-4 (MIU), and 9.1 (stylized facts) or Walsh 2.3.1 (MIU)
3. Romer 9.8 (cost of inflation)
4. Lucas (2000) (cost of inflation)
5. BF 4.2 and 4.5 (MIU)

Soderlind, P., 2000, Monetary Policy and the Fisher Effect, Journal of Policy Modeling, Forthcoming, also available as Working Paper No. 159, Stockholm School of
Economics.

Keywords: money multiplier, central bank intervention, traditional money demand


equations, money in the utility function, Friedmans rule.

Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

0.2

Price Level and Nominal Assets

1. Lecture notes
2. OR 8.2 (seignorage) or Walsh 4.3 (seignorage)
3. OR 8.4.1 and 8.4.3-4 (exchange rates)
4. Meese (1990) (exchange rate regressions)
5. BF 4.7 and 5.1 (seignorage)

122

123

6. Romer 9.7 (seignorage)

Keywords: general equilibrium in static model of monpolistic competition (BlanchardKiyotaki), menu costs, Nash equilibria with/without sticky prices, importance of real
rigidities.

7. Isard (1995) 4 (PPP), 5 (UIP), and 8 (exchange rate regressions)


Keywords: price level in general equilibrium, superneutrality of money, Cagans
model with rational expectations, seignorage, exchange rate determination from money
demand and UIP

0.3

0.5

Sticky Prices

1. Lecture notes
2. Rotemberg (1987) (adjustment costs for prices) or Walsh 5.5 (inflation persistence)

Money and Prices in RBC Models

1. Lecture notes
3. Roberts (1995)
2. Benassy (1995) (Long-Plosser model with money and predetermined wages)
4. Romer 6.5-9 (staggered prices, Caplin-Spulber)
3. Cooley and Hansen (1995) (RBC model with money and predetermined wages)
5. BF 8.2-4 (staggered prices, Ss, Caplin-Spulber)
4. King and Plosser (1984) (reverse causality)
Keywords: quadratic adjustment costs for prices (Rotemberg), Calvos model of price
changes, combining with flex price from monopolistic competition. (Advanced: Ss
rules)

5. Walsh 3.3 (CIA) and 5.3.1 (wage rigidity in MIU models)


Keywords: RBC model with money and sticky prices, reverse causality, CIA.

0.4

0.6

Money and Monopolistic Competition

Monetary Policy

1. Lecture notes

1. Lecture notes

2. OR 10.1 (GE with monopolistic competition)

2. OR 9.4-5 (Barro-Gordon model) or Walsh 8.1-2 (Barro-Gordon)

3. Romer (1993) (overview of New Keynesian models)

3. Taylor (1995) (Empirical analysis of transmission mechanism)

4. King (1993) (overview of New Keynesian models)

4. Fischer (1996) (Costs of inflation)

5. BF 8.1 (GE with monopolistic competition)

5. Obstfeld and Rogoff (1995) (Mirage of fixed exchange rates)

6. Romer 6.6 and 6.10-15 (monopolistic competition, real rigidities)

6. Bernanke and Mishkin (1997) (inflation targeting)

7. BF 9.5 (real rigidities)

7. Fuhrer (1997) (inflation/output trade-off)

8. Walsh 5.3.2 (imperfect competition and price stickiness)

8. Romer 5.1-5 (IS-LM) and 9.6 (What can policy accomplish?)

124

125

Bibliography

9. BF 11.2 (traditional monetary policy issues) and 11.4 (Barro-Gordon model)


10. Walsh 5.4 (macro model for policy analysis), 8.4-5 (institutions), and 10.5 (macro
model for policy analysis again)

Benassy, J.-P., 1995, Money and Wage Contracts in an Optimizing Model of the Business
Cycle, Journal of Monetary Economics, 35, 303315.

11. OR 9.2-3 (Dornbusch model)

Bernanke, B., 1983, Nonmonetary Effects of the Financial Crisis in the Propagation of
the Great Depression, American Economic Reveiw, 73, 257276.

Keywords: Mundell-Flemming model and choice of exchange rate regime, BarroGordon model of monetary policy (discretion and commitment), recent models for monetary policy,
Dornbusch model.

0.7

Bernanke, B. S., and M. Gertler, 1995, Inside the Black Box: The Credit Channel of
Monetary Policy Transmission, Journal of Economic Perspectives, 9, 2748.
Bernanke, B. S., and F. S. Mishkin, 1997, Inflation Targeting: A New Framework for
Monetary Policy, Journal of Economic Perspectives, 11, 97116.

Empirical Measures of the Effect of Money on Output


Blanchard, O. J., and S. Fischer, 1989, Lectures on Macroeconomics, MIT Press.

1. Lecture notes
Burda, M., and C. Wyplosz, 1997, Macroeconomics - A European Text, Oxford University
Press, 2nd edn.

2. Walsh 1 (overview)
3. Romer 5.6 (selective overview)
Keywords: interpreting money-output correlations, anticipated or unanticipated money,
VAR studies, case studies (Romer-Romer).

0.8

The Transmission Mechanism from Monetary Policy to Output

Cooley, T. F., and G. D. Hansen, 1995, Money and the Business Cycle, in Thomas F.
Cooley (ed.), Frontiers of Business Cycle Research, Princeton University Press, Princeton, New Jersey.
Fischer, S., 1996, Why Are Central Banks Pursuing Long-Run Price Stability, in
Achieving Price Stability, pp. 734. Federal Reserve Bank of Kansas City.
Fuhrer, J. C., 1997, Inflation/Output Variance Trade-Offs and Optimal Monetary Policy,
Journal of Money, Credit, and Banking, 29, 214234.

1. Mishkin (1995)
2. Walsh (1998) 7

Isard, P., 1995, Exchange Rate Economics, Cambridge University Press.

3. Bernanke and Gertler (1995)

King, R. G., 1993, Will the New Keynesian Macroeconomics Resurrect the IS-LM
Model?, Journal of Economic Perspectives, 7, 6782.

4. Bernanke (1983)

King, R. G., and C. Plosser, 1984, Money, Credit and Prices in a Real Business Cycle
Model, American Economic Review, 74, 363380.

5. Romer and Romer (1990)


6. Stiglitz and Weiss (1981)

Lucas, R. E., 2000, Inflation and Welfare, Econometrica, 68, 247274.


Keywords: money view, lending view, credit rationing (Stiglitz-Weiss).
126

127

Meese, R. A., 1990, Currency Fluctuations in the Post-Bretton Woods Era, Journal of
Economic Perspectives, 4, 117133.
Mishkin, F. S., 1995, Symposium on the Monetary Transmission Mechanism, Journal
of Economic Perspectives, 9, 310.
Obstfeld, M., and K. Rogoff, 1995, The Mirage of Fixed Exchange Rates, Journal of
Economic Perspectives, 9, 7396.
Obstfeldt, M., and K. Rogoff, 1996, Foundations of International Macroeconomics, MIT
Press.
Roberts, J. M., 1995, New Keynasian Economics and the Phillips Curve, Journal of
Money, Credit, and Banking, 27, 975984.
Romer, C. D., and D. H. Romer, 1990, New Evidence on the Monetary Transmission
Mechanism, Brookings Papers on Economic Activity, 1, 149213.
Romer, D., 1993, The New Keynesian Synthesis, Journal of Economic Perspectives, 7,
522.
Romer, D., 1996, Advanced Macroeconomics, McGraw-Hill.
Rotemberg, J. J., 1987, New Keynesian Microfoundations, in Stanley Fischer (ed.),
NBER Macroeconomics Annual . pp. 69104, NBER.
Stiglitz, J. E., and A. Weiss, 1981, Credit Rationing in Markets with Imperfect Information, American Economic Review, 71, 393410.
Taylor, J. B., 1995, The Monetary Transmission Mechamism: An Empirical Framework, Journal of Economic Perspectives, 9, 1126.
Walsh, C. E., 1998, Monetary Theory and Policy, MIT Press, Cambridge, Massachusetts.

128

Вам также может понравиться