Chapter 8

Sticky Prices
Carlos A. Végh
University of Maryland and NBER
Current draft: March 31, 2011

1 Introduction
After studying the frictionless world of Chapter 5 in which monetary and
exchange rate policy had no e¤ects on the real sector, Chapters 6 and 7
introduced frictions into the model – no interest-bearing bonds in the case
of Chapter 6 and links between nominal interest rates and consumption in
the case of Chapter 7 –that allowed us to focus on some of the key channels
whereby monetary and exchange rate policy a¤ect the real economy in a
small open economy. This chapter introduces the last – and perhaps best-
known – friction in the model of Chapter 5: sticky prices. This chapter’s
model can thus be viewed as an optimizing version of the Mundell-Fleming
model.1
Comments very welcome. This chapter is part of a graduate textbook on “Open Econ-
omy Macroeconomics in Developing Countries”, currently under preparation by the author
(and to be published by MIT Press) and should be cited accordingly. I am very grateful
to Daniel Hernaiz, Fernando Im, Pablo Lopez Murphy, Rong Qiang, Agustin Roitman,
Belen Sbrancia, Guillermo Tolosa, and Igor Zuccardi for their help in the preparation of
this chapter and to Rajesh Singh and Yossi Yakhin for helpful comments.
1
The Mundell-Fleming model is named after the contributions of Mundell (1963, 1964)
and Fleming (1962). This was one of the two major contributions that earned Robert
Mundell the 1999 Nobel Prize in Economics. The Mundell-Fleming model essentially
refers to an open economy model with perfect capital mobility and sticky prices. Modern
versions of the Mundell-Fleming model –couched in terms of Obstfeld and Rogo¤’s (1995)
in‡uential paper –have come to be known as “new open-economy macroeconomics”. Since,

1
 The main motivation for introducing sticky prices is that permanent
changes in the level of the nominal money supply have no real e¤ects in
the context of the basic ‡exible-prices model introduced in Chapter 5 and
modi…ed in Chapters 6 and 7.2 Once sticky prices come into the picture,
however, monetary policy will have real e¤ects and can thus potentially ex-
plain some key features of the real world. In particular, an expansion in the
money supply will (i) lead to higher aggregate demand and hence output –
since output is typically demand-determined in sticky-prices models – thus
providing a possible explanation for business-cycle ‡uctuations; (ii) gener-
ate a positive co-movement between nominal and real exchange rates, thus
explaining a well-documented stylized fact in open economies (see Mussa
(1986)); and (iii) possibly result in exchange rate overshooting, in the sense
that the short-run increase in the nominal exchange rate is larger than in the
long-run, which could explain the high volatility of nominal exchange rates.
In contrast to ‡exible exchange rates, a permanent devaluation did have
real e¤ects in the world of Chapter 6. Indeed, by reducing real money bal-
ances on impact, a devaluation led to a fall in consumption in order to gener-
ate the trade surpluses required to replenish real money balances over time.
In sharp contrast, we will see that, under sticky prices, a devaluation leads to
higher output and consumption. These contrasting results illustrate a long-
standing debate in development macroeconomics regarding the real e¤ects of
a devaluation, which is reviewed in Box 1.
The chapter proceeds as follows. Section 2 lays out the groundwork by in-
troducing sticky prices into the model of Chapter 5. With the model in hand,
Section 3 analyzes the e¤ects of monetary policy (i.e., permanent changes in
the money supply and the rate of growth of the money supply). In particular,
a permanent increase in the level of the money supply leads to an expansion
conceptually, there is clearly nothing “new” in these models, this label presumably refers
to the fact that explicit microfoundations are introduced both on the demand and the
supply side. In fact, the title of Obtsfeld and Rogo¤’s paper (“Exchange rate dynamics
redux”) –“redux” meaning “redone” or “brought back” –pays hommage to Dornbusch’s
1976 paper titled “Expectations and exchange rate dynamics”, his most famous paper and
arguably the most in‡uential paper in the Mundell-Fleming tradition (see Rogo¤ (2002)).
Rudi Dornbusch himself – a student of Robert Mundell at the University of Chicago –
would have been a likely recipient of the Noble Prize in Economics had it not been for his
untimely death of cancer in 2002 at the age of 60.
2
There could be real e¤ects, of course, if there is an endogenous labor/leisure choice
or money were introduced through a transactions costs technology, but the adjustment
would be instantaneous and there would be no dynamics.

2
in the non-tradables goods sector, higher in‡ation, a real depreciation of the
domestic currency, and a fall in the domestic real interest rate. Intuitively,
the presence of sticky prices prevents the price index (which comprises the
prices of both tradables and non-tradable goods) to fully respond on impact
to the increase in the money supply. As a result, there is an incipient excess
supply in the money market, which requires an increase in consumption of
non-tradable goods. This higher consumption is brought about by a fall in
the relative price of non-tradable goods (i.e., a real depreciation).
In Section 4, we turn our attention to predetermined exchange rates.
The main result is that, under sticky prices, a devaluation is expansionary.
Intuitively, sticky prices imply that, on impact, a nominal devaluation leads
to a real devaluation (i.e., a fall in the relative price of non-tradable goods).
As a result, demand for non-tradable goods increases, which results in an
output expansion.
Finally, in Section 5, we focus on one of the more in‡uential results to
come out of the Mundell-Fleming tradition: Dornbusch’s (1976) exchange
rate overshooting. In the model developed in Section 3, the nominal ex-
change rate increases on impact by the same proportion as the money supply
and hence by the same amount that it will increase in the long-run. In other
words, there is no overshooting in the sense of Dornbusch (1976). Section 5
studies a slightly more general version of the model of Section 3 which, by
generating a money demand with a consumption-elasticity that is not nec-
essarily equal to one, can yield both overshooting or undershooting of the
nominal exchange rate. In fact, under the more relevant parameter con…gu-
ration, the nominal exchange rate “overshoots”in the short-run its long-run
level. This is a remarkable result because it can explain short-run volatility
in nominal exchange rates that goes beyond that of the “fundamentals” (in
this case the money supply).

3
2 A sticky prices model
We now incorporate sticky prices into the model of Chapter 5.3 Sticky prices
enter the picture through the supply side so there are no modi…cations of
substance on the consumer side.
Consider a small open economy inhabited by a large number of identi-
cal, in…nitely-lived consumers, who are endowed with perfect foresight. The
economy is perfectly integrated with the rest of the world in both goods and
capital markets. There exist two physical goods (one tradable and the other
nontradable, both nonstorable). The law of one price holds for the tradable
good and the foreign price of the tradable good is assumed to be one; hence,
P T = E. The economy can borrow/lend at a constant world real interest
rate, r. As in Chapter 5, money is introduced in the utility function.

2.1 Consumers
The only two minor modi…cations to the consumer side of Chapter 5 are the
adoption of logarithmic preferences and the introduction of a non-tradable
good. Preferences are thus given by
Z 1
log(cTt ) + log(cN
t ) + log(zt ) e
t
dt; (1)
0

where cTt
and cN
tdenote consumption of tradable and non-tradable goods,
respectively, and zt ( Mt =Pt ) denotes real money balances in terms of the
price index P , given by4
p
P PTPN: (2)
Since money balances enter the utility function to capture the liquidity ser-
vices provided by money, it seems natural to de‡ate nominal balances by a
3
Here we follow Calvo and Vegh (1993). A similar model – but in discrete time and
with a more re…ned supply-side –can be found in the appendix of Obtsfeld and Rogo¤’s
(1995) paper. The short-run dynamics behind the two models –including the conditions
for overshooting/undershooting examined in Section 5 –are the same because both models
assume that output is demand-determined.
4
Notice that, to be consistent with our notation in earlier chapters, we will use z to
denote real money balances in terms of the price index and continue to use m to denote
real money balances in terms of tradable goods. (We will continue to use tradable goods
as the numeraire.)

4
price index because the consumer consumes both goods.5
Let at ( mt + bt ) denote real …nancial assets in terms of tradable goods,
where bt denotes net foreign bonds held by consumers. The consumer’s ‡ow
constraint is then given by

ytN cN
t zt
a_ t = rat + ytT + + t cTt it p ; (3)
et et et
where ytT and ytN denote output of tradable and non-tradable goods, respec-
tively, et ( P T =P N ) is the real exchange rate, and it is the nominal interest
rate. To understand the last term on the RHS of constraint (3), notice that
since we continue to use tradable goods as our numeraire, the opportunity
cost of holding real money balances –given by i(M=E) –can be expressed as
i(M=P )(P=E) which, taking into account (2) and the de…nition of the real
p
exchange rate, equals iz= et .
Integrating (3) and imposing the appropriate transversality condition, we
obtain

Z 1 Z 1
yN cN zt
a0 + ytT + t + t e rt
dt = cTt + t
+ it p e rt
dt: (4)
0 et 0 et et

The consumer chooses cT , cN , and z to maximize (1) subject to the intertem-

poral constraint (4). In terms of the Lagrangian:

Z 1
L = [log(cTt ) + log(cN
t ) + log(zt )]e
t
dt
0
Z 1 Z 1
ytN cN zt
+ a0 + T
yt + + t e dt rt
cTt + t + it p e rt
dt :
0 et 0 et et

The …rst-order conditions with respect to cT , cN , and z are given, respectively,

by (assuming, as usual, that = r)
5
This price index corresponds to the minimum nominal expenditure required to achieve
a given level of utility; see Appendix 7.3 in Chapter 6 for the derivation. As also discussed
in Chapter 6, in the case of logarithmic preferences it does not make a di¤erence whether
we introduce z or m in the utility function. We choose the speci…cation with z because
the model will be generalized to a CES speci…cation in z below.

5
 1
= ; (5)
cTt
1
= ; (6)
cN
t et
1 it
= p : (7)
zt et

Combining (5) and (6), we obtain the condition :

cN
t
= et : (8)
cTt
This condition should, of course, be familiar from Chapter 4 and says that,
at an optimum, the marginal rate of substitution between tradables and non-
tradables equals the relative price.
To obtain the money demand, combine (6) and (7) to obtain:

cN
zt = p t . (9)
et it
The demand for real money balances in terms of the price index, z, depends
positively on home goods consumption expressed in terms of the price index
p
–given by cN P N =P = cN = et –and negatively on the nominal interest rate.
For further reference, let us also derive the money demands in terms of
tradable goods and non-tradable goods. By combining (5) and (7), we obtain
the demand for real money balances in terms of tradable goods:

cTt
mt = ; (10)
it
which is, of course, familiar from previous chapters.
Finally, to derive the money demand in terms of non-tradable goods,
p
notice that z et = M=P N and rewrite (9) as:

cNt
nt = , (11)
it
where nt Mt =PtN . The usefulness of de…ning a money demand in terms of
non-tradable goods will become apparent below.

6
2.2 Supply side
The supply of tradable goods is assumed to be constant over time and equal
to y T . The major departure from the frictionless world of Chapter 5 is that
prices in the non-tradable goods sector (P N ) are assumed to be sticky (i.e.,
they cannot change at any given point in time but can, of course, change
over time). Since prices are given at any point in time, they will not be
able to adjust to clear the market in response to shocks that may lead to
excess supply or demand of non-tradable goods. Instead, we will assume
that quantities adjust to clear the market for non-tradable goods. More
speci…cally, we will assume that output of non-tradable goods is demand-
determined and hence supply always adjusts to demand.
Formally, sticky prices are introduced into the model via Calvo’s (1983)
staggered prices formulation, which is a continuous-time version of overlap-
ping contracts models à la Fischer (1977)-Taylor (1979, 1980). In this formu-
lation, the rate of change of the in‡ation rate is a negative function of excess
aggregate demand:

_t = (ytN yfN ); > 0; (12)

where ytN is aggregate demand and yfN is the “full-employment” level of
output. In this formulation, the price level is sticky (i.e., it is predetermined
at each instant in time), but the in‡ation rate is fully ‡exible because it
is a forward-looking variable.6 Equation (12) can be derived by assuming
that …rms set prices in a non-synchronous manner taking into account the
future path of aggregate demand and the average price level prevailing in the
economy (see Appendix 8.1 for the derivation of equation (12)).
Intuitively, think of …rms as being able to change their individual prices
only if they receive some random signal. In this context, suppose that there
is an increase in aggregate demand. Then some …rms – those which do
receive the random signal –will be able to change their individual price (and
hence in‡ation will rise). Most …rms, however, will not be able to change
6
If this is the …rst time that you encounter Calvo’s (1983) staggered-prices formulation,
you may be somewhat surprised to see that the change in the rate of in‡ation is a negative
function of excess aggregate demand. This, however, makes perfect sense in light of the
intuition given below. The key is that in this set-up the in‡ation rate itself is fully ‡exible.
If the in‡ation rate were sticky, then one would need to assume that the change in the
in‡ation rate is a positive function of excess aggregate demand for the problem to be
well-de…ned.

7
prices and hence the price level itself will not change. Next instant, there
will be some …rms that could not change their prices before that will be able
to do so. But since the random signal follows an exponential distribution,
the number of …rms changing prices will be smaller than those that changed
prices immediately after aggregate demand went up. Hence, the in‡ation
rate will rise by less (relative to the pre-shock situation) than before, which
explains why the rate of in‡ation falls over time.7

2.3 Government
The government is unchanged relative to Chapter 5. Its ‡ow constraint is
therefore given by:

h_ t = rht + m
_ t + "t m t t. (13)
The corresponding intertemporal constraint is
Z 1 Z 1
rt rt
h0 + (m
_ t + "t mt )e dt = te dt. (14)
0 0

2.4 Equilibrium conditions

Since perfect capital mobility prevails, the interest parity condition holds:

it = r + " t . (15)
Equilibrium in the non-tradable goods market requires that

cN N
t = yt : (16)
Recall that, in the current set-up, output of non-tradable goods is demand-
determined, so “equilibrium” in the non-tradable goods market holds by
construction.
By de…nition, e = E=P N . Hence,
e_ t
= "t t: (17)
et
7
But how sticky are prices in practice? Box 2 reviews the available empirical evidence.

8
This dynamic equation simply states that if tradable goods in‡ation (given
by "t ) is, say, larger than non-tradables goods in‡ation ( t ), then the relative
price of tradable goods will be increasing over time (i.e., e_ t > 0).
As in Chapter 4, let us now de…ne the real interest rate in terms of non-
tradable goods, rd , as
e_ t
rtd r+ . (18)
et
If, say, the relative price of tradable goods is rising over time (i.e., e_ t > 0),
then rd > r. Intuitively, if you have invested in a tradable bond, your return
in terms of non-tradable goods will be higher if you are able to purchase more
non-tradable goods later on.
To derive the Euler equation for non-tradable goods, totally di¤erentiate
(6) and use (18) to obtain

c_N
t
= rtd r.
cNt

As discussed in Chapter 4, if, say, rd > r, today’s consumption of non-

tradable goods is lower than “tomorrow’s”because the return on postponing
consumption (rd ) is higher than the utility cost of deferring consumption ( ,
which equals r).
Combining the consumers’‡ow constraint (given by (3)) with the govern-
ment’s (given by (13)) and imposing equilibrium in the non-tradable goods
market, we obtain

k_ t = rkt + y T cTt ;
where k( b + h) denotes the economy’s total net foreign assets.
By the same token, combining the consumer’s and the government’s in-
tertemporal constraints –given by (4) and (14), respectively –and imposing
equilibrium in the home goods market, we obtain
Z 1
yT
k0 + = cTt e rt dt: (19)
r 0

9
3 Flexible rates
3.1 Perfect foresight equilibrium
Suppose that the economy is operating under ‡exible exchange rates. Hence,
we assume that ht = 0 for all t. We will now solve for the perfect foresight
equilibrium path for a constant rate of money growth, . To this e¤ect, we
proceed in three stages. In the …rst stage, we will show that consumption
of tradables is constant (and, in fact, independent of monetary policy). In
the second stage – and as we have done already in Chapters 5 and 7 – we
will show that the path of m is governed by an unstable di¤erential equation
and argue that a convergent perfect foresight equilibrium path requires that
m be constant over time. In the third and …nal stage, we will set a dynamic
system in n (real money balances in terms of non-tradable goods) and to
solve for the rest of the model.

3.1.1 Consumption path of tradable goods

First-order condition (5) makes clear that cT will be constant over time.
Using the resource constraint (19), this constant value, denoted by cT , will
be given by

cT = rk0 + y T . (20)
For further reference, notice that the path of consumption of tradables will
be given by (20) regardless of the path of the money supply and, furthermore,
will not a¤ected by any (anticipated or unanticipated) change in monetary
policy. Hence, from (5), the same is true of the multiplier .

3.1.2 Real money balances

By de…nition, m = M=E. (Recall that, by assumption, the foreign price of
the tradable good is unity.) Hence,
m
_t
= "t : (21)
mt
Solving for "t from the interest parity condition (15) and using (10) –taking
also into account (20) –we obtain

10
 cT
"t = r:
mt
Substituting this last expression into (21), we obtain a linear di¤erential
equation for m:

m
_ t = (r + )mt cT ; (22)
where cT is a constant given by (20). Given that
@m
_t
=r+ > 0;
@mt
the di¤erential equation (22) is unstable. It follows that for mt to follow a
convergent path, m_ t = 0 for all t 0. Hence, along a perfect foresight path
with a constant , m will be constant and equal to

cT
m= . (23)
r+
An important implication is that, in response to unanticipated and perma-
nent changes in , real money balances will need to adjust instantaneously
to their new steady state. If they did not, they would diverge over time.
Finally, notice that since m_ t = 0 along a perfect foresight path with
constant , it follows from (21) that "t will also be constant:

"= .
Hence, the nominal interest rate is also constant and given by

{=r+ .

3.1.3 Dynamic system

To solve for the rest of the system, we will set up a dynamic system in n
(real money balances in terms of non-tradable goods, as de…ned in (11)) and
8
t . The variable n is a predetermined variable since M is exogenous (i.e.,

8
The reader may wonder why we need to introduce a di¤erent measure of real money
balances for the purposes of setting up the dynamic model. The answer is that when it
comes to setting up a dynamic system, it is convenient to have a predetermined variable.
In this case, n is such a variable.

11
controlled by the monetary authority) and prices of non-tradable goods are
sticky (i.e., cannot jump at any point time). Since, by de…nition, n M=P N ,
it follows that

n_ t = nt ( t ); (24)
where, by de…nition, t ( P_ N =P N ) is the rate of in‡ation of non-tradable
goods.
To derive our second dynamic equation, …rst use (16) to rewrite (12) as

_t = yfN cN . (25)
Hence, in‡ation will be rising if consumption of non-tradable goods is below
its full-employment level and viceversa.
Taking into account (8) and noting that et = nt =mt , equation (25) can
be rewritten as as:
!
c T
_t = yfN nt : (26)
m
Equations (24) and (26) constitute a dynamic system in n and , for given
values of and m.9
To characterize the steady-state of the dynamic system, set n_ t = _ t = 0
in (24) and (26), respectively, to obtain:

ss = ; (27)
yfN m
nss = : (28)
cT
Linearizing the system around the steady-state, we obtain:
:
nt 0 nss nt nss
: = cT :
t m
0 t

The determinant of the matrix associated with the linear approximation is

negative:
9
Notice that, as far as the dynamic system is concerned, m is like a parameter since it
will adjust instantaneously in response to permanent changes in .

12
 nss cT
= < 0:
m
The system has therefore one positive and one negative root and thus exhibits
saddle-path stability.
As in Chapter 6, we now proceed to characterize the qualitative behavior
of this dynamic system by resorting to a phase-diagram. To construct the
phase diagram, we …rst draw the n_ t = 0 and _ t = 0 loci. To obtain these
curves, set n_ t = 0 in equation (24) and _ t = 0 in (26) to obtain, respectively,

t = ;
yfN m
nt = :
cT
Hence, the n_ t = 0 locus shows up in the phase diagram (Figure 1) as a hori-
zontal line, while the _ t = 0 locus shows as a vertical line. The intersection
of both loci at point A determines the steady-state of the system. As we saw
in Chapter 6, the n_ t = 0 and _ t = 0 curves de…ne four regions. Proceeding
in analogous fashion, we can draw the arrowheads shown in Figure 1 and
conclude that the saddle path will be positively-sloped.

[Figure 1]
The variable n is predetermined in the sense that it cannot jump in an
endogenous way. Hence, if n at time 0 were above nss (denoted by n0 in
Figure 1), then in‡ation at 0 would also be above (at a value given by 0
in Figure 1) so as to position the system along the saddle path at a point like
B. Hence, both and n would fall during the adjustment process towards
point A. Since _ < 0 during the adjustment, cN would be above its full-
employment level, as equation (25) makes clear. Conversely, if n at time 0,
were below nss , (at a value given by, say, n;0 in Figure 1), then the rate of
in‡ation would adjust endogenously to a value below ( ;0 in Figure 1) so
as to position the system on the saddle path at a point like C. From then on,
the system would travel along the saddle path towards point A. Both and
n would thus increase during this adjustment process. Since _ > 0 during
the adjustment to the steady-state, cN would be below its full-employment
level, as equation (25) makes clear.

13
 We thus conclude that the model endogenously generates a Phillips-curve
type relationship in the sense that when in‡ation is above , the economy is
operating above its full-employment level (i.e., cN N
t > yf ) and when in‡ation is
below , the economy is operating below its full-employment level (i.e., cN t <
yfN ). While traditional sticky-prices model postulate such a relationship, here
it arises as an equilibrium phenomenon.10

3.2 Permanent increase in the money supply

Suppose that, just before t = 0, the economy is in the stationary equilibrium
corresponding to = 0 and point A in Figure 1. At time 0, there is an unan-
ticipated and permanent increase in the stock of money supply, M (Figure
2, Panel A). How does the economy react?

[Figure 2]
As we have already established, cT will not change. It is also the case
that an increase in M will not a¤ect the steady-state value of m, as (23)
makes clear. Hence, the nominal exchange rate will increase by the same
proportion as the nominal money supply so as to keep m constant.
In terms of the dynamic system, the increase in M will not a¤ect the
steady-state values of n and , as follows from equations (27) and (28).
Hence, the steady-state of the system will remain at point A in Figure 1.
On impact, however, n will increase to a value like n0 in Figure 1 because
the nominal money supply goes up and the price of non-tradables goods is
sticky. Given n0 , the in‡ation rate will have to jump to 0 so as to position
the system on the saddle path (point B in Figure 1). After this initial jump
from point A to point B, the system travels back to point A along the saddle
path. The corresponding paths of n and as a function of time are depicted
in Figure 2, Panels B and C.
To …nd out the path of the real exchange rate, recall that e = n=m. Since
m does not change, e will behave in the same way as n, jumping up on impact
(real depreciation) and then falling back to its initial steady-state (Figure 2,
Panel D).
10
This is true, of course, as long as the system is on the saddle-path. Interestingly, if
the economy is not on the saddle-path (which can happen in response to a temporary
shock), then this Phillips-curve relationship will not necessarily hold, which could explain,
for example, periods of “stag‡ation”(i.e., high in‡ation and underutilization of resources)
as illustrated by Exercise 1 at the end of this chapter.

14
 Given condition (8) and the fact that cT does not change, the behavior
of cN will mimic that of e, as illustrated in Figure 2, Panel E.
Finally, what will happen to the domestic real interest rate, rtd it t?
d
Since the nominal interest rate remains constant, the behavior of rt will be
dictated by the behavior of in‡ation. Hence, the domestic real interest rate
falls on impact and then gradually reverts back to its initial steady-state, r
(Figure 2, Panel F).
In sum, a permanent increase in the stock of the money supply leads to
higher in‡ation, a real depreciation, an expansion in the non-tradable goods
sector, and a fall in the domestic real interest rate. This is, of course, in
sharp contrast to the world of Chapter 5 where a permanent increase in the
money supply had no real e¤ects.
What is the economic intuition behind these e¤ects? For these purposes
–and using (11) –let us focus on the money market equilibrium for n at time
0:

M0 cN
0
= : (29)
P0N i0
|{z} |{z}
real money supply real money demand

As indicated below the equation, think of the LHS as real money supply and
the RHS as real money demand. The critical implication of price stickiness
is that an increase in the nominal money supply translates into an increase
in the real money supply. Hence, for unchanged real money demand, there
would be an incipient excess supply in the money market. In their attempt to
get rid of unwanted money balances by purchasing foreign bonds, households
would bid up the domestic price of foreign bonds (E). Since P N is sticky,
this nominal depreciation translates into a real depreciation (i.e., a fall in the
relative price of non-tradable goods). As a result, demand for non-tradables
increases which – given that output is demand-determined – leads to an
output expansion in the non-tradable sector.
In the long-run, however, real money demand will not change relative
to its pre-shock value. Hence, real money supply must fall over time to its
pre-shock value. For this to happen, in‡ation of non-tradable goods, , must
increase on impact above the (unchanged) rate of money growth, . This
high in‡ation –coupled with no changes in the nominal exchange rate after
the initial jump –explains the real appreciation that takes place over time.
Finally, notice that in this model the nominal exchange rate increases

15
by the same proportion as the nominal money supply. There is thus no
overshooting in the sense of Dornbusch (1976). Overshooting would occur
if, on impact, the nominal exchange rate increased by more than the money
supply does. Why does this model not generate overshooting? The answer to
this question will become clear once we show in Section 5 how a version of this
model with more general preferences can indeed generate both overshooting
and undershooting.

3.3 Permanent reduction in the rate of money growth

Suppose, once again, that an instant before time 0, the system is in the
steady state characterized above (with the rate of money growth given by
H
). At t = 0, there is an unanticipated and permanent reduction in the rate
of money growth from H to L ( L < H ) (Figure 3, Panel A). Consumption
of tradable goods – given by (20) – does not change because, as discussed
above, its level is independent of monetary policy. From (23), we can see that
real money balance in terms of tradable goods (m) will be higher in the new
stationary state because the opportunity cost of holding money has fallen.
Given the instability of the di¤erential equation governing the behavior of
real money balances, m must adjust instantaneously to its higher value. If it
did not, it would diverge over time. The rate of depreciation will therefore
also fall down immediately to its new steady-state value and so will the
nominal interest rate.

[Figure 3]
In terms of the dynamic system, suppose that the system is initially at
point A in Figure 4. At point A, the steady-state rate of in‡ation of non-
tradable goods is equal to H and the corresponding real money balances are
nss ( H ). In the new steady-state –and as (27) and (28) make clear –n will
be higher and will be lower (point B in Figure 4). Hence, on impact the
system must jump from point A to point C and then travel along the saddle-
path towards point B. The corresponding paths of n and are illustrated in
Figure 3, Panels B and C, respectively. On impact, the in‡ation rate falls by
more than it will in the long-run.

[Figure 4]

16
 To derive the path of the real exchange rate, recall that e = n=m. Since
m increases at time 0, e will jump down at time 0 (real appreciation), as
illustrated in Figure 3, Panel E. Given that the real exchange rate does not
change across steady-states, it will need to gradually rise back to its initial
steady-state. Given equation (8), consumption of non-tradable goods follows
the same path as the real exchange rate (Figure 3, Panel D). Finally, the
domestic real interest rate will increase on impact because the fall in in‡ation
is larger than the fall in the nominal interest rate (Figure 3, Panel F). It then
falls gradually to its unchanged steady-state.
Summing up, an unanticipated reduction in leads to a recession, real
exchange rate appreciation, and higher real interest rates. Interestingly –and
as documented in Chapter 14 –these are the main stylized facts associated
with money based stabilizations.
The economic intuition behind the results just discussed is as follows.
Think again in terms of the money market equilibrium described by equation
(29). The reduction in the money growth rate lowers the nominal interest
rate and hence increases real money demand. Real money supply, however,
does not change on impact. Hence, for the initial level of cN , there is an
excess demand for money. As a result, the public will try to get rid of foreign
bonds in order to acquire money, which pushes down the domestic price of
foreign bonds, E. Since the price of non-tradable goods is sticky, the fall
in E translates into a reduction in e. This increase in the relative price of
non-tradable goods reduces their demand, which leads to an output fall.

4 Predetermined exchange rates

We now solve the model for the case of predetermined exchange rates and
use it to ask the question: How does the economy respond to a permanent
devaluation? Contrary to our results in Chapter 6 – where a devaluation
led to a reduction in consumption –in this sticky-prices model a devaluation
will lead to an expansion in aggregate demand for non-tradable goods and
therefore output.
We begin by solving for the perfect foresight equilibrium corresponding
to a constant rate of devaluation. We then turn our attention to a permanent
devaluation and a permanent reduction in the rate of devaluation.

17
4.1 Perfect foresight equilibrium
Let us now characterize the perfect foresight equilibrium for a constant value
of the rate of devaluation, ".
First, notice that, as in the ‡exible exchange rates case, consumption of
tradables will be constant and given by (20). Furthermore, consumption of
tradables will not be a¤ected by changes in the rate of devaluation.
From the interest parity condition (15), the nominal interest rate will be
constant as well and given by

{ = r + ".
Given that both cT and { are constant, the money demand equation (10) tells
us that m will be constant as well.
To solve for the rest of the variables, we need to set up a di¤erent dy-
namic system from the one we used above for ‡exible exchange rates for
the following reasons. For starters, that system has the variable which
is now an endogenous variable. Furthermore, real money balances in terms
of non-tradables goods (n) are no longer a predetermined variable under
predetermined exchange rates because the nominal money supply is an en-
dogenous variable. As a methodological matter, it would not be wise to set
a dynamic system with two jumping variables because it would be harder to
solve. We thus need to …nd a variable that will be predetermined under pre-
determined exchange rates and sticky prices. A moment’s re‡ection should
reveal that the obvious candidate is the real exchange rate, e. Since, by de-
…nition, e = E=P N , the real exchange rate will be a predetermined variable
in standard models of predetermined exchange rate under sticky prices.
We will therefore set up a dynamic system in and e. To obtain the …rst
dynamic equation, substitute (8) into (25) to obtain:

_t = yfN et cT : (30)
The second dynamic equation is given by (17).
The system’s steady state is given by:

ss = "; (31)
yfN
ess = : (32)
cT

18
Linearizing the system around the steady-state, we obtain:

e_ t 0 ess et ess
: = :
t cT 0 t "
The determinant of the matrix associated with the linear approximation
of the system is given by

= cT ess < 0;
which implies that the system is saddle-path stable.
Proceeding as before, it is easy to construct the phase diagram illustrated
in Figure 5. As depicted, the saddle-path is positively sloped.

[Figure 5]
The path of the remaining variables along a perfect foresight equilibrium
will depend on the initial value of the real exchange rate. Suppose that the
initial value of the real exchange rate were given by e0 in Figure 5. Then the
in‡ation rate would need to be 0 so as to position the system at point B
in Figure 5. Both the real exchange rate and in‡ation would fall over time
towards the steady-state, given by point A. Given condition (8), consumption
of non-traded goods would behave in the same way as the real exchange rate
and hence fall over time.

4.2 Permanent devaluation

Suppose that, just before t = 0, the economy is in the steady-state character-
ized above. At t = 0, there is an unanticipated and permanent devaluation
(Figure 6, Panel A). How does the economy react?

[Figure 6]
In terms of the dynamic system, it is clear from equations (31) and (32)
that the devaluation does not a¤ect the steady-state values of the in‡ation
rate of non-tradable goods or the real exchange rate. On impact, however,
the real exchange rate will increase. In other words, in this particular model,
a nominal depreciation leads on impact to a real depreciation.11 The real
11
The model is thus able to explain the high correlation between nominal and real
exchange rates, as documented by Mussa (1986). See Chari, Kehoe, and McGrattan (2002)

19
exchange rate thus jumps on impact to a value such as e0 in Figure 5. The
in‡ation rate must then adjust so that the system positions itself along the
saddle path (at point B in Figure 5). The system then travels along the
saddle path back to its initial steady-state, point A. The corresponding path
of and e are illustrated in Figure 6, Panels C and D.
The path of consumption of non-tradable goods follows from condition
(8). Consumption of non-tradable goods increases on impact and then falls
back towards its full-employment level (Figure 6, Panel E). The domestic
real interest rate falls on impact as a result of the increase in in‡ation and
then gradually reverts back to its initial steady-state (Figure 6, Panel F).
The path of n –illustrated in Figure 6, Panel B –follows from the fact that
n = em.
We thus conclude that a devaluation is expansionary. Intuitively, the key
is that, due to sticky prices, a nominal devaluation translates into a real
devaluation. The increase in the relative price of tradable goods induces
consumers to switch expenditures towards non-tradable goods. Since output
is demand-determined, output of non-tradable goods responds immediately.
The expansionary e¤ects of a devaluation under sticky prices stand in
sharp contrast to the results that we obtained in Chapter 6 where a deval-
uation – by reducing real money balances and forcing consumers to reduce
consumption to replenish real money balances –was actually contractionary.
We have thus illustrated two possible channels through which a devaluation
may impact the real economy. In fact –and as argued in Box 1 –there are
other plausible channels, which makes the overall impact of a devaluation an
empirical matter.

4.3 Permanent reduction in devaluation rate

Suppose that the economy is initially in a situation of high in‡ation at a
point like C in Figure 5. At t = 0, there is an unanticipated and permanent
reduction in the rate of devaluation. In the new steady-state (point A in
Figure 5), the rate of in‡ation is lower and the real exchange rate remains
unchanged. How will the economy adjust from point C to point A? A mo-
ment’s re‡ection reveals that the economy must adjust instantaneously to
its new steady state. Since e is a predetermined variable, the system must
for a quantitative analysis. Remember, however, from Chapter 6 that other frictions could
generate the same outcome.

20
remain along the vertical line corresponding to ess at time 0. But if the sys-
tem placed itself at any point other than A along that vertical line, it would
diverge over time. Hence, the only possible equilibrium is for the system to
jump from point C to point A. The reduction in the devaluation rate is thus
super-neutral.
This is a remarkable result because it says that, even if prices are sticky, a
permanent reduction in the devaluation rate might reduce in‡ation at no real
costs. This model may thus be used to think about the end of hyperin‡ations
which, by and large, have involved large and sudden reductions in in‡ation
at little real costs (see Chapter 14).12

5 Overshooting
We mentioned earlier that our basic sticky–prices model does not generate
over/undershooting. This section analyzes a more general version of the
model in which both overshooting and undershooting are possible.

5.1 Consumers
The only change in the model is that the sub-utility for real money balances
now takes a CES form:
Z 1" 1 1=
#
zt 1
log(cTt ) + log(cN
t )+ e t dt; (33)
0 1 1=
where is a positive parameter that, as will become clear below, will capture
the consumption and interest-rate elasticity of real money demand.
The intertemporal constraint remains given by (4). The …rst-order con-
ditions for cT and cN continue to be given by (5) and (6). The …rst-order
condition for z now reads:

1= it
zt = p : (34)
et
12
We should notice the economy’s very di¤erent response to a permanent reduction in
the rate of devaluation compared to the response to a permanent reduction in the rate
of money growth. Interestingly enough – and as shown in Exercise 2 at the end of this
chapter –under logarithmic preferences the response to a change in …scal policy would be
the same under either regime.

21
 Using this …rst-order condition to solve for z –taking into account (6) –
we obtain the real money demand (in terms of the price index):

cN
zt = pt . (35)
et it
The parameter thus denotes the consumption and interest-rate elasticity
of money demand. When = 1, the model reduces to the case analyzed in
Section 3 (recall equation (9)).
For further reference, it is also convenient to obtain the real money de-
mand in terms of tradable goods, m. Recalling that m M=E and taking
into account (2), equation (35) can be rewritten as:
" #
(1=2)(1 1= ) T
et ct
m= : (36)
it
Once again, notice that when = 1, equation (36) reduces to (10).

5.2 Dynamic system

We now proceed to solve this model for the case of ‡exible exchange rates in
the same way as we did before. A key di¤erence, however, will be that, due
to the CES preferences for money, the resulting dynamic system in m, n, and
will not be block recursive. In other words, with logarithmic preferences
for z, the system breaks down into a single di¤erential equation for m and a
system of two di¤erential equations for n and . This will not be the case
for CES preferences. As a result, we will have no choice but to set up a
three-equation di¤erential-equation system in m, n, and . p
For starters, solve for it from (34) –taking into account that Et =Pt = eto
obtain:
(1=2)(1 1= )
et
it = 1=
:
mt
Substituting this last equation into (21) – taking into account the interest
parity condition–we obtain
" #
(1=2)(1 1= )
n t
m
_ t = mt + r cT 1= +(1=2)(1 1= ) ; (37)
mt

22
where cT is the constant value of consumption of tradable goods given by
(20). This di¤erential equation, together with (24) and (26), constitute a
dynamic system of three di¤erential equations in n, , and m.
To characterize the system’s steady-state, set m
_ t = n_ t = _ t = 0 in (24),
(26), and (37) to obtain:

2 1
3
6 cT yfN 7
mss = 4 5 ; (38)
+r

yfN mss
nss = ; (39)
cT
ss = ; (40)
where

1 1
1 (41)
2
is a parameter that will be critical for the dynamics of mt .
Linearizing the system around the steady-state, we obtain:

0 1 0 1 10 1
m
_t (1= + )cT 1=nss+ cT 1= +
nss
1 0 mt mss
mss mss
@ nt A = B
:
@ 0 0 nss
C@
A nt nss A :
:
cT nss cT
t
m2ss mss
0 t

Taking into account (38) and (39), we can rewrite this dynamic system as

0 1 0 mss
10 1
m
_t (1= + )( + r) ( + r) nss
0 mt mss
: B
@ nt A = @ 0 C@
0 nss A nt nss A :
: yfN yfN
t
mss nss
0 t

The trace and the determinant of the matrix associated with the linear
approximation are given by, respectively,

T r = (1= + )( + r) > 0;
= ( + r) yfN < 0;

23
where we have used (39) to simplify the expression for the determinant.
Since the determinant is negative (recall that the determinant is equal to the
product of the roots), the system could have either three negative roots or
one negative and two positive roots. The fact that the trace (which equals
the sum of the roots) is positive, however, rules out the case of three negative
roots. We thus conclude that the system has one negative and two positive
roots.
Let denote the negative root associated with this system. Denoting by
(h1 ; h2 ; h3 ) the characteristic vector associated with the root , we can write:

0 mss
10 1 0 1
(1= + )( + r) ( + r) nss
0 h1 0
B 0 C@ A @
@ N nss A h2 = 0 A
yf yfN
h3 0
mss nss

From the …rst and second rows, respectively, it follows that

h1 ( + r) mss
nss
= ;
h2 (1= + )( + r)
h2 nss
= > 0:
h3
Notice that, as discussed in detail below, the sign of h1 =h2 depends on the
sign of and hence on .
Setting to zero the constants corresponding to the positive roots, the
solution to the dynamic system is given by:

mt mss = !h1 e t ; (42)

nt nss = !h2 e t ; (43)
t = !h3 e t : (44)
Combining (43) and (44), we get
nt nss h2
=> 0; (45)
t h3
which tells us that, along a perfect foresight path, nt and will move in the
same direction. This should not come as a surprise since, in the logarithmic
version of this model studied above, this was also the case.

24
 Combining (42) and (44), we obtain
mt mss h1
= .
nt nss h2
Based on this expression, we can distinguish three possible cases:

1. = 1. In this case, = 0, as follows from (41). Hence h1 =h2 = 0,

which implies that m is always equal to its steady-state value. This is,
of course, the case analyzed earlier in this chapter.

2. < 1. In this case, < 0, as follows from (41). Hence, h1 =h2 < 0.
This implies that, along a perfect foresight path, mt and nt will move
in opposite directions. Together with (45), this implies that mt and t
also move in opposite directions.

3. > 1. In this case, > 0, as follows from (41). Hence, h1 =h2 > 0.
This implies that, along a perfect foresight path, mt and nt will move
in the same direction. Together with (45), this implies that mt and t
also move in the same direction.

5.3 Permanent increase in the money supply

Suppose that the economy is initially at the steady-state given by (38), (39),
and (40), with = 0. At t = 0, there is an unanticipated and permanent
increase in the nominal money supply. To …x ideas, suppose that the money
supply doubles from M to 2M . How will this economy react?
The …rst observation is that such a change does not alter the system’s
steady-state.
On impact, n will increase because P N is a sticky variable. Further, since
the system has only one negative root, it will adjust monotonically to its
unchanged steady-state. To show this formally, normalize h2 to one, and
evaluate (43) at t = 0 to obtain:

n0 nss = ! > 0:
Substituting this last piece of information into (43) and di¤erentiating with
respect to time, we obtain:

n_ t = (n0 nss ) e t < 0.

25
We have already established that will move in the same direction as n.
Hence, will increase on impact and fall gradually over time.
How will m behave? We need to consider three cases:

1. = 1. In this case, mt = mss for all t. This case corresponds to the

one analyzed in Subsection 3.2. Since m does not change on impact,
the nominal exchange rate increases by the same proportion as the
nominal money supply. In terms of Figure 7, the nominal exchange
rate will double on impact from E to 2E (point A) and stay there.
Since the rate of devaluation is zero, there is no change in the nominal
interest rate.

2. < 1. In this case – and as established above – mt and t move in

opposite directions. It follows that mt will fall on impact and then rise
over time. The fall on impact in mt implies that the nominal exchange
rate rises by more than the nominal money supply. In the long run,
however, the nominal exchange rate increases by the same proportion.
Hence, on impact, the nominal exchange rate overshoots its long-run
level. In terms of Figure 7, the nominal exchange rate jumps on impact
to a point such as B and then falls over time to its long-run level. Since
the nominal exchange rate falls over time, "t < 0, which implies that
the nominal interest rate falls on impact.

3. > 1. In this case –and as established above –mt and t move in the
same direction. It follows that mt will increase on impact and then fall
over time. Hence, on impact, the nominal exchange rate increases by
less than the nominal money supply. The exchange rate thus under-
shoots its long-run level. In terms of Figure 7, the nominal exchange
rate jumps to a point like C and then increases over time. Since the
nominal exchange rate increases over time, "t > 0, which implies that
the nominal interest rate increases on impact.

[Figure 7]
What is the intuition behind the over/undershooting results? Recall the
real money demandp equation given by (35) – rewritten below taking into
N N N
account that c = e = P c =P –and, once again, interpret it as the equi-
librium condition in the money market with the LHS capturing real money

26
supply and the RHS denoting real money demand:

Mt PtN cNt =Pt

= . (46)
Pt it
|{z} | {z }
real money supply real money demand

At t = 0, the nominal money supply doubles. Suppose that, in response

to this increase in the money supply, the nominal exchange rate also doubled
(i.e., E^ = M
^ ), thereby adjusting instantaneously to its long-run equilibrium
level.13 This implies, of course, that the rate of depreciation would be zero
and that the nominal interest rate would not change. Would the money
market p be in equilibrium? To answer this question, …rst notice that since
P = P T P N and P N cannot jump, the real money supply would increase
by M ^ E=2 ^ =M ^ =2. In other words, the real money supply would increase
by 50 percent.
To …nd out the change in real money demand, we need to establish the
change in the demand for non-tradable goods. Since P N does not change,
the real exchange rate, et ( P T =P N ), increases by the same proportion as
the nominal money supply (i.e., it doubles). Given equation (8), the demand
for non-tradable goods also doubles (recall that cT is invariant). In terms of
the price index, however, real consumption of non-traded goods increases by
M^ E=2 ^ =M ^ =2. In other words, real consumption of non-traded goods less
than doubles. Since the consumption-elasticity is , the resulting increase in
real money demand is therefore M ^ =2.
We thus conclude that, if the nominal exchange rate increased by the
same proportion as the money supply, real money supply would increase by
M^ =2 and real money demand by M ^ =2. Hence, the excess supply in the
money market would be given by M ^ =2 M^ =2, which is simply (1 ^ =2.
)M
Based on this simple expression, it follows immediately that:

If = 1, both real money supply and real money demand increase

by the same amount and therefore the equiproportional increase in
the nominal exchange rate (i.e., no overshooting or undershooting) is
consistent with money market equilibrium.

If < 1, real money supply would increase by more than real money
demand and there would be excess supply in the real money market.
13
A “hat” over a variables denotes proportional change.

27
 This, of course, is not an equilibrium. The excess supply of money
requires a fall in the nominal interest rate. For this to happen, the
rate of depreciation must become negative (i.e., agents must expect a
nominal appreciation of the currency). For the rate of depreciation to
become negative, the nominal exchange rate must overshoot its long-
run level and fall over time.

If > 1, real money supply would increase by less than real money de-
mand and there would be excess demand for money. To equilibrate the
money market, the nominal interest rate needs to increase. From the
interest parity condition, this requires a depreciation of the currency.
For this to happen, the nominal exchange rate must jump by less than
its long-run level.
Finally, two observations are worth making. First, in general equi-
librium, a sticky-prices model does not necessarily lead to a liquidity
e¤ect (i.e., an increase in M leading to a fall on impact of the nom-
inal interest rate). In fact –and as we just saw– an increase in M is
consistent with i falling, increasing, or remaining unchanged. Second,
in the model the co-movement on impact between the nominal interest
rate and the level of the exchange rate is also ambiguous. The model
does not support the notion –typically found in the …nancial press and
undergraduate textbooks –that a depreciation of the currency will be
necessarily associated with a fall in nominal interest rates.

6 A model of sticky wages

While the model with sticky-prices developed above is extremely useful to
ask a myriad of positive (as opposed to normative) questions –such as what
happens when there is an increase in the money supply or a devaluation –it
is not well-suited to ask normative questions. The reason is that, outside the
steady-state, output is demand-determined and hence the present discounted
value of output (which would determine the average level of consumption)
does not obey any physical constraints. To address this shortcoming, this
section develops a model with a fully-speci…ed supply side in which nominal
wages –rather than prices –are sticky.14 This will provide us with a model
14
The model is a simpler version of Lahiri and Vegh (2002). See Barro and Grossman
(1971) for an early and highly in‡uential contribution on disequilibrium models.

28
that, in addition to capturing the key dynamics of an economy with nominal
rigidities, is perfectly suited to respond to normative questions.
What are the kind of normative questions that we would like to ask? Per-
haps one of the most important –and most hotly debated by policy-oriented
academic economists –is under what circumstances a country would …nd it
optimal to devalue its currency. In times of low growth and trade de…cits,
economists often call for a devaluation to address such imbalances. An excel-
lent case in point is Rudiger Dornbusch’s forceful advocacy of a devaluation
in Mexico during 1994 (see Box 3). In fact, the model developed in this
section may be viewed as the best-case scenario for a Dornbusch-type argu-
ment since, for conceptual clarity, the model focuses exclusively on nominal
rigidities and ignores other potential problems that may be associated with a
devaluation (like credibility problems). In fact, we will see that in this model
it is optimal to devalue in response to a negative real shock.
To simplify the presentation, we will consider a one-good model and
thus abstract from non-tradable goods. We will, however, introduce a la-
bor/leisure choice and, hence, endogenous production. The law of one price
holds for the only good (i.e., Pt = Et Pt ). There is no foreign in‡ation
and, to simplify notation, the foreign nominal price is taken to be unity (i.e.,
P = 1). Hence, Pt = Et . (Unless otherwise noticed, the notation is the same
as above.) The economy operates under predetermined exchange rates and,
for simplicity, the rate of devaluation is taken to be zero (i.e., the exchange
rate is …xed).

6.1 Households
Preferences are now given by
Z 1
flog[ct (`st ) ] + log(mt )g e t
dt; > 0; > 1: (47)
0

where c is consumption of tradable goods (the only good in this world), `s

denotes labor supply, and m denotes real money balances (m M=E).15
These are the so-called GHH preferences – after the paper by Greenwood,
Hercowitz and Hu¤man (1988) –that generate a labor supply function that
15
In this model, it is critical to distinguish between labor supply and labor demand
because, as discussed in detail below, the labor market may be in disequilibrium (i.e.,
labor supply may not be equal to labor demand at all points in time).

29
depends only on the real wage. In other words, there is no wealth e¤ect on
leisure, which greatly simpli…es the solution of the model.16
The household’s ‡ow budget constraint is given by

a_ t = rat + wt `st + t + t ct it m t ; (48)

where a( m + b) denotes real …nancial wealth, w( W=E) denotes the real
wage, and t are dividends from …rms (which are owned by households). The
corresponding lifetime constraint reads as
Z 1 Z 1
s rt
a0 + (wt `t + t + t ) e dt = (ct + it mt ) e rt dt: (49)
0 0
The household chooses fct ; `st ; mt g1
to maximize (47) subject to lifetime
t=0
constraint (49). In terms of the Lagrangian,

Z 1
L = [log[ct (`st ) ] + log(mt )] e t
dt
0
Z 1 Z 1
+ a0 + (wt `st + t+ t) e
rt
dt (ct + it mt ) e rt
dt :
0 0

The …rst-order conditions with respect to ct ; `st ; and mt are given by, respec-
tively,

1
= ; (50)
ct (`st )
(`st ) 1
= wt ; (51)
ct (`st )
1
= it : (52)
mt
Condition (50) is the familiar condition in models with no intertemporal dis-
tortions that says that, along a perfect foresight path, the marginal utility
of consumption will be constant. In this formulation, however, the mar-
ginal utility of consumption depends on labor supply. Hence, consumption
smoothing will not necessarily obtain in this model.17
16
You may recall that we have already encountered these preferences in Exercise 6 at
the end of Chapter 1.
17
As you may recall, this point was emphasized in Exercise 6 in Chapter 1.

30
 Substituting equation (50) into equation (51), we obtain:18

(`st ) 1
= wt . (53)
To obtain the labor supply schedule, solve for `s from the last equation to
obtain:
1
wt 1
`st = . (54)

As anticipated, labor supply depends solely on the real wage, w. Labor

supply is an increasing function of the real wage, with the elasticity given by
1=( 1).
The money demand follows from combining (50) and (52):

ct (`st )
mt = : (55)
it
GHH preferences thus generate a non-standard money demand, in that the
scale variable is consumption net of the disutility of labor.

6.2 Supply side

Firms produce tradable goods according to the following technology:

yt = `dt ; < 1; (56)

where `d denotes labor demand.
The representative …rm’s pro…ts are given by:

t = yt wt `dt . (57)
Substituting (56) into (57) yields:

t = `dt wt `dt . (58)

Firms choose `dt to maximize (58). The …rst-order condition is given by
1
`dt = wt . (59)
18
As discussed in detail below, in this disequilibrium model, actual employment may
not coincide with labor supply, in which case condition (51) will not hold. In that case,
the marginal disutility of labor will fall short of the real wage.

31
Production e¢ ciency requires that the marginal productivity of labor be
equated to the real wage. Solving for `dt from equation (59) yields the labor
demand equation:
1
1
`dt = . (60)
wt
Labor demand is a decreasing function of the real wage because a higher real
wage induces …rms to shed labor to increase its marginal productivity.

6.3 Labor market

The key action in this model takes place in the labor market. To focus the
discussion, Figure 8 illustrates the labor market. Labor demand –given by
equation (60) – is shown as a decreasing function of the real wage whereas
labor supply –given by equation (54) –is an increasing function of the real
wage. As a benchmark, we will …rst discuss the ‡exible-wages case and then
turn to the sticky-wages case.

[Figure 8]

6.3.1 Flexible wages

In the ‡exible-wages version of this model, labor market equilibrium requires
that

`s = `d .
Graphically, the economy is always at point A in Figure 8. In other words,
the economy always enjoys full-employment in the sense that all workers
that are willing to work at the prevailing real wage are indeed employed. As
indicated in Figure 8, we will denote the ‡exible-wages equilibrium values of
labor and the real wage by, respectively, `f and wf . Formally, using (54) and
(60), it follows that
h 1 1
i( 1)(1 )

wf = ( )1 ( ) 1 . (61)

Using (54), equilibrium labor can be expressed as:

32
 1

f wf 1
` = . (62)

Hence, in equilibrium, both the real wage and labor are an increasing function
of the productivity parameter, .

6.3.2 Sticky wages

If the nominal wage is sticky (i.e., it is a predetermined variable), the labor
market will not necessarily be in equilibrium.19 The reason is that, since the
economy is operating under a …xed exchange rate, a sticky nominal wage
implies a sticky real wage. To illustrate the concept of labor market dise-
quilibrium, suppose that the prevailing real wage is above the equilibrium
real wage and given by wH in Figure 8. At the real wage wH , labor supply
(point C) exceeds labor demand (point B). Graphically, the excess supply of
labor is given by the segment BC in Figure 8. Conversely, suppose that the
prevailing real wage is below the equilibrium real wage and given by wL in
Figure 8. At this real wage, there is excess demand for labor –given by the
segment DE.
If there is excess labor supply or demand, what will actual employment
be? The more natural assumption –and the one commonly adopted in the
literature –is that the short-end of the market prevails. Speci…cally, if labor
demand falls short of labor supply, actual employment is given by labor
demand (point B in Figure 8). If, on the other hand, labor supply falls short
of labor demand, actual employment is given by labor supply (point D in
Figure 8). Formally, denoting actual labor by `a , we have that:
8 f
< `t ; if `dt = `st ;
`at = `d ; if `dt < `st ;
: st
`t ; if `dt > `st :
As a result, if the real wage is wH , there is involuntary unemployment
in the sense that not all workers willing to work at the prevailing wage are
19
To …x ideas, the discussion will assume that the nominal wage is sticky both upwards
and downwards. An alternative assumption would be that the nominal wage is sticky
downwards but not upward. In that case, sticky wages would not be a binding constraint
for the economy’s response to any shock that requires an increase in the equilibrium real
wage.

33
employed (i.e., `at < `st ). If the wage is wL , …rms would not be able to hire all
the workers that they would like at the prevailing real wage (i.e., `at < `dt ).
In either case, the labor market is in disequilibrium because demand and
supply are not the same. This is, of course, equivalent to saying that one
of the two marginal conditions related to the labor market is not holding.
If `at = `dt < `st , …rms are operating on their demand curve (i.e., marginal
condition (59) holds) but households are not on their labor supply curve
(i.e., condition (53) does not hold because the marginal disutility of labor,
evaluated at `at , falls short of the real wage). Conversely, if `at = `st < `dt ,
then households are on their labor supply curve, but …rms are not on their
demand curve (in fact, the marginal productivity of labor evaluated at `at
exceeds the real wage).
Finally, we need to ask: if the labor market is in disequilibrium, how will
it adjust over time to reach equilibrium? A natural assumption regarding
the adjustment in the labor market is to posit that the nominal wage evolves
according to the deviation of the actual real wage from the full-employment
real wage:

_t= Wt
W wf ; > 0; W0 given. (63)
Et
Hence, if the prevailing real wage is above the full-employment level (i.e.,
Wt
Et
> wf ), the nominal wage falls over time. The idea is that the excess
labor supply leads to a gradual fall in nominal wages, as unemployed workers
become more willing over time to take jobs at lower nominal wages. On the
other hand, if the prevailing real wage is below the full-employment level
(i.e., W
Et
t
< wf ) and there is thus excess demand for labor, the nominal wage
increases over time re‡ecting the willingness of …rms to pay higher nominal
wages due to the tight labor market conditions.

6.4 Government
The government’s budget constraints are unchanged relative to our previous
model and continue to be given by equations (13) and (14).

34
6.5 Equilibrium conditions
Given perfect capital mobility and a …xed exchange rate, it follows that

it = r: (64)

Before aggregating the households’and …rms’constraints, we need to take

into account that when the labor market is in disequilibrium, the relevant la-
bor variable for either households’wage income or …rm’s productive purposes
is `a . Hence, we can replace `s with `a in the households’budget constraint
(48) and `d with `a in the …rms’pro…ts (58). We can then combine both to
obtain:

a_ t = rat + (`at ) + t ct it m t :
Combining this constraint with the government’s ‡ow constraint –given by
(13) –yields:

k_ t = rkt + (`at ) ct .
Proceeding in the same way, we can combine the household’s intertempo-
ral constraint (49) with the …rm’s constraint and then with the government’s
lifetime constraint (14) to obtain:
Z 1 Z 1
a rt
k0 + (`t ) e dt = ct e rt dt: (65)
0 0

6.6 Initial stationary equilibrium

Consider an initial perfect foresight equilibrium in which the economy is in a
full-employment equilibrium. The equilibrium real wage and labor are thus
given by expressions (61) and (62), respectively. From (56), it follows that
full-employment output is given by

yf = `f . (66)

From the resource constraint (65), it follows that full-employment consump-

tion is given by

cf = rk0 + `f : (67)
Finally, we can derive the real money demand from (55) and (64):

35
 cf (`f )
mf = : (68)
r

6.7 Permanent fall in productivity

Suppose that up to time t = 0 the economy is in the full-employment
stationary state just described. At time 0, there is an unanticipated and
permanent reduction in . As a benchmark, we will …rst analyze the ad-
justment that would take place under ‡exible wages and then focus on the
sticky-wages case.

6.7.1 Flexible wages

As (60) makes clear, the fall in reduces labor demand for a given level of
the real wage. In terms of Figure 9, the initial equilibrium is given by point
A. The fall in productivity shifts the labor demand to the left (from `d to
0
`d ). The labor market adjusts instantaneously from its initial equilibrium,
point A, to the new equilibrium, point B, at which both the real wage (wf )0
0
and employment are lower `f . Since the exchange rate is …xed, the fall in
the real wage is e¤ected through a fall in the nominal wage, W . Both output
and consumption fall, as follows from (66) and (67). (This adjustment is
captured by the full lines in Figure 10.)

[Figure 9]
[Figure 10]

Although it seems intuitive that, in response to the fall in consumption,

real money demand should fall, this is not formally obvious from just glancing
at equation (68). To formally show this, we compute the change in cf (`f )
in response to a small change in which, using (67), leads to:

d cf (`f ) d`f h 1
i
= `f + `f (`f ) 1
:
d d
The term in square brackets on the RHS is simply the di¤erence between the
marginal productivity of labor and the marginal disutility of labor, which is
always zero under ‡exible wages. Hence:

36
 d cf (`f )
= `f > 0: (69)
d
Intuitively, this is an envelope condition that says that, at an optimum,
consumption net of the disutility of labor falls by the direct e¤ect of the fall
in productivity on output since, at the margin, production e¢ ciency always
implies that the marginal productivity of labor is equated to the marginal
disutility of labor. In light of (69), it follows that real money demand falls.
Finally, we verify that, as one should expect, welfare falls. Given that
the economy jumps from one stationary state to the next, we infer from (47)
that the change in welfare depends on the change in cf (`f ) and in m.
Since we have shown that both fall, welfare also falls.

6.7.2 Sticky wages

Suppose now that the nominal wage is sticky. Since the economy is operating
under a …xed exchange rate, the real wage cannot change on impact. Hence,
on impact, the economy …nds itself at point C in Figure 9. At the prevailing
wage, wf , there is excess supply of labor and actual employment, `a , is
given by labor demand at the level `0 . We thus see how, in the presence
of sticky nominal wages, this negative supply shock leads to involuntary
unemployment, in the amount CA in Figure 9.
How will the nominal wage evolve over time? At t = 0, we know, using
(63),

_0= 0 W0
W wf < 0.
E
The nominal wage begins to fall at time 0 and, in fact, will continue falling
over time. Since the exchange rate is …xed, the real wage, which does not
change on impact, also falls over time towards its full-employment level (Fig-
ure 10, Panel B). As the real wage falls over time, actual employment will
continue to be given by the short-end of the market. In terms of Figure 9,
this means that actual employment will increase over time along the arrowed
path from point C to point B as the nominal – and hence real – wage fall
over time. Formally, it follows from (60) that

`dt
`_at = `_dt = w_ t > 0.
(1 )wt

37
The path of labor is thus given by Figure 10, Panel C. On impact, there-
fore, the fall in employment is larger than under ‡exible wages. Given the
production function (56), the path of output follows that of labor (Figure
10, Panel D). The initial fall in output is thus also larger under sticky wages
than under ‡exible wages.
Let us turn to the behavior of consumption.20 From …rst-order condition
(50) and the fact that labor increases over time, it follows that consumption
also increases over time:
1 _a
c_t = (`at ) `t >0 (70)
To …nd out the change on impact, we need to look at the resource constraint.
The output path illustrated in Figure 10, Panel D, indicates that the present
discounted value of output falls. Hence, consumption can neither increase
nor stay the same for, if it did, the present discounted value of consumption
would increase and thus violate the resource constraint. We thus conclude
that consumption falls on impact. Figure 10, Panel E illustrates the path of
consumption.21
What happens to the trade balance? By de…nition, T Bt = yt ct . Hence,
taking into account (56) and (70),
:
T Bt = (`at ) 1
(`a ) 1
`_at > 0;
where the sign follows from the fact that at point C in Figure 9 (and any
other point along the arrowed path CB), the marginal productivity of labor is
greater than the marginal disutility of labor. The trade balance thus improves
over time. It immediately follows that, on impact, the trade balance must
fall. If it did not change or increase on impact –and since it then increases
20
Under sticky wages, one way to think about the household’s optimization problem
is that the household is now taking as given the path of labor illustrated in Figure 10,
Panel C, and optimally choosing consumption and real money balances (see Barro and
Grossman (1971) for a detailed discussion). In other words, while …rst-order conditions
(50) and (52) continue to hold, …rst-order condition (51) does no longer hold. In fact,
when actual labor is dictated by labor demand (i.e., there is an excess supply of labor),
the marginal disutility of labor will be lower than the prevailing real wage (indicating an
unsatis…ed desire to work).
21
To show that consumption ends up below its full-employment level, notice that, as
shown below, net consumption (i.e., c (`at ) ) is lower under sticky wages than under
‡exible wages. Since labor eventually converges to the same value in both cases, consump-
tion must be lower in the long-run under sticky wages than under ‡exible wages.

38
over time – it would violate the resource constraint. The path of the trade
balance is illustrated in Figure 10, Panel F (assuming k0 = 0). It follows
that there is some consumption smoothing taking place as the trade de…cit
is largest early on when the production is at its lowest point.
We now turn to the path of real money balances which, from (55) and
(64), is given by

ct (`at )
mt = : (71)
r
We know from …rst-order condition (50) that ct (`at ) will be constant
along the new perfect foresight path. Further, as shown in Appendix 8.2,
ct (`at ) falls on impact. Hence, real money demand will fall on impact as
well and remain at that level thereafter.
Finally, we turn to the issue of welfare. It should be clear that the econ-
omy’s adjustment under sticky wages is costlier than under ‡exible wages.
In fact, since there are no distortions of any kind in the ‡exible wages case,
the economy’s adjustment constitutes the …rst-best response. In other words,
given that the economy is poorer, it is optimal to adjust immediately to the
new reality. It follows that the adjustment under sticky nominal wages –
which deviates from the …rst-best adjustment –is costlier.
It is, in fact, easy to verify that welfare falls by more in the sticky wage
than in the ‡exible wage case. As shown in Appendix 8.2, ct (`at ) –and
hence real money balances –fall by more in the sticky wage case than in the
‡exible wage case. Hence, welfare will also be lower.

6.8 Optimal devaluation

As illustrated in Figure 10, we have concluded from our analysis that sticky
nominal wages interfere with the optimal adjustment of the economy to a
negative shock. Since the economy has become poorer as the result of a per-
manent fall in productivity, the optimal adjustment consists in an immediate
reduction in the real wage (from point A to point B in Figure 9), which en-
sures that full employment continues to prevail (albeit at a lower level given
that the economy is now less productive). Under …xed exchange rates, the
adjustment in the real wage should take place through a fall in the nominal
wage. By preventing the real wage from adjusting, wage stickiness causes
the economy to undergo a costlier adjustment. In fact, the economy must go
through a protracted period of unemployment before it …nally reaches the

39
long-run equilibrium, As a result, welfare is lower than in the ‡exible wage
equilibrium.
Is there anything policymakers can do to ease the economy’s adjustment
under sticky wages? They certainly can. In fact, by devaluing when the
0
negative shock hits, policymakers can reduce the real wage from wf to wf
in Figure 8 thus managing to take the economy from point A to point B in
spite of sticky wages. The reduction in the real wage is achieved through
a devaluation rather than through a fall in the nominal wage. Clearly, the
devaluation is the …rst-best response to this negative shock as it reproduces
the outcome that would prevail under ‡exible wages.22 23
Even if policymakers do not devalue as soon as the negative shock hits, it
would still be optimal to devalue at any point in time as the economy travels
from point C to point B in Figure 9 and take the economy immediately
to point B, rather than letting the adjustment take its natural course. In
fact, the adjustment along the segment CB in Figure 9 is characterized by low
output and (initial) trade de…cits, symptoms that are often seen as requiring a
discrete devaluation. Needless to say, however, to make the best-case scenario
for a devaluation, we have ignored many other important aspects of reality,
such as credibility problems, that may play an important role in practice (see
Box 3).

7 Conclusions
This chapter has removed the veil from our monetary model in Chapter 5 by
incorporating sticky prices, by far the most popular friction in open economy
models. In such a context, permanent changes in monetary policy (both in
the level and in the rate of growth of the money supply) are expansionary
as they lead to higher aggregate demand and output. Sticky prices also al-
lowed us to rationalize the high volatility of nominal exchange rates (i.e., the
overshooting phenomenon). In a similar vein, a devaluation leads to higher
22
For the same reasons, it is easy to see that, under ‡exible exchange rates, the economy
would adjust instantaneously from point A to point B in spite of sticky wages! This
suggests that in a world with nominal rigidities –and in response to real shocks –‡exible
rates are better than …xed rates. The opposite, however, will be true for monetary shocks.
We will study these issues in detail in Chapter 12 on optimal exchange rate regimes.
23
Notice, however, that a change in the nominal exchange rate (either a devaluation or
a revaluation) starting from a steady-state would always lead to a fall in labor and output,
as analyzed in Exercise 3 at the end of this chapter.

40
output and consumption, in contrast to the model of Chapter 6, in which
a devaluation reduces consumption. We have also studied a disequilibrium
model of sticky wages, a slightly more complicated theoretical set-up but
arguably a more insightful representation of a world with nominal rigidi-
ties. In particular, we saw how such a model can rationalize the need for a
devaluation of the domestic currency in response to a negative shock.
This chapter concludes our …rst incursion into monetary models. We
studied the basic monetary model in Chapter 5 –in which money is a veil –
and then removed the veil by abstracting from interest-bearing bonds (Chap-
ter 6), introducing a link between nominal interest rates and consumption
(Chapter 7) and sticky prices (Chapter 8). We now move to Part II of the
book in which we will put all these tools to work in our quest for understand-
ing important macroeconomic policy issues.

8 Appendices
8.1 Calvo’s (1983) staggered prices
Calvo (1983) developed an extremely useful continuous-time version of the
staggered-prices models à la Taylor (1979, 1980) and Fischer (1977). Sup-
pose that there is a large number (technically, a continuum) of …rms in the
[0; 1] interval. Total number of …rms is therefore one. Each …rm produces
a non-storable good at zero variable cost, the quantity of which is demand-
determined. Each …rm may change its price only when it receives a random
price signal. The probability of receiving a signal follows an exponential
distribution. When a …rm changes its price, it takes into account the ex-
pected average price and the level of excess aggregate demand (A) expected
to prevail in the future.
The probability of receiving a price signal j periods from now is e j ,
where > 0. The …rm’s price setting rule is assumed to be given by:
Z 1
log(Vt ) = [log(Ps ) + !As ]e (s t) ds; ! > 0; (72)
t

where Vt is the price quotation set at t, Ps is the price level (to be de…ned
below) and As denotes excess aggregate demand. Note that V may jump if
an unexpected change in, say, A takes place.

41
 If price changes are stochastically independent across …rms, the propor-
tion of prices set at time s that have not been modi…ed as of time t is given
by e (t s) . The (logarithm of) the price level is de…ned as the weighted
average of prices currently quoted. Hence:
Z t
log(Pt ) = log(Vs )e (t s) ds: (73)
1

An important observation is that, unlike Vt , Pt is a predetermined variable

because it is given by past price quotations. Along paths where Pt and
At are uniquely determined, however, Vt is a continuous function of time.
Di¤erentiating equation (73) with respect to time yields (using Leibnitz’s
rule):

t = [log(Vt ) log(Pt )] ; (74)

where P_t =Pt .24 Notice that anticipated changes in At will not a¤ect
. In other words, along a perfect foresight path, t will be a continuous
function of time.
At points in time at which At is continuous, we can di¤erentiate equation
(72) to obtain (again, using Leibnitz’s rule)

V_ t
= [log(Vt ) log(Pt ) !At ] : (75)
Vt
It follows from (74) and (75) that (at points in time at which At is continuous)

_t = At ; (76)
2
where ! > 0. Equation (76) is thus a “higher” order inverse Phillips
curve which indicates that the change in the in‡ation rate is negatively related
to excess demand.
24
In its more general formulation, the Leibnitz’s rule states that if we have a function,
F (t), de…ned as
Z x(t)
F (t) = f (s; t)ds,
g(t)

its derivative is given by

Z x(t)
0 0 0 @f (s; t)
F (t) = f (s; x(t))x (t) f (s; g(t))g (t) + ds.
g(t) @t

42
8.2 Sticky wage model
This appendix analyzes the behavior of ct (`at ) in response to the perma-
nent fall in the productivity parameter, . First-order condition (50) –with
`at in lieu of `st – indicates that ct (`at ) will be constant along the new
perfect foresight path. To pin down the level, we need to compute the present
discounted value of ct (`at ) . Using the economy’s resource constraint, we
can write it as:
Z 1 Z 1
a rt
P DV [ct (`t ) ] e dt = k0 + [ (`at ) (`at ) ] e rt dt:
0 0

We now compute the change in the present discounted value as a result of a

small change in :

Z 1
dP DV 1 d`at
= (`at ) + (`at ) (`at ) 1
e rt
dt > 0. (77)
d 0 | {z }d
+ |{z}
+

As indicated, the term in curve brackets is positive because it is the di¤erence

between the marginal productivity of labor and the marginal disutility of
labor. (Notice that this term is zero in the ‡exible wages case.) Equation
(77) thus indicates that in response to a fall in , the PDV of ct (`at ) also
falls and, in fact, falls by more than in does in the ‡exible wages case.
We thus conclude that ct (`at ) falls on impact and remains at that
level thereafter. From (71), it follows that m falls and, in fact, falls by more
than in the ‡exible wage case. It follows that welfare follows by more in the
sticky wage case than in the ‡exible wages case because both ct (`at ) and
m fall by more.

43
 Exercises25
1. Temporary reduction in money growth rate
The purpose of this exercise is to show that the sticky-prices model
developed in the text is capable of explaining situations of “stag‡a-
tion”(i.e., the co-existence of high in‡ation and output below the full-
employment level).
In the context of the model developed in Section 2:

(a) Analyze the e¤ects of a temporary reduction in the money growth

rate.
(b) Explain the intuition behind the results.

2. Fiscal policy in a sticky prices model

This exercise incorporates …scal policy into the sticky prices model
analyzed in this chapter and studies the e¤ects of a permanent increase
in government spending on non-tradable goods under both ‡exible and
predetermined exchange rates.
Speci…cally, suppose preferences are given by
Z 1
log(cTt ) + log(cN
t ) + log(mt ) e
t
dt;
0

The consumer’s intertemporal constraint is given by

Z 1 Z 1
yN cN
a0 + ytT + t + t e rt
dt = cTt + t
+ it m t e rt
dt:
0 et 0 et

The government’s ‡ow budget constraint is given by

gtN
h_ t = rht + m
_ t + "t m t t .
et
The corresponding intertemporal constraint is given by
Z 1 Z 1
rt gtN rt
h0 + (m
_ t + "t mt )e dt = + t e dt.
0 0 et
25
An answer key is available from the author upon request.

44
 Equilibrium in the non-tradable goods market dictates that

ytN = cN N
t + gt :

The rest of the model remains the same as in the text.

In the context of this model:

(a) Analyze the e¤ects of a permanent and unanticipated increase

in government spending on non-tradable goods under ‡exible ex-
change rates.
(b) Analyze the e¤ects of a permanent and unanticipated increase in
government spending on non-tradable goods under predetermined
exchange rates.
(c) Explain intuitively why the response is the same. Do you think
that it would continue to be the same under non-logarithmic pref-
erences? 26

3. Devaluation/revaluation in a sticky wages model

In the context of the sticky wages model of Section 6, show that both
a devaluation and a revaluation of the currency (i.e., an increase and
a decrease in the nominal exchange rate) lead to a fall in actual labor
and output. Explain the intuition behind the results. (Lahiri and Vegh
(2002) use this feature of the sticky-wages model to think about the
costs of a ‡uctuating nominal exchange rate.)

26
In fact, the data suggest that output responds di¤erently under predetermined and
‡exible exchange rates (see Ilzetzki, Mendoza, and Vegh (2010)).

45
References
[1] Agenor, Pierre-Richard. 1991. Output, devaluation and the real ex-
change rate in developing countries. Review of World Economics 127
(1): 18-41.

[2] Agenor, Pierre-Richard and Peter J. Montiel. 1999. Development macro-

economics, 2nd Edition. Princeton University Press.

[3] Álvarez, Luis J. and Ignacio Hernando. 2006. Price setting behaviour in
Spain. Evidence from consumer price micro-data. Economic Modelling
23 (4): 699-716.

[4] Aucremanne, Luc and Emmanuel Dhyne. 2004. How frequently do prices
change? Evidence based on the micro data underlying the Belgian CPI.
ECB Working Paper 331. European Central Bank, Frankfurt, Germany.

[5] Baharad, Eyal, and Benjamin Eden. 2004. Price rigidity and price dis-
persion: Evidence from micro data. Review of Economic Dynamics 7
(3): 613-641.

[6] Barro, Robert J. and Herschel I. Grossman. 1971. A general disequilib-

rium model of income and employment. American Economic Review 61
(1): 82-93.

[7] Barros, Rebecca, Marco Bonomo, Carlos Carvalho, and Silvia Matos.
2009. Price setting in a variable macroeconomic environment: Evidence
from Brazilian CPI. Unpublished paper. Getulio Vargas Foundation and
Federal Reserve Bank of New York.

[8] Baumgartner, Josef, Ernst Glatzer, Fabio Rumler, and Alfred

Stiglbauer. 2005. How frequently do consumer prices change in Austria?
Evidence from micro CPI Data. ECB Working Paper 523. European
Central Bank, Frankfurt, Germany.

[9] Bils, Mark, and Peter Klenow. 2004. Some evidence on the importance
of sticky prices. Journal of Political Economy 112 (5): 947-985.

[10] Blinder, Alan S., Elie R. D. Canetti, David F. Lebow, and Jeremy B.
Rudd. 1999. Asking about Prices: A New Approach to Understanding
Price Stickiness. Review of Industrial Organization 15 (1): 97-101.

46
[11] Bunn, Philip and Colin Ellis. 2009. Price-setting behaviour in the United
Kingdom: A microdata approach. Bank of England Quarterly Bulletin
2009 Q1.

[12] Burstein, Ariel, Martin Eichenbaum, and Sergio Rebelo. 2005. Large
devaluations and the real exchange rate. Journal of Political Economy
113 (4): 742-784.

[13] Bu¢ e, Edward and Yongkul Won. 2001. Devaluation and investment in
an optimizing model of the small open economy. European Economic
Review 45 (8): 1461-1499.

[14] Calvo, Guillermo A. 1983. Staggered prices in a utility-maximizing

framework. Journal of Monetary Economics 12 (3): 383-398.

[15] Calvo, Guillermo A. and Carlos A. Végh. 1993. Exchange rate-based

stabilisation under imperfect credibility. In Open-Economy Macroeco-
nomics, ed. H. Frisch and [Link], 3-28. London: MacMillan Press.

[16] Calvo, Guillermo A., and Carlos A. Végh. 1999. In‡ation stabilization
and BOP crises in developing countries. In Handbook of Macroeconomics,
Volume C, ed. John Taylor and Michael Woodford, 1531-1614. North
Holland.

[17] Carlton, Dennis W. 1986. The Rigidity of Prices. American Economic

Review 76 (4): 637-658.

[18] Cavallo, Alberto. 2010. Scraped data and sticky prices: Frequency, haz-
ards and synchronization. Mimeograph, Harvard University.

[19] Cavallo, Michele, Kate Kisselev, Fabrizio Perri, and Nouriel Roubini.
2004. Exchange rate overshooting and the costs of ‡oating. Mimeo. New
York University, New York.

[20] Cecchetti, Stephen. 1986. The frequency of price adjustment: A study

of the newsstand prices of magazines. Journal of Econometrics 31 (3):
255–274.

[21] Cespedes, Luis F., Roberto Chang, and Andres Velasco. 2004. Balance
sheets and exchange rate policy. American Economic Review 94 (4):
1183-1193.

47
[22] Chari, V.V., Patrick K. Kehoe, and Ellen R. McGrattan. 2002. Can
sticky price models generate volatile and persistent real exchange rates?
Review of Economic Studies 69 (3): 533-563.

[23] Connolly, M. 1983. Exchange rates, real economic activity and the bal-
ance of payments: Evidence from the 1960s. In Recent Issues in the
Theory of the Flexible Exchange Rate, ed. E. Classen and P. Salin, 129-
143. Amsterdam: Elsevier.

[24] Cooper, Richard N. 1971. Currency devaluation in developing countries.

In Government and Economic Development, ed. G. Ranis. New Haven:
Yale University Press.

[25] Coricelli, Fabrizio and Roman Horváth, 2010. Price setting and market
structure: An empirical analysis of micro data. Managerial and Decision
Economics 31 (2-3): 209-233.

[26] Creamer, Kenneth and Neil A. Rankin. 2008. Price setting in South
Africa 2001–2007 - Stylised facts using consumer price micro data. Jour-
nal of Development Perspectives 1 (4): 93-118.

[27] Dias, Mónica, Daniel Dias, and Pedro D. Neves. 2004. Stylised features
of price setting behaviour in Portugal: 1992-2001. ECB Working Paper
332. European Central Bank, Frankfurt, Germany.

[28] Diaz Alejandro, Carlos F. 1963. A note on the impact of devaluation

and the redistributive e¤ect. Journal of Political Economy 71: 577-580.

[29] Dhyne, E., L. J. Álvarez, H. Le Bihan, G. Veronese, D. Dias, J. Ho¤-

mann, N. Jonker, P. Lünnemann, F. Rumler and J. Vilmunen. 2006.
Price changes in the Euro Area and the United States: Some facts from
individual consumer price data. Journal of Economic Perspectives 20
(2): 171-192.

[30] Dornbusch, Rudiger. 1976. Expectations and exchange rate dynamics.

Journal of Political Economy 84 (6): 1161-1176.

[31] Dornbusch, Rudiger and Alejandro Werner. 1994. Mexico: Stabiliza-

tion, reform, and no growth. Brookings Papers on Economic Activity 25
(1994-1): 253-316.

48
[32] Dutta, Shantanu, Mark Bergen, Daniel Levy, and Robert Venable. 1999.
Menu costs, posted prices, and multiproduct retailers. Journal of Money,
Credit, and Banking 31: 683-703.
[33] Edwards, Sebastian. 1986. Are Devaluations Contractionary? Review of
Economics and Statistics 68 (3): 501-508.
[34] Eichenbaum, Martin, Nir Jaimovich and Sergio Rebelo. 2009. Reference
prices and nominal rigidities. American Economic Review (forthcom-
ing).
[35] Fabiani, Silvia, Angela Gattulli, Roberto Sabbatini and Giovanni
Veronese. 2006. Consumer price setting in Italy. Giornale degli Econo-
misti e Annali di Economia 65 (1): 31-74.
[36] Fischer, Stanley. 1977. Long-term contracts, rational expectations,and
the optimal money supply rule. Journal of Political Economy 85 (1):
191-205.
[37] Fleming, Marcus J. 1962. Domestic …nancial policies under …xed and
under ‡oating exchange rates. International Monetary Fund Sta¤ Papers
9 (3): 369-379.
[38] Frenkel, Jacob A., and Assaf Razin. 1987. The Mundell-Fleming model
a quarter century later: A uni…ed exposition. International Monetary
Fund Sta¤ Papers 34 (4): 567-620.
[39] Gabriel, Peter and Adam Rei¤. 2010. Price setting in Hungary –a store-
level analysis. Managerial and Decision Economics 31 (2-3): 161-176.
[40] Gagnon, Etienne. 2009. Price setting during low and high in‡ation: Ev-
idence from Mexico. Quarterly Journal of Economics 124: 1221 –1263.
[41] Gouvea, Solange. 2007. Price rigidity in Brazil: Evidence from CPI micro
data. Central Bank of Brazil Working Paper 143. Central Bank of Brazil.
[42] Greenwood, Jeremy, Zvi Hercowitz and Gregory Hu¤man. 1988. Invest-
ment, capacity utilization, and the real business cycle. American Eco-
nomic Review 78 (3): 402-417.
[43] Gylfason, Thorvaldur and Michael Schmid. 1983. Does devaluation cause
stag‡ation? Canadian Journal of Economics 16 (4): 641-654.

49
[44] Gupta, Poonam, Deepak Mishra, and Ratna Sahay. 2007. Behavior of
output during currency crises. Journal of International Economics 72
(2): 428-450.

[45] Hansen, Bo William and Niels Lynggard Hansen. 2006. Price setting
behavior in Denmark: A study of CPI micro data 1997-2005. Danmarks
Nationalbank Working Paper 39. Denmark.

[46] Ho¤mann, Johannes and Jeong-Ryeol Kurz-Kim. 2006. Consumer price

adjustment under the microscope: Germany in a period of low in‡ation.
ECB Working Paper 652. European Central Bank, Frankfurt, Germany.

[47] Ilzetzki, Ethan, Enrique Mendoza and Carlos Vegh. 2010. How big
(small?) are …scal multipliers? NBER Working Paper No. 16479.

[48] Jonker, Nicole, Carsten Folkertsma, and Harry Blijenberg. 2004. Empir-
ical analysis of price setting behaviour in the Netherlands in the period
1998-2003 using micro data. ECB Working Paper 413. European Central
Bank, Frankfurt, Germany.

[49] Kamin, Steven B., and Marc Klau. 1998. Some multi-country evidence
on the e¤ects of real exchange rates on output. Federal Reserve Board
International Finance Discussion Paper No. 611.

[50] Kashyap, Anil. 1995. Sticky prices: New evidence from retail catalogs.
Quarterly Journal of Economics 110 (1): 245-274.

[51] Klenow, Peter and Oleksiy Kryvtsov. 2008. State-dependent or time-

dependent pricing: Does It matter for recent U.S. in‡ation? Quarterly
Journal of Economics 123 (3): 863-904.

[52] Klenow, Peter and Benjamin Malin. 2010. Microeconomic evidence on

price setting. In Handbook of Monetary Economics, ed. Benjamin Fried-
man and Michael Woodford.

[53] Kovanen, Arto. 2006. Why do prices in Sierra Leone change so often?
A case study using micro-level price data. IMF Working Paper 06/53.
International Monetary Fund. Washington D.C.

[54] Krugman, Paul and Lance Taylor. 1978. Contractionary e¤ects of deval-
uation. Journal of International Economics 8 (3): 445-456.

50
[55] Lach, Saul, and Daniel Tsiddon. 1992. The behavior of prices and in-
‡ation: An empirical analysis of disaggregated price data. Journal of
Political Economy 100 (2): 349-389.
[56] Lahiri, Amartya and Carlos A. Végh. 2002. Living with the fear of ‡oat-
ing: An optimal policy perspective. In Preventing Currency Crises in
Emerging Markets, ed. Sebastian Edwards and Je¤ Frankel, 663-704.
University of Chicago Press.
[57] Levy, Daniel, Mark Bergen, Shantanu Dutta, and Robert Venable. 1997.
The magnitude of menu costs: Direct evidence from large U.S. super-
market chains. Quarterly Journal of Economics 112: 791-825.
[58] Lizondo, J. Saul, and Peter J. Montiel. 1989. Contractionary devalua-
tion: An analytical overview. International Monetary Fund Sta¤ Papers
36: 182-227.
[59] Lünnemann, Patrick and Thomas Y. Mathä. 2005. Consumer price be-
haviour in Luxembourg: Evidence from micro CPI data. ECB Working
Paper 541. European Central Bank, Frankfurt, Germany.
[60] Magendzo, Igal L. 2002. Are devaluation really contractionary? Working
Paper No. 82 (Central Bank of Chile).
[61] Medina, Juan Pablo, David Rappoport, and Claudio Soto. 2007. Dy-
namics of price adjustment: Evidence from micro level data for Chile.
Central Bank of Chile Working Paper 432. Central Bank of Chile.
[62] Midrigan, Virgiliu. 2008. Menu-costs, multi-product …rms and aggregate
‡uctuations. Econometrica (forthcoming).
[63] Milesi-Ferretti, G.M. and A. Razin. 2000. Current Account Reversals
and Currency Crises: Empirical Regularities. In Currency Crises, ed.
Paul Krugman. Chicago, IL: University of Chicago Press.
[64] Morley, Samuel A. 1992. On the e¤ect of devaluation during stabilization
programs in LDCs. The Review of Economics and Statistics 74 (1): 21-
27.
[65] Mundell, Robert A. 1963. Capital mobility and stabilization policy under
…xed and ‡exible exchange rates. Canadian Journal of Economic and
Political Science 29 (4): 475-485.

51
[66] Mundell, Robert A. 1964. A reply: Capital mobility and size. Canadian
Journal of Economic and Political Science 30: 421-431.
[67] Mussa, Michael. 1986. Nominal exchange rate regimes and the behavior
of real exchange rates: Evidence and implications. Carnegie-Rochester
Conference Series on Public Policy 25 (1): 117-214.
[68] Nakamura, Emi and Jón Steinsson. 2008. Five facts about prices: A
reevaluation of menu cost models. Quarterly Journal of Economics 123
(4): 1415-1464.
[69] Obstfeld, Maurice and Kenneth Rogo¤. 1995. Exchange rate dynamics
redux. Journal of Political Economy 103 (3): 624-660.
[70] Rogo¤, Kenneth. 2002. Dornbusch’s overshooting model after twenty-
…ve years. Mundell-Fleming Lecture at the IMF’s Second Annual Re-
search Conference. Washington D.C.
[71] Saita, Yumi, Izumi Takagawa, Kenji Nishizaki, and Masahiro Higo. 2006.
Price setting in Japan: Evidence from individual retail price data. Bank
of Japan Working Paper Series 06-J-02. Bank of Japan (in Japanese).
[72] Taylor, John B. 1979. Staggered wage setting in a macro model. Amer-
ican Economic Review Papers and Proceedings 69 (2): 108-13.
[73] Taylor, John B. 1980. Aggregate dynamics and staggered contracts.
Journal of Political Economy 88 (1): 1-23.
[74] Uribe, Martin. 1999. Comparing the welfare costs and initial dynam-
ics of alternative in‡ation stabilization policies. Journal of Development
Economics 59 (2): 295-318.
[75] Van Wijnbergen, Sweder. 1986. Exchange rate management and stabi-
lization policies in developing countries. Journal of Development Eco-
nomics 23 (2): 227-247.
[76] Végh, Carlos A. 1992. Stopping High In‡ation: An Analytical Overview.
International Monetary Fund Sta¤ Papers 39: 626-695.
[77] Vilmunen, Jouko and Helinä Laakkonen. 2005. How often do prices
change in Finland? Micro-level evidence from the CPI. Unpublished
paper, Bank of Finland.

52
[78] Wulfsberg, Fredrik. 2009. Price adjustments and in‡ation: Evidence
from consumer price data in Norway 1975-2004. Norges Bank Working
Paper 2009/11.

53
 Box 1. Are devaluations expansionary or
contractionary?
As argued in the text, the real e¤ects of a currency devaluation are the-
oretically ambiguous. In the model laid out in Chapter 6 –with no interest-
bearing bonds – a devaluation is contractionary, as private agents reduce
consumption to rebuild their stock of real money balances. In this chapter’s
model with sticky prices, a devaluation is expansionary since it reduces on
impact the relative price of non-traded goods, thus leading to an increase in
aggregate demand and output. In addition to these two channels, the liter-
ature has explored other, mostly contractionary, channels. Among the most
relevant ones:27

E¤ects on imported inputs/investment. In models with imported in-

puts, a devaluation will rise the domestic price of such inputs and
thus be contractionary (see, for example, Gylfason and Schmid (1983),
van Wijnbergen (1986) and Edwards (1986)). In a similar vein, Bu¢ e
and Wong (2001) show how a devaluation is likely to reduce aggre-
gate investment and output in a model in the spirit of Chapter 6 (i.e.,
consumers hold no foreign bonds) but with a fully-‡edged two-sector
supply side. In their model, the resulting real devaluation increases
the relative price of capital in the non-tradable sector but decreases
it in the tradable sector. Furthermore, the marginal productivity of
capital increases in the tradable sector and falls in the non-tradable
sector. Based on these two e¤ects, investment would rise in the trad-
able sector and fall in the non-tradable sector. A third e¤ect – the
real money balances e¤ect examined in Chapter 6 – tends to reduce
investment in both sectors as households switch from capital to money
to rebuild their real money balances. Calibrations of the model suggest
that overall investment will fall.

Income redistribution. Pro…ts will be boosted in export and import

competing industries as devaluations lead to higher relative prices for
traded goods. When this increased price level leads to lower real wages,
national spending is likely to shrink since the marginal propensity to
27
The following list is far from exhausive. The reader is referred to Lizondo and Montiel
(1989) and Agenor and Montiel (1999, Chapter 8) for a more detailed discussion.

54
 save from pro…ts exceeds that from wages (see, in particular, Diaz-
Alejandro (1963), Cooper (1971), and Krugman and Taylor (1978)).

Balance sheet e¤ects. The risk premium typically increases if public and
private sector’s net worth is a¤ected by real exchange rate movements
(see, for instance, Cespedes, Chang and Velasco (2004)).

The theory thus identi…es several contractionary channels that could, in

principle, more than outweigh the traditional expansionary e¤ects empha-
sized by sticky-prices model. The overall e¤ect thus becomes an empirical
question.
What do the data say? Table 1 summarizes the results of some notable
contributions on the subject. An early study by Gylfason and Shmid (1983)
reported a mostly positive e¤ect of a devaluation on output. Subsequent
studies, however, tended to …nd a weak negative impact, if any, of devalua-
tions on output. A recent study by Gupta, Mishra, and Sahay (2007), which
relies a on large dataset comprising 91 developing countries and spanning al-
most 30 years, documents that, on average, output has fallen when currency
crises have taken place28 Still, a signi…cant number of episodes (43 percent)
have been associated with output increases.

[Table 1]
What could explain di¤erent responses of output to a devaluation? Ac-
cording to some authors, the output e¤ects may be conditional on other
relevant variables. Gupta, Mishra, and Sahay (2007), for instance, …nd that
the output e¤ect is positively associated with commercial integration with
the rest of the world but negatively associated with …nancial integration and
previous periods of capital in‡ows. They also report that large emerging
economies are more likely to su¤er contractionary devaluations than small
ones. On the other hand, Cavallo et al (2004) stress the empirical importance
of balance sheet mismatches. They claim that, among developing economies,
output contraction has been greater in relatively large and more developed
economies than in smaller and less developed economies.
28
It should be noted that the focus of this paper is on currency crises (the de…nition
of which typically includes increases in the exchange rate but also includes changes in
reserves). Ideally, one would like to see a similarly large dataset used to study the e¤ects
of devaluations only. The overlapping, of course, would be large.

55
 In sum, the jury is still out on whether, in practice, devaluations are
expansionary or contractionary. If anything, the often-mixed evidence sug-
gests that the real e¤ects of devaluation likely depend on the circumstances
surrounding it. For instance, it stands to reason that a devaluation carried
out in the middle of a full-‡edged balance of payments and …nancial crisis
and possibly accompanied by tighter …scal and monetary policies is much
more likely to be contractionary than a devaluation undertaken as part of
an orderly adjustment to, say, a perceived exchange rate “misalignment”.
Unfortunately, controlling for these factors is not a trivial empirical task.

56
 Box 2. How sticky are prices?
As mentioned in the Introduction, price stickiness is, by far, the most
common friction used in models that study the real e¤ects of monetary and
exchange rate policies. Many closed and open economy models assume the
existence of a large number of di¤erentiated goods and introduce price stick-
iness a la Calvo (1983) so that, at every point in time, there is a fraction
of …rms that are unable to change their price. But how sticky are prices in
practice? Empirically, price stickiness is typically measured by analyzing
time series for prices of di¤erent goods, estimating the frequency of price
changes for each series, and aggregating the results by appropriately weight-
ing each price in order to obtain an estimate of the average time between
price changes for di¤erent categories of goods.
Early empirical work measuring the frequency of price changes in retail
and wholesale prices established that many prices often go unchanged for
many months. However, these estimates were usually based on relatively
narrow sample of goods.29 In recent years, the availability of richer datasets
has allowed researchers to come up with a much clearer picture of the be-
havior of individual prices. In an in‡uential paper, Bils and Klenow (2004)
examined the frequency of price changes for 350 categories of goods and ser-
vices covering about 70 percent of the U.S. Consumer Price Index (CPI).
Surprisingly, they found a median time between price changes between 4.3
to 5.5 months, well below previous estimates. They also found a very large
degree of heterogeneity in the behavior of price changes across goods.
More recently, Nakamura and Steinsson (2008), Klenow and Kryvtsov
(2008), and Klenow and Malin (2010) have also studied the U.S. CPI but us-
ing datasets that allowed them to di¤erentiate between regular price changes
and those related to temporary price sales, …nding that the inclusion or ex-
clusion of price sales has sizable e¤ects on the estimated frequency of price
changes. For example, Nakamura and Steinsson (2008) report a median du-
ration of price changes between 4.4 to 4.6 months when sale-related prices
are included in the data, and a median duration between 8 to 11 months
when sale-related price changes are excluded from the data. By now, many
studies have applied approaches similar to Bils and Klenow’s to the analysis
of the frequency of price changes in other countries, as reported in Table 2.
29
See, for example, Carlton (1986), Cecchetti (1986), Kashyap (1995), Levy et al (1997)
and Blinder et al (1999).

57
 [Table 2]
The availability of scanner data has allowed researchers to further probe
into the issue of price stickiness. For example, using scanner data from a
large grocery store chain in Chicago, Midrigan (2008) reports the presence
of many small and short lived price changes and uses this as a motivation
to construct a menu cost model where …rms face economies of scale when
adjusting their prices. Eichenbaum et. al. (forthcoming), on the other hand,
use scanner data from a large U.S. retailer and …nd that nominal rigidities
take the form of inertia in reference prices, with weekly prices ‡uctuating
around reference values that tend to remain constant over extended periods
of time.30 A newer and equally promising source of information on high fre-
quency price movements is scraped online data (data extracted from internet
websites)31 .
Finally, on a line of work more related to open economy issues, Burstein et
al (2005) use their own survey data for Argentina and examine the e¤ects of
large devaluations on the real exchange rate. These authors present evidence
that suggests that the large decline in the real exchange rate is caused by
the sluggish adjustment of the prices of non-tradable goods and not from
deviations from the law of one price, thus providing support for set-up used
in this paper that posits fully ‡exible tradable goods prices and sticky non-
tradable goods prices.
30
A reference price is de…ned as the most common price in a given time window. Eichen-
baum et al (2009) use a quarter.
31
Using scrapped data for Argentina, Brazil. Chile and Colombia, Cavallo (2010) …nds
that the distributions of the size of price changes in these countries are bimodal, that
hazard functions are upward-sloping, and that there is strong daily price synchronization
within narrow categories of goods, suggesting that strategic complementarities play an
important role in price-setting decisions.

58
 Box 3. To devalue or not to devalue? That
is the question
In December 1987, annual in‡ation in Mexico had reached 160 percent.
In response, the Mexican government instituted an exchange-rate based sta-
bilization program initially based on a …xed exchange rate and income and
wages policies. As expected –and analyzed in detail in Chapter 13 –the ex-
change rate-based stabilization program led to a signi…cant real appreciation
of the currency (Figure 11). By April 1994, the continuing real apprecia-
tion accompanied by slow growth and a widening current account de…cit had
called into question the stabilization strategy. Further, a speculative attack
on the peso and a sharp upturn in interest rates were taken as a ominous
sign of a future crisis.
Two sharply contrasting assessments of Mexico’s situation at the time
became the subject of intense public debate. On the one hand, some observes
and, in particular, the authorities at the time put forward the view that the
real appreciation simply re‡ected an equilibrium phenomenon resulting from
the reforms that had been undertaken, including budget, trade, and wealth
e¤ects resulting from the exchange rate based stabilization plan. The logical
corollary of this view was that no policy remedies were needed. On the
other hand, there was a disequilibrium view which, even if it accepted the
potential bene…ts of reforms, liberalization, and the presence of NAFTA,
argued that the overvaluation of the domestic currency was a policy mistake
that could and should be remedied. Rudy Dornbush, particularly in a joint
paper with Alejandro Werner, was the most famous proponent of this view.
Their idea was that the interaction of the exchange rate-based stabilization
and incomes policy had been responsible for the overvaluation. Using a model
very similar to the sticky-in‡ation model of Chapter 13, they argued that
…xing the nominal interest rate immediately reduced the nominal interest
rate. Inertia in the in‡ation process, however, would imply that in‡ation
would fall only slowly over time, thus leading to a real appreciation. Further,
the resulting fall in real interest rates would push up aggregate demand
and reinforce the process of real appreciation. In Dornbusch and Werner’s
view of the world, this real appreciation would slow growth and increase
unemployment. In their minds, therefore, the policy remedy was clear: a
once-and-for-all devaluation of about 20 percent, which would take care of
the real appreciation.
In his discussion of the Dornbush-Werner piece, Guillermo Calvo ve-

59
hemently disagreed. “In my opinion, this is not the time to implement
a Dornbusch-Werner devaluation. The forces that have held together the
“good”equilibrium may dissipate overnight.”He pointed out that the biggest
problem for Mexico is credibility. He argues that it is the lack of credibility of
authorities’policies what caused the boom-bust cycle (as in Chapter 7). The
intertemporal substitution caused by an intertemporal distortion (imperfect
credibility) gives rise to a socially costly consumption boom. Individuals are
forced to cut consumption in the future to satisfy budget constraint. Thus,
future real wages will fall and real exchange rate will rise. In this context,
downward price-wage in‡exibility will result in higher unemployment and
excess capacity. A devaluation a la Dornbusch and Werner may solve the
overappreciation problem in the short run, but it will also cause a more pro-
nounced appreciation and in‡ation in the future. “Authorities could have
revealed their taste for discretionary policy, and people may come to believe
that it could happen again. Therefore, the same mechanism that provoked
the present misalignment will be set in motion again.”
In December 1994, Mexico devalued the peso by 15 percent. The devalu-
ation set o¤ a …restorm: since reserves were low before the devaluation, there
was an immediate attack on the Mexican peso setting o¤ a more substantial
fall in reserves. Almost immediately, the government was forced to allow the
peso to ‡oat. From late 1994 to early 1995 the peso depreciated by almost 80
percent and the yield on CETES (Mexican T-bills) more than tripled (Figure
12). The inescapable conclusion is that Calvo was right. Models such as the
one in Section 6 that ignore credibility problem may yield the wrong policy
prescription!

60
 Figure 1. Phase diagram

.
p p=0

p0 .
B

m
A
. .
n=0

p0’ .
C

n’0 nss n0 n
 Figure 2. Permanent increase in money supply

A. Money supply B. Real money balances

M n

nss

0 time time
0

C. Inflation rate D. Real exchange rate

p e

m ess

0 time 0 time

E. Consumption of home goods F. Domestic real interest rate

cN rd

r
N
y

0 time 0 time
 Figure 3. Permanent reduction in rate of money growth

A. Rate of monetary growth B. Real money balances

m n

mH

nss
m
L

0 time 0 time

C. Inflation rate D. Consumption of home goods

p cN

mH y
N

m
L

time 0 time
0

E. Real exchange rate F. Domestic real interest rate

e rd

ess

0 time 0 time
 Figure 4. Permanent reduction in money growth rate: Phase diagram

.
p p=0

mH .A

mL
B . .
n=0

.
C

nss(mH) nss(mL) n
 Figure 5. Phase diagram: Predetermined exchange rates

p .
p=0

.C
p0 .
B

.
e
A . e=0

ess e0 e
 Figure 6. Permanent devaluation

A. Exchange rate B. Real money balances

E n

nss

0 time time
0

C. Inflation rate D. Real exchange rate

p e

e ess

0 time 0 time

E. Consumption of home goods F. Domestic real interest rate

cN rd

r
N
y

0 time 0 time
 Figure 7. Overshooting

E
B .
s<1

2E A . s=1

C . s >1

0 time
 Figure 8. Labor market

w
w
ls
B C
wH

A
wf

D E
wL

ld
lf s
l ,l
d
 Figure 9. Labor market: Fall in productivity

w
w
ls

C A
wf
f
(w )’ B

(ld)’ ld
l0 (lf)’ lf s
l ,l
d
 Figure 10. Permanent fall in productivity

A. Productivity parameter B. Real wage

y w

0 time 0 time

C. Labor D. Output

la y

0 time 0 time

E. Consumption F. Trade balance

c TB

0 time 0 time

Note: A full line indicates the adjustment under flexible wages; a dashed-line under sticky wages.
 Figure 11 - Real Effective Exchange Rate (1982=100)

160

150

140

130

120

110

100

90

80
198219831984198519861987198819891990199119921993199419951996

Source: Banco Central de Mexico

 Figure 12 – CETES yields in percent

120

100

80

60

40

20

0
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996

CETES_28 days CETES_91 days

Source: OECD database

 Table 1: Empirical studies on output effect of devaluation

Author/s (year) Period/ Sign of effect Methodology and controls

countries
Gylfason and 1959-1977 Positive in 8 out of the10 countries Parameter estimation based on
Schmid (1983) 5 industrial and (effect after two to three years) dynamic single-equation models
5 developing
countries
Edwards (1986) 1965-80 Negative in the short run (effect is Panel data (fixed effects)
12 developing offset in the second year) Government spending, money
countries growth
Agenor (1991) 1978-87 Negative (for anticipated Panel data (pooled OLS)
23 developing movements of RER) Government spending, money
countries Positive (for unanticipated supply, foreign income
movements of RER)
Morley (1992) 1974-84 Negative (short run) Cross section
28 developing Terms of trade, import growth,
countries the money supply and the fiscal
balance

Kamin and Klau 1970-96 Very weakly negative (short run) Panel data
(1998) 27 countries Insignificant (long run) Output gap, short term interest
rate, fiscal balance to GDP, terms
of trade, capital account to GDP
ratio, U.S. interest rate. Two
stages least squared

Milesi-Ferreti and 1970-1996 Negative (short run) Panel data.

Razin (2000) 105 low –and Insignificant (long run) Macroeconomic, external, debt,
middle –inome financial, foreign and regional
countries variables.
Magendzo (2002) 1970-99 Insignificant Matching estimators
155 countries
non OECD
countries
Cavallo et al. (2004) 1992-2002 Negative (2-year period), the effect Cross section analysis using OLS,
24 countries is stronger in presence of large IV and three-stage least square.
foreign debt.
Gupta, Mishra, and 1970-98 57 percent of crises contractionary; Frequency distribution and OLS
Sahay (2007) 91 developing 43 percent expansionary Change in external long-term
countries Outcome influenced by factors such debt, cumulative flow of external
as capital account liberalization and private capital, capital controls,
pre-crisis level of economic activity currency crises, business cycles,
and capital flows per capita GDP, banking crises,
short-term debt to reserves,
exchange rate overvaluation,
proxies for monetary and fiscal
policies, openness, competitive
devaluations by others, economic
size, external factors
                                     Table 2: Monthly Mean Duration of CPI Price Changes  
Country  Paper  Mean Duration* 
Excluding Sales 
Austria  Baumgartner et al. (2005)   6.1  ‐ 
Belgium  Aucremanne and Dhyne (2004)   5.4  ‐ 
Brazil  Barros et al. (2009)   2.1  ‐ 
 Gouvea (2007)   2.2  ‐ 
Chile  Medina et al.  (2007)  1.6  ‐ 
Denmark  Hansen and Hansen (2006)   5.3  ‐ 
Euro Area   Dhyne et al. (2006)   6.1  ‐ 
Finland  Vilmunen and Laakkonen (2005)   5.5  ‐ 
France  Baudry et al. (2007)   4.8  ‐ 
Germany  Hoffmann and Kurz‐Kim (2006)   8.3  ‐ 
Hungary  Gabriel and Reiff (2010)   6.1  ‐ 
Israel  Baharad and Eden (2004)   3.6  ‐ 
Italy  Fabiani et al. (2006)   9.5  ‐ 
Japan  Saita et al. (2006)   3.8  ‐ 
Luxembourg  Lunnemann and Matha (2005)   5.4  ‐ 
Mexico   Gagnon (2009)   2.9  ‐ 
Netherland  Jonker et al. (2004)   5.5  ‐ 
Norway   Wulfsberg (2009)   4.0  4.2 
Portugal  Dias et al. (2004)   4.0  ‐ 
Sierra LeonE  Kovanen (2006)   1.4  ‐ 
Slovakia  Coricelli and Horvath (2010)   2.4  ‐ 
South Africa  Creamer and Rankin (2008)   5.7  ‐ 
Spain  Álvarez and Hernando (2006)  6.2  ‐ 
United Kingdom  Bunn and Ellis (2009)   4.7  6.2 
United States  Bils and Klenow (2004)   3.3  ‐ 
Klenow and Kryvtsov (2008)   2.2  2.8 
   Nakamura and Steinsson (2008)   3.2  4.2 
Source: Klenow and Malin (2010) 
*Klenow and Malin report the frequency of price changes. The duration is computed as ‐1/ln(1‐x) 
where x is the frequency.