November 2002

A fiscal perspective on currency crises and the original sin

Abstract
Maturity and currency denomination of public debt are crucial determinants of macroeconomic dynamics
in an open economy. The focus of this paper is on dynamics in reaction to a shock that increases the real
value of debt relative to governments primary surpluses. In response to such a shock, the rate of price
level inflation and exchange rate devaluation depend on denomination of debt, whereas their timing on
nominal debts maturity. The adjustment to a fiscal imbalance can take the form of a real appreciation
and a current account deficit if the exchange rate parity is kept temporarily fixed, followed by a real
depreciation and a current account surplus coincident with devaluation. Our analysis suggests a broader
view of: (1) fiscal consequences of inflation and devaluation, (2) macroeconomic eects of fiscal shocks, and
(3) interations between fiscal and monetary policy, relative to standards in the open-economy literature.

Giancarlo Corsetti
University of Rome III, Yale University and CEPR
Bartosz Mackowiak
Humboldt University
Paper prepared for the conference Currency and Maturity Mismatching: Redeeming Debt from Original Sin,
organized by the Inter-American Development Bank and Harvard University with the support of the University of
California, Berkeley, to be held in Washington DC on November 21-22, 2002. Correspondence: corsetti@yale.edu
and bartosz@wiwi.hu-berlin.de.

Introduction

Why do governments in emerging markets find it so hard to issue non-indexed long-term debt
denominated in domestic currency? The literature addresses this question from dierent angles,
but a recurrent theme is the lack of credibility of policies that would guarantee price level and
currency stability. Insofar as governments have an incentive to engineer unexpected inflation and
any promise not to do so lacks credibility, a markets assessment of risk in lending long-term in
domestic currency translates into prohibitively high premia.
While this view is common in policy debates, most formal studies provide only a partial
analysis of the reason why surprise inflation can bring ex-post benefits to the government. The
most popular approach focuses on seigniorage revenues. The vast literature on currency crises after
the classic contributions of Krugman (1979) and Flood and Garber (1984), for instance, assumes
that a fiscal imbalance must be matched with revenues from money creation.1 According to this
approach, what ultimately undermines the credibility of macroeconomic stabilization policies is
either the attractiveness of or the lack of alternatives to the option to financing the deficit by
printing money: seigniorage is the safety valve of emerging markets with chronic fiscal deficiencies.
This view of currency crises has received renewed attention after the Asian crisis in the late 1990s,
in which governments with traditionally low or even negative budget deficits reacted to private
sector bankruptcies by providing massive support to firms (see for example Corsetti, Pesenti and
Roubini (1999a)).
Yet revenues from seigniorage are actually quite moderate in many crisis episodes: the adjustment mechanism at the center of these models is not apparent even when crises have a striking
fiscal dimension. Depending on the interest elasticity of money demand, some studies even ques1 The models of Krugman (1979) and Flood and Garber (1984) build on Salant and Henderson (1978). The
first-generation literature includes for example: Obstfeld (1986), Buiter (1987) and Calvo (1987). Buiter, Corsetti,
and Pesenti (1998), Calvo (1998), Cavallari and Corsetti (2000), Flood and Marion (1999) and Jeanne (2000) are
recent discussions of the literature on currency crises. Corsetti, Pesenti, and Roubini (1999a, 1999b) adopt the
first-generations setup as a starting point to interpret crises in emerging markets. In those papers, dierent from
the Krugman model, a fiscal deficit can emerge after rather than before an exchange rate collapse.

tion the possibility that seigniorage revenues are relevant at all.2 Most important, the recent rise
of a general equilibrium literature on price level determination the Fiscal Theory of the Price
Level (henceforth FTPL) provides an alternative model of the link between fiscal imbalances
and currency and price instability, stressing the eect of inflation on the real value of nominal
public liabilities.3

By now, the literature recognizes that seigniorage is not the only mechanism

through which inflation and devaluation can produce fiscal revenues and that it may not be the
one most relevant empirically Burnside, Eichenbaum and Rebelo (2001a, b) are a well-known
articulation of this view.
In this paper we build a simple framework to study a currency crisis associated with a fiscal
imbalance.4 We analyze how the lack of fiscal discipline and flexibility undermines currency and
price level stability, independently of any need for seigniorage. Relative to the existing literature,
we refine the definition of shocks generating fiscal imbalances (including foreign nominal shocks),
we clarify the interaction of fiscal and monetary policy in determining the dynamics of a crisis,
and we analyze the behavior of the real exchange rate during the adjustment process (extending
previous models to include nontraded goods).
Our starting point is the experiment of Krugman and Flood and Garber: in a small open
economy with a fixed exchange rate regime, an exogenous shock decreases the present value of
governments primary surpluses relative to the real value of its outstanding liabilities. Key to this
experiment is that the government cannot or is not willing to adjust its plans so as to correct
the eect of the shock on its fiscal stance. The shock causing fiscal distress, we observe, need not
2 See the working paper version of Corsetti and Ma
ckowiak (2002). Lahiri and Vegh (1998) study in greater
detail how the present value of seigniorage depends in first-generation models on the interest elasticity of money
demand.
3 Among papers on FTPL are, for example: Benhabib, Schmitt-Groh, and Uribe (2001), Cochrane (1999a,
1999b, 2001), Leeper (1991), Sims (1994, 1997, 1999) and Woodford (1994, 1995, 1996, 1998). Woodford (2000)
is a recent review with further references, while Bergin (2000), Dupor (2000), Canzoneri, Cumby, and Diba (1998)
and Loyo (1998) are open-economy applications.
4 It has been observed (see Sims (1997)) that the recent Fiscal Theory of Price Level determination and the
first-generation models of currency crises are close cousins Daniel (2001) and Corsetti and Mackowiak (2002)
are recent papers that study this link in detail. Mackowiak (2002) is a related stochastic equilibrium model in
which fiscal policy regime changes under certain conditions, causing devaluation.

be limited to the path of real primary surpluses, but can be a nominal disturbance such as foreign
deflation. This is because nominal shocks aect the real value of public liabilities and are thus
capable per se of causing fiscal strain.
In our framework the solved-forward government budget constraint by itself implies that the
exchange rate must depreciate or public bond prices must fall, in such a way that the resulting
wealth transfer to the government finances the imbalance. To appreciate our analysis, it is useful to
keep in mind that governments have many imperfectly indexed liabilities other than publicly traded
bonds, including long-term spending commitments like pensions and other transfer programs. For
instance, in the aftermath of an exchange rate crisis governments sometimes freeze spending in
nominal terms. Even when temporary, imperfect indexation of spending (including public sector
wages) can generate large fiscal gains. Moderate changes in the price level can bring about
substantial wealth transfers that help resolve a fiscal crisis in this respect, the simulations on
the fiscal impact of moderate inflation in Sweden by Persson, Persson and Svensson (1998) provide
an instructive example. Consistent with that study, one should interpret nominal public debt
in our model in a broad sense.
We summarize our results as follows:
1. Nominal non-indexed debt acts as a cushion for a government in the presence of a fiscal
shock devaluation and short-run inflation rates are larger when foreign-currency debt is
a higher fraction of all public liabilities.
2. Within the FTPL literature Cochrane (2001) and Woodford (1998) point out in a closedeconomy context that debt maturity matters for dynamics of macroeconomic variables in
reaction to exogenous disturbances. In particular, long-term nominal debt smooths out
eects of fiscal shocks on the equilibrium rate of inflation. The point stressed by Cochrane
and Woodford applies in an open-economy context: a large stock of nominal long-term
liabilities allows the government to prevent fiscal shocks from translating into immediate

pressure on the exchange rate and the price level the government can therefore buy time
and delay nominal adjustment. This is because the current price of long-term bonds decreases
as the imbalance appears, reflecting expectations of future devaluation and inflation. The
option to delay adjustment is therefore available only to countries with a suciently large
stock of long-term nominal liabilities.
3. Interest rate policies that deliver low long-run inflation increase the required adjustment in
the exchange rate and price level in the short run. This result lets us understand why low
chronic depreciation is consistent with a one-time devaluation of dramatic size.
4. Seigniorage revenues need not play any role in financing a fiscal imbalance. To stress this
point, our model abstracts from the monetary base altogether. We study devaluation scenarios in which none of the imbalance is financed by anticipated money growth, and all of
it is financed by unanticipated wealth transfers from holders of nominal liabilities.
5. Devaluation and short-run inflation are highly nonlinear in the share of foreign-currency and
indexed debt in total debt. Relatively small changes in the extent of dollarization of public
liabilities can thus lead to large dierences in the magnitude of an exchange rate and price
level adjustment. This result lets us understand why a government that borrows heavily in
a foreign currency, like some emerging markets, is exposed to a large crisis.
6. A standard small open-economy model with traded and nontraded goods behaves in the
same way as our baseline one-good model: unless the stock of long-term public liabilities is
suciently large, the crisis occurs at the time in which the fiscal imbalance occurs.
7. A simple modification of the standard two-good model can bring the theory more closely into
line with two key stylized facts: a real appreciation and a current account deficit precede
an exchange rate crisis, while a real depreciation and a current account surplus coincide
with a crisis. The amendment of the model consists in allowing for utility-generating public

spending aecting marginal utility of consumption and assuming that spending is cut
at the time of a crisis in proportion to the magnitude of the crisis.
The key conclusion from our analysis is that the dynamics and intensity of currency crises
crucially depend on the currency and maturity structure of public liabilities. While we do not
solve for the currency and maturity structure of public debt endogenously (they are given initial
conditions in our model), our results suggest that the fiscal benefits of inflation are increasing in
the stock of non-indexed public liabilities denominated in local currency. An important topic for
future research is the study of time-consistent equilibria, in which the original sin emerges as an
endogenous equilibrium feature.
This paper is organized as follows. The next section sets up a single-good version of the model.
Section 3 analyzes shocks causing a fiscal imbalance as well as describes post-devaluation interest
rate policy. In section 4, we discuss determinants of the magnitude of inflation and devaluation
at the time of a crisis. Section 5 analyzes macroeconomic dynamics after the emergence of a fiscal
shock. Section 6 extends the analysis to an economy with a traded and a nontraded good, looking
at the dynamics of the real exchange rate and the current account. Concluding remarks are in
section 7. The first appendix describes the behavior of the model when post-devaluation interest
rate policy is more general than a simple peg we consider in the main text. The second appendix
puts together equations of the model with a nontraded good loglinearized about its steady state.
It also lists values of parameters and steady state variables used in numerical solutions of this
model.

A single-good model

This section presents what we think of as the simplest fully-specified model to illustrate the mechanism through which a fiscal imbalance generates currency and price instability. The setup draws
on Corsetti and Mackowiak (2002), where we postulate a small open-economy with a government
and a representative individual who receives an endowment of a single, tradable and perishable
5

consumption good. In the single-good world economy with costless trade, the domestic price level
P and the foreign (dollar) price level P are linked by the law of one price: P = EP , where E
denotes the exchange rate. The economy takes exogenously the foreign real interest rate (equal
to the dollar nominal rate). We assume that the interest rate is equal to a constant r such that
(1 + r) = 1, where is the representative individuals discount factor.

2.1

The representative individuals optimum problem

The representative individual pays lump-sum taxes and holds: one-period nominal bonds B paying
an interest rate denoted by it , nominal perpetuities L selling at a price t , and one-period dollar
bonds B . We assume that B and L are each strictly greater than zero, issued by the domestic
government and held only by the representative individual in the domestic economy, whereas
dollar bonds can be traded internationally.5
The representative individual maximizes:

t ln Ct

(1)

t=0

subject to:

Bt Et Bt
(1 + t ) Lt1 (1 + it1 ) Bt1 Et (1 + r) Bt1
t Lt
+
+

+
+
+ Yt t Ct
Pt
Pt
Pt
Pt
Pt
Pt

in every period, as well as to:

lim

1
1+r

t Lt Bt Et Bt
+
+
Pt
Pt
Pt

where Y and C denote endowment and consumption of the single good, respectively, and are
real lump-sum taxes (or transfers, if negative).
5 Notice we think of the price level, P , as the rate at which a newly issued, one-period nominal bond trades
for the real commodity. Analogously, of the nominal interest rate as the rate at which the domestic government
exchanges old, one-period nominal bonds for new ones; and of the perpetuity as an asset promising one unit of
short-term bonds in every period from now on. Also, a comment is in order regarding the assumption that foreign
investors do not own domestic nominal assets. This assumption simplifies the analysis, allowing a sharp focus on the
dynamics of currency collapse this paper aims to illustrate. The distribution of nominal claims on the government
between domestic agents and foreigners matters for a welfare evaluation of a crisis, something we leave for future
research.

Let t equal the shadow value of the individuals budget constraint. The first order conditions
for an optimum in the individuals problem are as follows (in turn, we display conditions with
respect to C, B , B and L):
1
= t
Ct
t+1 Et+1
t Et
= (1 + r)
Pt
Pt+1
(1 + it ) t+1
t
=
Pt
Pt+1
(1 + t+1 ) t+1
t t
=
Pt
Pt+1
The transversality condition is:
lim

1
1+r

t Lt Bt Et Bt
+
+
Pt
Pt
Pt

=0

(2)

The first order conditions with respect to B and B imply that the uncovered interest rate
parity holds:
1 + it = (1 + r)

Et+1
Et

(3)

and the first order condition with respect to L yields an intertemporal relationship between perpetuity prices:
t (1 + it ) = 1 + t+1

(4)

We solve (4) forward, arriving at a relationship between the perpetuity price and future nominal
interest rates:6
t =

" s

X
Y
s=0

k=0

1
1 + it+k

(5)

6 In our forward solution, we restrict attention to competitive equilibria in which the following terminal condition
holds:
" s
#
Y
1
lim
t+s+1 = 0
s
1 + it+k
k=0

2.2

The government budget constraint

The period-by-period government budget identity is:

(1 + t ) Lt1
(1 + it1 ) Bt1 Et (1 + r) Ft1
t Lt Bt Et Ft
+
+
=
+
+
t
Pt
Pt
Pt
Pt
Pt
Pt

(6)

where F denotes one-period dollar bonds issued by the domestic government. Notice our assumption the government can borrow in dollars at the world interest rate r, implying that it does
not default on its dollar bonds.7
terminal condition:
lim

Using the individuals first order conditions and the following

1
1+r

Et Ft
t Lt Bt
+
+
Pt
Pt
Pt

=0

(7)

we obtain a solved-forward version of the government budget constraint (6):

(1 + t ) Lt1 + (1 + it1 ) Bt1 Et (1 + r) Ft1 X

+
=
Pt
Pt
s=0

1
1+r

t+s

(8)

In equilibrium, the real value of public liabilities equals the present discounted value of future real
primary surpluses.8

2.3

Equilibrium

A competitive equilibrium in this small open-economy is a specification for a time path of the vector
{Y, r, C, B , B, L, F, , i, , E, P , P } such that: (1) when the representative individual takes the
{Y, r, , i, , E, P , P } part of the path as given, {C, B , B, L} solves her optimum problem; (2)
the solved-forward government budget constraint (8) holds.9
7 It is not realistic to rule out the possibility of government default, but there is a good reason for doing so in
a simple model. Outright default and the kind of implicit default on nominal government liabilities that occurs
via unanticipated inflation are dierent phenomena. A holder of a government bond on which outright default is
possible is exposed not just to to aggregate default uncertainty, but also to negotiation and legal costs as well as
additional uncertainty over which creditors would bear the default cost. See Sims (1997) for a discussion.
8 In a closed economy model, this derivation would simply make use of (2) as the terminal condition we
can see this by imagining a closed economy model with an identical setup except that B (or F ) would denote
indexed, or real, debt. An open-economy model like ours, however, introduces a subtle problem: it is theoretically
possible that, with perfect insurance markets between countries, debt of one government can grow without a bound
corresponding to an explosive growth of assets of another government. Private sectors transversality generates
a terminal condition for solved-forward budget constraints of all governments taken together, rather than for the
constraint of each government separately. Like Bergin (2000) and Sims (1999), we find it realistic to consider
equilibria that emerge when the government in an open economy faces a terminal condition in the form of (7).
9 Recall that we postulate (8) as an additional equilibrium condition. In a closed economy model, an equivalent
condition would be implied by (6) and private sectors optimization. Note also that in this open-economy model
F , determined by government supply, need not equal B .

On the nature of shocks causing a fiscal imbalance

Suppose that the foreign price level is equal to a constant, P , and that the government fixes the
exchange rate at E. For a fixed exchange rate policy to be sustainable, the government must run a
fiscal policy consistent with its solved-forward budget constraint (8) holding at the current parity.

Define { t+s }s=

s=0 as a path of primary surpluses such that (8) holds given P and initial debt

s 0. By
(Bt1 , Lt1 and Ft1 ) with the exchange rate permanently fixed at E, that is Et+s = E,
definition, then, { t+s }s=
s=0 is a sequence such that:
s

X
(1 + 1/r) Lt1 + (1 + r)Bt1 (1 + r) Ft1
1
t+s =
+

1
+
r
P
EP
s=0

(9)

To satisfy the above condition, a sequence of primary surpluses must eventually imply a strong
feedback from debt to s. In other words, in response to shocks hitting the economy, policy must
The FTPL literature refers
set a sequence of s so as to insure that they fully back debt given E.
to such a fiscal policy as passive or Ricardian.
Recent contributions begin their analysis of a currency crisis, in the spirit of Krugman (1979),
by assuming that the government fails to stick to Ricardian policy (i.e. Burnside, Eichenbaum and
Rebelo (2001a), Daniel (2001), Corsetti and Mackowiak (2002)). As a complement to focusing
directly on a deterioration of primary surpluses, it is appealing to motivate a fiscal imbalance
as a consequence of an external shock. There is a vector-autoregressive literature documenting
that a substantial fraction of macroeconomic variation in small economies originates abroad (see
for example Calvo, Leiderman and Reinhart (1993), Cushman and Zha (1997) and Mackowiak
(2001)). Most economists would probably find it intuitive that a change in real international
prices exogenous to a small economy, like the terms-or-trade or the world real interest rate, can
result in fiscal slippage. What is not readily apparent, and what this model makes clear, is that
a foreign nominal shock can also lead to a fiscal imbalance with critical consequences for a fixed
exchange rate policy.
Suppose the foreign price level changes in period t to a new, permanent level P where P P .
9

Provided the inequality is strict, the shock increases on impact the real value of both nominal
and dollar debt. Foreign deflation increases the quantity of the real commodity the government
promises to bondholders, regardless of whether their claims are denominated in domestic or foreign
currency. Given the definition of { t+s }s=
s=0 , we now write the solved-forward government budget
constraint (8) in period t as follows:
"
#
s
X
(1 + r) Ft1
1
(1 + t ) Lt1 + (1 + r) Bt1
+
=
t+s
Et P
P
1+r
s=0

(10)

where Et may or may not equal E in equilibrium. In the above expression, is a measure of
the extent of fiscal adjustment the government undertakes in reaction to the shock, with < 0
corresponding to reform and > 0 to deterioration beyond the immediate eect of the shock.10
A restriction we place on is that it satisfies:
"
# "
#
s
X
(1 + 1/r) Lt1 + (1 + r)Bt1 (1 + r) Ft1
1
>
t+s
+

1+r
P
EP

s=0
Thus we in eect assume that the shock coincides with fiscal policy becoming active or nonRicardian, in the terminology of the FTPL literature.
Two realistic scenarios make the case P < P interesting. The first is deflation in the euro area
or the United States, or an appreciation of the dollar that decreases dollar prices of commodities
traded in the world market. The second is the possible role of Brazils devaluation in 1999 in
the recent crisis in Argentina. A devaluation in Brazil decreases international prices faced by
Argentina. While this is an improvement in Argentinas terms-of-trade or a relative price change,
the impact of the shock on the real value of Argentinas public debt is equivalent to that of a
decrease in P or a purely nominal shock.11
10

For example, one can assume that from time Td onwards primary surpluses change by a constant d:

Td t
(1 + r) d
1

r
1+r

Letting Td t allows for the possibility that primary surpluses change at a future date.
1 1 A more articulated model, with a nominal rigidity, might predict a decrease in Argentinas output in reaction to
Brazils devaluation, and a consequent deterioration in Argentinas primary surpluses. See for example Hausmann
and Velasco (2002) for an analysis of the recent crisis in Argentina, including a discussion of the impact of Brazils
devaluation on its neighbor.

10

While we think of the case P < P as a realistic example, our analysis obviously allows for
fiscal imbalances to emerge also when P = P . The point is that the adverse shock underlying
can stem from a variety of external and domestic sources. Once this disturbance occurs, the
government may attempt to adjust the path of its budgets in an eort to reverse the negative
impact of the shock. What we assume is that the government cannot (or is not willing to)
implement reforms such that its debt continues to be backed fully with real taxes and (8) holds
Thus is defined as the net change in the present
with the exchange rate permanently fixed at E.
value of real primary surpluses after the government has taken all measures it deemed feasible to
reverse the impact of the shock. Note that the restriction on implies that > 0 so long as
P = P .
So long as the exchange rate is fixed at E, the nominal interest rate is determined by the
To complete the description of equilibrium,
uncovered interest rate parity (3) evaluated at Et = E.
we must specify an interest rate policy in the event of a devaluation. We posit that the interest
rate is set according to the reaction function:
1 + iT +s = 0 + 1

ET +s
ET +s1

(11)

where T t is the period in which the current peg is abandoned, and s 0. According to this rule,
the (gross) nominal interest rate is a linear function of the (gross) depreciation rate. We assume
that 1 < 1, that is, policy makes the nominal interest rate respond weakly to depreciation (or
inflation).12

Thus our rule is an instance of passive monetary policy familiar from the FTPL

literature. To derive closed-form solutions, in the main text we specialize to a simple interest rate
peg such that iT +s = ip (that is, 1 = 0 and 0 1 = ip ), s 0, and examine the general rule
(11) in the appendix. Note that the choice of ip determines the rate of chronic depreciation in the
post-devaluation period.
1 2 We also assume that 1 so that the post-devaluation (gross) chronic depreciation rate converges
0
1
to a number greater than or equal to 1. This assumption is inessential for substantive results of this paper, but
allows us to rule out persistent deflation after devaluation.

11

Nominal debt and the severity of currency crises

4.1

Devaluation and inflation with foreign nominal shocks

In this section, we analyze the adjustment to fiscal imbalance under the assumption that the
government abandons the peg in the same period in which the imbalance appears thus causing
an unanticipated jump in the exchange rate at t. Unanticipated devaluation and inflation reduces
the real value of nominal public liabilities B and L, resulting in a wealth transfer from their
holders. At the same time, the post-devaluation interest rate rule determines the equilibrium
price of the perpetuity, t . By (5), a higher post-collapse inflation (and higher nominal interest
rates) reduce t , therefore creating an additional wealth transfer from holders of perpetuities to
the government. In equilibrium, the wealth transfers due to jumps in both the exchange rate and
the price of perpetuities finance the fiscal imbalance.
Formally, once policy chooses the post-collapse nominal interest rate it+s = ip , s 0, the
equilibrium condition (5) yields the perpetuity price t = (1/ip ). To solve for the equilibrium

devaluation rate Et /E , we make use of the solved-forward government budget constraint

combining (9) with (10) we obtain the unique solution:

[1 + (1/ip )] Lt1 + (1 + r) Bt1

EP
Et

=
E
P P (1 + r) Ft1
[1 + (1/r)] Lt1 + (1 + r) Bt1

P P
EP

(12)

The equation (12) highlights several properties of the equilibrium devaluation rate Et /E . To

start with, the jump in the exchange rate is increasing in the size of the fiscal imbalance: Et /E
is decreasing in P and increasing in . The exchange rate change is determined both by the

size of the shock, and the policy reaction to it. But for any given size of the imbalance, dollar
debt acts as leverage the higher the fraction of debt denominated in dollars, the larger the
devaluation rate. An economy in which the government borrows heavily in dollars, like many
emerging markets, sees a larger jump in the exchange rate for any given fiscal imbalance. Note

that post-devaluation interest rate policy influences Et /E , because of its eect on the long-term
12

bond price. Specifically, a higher ip implies a larger jump in t and consequently a smaller one

time devaluation rate Et /E . Thus a currency crisis followed by little chronic inflation, like in
some recent episodes, is associated with a larger initial exchange rate adjustment. In fact, setting
ip = r implies that the depreciation rate in the post-devaluation steady state is zero. If no further
shocks hit the economy, the dynamics involve a (relatively larger) one-time jump in the exchange
rate to a new, constant level.13
It is straightforward to check that the unique solution to the solved-forward government budget constraint satisfies the other equilibrium conditions, and therefore is the unique equilibrium
exchange rate. In particular, given ip and Et , the uncovered interest rate parity (3) determines
the anticipated (gross) rate of depreciation (Et+1 /Et ). The result that non-Ricardian fiscal policy
and passive monetary policy deliver a uniquely determinate price level and a stationary inflation
rate is familiar from the FTPL literature.14
Subsequent algebraic derivations are easier to follow if we assume that P = P . The solution

for Et /E makes is clear that this assumption does not aect generality of the argument we can

simply think of as a complete measure of the fiscal impact of the shock, whatever such shock

may have been. So long as P = P , we can normalize P to one and think of the domestic price
level as synonymous with the exchange rate. Given our earlier assumption that (1 + r) = 1, the
first order condition with respect to B now implies that the shadow value of the budget constraint
is constant at . We then know that optimal consumption is constant and equal to 1/.

4.2

The shadow exchange rate

An important question is whether the government can choose to delay the exchange rate and
price adjustment past period t. For a delay to be consistent with equilibrium, the solved-forward
government budget constraint must hold with Et = E exclusively by virtue of a wealth transfer
13

We assume that ip is chosen so that Et > E.

1 4 To determine the equilibrium level of consumption, so long as P < P , we would need to make an arbitrary

assumption regarding the small economys net international asset position in dollar bonds.

13

from holders of long-term debt due to an unanticipated jump in t . But this is only possible if
there are enough perpetuities outstanding at the time of the fiscal shock. Namely, it must be true
that:

(1/r) Lt1 /E

(13)

where the expression on the right-hand side of the inequality is the maximum wealth transfer
through an initial jump in the long-term bond price. If the above expression does not hold,
the government has no choice but to abandon the fixed exchange rate at t in equilibrium
the exchange rate must depreciate at t. Note that, while we consider perpetuities for simplicity,
our argument applies to the case of a large enough stock of nominal debt of sucient maturity.
Thus the ability to postpone devaluation requires that long-term nominal debt be of sucient size
relative to an imbalance a condition emerging markets are less likely to satisfy than developed
economies.
Assuming that (13) holds as a strict inequality, we can now study the determinants of the
devaluation rate if the government does not abandon the peg at the time when the fiscal imbalance
materializes. First, combining (9) and (10) with Et = E, we solve for the equilibrium perpetuity
price in period t:
t =

E
1
>0

r
Lt1

(14)

Note that t is uniquely defined, because an initial jump in the bond price must guarantee that
the solved-forward government budget constraint holds with Et = E. Second, let T > t denote the
date in which the government abandons the peg: using equilibrium relations (3) and (4) as well
as the policy rule iT +s = ip , s 0, we obtain an intertemporal relationship between perpetuity
prices in periods t and T :
1
t = +
r

1
1+r

T t1

1
1 + (1/ip )

r
(1 + r) ET /E

(15)

Putting together (14) and (15), we arrive at the unique solution for the equilibrium devaluation

14

rate ET /E :15

ET
=
E

Lt1 /E

ip
1+r Lt1
ip
(1 + r)T t

p
1+ip
r
1+i
E

(16)

Some properties of this rate and of the perpetuity price t are by now familiar from our analysis
of devaluation in period t: (1) A larger imbalance requires a larger jump in the bond price
and a larger devaluation rate at any date. (2) A larger Lt acts as a cushion for the government,
being associated with a smaller jump in the bond price and a smaller devaluation rate. (3) An
increase in post-devaluation interest rates (or in chronic inflation) decreases the devaluation rate.
In addition, we can now see that delaying the collapse further into the future (i.e. an increase in
T ) raises the equilibrium devaluation rate.16
After Flood and Garber (1984) we refer to the exchange rate conditional on abandoning the
peg, determined by equation (16), as the shadow exchange rate. The shadow rate can be
calculated for all dates s > t and shares the properties of ET just discussed.

4.3

Numerical examples

Properties of the equilibrium devaluation rate are best appreciated with the help of numerical
simulations. Figure 1 presents several plots of the devaluation rate in an economy where total
public debt is 50 percent of GDP, and unless stated otherwise the post-collapse domestic
interest rate ip is set equal to the international rate r at 5 percent.
Consider the left-hand side graph in the top row, which assumes that dollar debt is 50 percent
of total debt. With this fraction of foreign-currency debt relatively low for many emerging
markets17 foreign deflation and a fiscal imbalance have a notable but moderate impact on the

rate of devaluation Et /E . In this example, foreign deflation causes a devaluation of approximately

the same rate (see the line corresponding to = 0). Deteriorating primary surpluses magnify the
15

Daniel (2001) derives a similar expression.

Following our discussion of devaluation in period t, it should be evident that the unique solution for ET /E
satisfies the other equilibrium conditions.
16

1 7 For example, Levy (2002) reports for Argentina a share of dollar bonds in government debt of 87-98 percent
between 1997 and 2001.

15

impact of foreign deflation on the devaluation rate. The same graph includes two lines drawn for
= 0.1 and = 0.2, corresponding to imbalances with present discounted value as high as 10
and 20 percent of GDP. These lines lie about 4 and 8 percentage points above the line for = 0,
respectively.
Things change dramatically when a large fraction of public liabilities is indexed to a foreign
currency. Not only is the devaluation rate much larger than the rate of foreign deflation, but the
relation between the two variables becomes highly nonlinear. Consider the right-hand side graph
in the top row. When the fraction of dollar debt (call it ) is 50 percent, a 10 percent negative
shock to foreign prices is associated with a 10 percent devaluation. When increases to 90 percent,
the same shock leads to a devaluation by more than 20 percent. Because of the nonlinearity due
to dollar debt, a moderate amount of foreign deflation can cause a large devaluation even without
a deterioration in primary surpluses and without a delay in the timing of devaluation. The left

hand-side graph in the bottom row illustrates the nonlinearity most clearly, showing Et /E as a

function of with foreign deflation being held constant at 10 percent. For example, the impact
of a combination of foreign deflation and = 0.1 more than doubles when increases from 80 to
90 percent. A relatively small change in the extent of dollarization of public liabilities can lead
to large dierences in the magnitude of exchange rate adjustment during a crisis.
The right-hand-side graph in the bottom row illustrates that the shadow devaluation rate

ET /E is a nonlinear increasing function of the time interval between the arrival of adverse fiscal
news and the abandonment of the peg. As discussed in the next section, the timing of a delayed

devaluation is bounded from above and devaluation rates approach infinity close to this bound.
Observe also that a larger stock of long-term liabilities (a higher value of in the graph) helps
containing the jump in the exchange rate for any given time of collapse. A higher post-devaluation
interest rate ip decreases the value of long-term liabilities in period t, lowering the shadow rate
for any given time of collapse T .

16

Macroeconomic dynamics

5.1

An upper bound to the date of the crisis

In this section, we delve into the question of understanding the dynamics of adjustment to fiscal
imbalances. We first note that, when calculating the equilibrium devaluation rate at T , we have
implicitly assumed that (1+r)T t is no larger than the maximum wealth transfer from long-term
bond holders:
(1 + r)T t <

Lt1
E

1+r
r

(17)

Otherwise, the adjustment cannot but be immediate. The above formula then suggests that, given
and Lt1 , a currency collapse must occur before the passage of time changes the sign of the
above inequality. It is straightforward to show that this happens at Tb such that:

Lt1 1 + r
log
log
r
E
Tb = t +
r

(18)

Thus conditional on Lt1 and , there exists a finite upper bound Tb to the date of devaluation.

The government can borrow to defend the peg until Tb.18

A currency crisis occurring at a date

close to this upper bound corresponds to an extreme scenario in which the government extracts
the maximum wealth transfer from holders of long-term bonds. We have seen above that the size
of devaluation is increasing in the dierence T t: evaluating expression (16) as T Tb, we find

that such a crisis is associated with extreme rates of devaluation and inflation.

While a fiscal imbalance makes devaluation inevitable, the model fails to pin down the date
of the collapse any period between t and Tb can be the devaluation date T .19

This is a

key dierence relative to the first-generation literature on currency crises. Here, unlike in first-

generation models, it is not true that a crisis must occur in the first period in which the shadow
1 8 Lahiri and Vegh (2000) introduce a government bond that is assumed to have utility value its price is
therefore not determined exclusively by the uncovered interest parity. The government can then borrow by issuing
such bonds at increasing interest rates. Buiter (1987) directly assumes an upward-sloping supply schedule for
reserves. Those considerations do not matter for the substance of our argument they merely act to decrease Tb as
borrowing increases.
1 9 If the government attempted to delay the devaluation past T
b, its solved-forward budget contraint would fail
to hold with Et = E, and the peg would collapse in period t.

17


exchange rate, uniquely defined in every period between t and Tb, is above E.

5.2

Interest rate policy and the timing of currency collapse

There are several reasons why an exchange rate adjustment may occur before Tb. One can appeal

to self-validating runs on debt, political factors, or interaction of fiscal and monetary policy. To
model each of these possible reasons, the basic specification must be appropriately modified. Our
preferred approach is as follows. Suppose that the government finds interest rate increases costly
to implement. Specifically, assume that there is a unique upper bound on the interest rate that
policy is unwilling to cross: the government abandons the peg in the first period T > t such that
keeping the exchange rate fixed would cause the interest rate to rise to or above i at T . The period
T 1 is then the last period in which the interest rate is below the policy-specified threshold:
T = min s such that is i

(19)

By the interest rate parity condition (3), the shadow exchange rate maps one-to-one into
the shadow nominal interest rate, defined as the nominal interest rate conditional on devaluation one period ahead. When monetary policy delays devaluation, the shadow interest rate rises
monotonically. Given an upper limit on the interest rate, the timing of a delayed devaluation is
unique within the interval between t and Tb. Thus the model predicts that fiscal policy makes

devaluation inevitable, but that its date is determined by interest rate policy. Other things equal,
policymakers willing to let the interest rate rise higher can defend the peg longer.20
The dynamics of macroeconomic adjustment after a fiscal imbalance are best captured by using
the shadow interest rate and the shadow exchange rate. When the adverse news arrives at t, the
2 0 Recall that when we examine implications of a fiscal imbalance for timing of the crisis, we find that an imbalance
per se implies only an upper bound for the timing. By contrast, the first-generation literature on currency crises
emphasizes the unique timing of devaluations caused by a fiscal imbalance indeed, one often refers to uniqueness
as one of the main insights from that literature. As is well known, the unique timing in the first-generation models
is due to an assumption of a minimum level of reserves models with debt in eect assume that investors impose
an exogenous borrowing constraint on the government (see for example Burnside, Eichenbaum and Rebelo (2001a)
and Daniel (2001)). In this section we have adopted a similar modeling strategy an exogenous interest rate
threshold that delivers the first-generations uniqueness result in our framework. It is easy to show that the
exogenous limit on reserves in Krugmans (1979) model is a special case of the constraint on interest rate policy
we use. If one abstracts from the possibility of government borrowing as Krugman does his assumptions of
an exogenous rate of domestic credit expansion and a lower threshold on reserves imply, taken together, a lower
bound on the monetary base, or an upper bound on the interest rate (see also Cavallari and Corsetti (2000)).

18

shadow exchange rate jumps discontinuously,21 then depreciates smoothly until period T , when it
coincides with the actual exchange rate ET . The shadow interest rate also jumps at t, but remains
below i at T 1. T is the first period in which the shadow interest rate is equal to or above i at
T . Clearly, the realized interest rate equals its shadow value in period T 1, when the speculative
attack leading to the exchange rate crisis occurs.22

A model with a traded and a nontraded good

6.1

An additional adjustment mechanism

The discussion of the adjustment to a fiscal imbalance has so far focused on the maturity of public
debt. We have shown that the point stressed in the FTPL literature applies in an open-economy
context: future inflation and devaluation reduce the current value of nominal long-term liabilities,
and thus may allow the government to maintain temporarily price level stability and the exchange
rate peg. But the option to delay a price level and exchange rate adjustment is only available
when the stock of long-term liabilities is large enough relative to the size of a fiscal imbalance.
A more realistic open-economy model with a traded and a nontraded good suggests
another possible mechanism, independent of debt maturity, via which price level inflation can
bring about a fiscal transfer consistent with a temporary exchange rate stability. Suppose that,
when an imbalance appears, the price of the nontraded good rises in such a way that the resulting
decrease in the real value of nominal government debt finances the imbalance. If this were the
2 1 The shadow exchange rate can depreciate or appreciate on impact. The latter case is possible when policy
chooses a large ip , generating a capital loss to holders of perpetuities that is more than enough to finance the
imbalance at the time of the shock. By contrast, the first-generation literature derives a famous result, replicated
in textbooks: the timing of a speculative attack is unique because a crisis must occur in the first period in which the
shadow exchange rate is above the old parity. Put dierently, if agents attacked the peg too early, the currency
would appreciate. We have always found it awkward that the classic model refers to appreciation an instant before
a crisis, when everyone knows the fiscal problem. The predictions of our model appear to us more convincing: the
shadow rate changes discontinuously at the time when the fiscal shock arrives and continues to depreciate smoothly
thereafter.
2 2 In discrete-time versions of the Krugman model, there are two types of attacks: the ones occurring before T ,
when reserves are lost and the final decisive attack in period T , when reserves are exhausted. What is awkward
in this setup is that, at T , the exchange rate is equal to the post -devaluation equilibrium value ET . However, in the
literature agents attacking the currency at T are able to buy dollars at the old parity, at an o-equilibrium price.
Obstfeld (1986) is a classic reference on this issue, pointing out that such models implicitly assume that speculators
receive a fiscal transfer at the time of the attack. Our specification with an interest rate threshold avoids any such
considerations.

19

case, the government could choose to postpone devaluation and the adjustment mechanism would
involve a real appreciation in anticipation of a speculative attack. Dierent from the single good
model, there would be real eects of anticipated devaluation, consisting of fluctuations of the real
exchange rate and the current account.
What makes this new channel of adjustment interesting is that the implied macroeconomic
dynamics could bring the model closer to empirical regularities from crisis episodes. The key
stylized facts are that a currency crisis is often preceded by a real appreciation and a current
account deficit, and coincides with a real depreciation and a current account surplus.23

The

question is whether a two-good version of our model can improve the match of the theory with
the stylized facts.
In what follows, we set up a version of our model with a traded and a nontraded good but without long-term debt, and reconsider the eects of an external deflationary shock. First, we prove
a negative result: in our benchmark specification there exists no equilibrium with a temporary
real appreciation and a delayed devaluation. Equilibrium involves a jump in the exchange rate at
the time of the shock, like in the single-good model, without any change in the real exchange rate.
We interpret this finding as a confirmation that debt maturity is critical for understanding when
a given fiscal shock aects the equilibrium exchange rate, as suggested by our single-good model.
Our nonexistence result is related to Uribes (2002) analysis of the price-consumption puzzle
of currency pegs. As Uribe points out, standard optimizing models imply that an anticipated real
appreciation must be accompanied by a declining path of consumption. In a similar fashion, our
model suggests that rates of return on assets are high in anticipation of a crisis, which implies a decreasing path for marginal utility of consumption. On the other hand, the adjustment mechanism
we conjecture involves a current account deficit at the time of a fiscal shock followed by a surplus
2 3 For a discussion of stylized facts see for example Burstein, Eichenbaum and Rebelo (2002), Corsetti, Pesenti
and Roubini (1999a) and Milesi-Ferretti and Razin (1998). Schneider and Tornell (2000) derive predictions for the
dynamics of the real exchange rate in a setup with a credit market imperfection and a government bailout guarantee.
Burnside, Eichenbaum and Rebelo (2001b) and Burstein, Eichenbaum and Rebelo (2002) present models accounting
for a large real depreciation at the time of an exchange rate crisis but not for a real appreciation prior to it.

20

at the time of a delayed crisis, implying a declining path for consumption. Those two implications
contradict each other if there is a monotone decreasing relationship between consumption and its
marginal utility.
Second, we explore a richer specification of our two-good model, introducing two modifications
to its basic setup. The first modification consists of allowing for government spending on utilityproviding public goods that aect marginal utility of consumption. Specifically, we postulate
a cut in government spending at the time of a crisis capable of decreasing marginal utility of
consumption even if consumption itself is also low. The idea is that a reduced supply of public
goods (think of policy and judicial services guaranteeing private property) decreases the enjoyment
of any given quantity of private goods. This modeling strategy is a way to break the monotone
relationship between consumption and its marginal utility.
The second modification consists of postulating a feedback from pre-devaluation endogenous
variables to post-crisis fiscal policy. The FTPL literature by now recognizes that uniqueness of
equilibrium in models like ours depends on beliefs about how the buildup of debt before the crisis
aects the scale of fiscal reform after the abandonment of the peg (see Mackowiak (2002)). The
reason is that an increase in it drives up government debt and, with insucient post-devaluation
reform, can cause a self-fulfilling increase in the equilibrium exchange rate.24
We combine both modifications by assuming that fiscal reform at the time of a crisis takes the
form of a decrease in utility-providing government spending, and that the extent of the reform is
increasing in the size of the speculative attack.25 In the unique equilibrium of our modified model,
the real exchange rate appreciates in the runup to the crisis, while the current account turns into
a deficit and consumption increases. The devaluation then coincides with a real depreciation and
2 4 Our single-good model has a uniquely defined exchange rate only because, once a fiscal imbalance appears,
the solution for the exchange rate depends solely on T but not on the stock of debt or the interest rate.
2 5 It is plausible that a crisis prompts some fiscal reform, and that the extent of the reform is related to the size of
the crisis. Crises and reform are costly, and reform may be delayed in a heterogeneous society until a crisis makes
it evident that costs of further fiscal inaction are suciently high. The political economy literature has formalized
the idea that crises are beneficial for economic reforms (for example, Drazen and Grilli (1993) and Zarazaga (1997)).
It is also plausible that fiscal reform in the context of a crisis begins with cuts in funding for law enforcement
that decrease marginal utility of private consumption.

21

a current account surplus. It is worth stressing that the model matches the dynamics of the real
exchange rate and the current account in the data, purely as part of a fiscal adjustment to a
foreign shock.

6.2

A new setup

This subsection presents formal details of the new model it diers from the one in section 2 in
two important ways: we allow for a nontradable good and we assume utility-providing government
spending. To save space, we do not restate expressions that are either unchanged or changed in
obvious ways.
In the new setup, the representative agent receives endowments of an internationally traded
and a nontraded good and maximizes:

t Gt ln Ct

t=0

where C is the Cobb-Douglas composite of consumption of the traded good, CT , and the nontraded
good CN :
C=

1
CT CN
(1 )1

(20)

0 < < 1. The price level depends on the price of the traded good, equal to EPT because of the
law of one price, and the price of the nontraded good PN :

P = (EPT ) PN1
The period-by-period budget constraint of the individual takes the form:

Et PT,t YT,t
(1 + it1 ) Bt1 (1 + r) Bt1
PN,t YN,t
Bt Et Bt
+
=
+
+
+
t Ct
Pt
Pt
Pt
Pt
Pt
Pt

Notice we drop the long-term bond from the menu of available assets.
Consider the first order conditions in the agents intertemporal optimization problem. The
condition with respect to C becomes:
Gt
= t
Ct
22

while the conditions with respect to B and B remain unchanged. When (1 + r) = 1, the three
intertemporal conditions simplify to the familiar uncovered interest rate parity (3) and:
Gt+1 Et+1
Gt Et
=
Ct Pt
Ct+1 Pt+1

(21)

The agents intratemporal optimization is governed by the condition:

CT EPT

CN PN

(22)

The period-by-period government budget identity is:

Bt
(1 + it1 ) Bt1
=
t + Gt
Pt
Pt

(23)

where now denotes real lump-sum taxes (not the primary surplus, which equals G). Since we
solve this model in loglinear form and we already know that the most interesting eect of dollar
debt is due to a nonlinearity, we assume here that all public debt is nominal.
To complete the description of the model, we state the current account identity:

Bt = (1 + r) Bt1
+ PT,t
(YT,t CT,t )

and define the equilibrium in the market for nontradables, assuming for simplicity that government
spending is limited to the nontraded good:
YN G = CN

(24)

It is convenient to illustrate our argument in a two-period version of this model (we denote these
periods with t and t + 1). Consider a steady state with a constant PT in which the government
fixes the exchange rate permanently at E, pursuing a Ricardian fiscal policy. We loglinearize the
model around its steady state, using the standard notational convention according to which X t
denotes the value of the variable X in the steady state in period t, while xt denotes the percentage
deviation of the current value of the variable Xt from X t , i.e. xt = log Xt log X t .
We assume that the external shock to the loglinearized model consists of an exogenous decrease
in the international price of tradables, pT,t < 0 (we allow for the shock to be temporary, i.e.
23

pT,t+1 = 0, or persistent.) Holding the exchange rate parity fixed, then, the law of one price
implies a fall in the domestic price of traded goods, raising the real value of government debt. We
suppose that fiscal policy switches to non-Ricardian coincident with the shock, that is the path of
primary surpluses fails to adjust fully to match the new, higher value of debt.
As we are interested in the macroeconomic dynamics in the process of adjustment, we also
suppose that policy keeps the exchange temporarily fixed, et = 0, letting the currency depreciate
in period t + 1 (i.e. et+1 > 0). We then solve the model expecting that pN,t > 0. The second
appendix lists all equations of the loglinearized model. An equilibrium of the loglinearized model
is a vector
(ct , ct+1 , cT,t , cT,t+1 , cN,t , cN,t+1 , et+1 , pt , pt+1 , pN,t , pN,t+1 , it )
that solves equations (27) - (38) for given values of pT,t and pT,t+1 , assuming et = 0.

6.3

Reconsidering the role of long-term debt

As a first step, we carry out our experiment excluding any change in taxation and spending, i.e.
setting t = t+1 = gt = gt+1 = 0, which is equivalent to = 0. The notable result here is
that the model has no solution for et = 0. Without long-term debt, equilibrium requires that the
exchange rate depreciate immediately just as in the specification with a single good.
To see why, suppose that in equilibrium et = 0. With the price of tradables falling together
with the foreign price, i.e. pT,t = pT,t < 0, the solved-forward government budget constraint
requires that the price of the nontraded good PN,t must increase to reduce the real value of public
debt consistent with an unchanged path of primary surpluses. Thus, since in equilibrium there is
no change in consumption of the nontraded good, cN,t = 0, nominal spending on this good must
increase, i.e. cN,t + pN,t > 0. Consider then the intratemporal optimality condition (22) or its
loglinear version (33). Since spending on the traded good is proportional to CN,t PN,t , it too must
increase. For cT,t + pT,t to be positive with pT,t < 0 it must be the case that domestic agents
consume more of the traded good, moving the current account toward a deficit. By (20), overall

24

consumption Ct increases in period t. In period t + 1, by contrast, overall consumption Ct+1

must fall: the nontraded goods equilibrium still implies that cN,t+1 = 0, but the current account
must move toward a surplus so that cT,t+1 < 0. This argument establishes that, if a solution
with et = 0 exists, it exhibits a downward sloping path for consumption, ct+1 ct < 0. Elements
responsible for this conclusion are valuation of public debt, intratemporal optimization and the
current account.
The crucial observation for establishing nonexistence of equilibrium with et = 0 is that a
decreasing path for consumption is inconsistent with intertemporal optimality. Consider the intertemporal condition (21) or its loglinear version (27), normalizing government spending to unity
(G = 1) in both periods. Since we study the consequences of a deflationary shock, we expect
that a solution satisfies et+1 (pt+1 pt ) > 0, that is the rate of exchange rate depreciation in
period t + 1 exceeds the cumulative rate of inflation in periods t and t + 1. But then intertemporal
optimization requires that ct+1 ct > 0 (rates of return on assets are high in anticipation of a
crisis, which implies a downward sloping path for marginal utility of consumption), generating a
contradiction.
This result shows that an important lesson from the single-good model carries over to an
environment with a traded and a nontraded good: debt maturity is critical for understanding
when a given fiscal shock aects the equilibrium exchange rate. Without long-term debt, the
adjustment to a fiscal imbalance must occur immediately.
Yet the structure of our argument above suggests that the model may have an equilibrium
with et = 0 if we modify its specification in such a way that marginal utility of consumption at
the time of an exchange rate crisis is relatively low. Notice the first order condition (21) implies
that a declining path of consumption may become consistent with intertemporal optimality for a
relatively small value of (Gt+1 /Gt ). We turn to this case next.

25

6.4

Dynamics of the current account and the real exchange rate

This subsection reconsiders our experiment under the assumption that an exchange rate crisis at
t + 1 coincides with a cut in utility-providing government spending, i.e. gt+1 < 0. Specifically,
we set the change in spending gt+1 equal to a unique value such that all equilibrium conditions
hold with et = 0. This amendment to our model implies that the marginal utility of consumption
falls between the two periods even if aggregate consumption follows a declining path, avoiding the
contradiction at the hearth of the nonexistence result in the previous subsection.
With gt+1 < 0, there is a unique value of PN,t consistent with equilibrium. It is also easy
to show that, in general, there are a continuum of solutions for the interest rate in period t and
the price level at t + 1. In fact, changes in it aect the stock of outstanding public debt (by
the government budget constraint (23)) and therefore can cause self-fulfilling changes in et+1 and
pN,t+1 . While those self-fulfilling equilibria may be interesting per se, we prefer to rule them out
by restricting fiscal reform so that the size of spending cuts at the time of a crisis depends on
the pre-devaluation interest rate. Specifically, we add gt+1 to the vector of endogenous variables,
appending gt+1 = 0 + 1 it (with 1 < 0) to the list of equations in the second appendix.
A feedback from devaluation expectations to the size of the fiscal retrenchment eliminate the
possibility that the magnitude of the crisis be determined by self-fulfilling expectations.
Numerical examples of the unique equilibrium are plotted in figures 2 and 3, with parameters
and steady state values listed at the end of the second appendix. Figure 2 is drawn for the case
where foreign deflation is temporary, pT,t = 0.01, and figure 3 for the case of a permanent shock,
pT,t = pT,t+1 = 0.01. In each figure, thick lines correspond to = 0.1 whereas thin lines to
a relatively more open economy with = 0.5. The real exchange rate is defined as the price
of the traded good divided by the price of the nontraded good, so that an increase in the ratio
corresponds to real depreciation. Furthermore, endowments of income are assumed to be constant
so that an increase in consumption of the traded good signifies a movement toward a current

26

account deficit.
The simulations suggest conclusions concerning the dynamics of the real exchange rate and
the current account around the time of a currency crisis. Our model provides examples in which
a foreign deflationary shock ultimately leads to an abandonment of a fixed exchange rate regime,
but devaluation is delayed. As the price of the nontraded good temporarily increases in the runup
to the crisis, the country experiences real appreciation, a current account deficit and an increase in
consumption. A subsequent devaluation coincides with a real depreciation and a current account
surplus. Thus the simulations match the dynamics of the real exchange rate and the current
account suggested by the data, purely as part of the fiscal adjustment to a foreign shock and
not as a result of, say, credit market imperfections or changes in output.
To an observer unaware of the fiscal implications of foreign deflation, the real appreciation
during the first period may suggest that the country is experiencing competitiveness problems,
and that these problems are responsible for the current account deficit. The same observer may
then feel that he/she is proven right by the subsequent sharp correction in the nominal and real
exchange rate, accompanied by declining prices for nontradables and a surplus in the current
account. It is also worth noting that, by causing real depreciation at the time of the crisis, the
fiscal adjustment increases the costs of servicing debt denominated in tradables. These costs are
relatively lower in the runup to the crisis.
Consider what the simulations suggest about eects of openness. First, it is interesting that,
since the shock causing fiscal stress is a nominal deflation abroad, the adjustment can be consistent
with domestic price level deflation in the first period. This is because, while the price of the
nontraded good unambiguously increases, the price of tradables must fall. The latter eect tends
to prevail in economies that are relatively open, the result being price level deflation. It is also
notable that the adjustment can be consistent with price level deflation at the time of a jumpup in the exchange rate, and even when the external deflationary shock is temporary. Deflation
coincident with devaluation is more likely to occur in a relatively closed economy.
27

Second, one might approach the simulations having the notion that a relatively less open
economy is insulated from foreign deflationary shocks. This is true in the sense that, if the value
of (a measure of the extent of openness in the model) is low, a given disturbance causes a
smaller increase in the real value of public debt, than if is large. Suppose, however, that the
foreign shock is adverse enough to create a fiscal imbalance. Then the subsequent adjustment is
inconsistent with the idea of insulation in a striking way: lower values of require in equilibrium
larger changes in the exchange rate and the current account. The intuition is, in essence, that
the fiscal adjustment has to do with movements in the price level not in the exchange rate per
se, and the exchange rate must change a lot when is small to be consistent with a given change
in P . Similarly, larger changes in consumption of the traded good and in the current account are
required to match a given change in overall consumption.

Concluding remarks

This paper derives implications of the currency and maturity structure of public debt for a macroeconomic adjustment to a fiscal imbalance. Our analysis moves away from a partial view of this
adjustment, focused only on seigniorage and therefore having little to say about the role of the
original sin in the dynamics of currency instability. In contrast with most of the literature on
speculative attacks and currency collapses, we find that the composition of public liabilities is a
key determinant of the macroeconomic dynamics around a crisis, as well as of the magnitude of
exchange rate adjustment to a fiscal imbalance.
Many aspects of our analysis need additional research. For instance, we do not explicitly model
factors that constrain fiscal reform and monetary policy after an adverse shock. Political economy
and concerns about financial stability are promising areas of analysis. We likewise do not to
solve for the original sin endogenously, as an equilibrium outcome in the presence of policymakers
lacking credibility. Lastly, many devaluations coincide with both fiscal stress and sharp changes in
output. The focus of this paper is on the former, while some recent models emphasize the latter.26

28

Our contribution should therefore be regarded as a building block toward future models that will
simultaneously account for both aspects of actual crises.
2 6 Examples are: Aghion, Bacchetta, and Banerjee (2001), Chang and Velasco (2001), Gertler, Gilchrist, and
Natalucci (2001) and Schneider and Tornell (2000).

29

Appendix

We discuss the behavior of the model with the more general post-devaluation interest rate rule
(11) and with P = P . There isnt now a closed-form solution for the equilibrium exchange rate,
the diculty being that the perpetuity price in the period of devaluation no longer equals (1/ip )
as in the main text. Nevertheless, it is interesting to examine eects of a more general policy
specification.
Given the interest rate rule (11) with 1 < 1, equilibrium condition (3) becomes a stable
dierence equation in the (gross) depreciation rate upon the jump in the exchange rate in period
T t:
ET +s+1
ET +s
= 0 + 1
ET +s
ET +s1

(25)

where s 0. Once we know the exchange rate at the time of the crisis ET , (25) determines the
post-devaluation evolution of the exchange rate, and (3) determines the post-devaluation path of
nominal interest rates. Notice that the (gross) depreciation rate converges to:
0
1 1
which is, given our assumption 0 1 1 , greater than or equal to 1. 1 = 0 is a pure interest
rate peg analyzed in the main text. With 1 = 0 and 0 = 1 , the exchange rate jumps in
period T and remains at ET permanently the steady state depreciation rate is zero. For other
parameter values, the exchange rate jumps in period T and then the depreciation rate converges to
its limiting constant value. Greater values of policy parameters 0 and 1 imply that the nominal
interest rate and the depreciation rate are higher in the post-devaluation steady state. A larger
1 causes more persistent eects of the initial jump in E on interest and depreciation rates.
Using equilibrium condition (3) and policy rule (11), we derive:

s
X
ET
1 + iT +s = 0
j1 j + 1 (1 )s
E
j=0

where T t is the collapse time, and s 0. We note that the (gross) nominal interest rate
30

converges to 0 / (1 1 ) as s . We substitute the above equation into equilibrium condition

(5) to obtain an expression for T in terms of ET and policy parameters 0 and 1 :

s
k

X Y
X j j
ET
k
0
T =
1 + 1 (1 )

s=0 k=0
j=0

(26)

where T t. Since policy raises interest rates in response to devaluation and this tends to reduce

the long-term bond price, the above expression is a decreasing relationship between ET /E and

T . For a given ET /E , higher values of 0 and 1 imply a smaller T .

Suppose the exchange rate jumps in period T = t. Putting together (9) and (10) yields an

increasing equilibrium relationship between Et /E and t , implied by the solved-forward govern-

ment budget constraint:

[1 + (1/r)] Lt1 + (1 + r) Bt1
(1 + t ) Lt1 + (1 + r) Bt1
Et

=
E
E
E

The above equation and (26) with T = t uniquely determine Et /E . Higher values of 0 and 1

act like a higher ip in the main text, and are associated with a lower Et /E in equilibrium.

Consider now a delayed devaluation, i.e. suppose the exchange rate jumps in period T > t.

The unique solution for ET /E is found by combining (14) and the following version of (15):
1
t = +
r

1
1+r

T t1

1
1 + T

r
(1 + r) ET /E

with (26) substituted for T . Again, higher values of 0 and 1 act like a higher ip in the main

text and are associated with a lower ET /E in equilibrium.

31

Appendix

This appendix lists equations of the model with a nontraded good loglinearized about its steady
state. We use the standard notation that xt = log Xt log X t , where X t is a value of variable X in
the steady state in period t. When a lower-case letter was used for a variable in the nonlinear model
(for example ), we now interpret as a percentage deviation from its steady state value . We
consider an exogenous shock pT,t < 0 and assume that policy sets et = 0 and t = t+1 = gt = 0.
The model is:
ct pt = gt+1 + et+1 ct+1 pt+1

(27)

et+1 = it

(28)

ct = cT,t + (1 )cN,t

(29)

ct+1 = cT,t+1 + (1 )cN,t+1

(30)

pt = pT,t + (1 )pN,t

(31)

pt+1 = et+1 + pT,t+1 + (1 )pN,t+1

(32)

cT,t + pT,t = cN,t + pN,t

(33)

cT,t+1 + et+1 + pT,t+1 = cN,t+1 + pN,t+1

(34)

cN,t = 0

(35)

Gt+1 gt+1 = C N,t+1 cN,t+1

(36)

Rt P t t Gt pt = P t+1 t+1 Gt+1 pt+1 Rt B t et+1 Gt+1 P t+1 gt+1

(37)

P T,t Y T,t C T,t pT,t P T,t C T,t cT,t = P T,t+1 Y T,t+1 C T,t+1 pT,t+1 +P T,t+1 C T,t+1 cT,t+1

(38)

All numerical examples plotted in figures 2 and 3 assume the following values of parameters and
steady state variables: 0 = 0, 1 = 1, = 1, Rt = 1, Y T,t = Y T,t+1 = 1, Y N,t = Y N,t+1 = 1,
Gt = Gt+1 = 0.3, t = t+1 = 0.6, B t1 = 0.6, B t = 0.3, C T,t = C T,t+1 = 1, C N,t = C N,t+1 = 1,

E t = E t+1 = 1, P N,t = P N,t+1 = 1, P T,t = P T,t+1 = 1.

32

References
Aghion, Philippe, Philippe Bacchetta, and Abhijit Banerjee (2001): A Corporate BalanceSheet Approach to Currency Crises. Working paper, Harvard University, University of
Lausanne, and MIT.
Benhabib, Jess, Stephanie Schmitt-Groh and Martin Uribe (2001): Monetary Policy and
Multiple Equilibria. American Economic Review, 91(1), 167-86.
Bergin, Paul R. (2000): Fiscal Solvency and Price Level Determination in a Monetary
Union. Journal of Monetary Economics, 45(1), 37-53.
Buiter, Willem H. (1987): Borrowing to Defend the Exchange Rate and the Timing and
Magnitude of Speculative Attacks. Journal of International Economics, 23, 221-239.
Buiter Willem H., Giancarlo Corsetti and Paolo Pesenti (1998): Financial Markets and
European Monetary Coordination: The Lessons from the 1992-93 ERM Crisis. Cambridge
University Press.
Burnside, Craig, Martin Eichenbaum and Sergio Rebelo (2001a): Prospective Deficits and
the Asian Currency Crisis. Journal of Political Economy, 109(6), 1155-97.
Burnside, Craig, Martin Eichenbaum and Sergio Rebelo (2001b): On the Fiscal Implications
of Twin Crises. Working paper, The World Bank and Northwestern University.
Burstein, Ariel, Martin Eichenbaum and Sergio Rebelo (2002): Why Is Inflation So Low
After Large Devaluations. Working paper, University of Michigan and Northwestern University.
Calvo, Guillermo A. (1987): Balance of Payments Crises in a Cash-in-Advance Economy.
Journal of Money, Credit and Banking, 19, 19-32.
Calvo, Guillermo A. (1998): Varieties of Capital-Market Crises. In Guillermo A. Calvo
and Mervin King, editors: The Debt Burden and its Consequences for Monetary Policy.
International Economics Association Series, New York: St. Martins Press.
Calvo, Guillermo, Leonardo Leiderman, and Carmen M. Reinhart (1993): Capital Inflows
to Latin America: The Role of External Factors. IMF Sta Papers, 40(1), 108-151.
Canzoneri, Matthew B., Robert E. Cumby, and Behzad T. Diba (1998): Fiscal Discipline
and Exchange Rate Regimes. Center for Economic Policy Research, Discussion Paper 1899.
Cavallari, Lilia and Giancarlo Corsetti (2000): Shadow Rates and Multiple Equilibria in
the Theory of Currency Crises. Journal of International Economics, 51, 275-286.
Chang, Roberto and Andres Velasco (2001): A Model of Financial Crises in Emerging
Markets. Quarterly Journal of Economics, 116(2), 489-517.
Cochrane, John H. (1999a): A Cashless View of U.S. Inflation. In: Ben Bernanke and Julio
J. Rotemberg, eds.: NBER Macroeconomics Annual 1998. The MIT Press, Cambridge,
MA.
Cochrane, John H. (1999b): Money as Stock: Price Level Determination with no Money
Demand. Working paper, University of Chicago.
Cochrane, John H. (2001): Long-term Debt and Optimal Policy in the Fiscal Theory of
the Price Level. Econometrica, 69(1), 69-116.
33

Corsetti, Giancarlo and Bartosz Mackowiak (2002): Fiscal Imbalances and the Dynamics
of Currency Crises Working paper, Yale University.
Corsetti, Giancarlo, Paolo Pesenti, and Nouriel Roubini (1999a): What Caused the Asian
Currency and Financial Crisis? Japan and the World Economy, 3, 305-373.
Corsetti, Giancarlo, Paolo Pesenti, and Nouriel Roubini (1999b): Paper Tigers? A Model
of the Asian Crisis. European Economic Review, 43, 1211-1236.
Cushman, David O. and Tao Zha (1997): Identifying Monetary Policy in a Small Open
Economy Under Flexible Exchange Rates. Journal of Monetary Economics 39, 433-448.
Daniel, Betty C. (2001): A Fiscal Theory of Currency Crises. International Economic
Review, 42(4), 969-88.
Drazen, Allan and Vittorio Grilli (1993): The Benefit of Crises for Economic Reforms.
American Economic Review, 83, 598-607.
Dupor, Bill (2000): Exchange Rates and the Fiscal Theory of the Price Level. Journal of
Monetary Economics, 45, 613-630.
Flood, Robert P. and Peter M. Garber (1984): Collapsing Exchange Rate Regimes: Some
Linear Examples. Journal of International Economics, 17, 1-13.
Flood, Robert P. and Nancy Marion (1999): Perspectives on the Recent Currency Crisis
Literature. International Journal of Finance and Economics, 4(1), 1-26.
Gertler, Mark, Simon Gilchrist, and Fabio Massimo Natalucci (2001): External Constraints
on Monetary Policy and the Financial Accelerator. Working paper, New York University
and Boston University.
Hausmann, Ricardo and Andrs Velasco (2002): The Argentine Collapse: Hard Moneys
Soft Underbelly. Working paper, Kennedy School of Government, Harvard University.
Jeanne, Olivier (2000): Currency Crises: A Perspective on Recent Theoretical Developments. Special Paper in International Economics no. 20, International Finance Section,
Princeton University.
Krugman, Paul R. (1979): A Model of Balance of Payments Crises. Journal of Money,
Credit and Banking, 11, 311-325.
Lahiri, Amartya, and Carlos A. Vegh (1998): The Feasibility of BOP Crises: Monetary vs.
Fiscal Approach. Working paper, UCLA.
Lahiri, Amartya, and Carlos A. Vegh (2000): Delaying the Inevitable: Optimal Interest
Rate Policy and BOP Crises. Working paper, UCLA.
Leeper, Eric (1991): Equilibria Under Active and Passive Monetary and Fiscal Policies.
Journal of Monetary Economics, 27, 129-147.
Levy, Aviram (2002): Asset and Liability Dollarization in Emerging Economies: Empirical
Evidence and Policy Issues. Working paper, Bank of Italy.
Loyo, Eduardo (1998): Going International with the Fiscal Theory of the Price Level.
Working paper, Harvard University.
Mackowiak, Bartosz (2001): Some VAR Evidence on the Eects of the World Economy on
Emerging Markets. Working paper, Yale University and Princeton University.
34

Mackowiak, Bartosz (2002): Macroeconomic Regime Switches and Speculative Attacks.

Working paper, Humboldt University.
Milesi-Ferretti, Gian Maria and Assaf Razin (1998): Current Account Reversals and Currency Crises: Empirical Regularities. Working paper 89, International Monetary Fund.
Obstfeld, Maurice (1986): Speculative Attack and the External Constraint in a Maximizing
Model of the Balance of Payments. Canadian Journal of Economics, 19, 1-22.
Persson, Mats, Torsten Persson, and Lars E. O. Svensson (1998): Debt, Cash Flow and
Inflation Incentives: A Swedish Example. In Guillermo A. Calvo and Mervin King, editors:
The Debt Burden and its Consequences for Monetary Policy. International Economics
Association Series, New York: St. Martins Press.
Salant, Stephen W., and Dale W. Henderson (1978): Market Anticipation of Government
Policy and the Price of Gold. Journal of Political Economy, 86, 627-48.
Schneider, Martin, and Aaron Tornell (2000): Balance Sheet Eects, Bailout Guarantees
and Financial Crises. NBER working paper 8060.
Sims, Christopher A. (1994): A Simple Model for Study of the Determination of the Price
Level and the Interaction of Monetary and Fiscal Policy. Economic Theory, 4, 381-399.
Sims, Christopher A. (1997): Fiscal Foundations of Price Stability in Open Economies.
Working paper, Yale University.
Sims, Christopher A. (1999): The Precarious Fiscal Foundations of EMU. De Economist,
147(4), 415-36.
Uribe, Martin (2002): The Price-Consumption Puzzle of Currency Pegs. Journal of Monetary Economics, 49, 533-569.
Woodford, Michael (1994): Monetary Policy and Price Level Determinacy in a Cash-inAdvance Economy. Economic Theory, 4, 345-380.
Woodford, Michael (1995): Price-Level Determinacy Without Control of a Monetary Aggregate. Carnegie-Rochester Conference Series on Public Policy, 43, 1-46.
Woodford, Michael (1996): Control of the Public Debt: A Requirement for Price Stability?
National Bureau of Economic Research, Working Paper 5684. Reprinted in: Guillermo Calvo
and Mervyn King, eds. (1998): The Debt Burden and its Consequences for Monetary
Policy. International Economics Association Series, New York: St. Martins Press.
Woodford, Michael (1998): Public Debt and the Price Level. Working paper, Princeton
University.
Woodford, Michael (2000): Fiscal Requirements for Price Stability. Working paper, Princeton University.
Zarazaga, Carlos (1997): Recurrent Hyperinflations in a Dynamic Game with Imperfect
Monitoring in the Appropriation of Seigniorage. Working paper, Federal Reserve Bank of
Dallas.

35

Figure 1: Numerical examples of the singlegood model

Devaluation with foreign deflation

Devaluation with foreign deflation

50

90

= 0.2

Dollar debt as a fraction of total debt = = 0.9

45

80
70
Devaluation rate, in percent

Devaluation rate, in percent

40
35
=0
30
= 0.1

25
20
15

50
= 0.7
40
= 0.5

30
20

10

10

5
0

60

5
10
15
20
Foreign deflation, measured in foreign currency, in percent

25

10

15

20

25

Foreign deflation, measured in foreign currency, in percent

Delayed devaluation with = 0.02 and dollar debt 0.7 of total debt

Devaluation rate with 10% foreign deflation

55

180

The shadow devaluation rate, in percent

45
40
35
30
25
20

= 0.75

140
120
100
80
60
40
20
p

=0

15

Longterm debt as a fraction of nominal debt = = 0.5

160

= 0.1

50

Devaluation rate, in percent

i =r
0
p

i = r + 1%
10

0.1

0.2

0.3
0.4
0.5
0.6
Dollar debt as a fraction of total debt

0.7

0.8

0.9

20

10

20

30

40
Time (Tt)

50

60

70

80

Figure 2: Numerical examples of the twogood model (thin lines: = 0.5, thick lines: = 0.1)

percent deviation from steady state

Price of tradables (in foreign currency)

Exchange rate

Price level

1.5

1.5

1.5

1.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

1.5

1.5

1.5

1.5

Real exchange rate

percent deviation from steady state

Price of nontradables

Consumption of tradables

Consumption of nontradables

1.5

1.5

1.5

1.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

1.5

1.5

1.5

1.5

1
time

1
time

1
time

Consumption

1
time

Figure 3: Numerical examples of the twogood model (thin lines: = 0.5, thick lines: = 0.1)
Price of tradables (in foreign currency)

Exchange rate

Price of nontradables

3
2.5

1.5

Price level

1.5

1.5

0.5

0.5

0.5

0.5

1.5

1.5

percent deviation from steady state

2
1
1.5
0.5

0.5
0

0.5

0.5
1
1
1.5

1.5

percent deviation from steady state

Real exchange rate

Consumption of tradables

Consumption of nontradables

1.5

1.5

1.5

1.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

1.5

1.5

1.5

1.5

1
time

1
time

1
time

Consumption

1
time