In: Economics
Write a long essay discussing and critically evaluating the literature on the first-generation models of currency crises. Give some examples of actual currency crises explained by these models.
The currency crises and sovereign debt crises that have occurred
with increasing frequency since the Latin American debt crisis of
the 1980s have inspired a huge amount of research. There have been
several 'generations' of models of currency crises he recent crises
in Mexico and in several Asian countries have left many
people
wondering what we really know about currency crises, what provokes
them and whether
they can be predicted or not. In fact the sheer occurrence of these
events, as well as the
magnitude of their sequels in other countries, has cast doubts on
the ability of economists,
traders and international institutions to explain currency crises,
let alone to forecast them.
Some have gone beyond, stating that we really understand very
little about why and when
currency crises occur. We believe that this idea is incorrect and
that it provides an
erroneous account of our current understanding of what causes a
currency crisis.
This paper develops a formal empirical analysis on the determinants
of currency
crises using a panel dataset with annual information for 30
countries during the period
1975-96. Explanatory variables are defined in close correspondence
with the factors
highlighted in the theoretical literature, so that we may evaluate
the current status of our
understanding of exchange rate crises.
In fact, the paper tests the main predictions of two
generations of models of currency crises and evaluates the
explanatory power of some of
the key variables proposed in the literature.
Our paper departs from previous empirical studies on this subject
in at least four
important ways. First, our sample is larger, richer and more
diverse than that of most
previous studies. Early studies of currency crises tended to focus
on individual countries,
while we exploit the higher variability associated to a
multi-country study. Compared to
other cross-country studies, this paper has the advantage that it
uses a more diverse group
of economies. Unlike studies that have focused exclusively on
either developed
or
developing countries,
for example, our paper uses a diverse group of 15 high-income
and 15 middle-income countries. Additionally, the number of
countries in our sample is the second largest among these
studies.
These sample characteristics make our results
more robust and reduce the risk of overemphasizing the generality
of the findings. Also,
by focusing on a sufficiently large period and number of countries
that includes
numerous currency crises, our paper avoids the risks of obtaining
results and deriving
conclusions from a single crisis episodeSecond, we use a definition
of currency crisis that focuses only on events that
lead to a collapse of the exchange rate system (after all, that is
why they are presumably
called currency crises). Thus, unlike other previous work,
we exclude episodes of
unsuccessful speculative attacks from our definition of crisis. As
discussed below, these
situations are especially hard to define and the standard way to
measure them has
undesirable predictive properties.Third, our use of a probit model
with random effects takes into account the cross-
section and time-series characteristics of our data. This method of
estimation exploits
optimally the within-units correlation that results from observing
the same countries
repeatedly over time. To the best of our knowledge, this is the
first study on currency
crises that exploits this feature of the data.Fourth, our empirical
approach differs from previous studies in that it represents the
first
attempt to simultaneously test the main predictions of both first
and second generation models of
currency crises. Thus, this paper attempts to provide an assessment
of the relative importance of
the variables emphasized in each type of models.
The paper is organized as follows. Section 2 reviews the
theoretical literature on currency
crises, while section 3 discusses the empirical literature on the
determinants of currency crises.
Section 4 presents the definition of crisis used in this paper and
describes our method of
estimation. The next section discusses the data and the explanatory
variables. Section 6
presents our main empirical results followed by a preliminary
evaluation of the predictive
power of the model in section 7. A conclusion closes the paper.
First-generation models
The first formal model of balance-of-payments crises was put
forward by Krugman
(1979), based on the work of Salant and Henderson (1978). Krugman
argued that crises
occur when a continuous deterioration in the economic fundamentals
becomes
inconsistent with an attempt to fix the exchange rate. The original
source of problems in
Krugman’s model is the excessive creation of domestic credit to
either finance fiscal
deficits or to provide assistance to a weak banking system. More
specifically, the model
assumes that the government has no access to capital markets,
thereby forcing it to
monetize its expenditures. In this context, an interest rate parity
condition would induce
capital outflows and a gradual loss of foreign exchange reserves.
Further down the road,
the economy eventually becomes the victim of a speculative attack
on its foreign
exchange reserves, which triggers the collapse of the fixed
exchange rate system. The
timing of the attack in Krugman’s model is determined by a critical
level in the amount of
reserves. Once reserves reach such threshold level, speculators are
induced to exhaust the
remaining reserves in a short period of time to avoid capital
losses.
Krugman’s work was later extended and simplified by several
authors. Flood and
Garber (1984) constructed a simplified linear model, introducing a
stochastic component.
Connolly and Taylor (1984) analyzed a crawling-peg regime and
stressed the behavior of
the relative prices of traded goods preceding the collapse of the
exchange rate regime. In
their analysis, the real exchange rate appreciates and the current
account deteriorates
prior to the collapse. In related contexts, Edwards (1989) has also
stressed the patterns of
currency overvaluation and current account deterioration that tend
to precede
devaluations, while Calvo (1987) has discussed overvaluation in a
cash-in-advance
model.
Krugman’s model has been further extended to the case of
speculative attacks in
target zones by Krugman and Rotemberg (1991). More recently, and
inspired by the
Mexican crisis of 1994, Flood, Garber and Kramer (1996) have
incorporated the role of
an active sterilization policy into the analysis.
The Krugman model and its extensions represent what has become
known as
first-generation models of balance-of-payments crises.
Their main insight is that crises
arise as a result of an inconsistency between an excessive public
sector deficit that
becomes monetized and the exchange rate system. In this sense, a
crisis is both
unavoidable and predictable in an economy with a constant
deterioration of its
fundamentals.
r we have examined the determinants of currency crises in a
broad sample of
countries between 1975 and 1996. We found that a relatively small
set of macroeconomic
variables play an important role in the empirical determination of
crises. More
specifically, we found that high rates of seignorage, current
account deficits, real
exchange rate misalignments, low foreign exchange reserves relative
to a broad measure
of money, negative terms of trade shocks, negative per capita
income growth, and a
contagion effect, all help to explain the presence of currency
crises in our sample. Our
results are robust to changes in the specification, definition of
variables, method of
estimation and country sample.
Interestingly, most of these variables had already been identified
and suggested
by the theoretical literature. In this sense, our results should be
interpreted as supporting
both first and second-generation models of currency crises. More
important, however, is
our finding that the insights developed by second-generation models
complement rather
than substitute the explanation provided by first-generation
models. Therefore, while
neither of the two types of models tells the whole story about
currency crises, they both
add to our understanding of this phenomenon. Moreover, our results
suggest that there
may be instances in which a country falls into a crisis by factors
purely associated to first-
generation models, while in other situations the relevant factors
might be those stressed
by second-generation models. These results explain why currency
crises are not all alike,
but they also explain the existence of common patterns across
various crises episodes.
Our results also indicate that currency crises have a significant
predictable
component. Using our empirical estimates, we are able to predict
correctly a majority of
the currency crises in our sample. This suggests that at least some
currency crises could
have been prevented with sounder macroeconomic policies, and that
certain policies
could have been implemented earlier to reduce the risk of a crisis.
In this sense, we
interpret our results as supporting the view that an early-warning
system could help
reduce the number of currency crises in both emerging and
industrialized economies.