WPS 2-2 7
POLICY RESEARCH WORKING PAPER   2267
Do  High  Interest Rates                               No - there is no systematic
association between interest
Defend  Currencies  during                             rates and the outcome of
Speculative  Attacks?                                  speculative attacks.
Aart Kraay
The World Bank
Development Research Group
Macroeconomics and Growth                                     U
January 2000                                                      a



t PQmI.CY RESEARCH WORKING PAPER 2267
Summary findings
Drawing on evidence from a large sample of speculative            The lack of clear empirical evidence on the effects of
attacks in industrial and developing countries, Kraay           high interest rates during speculative attacks mirrors the
argues that high interest rates do not defend currencies        theoretical anmbiguities on this issue.
against speculative attacks. In fact, there is a striking lack
of any systematic association between interest rates and
the outcome of speculative attacks.
This paper - a product of Macroeconomics and Growth, Develcpment Research Group - is part of a larger effort in the
group to study the causes and consequences of financial crises. Copies of the paper are available free from the World Bank,
1818 H Street,N'W, Washington, DC 20433. Please contact Rina Bonfield, room MC3-354, telephone 202-473-1248, fax
202-522-3518, email address abonfield@worldbank.org. Policy Research Working Papers are also posted on the Web a!t
www.worldbank.org/research/workingpapers. The author may be contacted at akraay@<iworldbank.org. January 200(0.
(45 pages)
The Policy Research Working Paper Series dissemninates the findings of work in progress to encourage the exchange of ideas about |
development issues. An objective of the series is to get the findings out quickiy, even if the presentations ar-e less than f7lly polished. Th2e
papers carry the names of the authors and should be cited accordingly. The findings, interpretations, aisd conclusions expressed in this
paper are entirely those of the authors. They do uot necessa-ily represent the view of the World Bank, its Executive Directors, or the
countries they represent.
Produced by the Policy Research Dissemination Center



Do High Interest Rates Defend Currencies
During Speculative Attacks?
Aart Kraay
The World Bank
1818 H Street, N.W. Washington, DC 20433, (202) 473-5756, akraay@worldbank.org.
The opinions expressed in this paper are the author's, and do not reflect those of the
World Bank, its executive directors, or the countries they represent. I would like to
thank Alan Drazen, Ilan Goldfajn, Patrick Honohan, Vickie Kraay, Maria Soledad
Martinez Peria, Sergio Schmukler, Jakob Svensson, Jaume Ventura and seminar
participants at MIT, the Tinbergen Institute, and the World Bank for helpful comments.






1. Introduction
According to conventional wisdom, currencies that come under speculative
attack can be defended with high interest rates. By raising interest rates high enough,
the conventional wisdom argues that the monetary authority can make it prohibitively
costly for speculators to take short positions in the currency under attack. High interest
rates are often also said to convey a positive signal regarding the commitment of the
monetary authority to maintaining a fixed exchange rate. To the extent that this signal
alters the expectations of foreign exchange market participants, high interest rates can
serve to strengthen the domestic currency. A classic example in support of the
conventional wisdom is the response to the attack on the Swedish krona in the summer
of 1992, shown in the top panel of Figure 1. Between July and August, speculative
pressures against the krona resulted in a loss of nearly one-quarter of the reserves of
the Swedish central bank. To stem the outflow, the central bank's marginal lending rate
was raised to an incredible 500 percent on September 17 and 18, and hovered in the
vicinity of 50 percent for the next week. Reserve losses were promptly halted, and the
krona's peg was maintained.
Recently, a contrarian view of the effects of high interest rates during speculative
attacks has emerged, which calls into question both tenets of the conventional wisdom.
First, it notes that interest rates have to be increased to very high annualized rates in
order to entice even risk-neutral investors to hold local currency-denominated assets in
the face of a small expected devaluation over a short horizon, and such extremely high
interest rates are rarely observed in practice.' Second, this view notes that the signaling
value of high interest rates is unclear. Although signals must be costly in order to be
credible, often they impose costs that are too high for the monetary authority to take in
stride. If market participants know that the monetary authority is concerned about the
contractionary effects of high interest rates on domestic economic activity, they are
unlikely to believe that rates will be kept high enough, and for long enough, to deter
speculation. Worse, as the costs of high interest rates mount, the monetary authority's
signal can become less credible over time, raising devaluation expectations. A vicious
1For example, a risk-neutral investor expecting only a 0.5 percent overnight depreciation would requiie an
overnight annualized rate of return of 500 percent on domestic currency to compensate for the expected
devaluation.
1



spiral can result, as expectations of a devaluation force higher interest rates, which in
turn impose greater costs on the economy.2 An example consistent with this contrarian
view of the effects of high interest rates is Korea in the second half of 1997, shown in
the lower panel of Figure 1. As the East Asian financial crisis spread from Thailand and
Malaysia, speculative pressures against the Korean won intensified and the reserves of
the Korean central bank fell from 35 billion to 25 billion US dollars between June and
November. Although the overnight call rate was raised from around 12 percent in early
November to over 30 percent by the end of December, the won fell by over 50 percent
during this period.
In light of these theoretical ambiguities and conflicting anecdotes, this paper
asks whether there is any systematic empirical evidence in support of the conventional
wisdom regarding the effects of high interest rates during speculative attacks. To
answer this question, I study the behaviour of interest rates around a large number of
successful speculative attacks (i.e. attacks that end in a sharp nominal devaluations)
and failed speculative attacks (i.e. attacks that did not end in a devaluation) in a sample
of 75 developed and developing countries over the period 1960-1997. I examine
whether interest rates rise during failed speculaltive attacks (i.e. whether raising interest
rates is necessary to prevent a speculative attack from ending in a devaluation), and
whether raising interest rates increases the probability that an attack fails (i.e. whether
raising interest rates is sufficient to prevent a speculative attack from ending in a
devaluation).
This empirical exercise faces three difficulties: measuring the policy response to
a speculative attack, accounting for possible non-linearities in the effects of the policy
response, and controlling for the endogeneity of the policy response. First, it is difficult
to disentangle the monetary policy response to a given speculative attack from other
sources of variation in observed market interest rates during the attack. For example,
increases in market interest rates during a speculative attack might reflect both a
tightening of domestic credit by the monetary aLuthority, and also an increase in the
2Drazen and Masson (1994) develop a model in which signals become less credible over time. Bensaid
and Jeanne (1997) formalize devaluation spirals. Radelet and Sachs (1998) and Furman and Stiglitz
(1998) discuss other reasons why tighter monetary policy can weaken, rather than strengthen, the currency
under attack.
2



devaluation expected by market participants. In order to obtain a direct measure of the
monetary policy response to speculative pressures, I rely primarily on changes in
interest rates under the control of the monetary authority (i.e. central bank discount
rates) as a measure of policy. A drawback of this measure is that discount rates are
only one of many instruments that the monetary authorities have at their disposal to
resist speculative pressures. I therefore also check the robustness of the results using
a variety of other noisier indicators of the stance of monetary policy.
Second, there may be important non-linearities in the effects of interest rates on
speculative pressures, and ultimately on the outcome of the attack. For example, the
credibility of the monetary authority's signal of its intent to defend the currency may
depend on the economy's ability to withstand the contractionary effects of tight monetary
policy, or on the quantity of reserves held by the monetary authority. In this case,
simple correlations between measures of monetary policy and the outcome of
speculative attacks may obscure any effects of policy present only in certain
subsamples of speculative attacks. I take into account the possibility of episode-specific
variation in the effects of monetary policy by splitting the sample along various
dimensions, and by interacting measures of monetary policy with episode-specific
characteristics.3
Third and perhaps most important, the policy decisions of the monetary authority
are themselves endogenous, and are likely to depend on both episode-specific
characteristics that determine speculative pressures, and on speculative pressures
themselves. Consider an economy that is vulnerable to a speculative attack, perhaps
because its real exchange rate is overvalued or its reserves are low relative to its short-
term obligations. If attacks on vulnerable currencies are both more likely to succeed,
and also are more likely to provoke a strong interest rate defense on the part of a
"tough" monetary authority committed to maintaining the fixed exchange rate, one might
expect to find large increases in interest rates during successful attacks, and
conversely, small increases in interest rates during failed attacks. This endogeneity
3A second possible source of non-linearities is in the time dimension, if, for example, the signaling value of
tight monetary policy becomes less credible over time. Since I will be relying on the relatively low-frequency
monthly data available for this large sample of speculative attacks, there is unlikely to be enough time
series variation in each episode to identify non-linearities over time in the effects of monetary policy.
3



problem may obscure the positive effects of high interest rates on investor confidence
and the probability that an attack fails. It is also possible that the endogeneity bias
exaggerates, rather than obscures, the conventional wisdom regarding effects of high
interest rates. For example, if the monetary authority is "realistic" and determines that it
is futile to try to defend a highly overvalued currency, but is willing to vigourously defend
the currency when it believes fundamentals are sound, there may be a positive
association between high interest rates and failed attacks driven by common
fundamentals. In this paper, I present a simple model which formalizes this endogeneity
problem, and motivates possible instruments for the monetary policy response. I then
use these to control for the endogeneity of policy in a probit specification which
expresses the probability that a speculative attack fails as a non-linear function of policy,
episode-specific characteristics, and interactions between the two.
The empirical results are not very supportive of the conventional wisdom that
high interest rates defend currencies during speculative attacks. I find no evidence that
interest rates systematically increase during failed speculative attacks, nor that raising
interest rates increases the probability that a speculative attack fails. I obtain the same
results if I consider alternative measures of monetary policy, as well as possible non-
linearities in the effects of monetary policy due to differences in a variety of episode-
specific characteristics. The lack of evidence orn the efficacy of monetary policy during
speculative attacks persists even after I control for possible biases induced by the
endogeneity of policy. Although there appears to be little evidence in support of the
conventional wisdom, there is also little evidence in support of the contrarian view that
raising interest rates weakens currencies under speculative attack. In fact, the main
finding of this paper is the striking lack of any association whatsoever between changes
in various measures of monetary policy and the outcome of speculative attacks.
This evidence contributes to a small but growing empirical literature on the role
of monetary policy during speculative attacks.4 Goldfajn and Gupta (1999) focus on the
There is of course a large literature on the effectiveness of interventions in foreign exchange markets (see
Edison (1993) for a survey). Various authors have also applied VAR methodologies to estimate the effects
of monetary policy shocks on exchange rates. These papers, which focus on normal times as opposed to
the periods of speculative pressures considered in this paper, find mixed results. Eichenbaum and Evans
(1995) and Cushman and Zha (1997) find that positive innovations to monetary policy lead to depreciations
of the domestic currency for the US and for Canada, respectively. In contrast, Sims (1992) and Grilli and
Roubini (1995) find mixed evidence in the G5 and G7 economies, respectively, with positive monetary
4



role of interest rates in the aftermath of large devaluations that result in an
undershooting of the real exchange rate. They ask whether high interest rates following
a devaluation increase the likelihood that real exchange rate equilibrium is restored
through a nominal appreciation rather than through higher inflation. They find that high
interest rates are effective in this sense only in countries with strong banking sectors.
Furman and Stiglitz (1998) examine daily data on interest rates and exchange rates in a
sample of nine developing countries during the 1 990s to identify episodes of sustained
high interest rates, and then ask whether these were followed by an appreciation of the
domestic currency. They find little evidence that this is the case. In contrast Dekle,
Hsiao and Wang (1 999a,b) study the relationship between interest rates and exchange
rates using weekly data for Korea, Malaysia and Thailand during 1997 and 1998, and
argue in favour of the conventional view. The main difficulty with all of these papers is
that they simply document reduced-form (partial) correlations between interest rates and
exchange rates. Without controlling for the endogeneity of the monetary policy
response, it is difficult to infer anything regarding the effects of high interest rates from
these papers. This paper makes a first attempt to take seriously the identification
problem, drawing on a much larger sample of successful and failed speculative attacks.5
The remainder of this paper proceeds as follows. In Section 2, I describe the
data and the methodology used to identify successful and failed speculative attacks. In
Section 3, I present some descriptive results, which provide scant evidence of any
association between changes in interest rates and the outcome of speculative attacks.
In Section 4, I develop a simple model to illustrate the endogeneity problem, and I use
this to motivate a set of probit regressions expressing the probability that speculative
attacks fail as a non-linear function of policies and fundamentals. After instrumenting
for the endogeneity of policy, I again find no evidence of a significant impact of high
interest rates on the outcome of speculative attacks. Section 5 offers concluding
remarks.
shocks leading to appreciations in some countries and depreciations in others. Finally there is a large
empirical literature documenting the properties of macroeconomic variables around speculative attacks
(e.g. example Eichengreen, Rose and Wyplosz (1994,1995,1996)), which to date has not focused on the
policy and non-policy determinants of successful and failed attacks.
This concern with the endogeneity of monetary policy is of course not new, and is a recurring theme in the
literature on the effects of monetary policy during normal times (as opposed to periods of speculative
pressures). See for example the discussion in Bemanke and Mihov (1998) and Christiano, Eichenbaum
and Evans (1998).
5



2. Identifying Speculative Attacks
I identify successful speculative attacks as large nominal depreciations preceded
by relatively fixed nominal exchange rates. I begin with an unbalanced panel of monthly
observations on nominal exchange rates (expressed in local currency units per US
dollar) and non-gold reserves. The sample consists of 75 middle- and high-income
countries with populations greater than I million, over the period January 1960 to April
1999. Details of the data can be found in the Appendix. I first ideritify all episodes in
which the one-month depreciation rate (i.e. the increase in the nominal exchange rate)
exceeds 10%, which is roughly two standard deviations above the mean monthly
depreciation rate for the entire pooled sample of monthly observations. In order for
these large depreciations to be meaningfully considered successful speculative attacks,
it is necessary that the exchange rate be relatively fixed prior to the depreciation itself.6
Accordingly, for each observation I construct an average over the previous twelve
months of the absolute value of percentage chainges in the nominal exchange rate. I
then eliminate all large depreciation episodes fcr which this average exceeded 2.5%, or
about one half of one standard deviation from the mean for the entire sample. In order
to avoid double-counting prolonged crises in which the nominal exchange rate
depreciates sharply for several months, I further eliminate successful attacks that were
preceded by successful attacks in any of the prior twelve months. Finally, I discard all
speculative attack episodes for which there is no data available on any of the variables I
will use to measure the monetary policy response to the attack. This results in 105
usable successful speculative attack episodes.
I identify failed speculative attacks using two indicators of speculative pressures:
sharp reserve losses, and sharp increases in nominal market interest rates. Specifically,
I consider all episodes in which the monthly decline in non-gold reserves measured in
US dollars (the increase in the nonminal money niarket rate spread over the US Federal
Funds rate) exceeds 20% (exceeds 5%), which is about two standard deviations above
the mean change for the entire sample. En order to restrict attention to speculative
6I only require the exchange rate to be 'relatively" fixed prior to the devaluation for two reasons. First, this
enables me to identify the abandonment of narrow target zone exchange rates regimes as well as of fully
fixed exchange rate regimes. Second, this allows me to identify currencies that are pegged against
currencies other than the US dollar whose value relative to the US dollar does not fluctuate much (e.g. the
German mark).
6



pressures against relatively fixed exchange rates, I eliminate all those episodes for
which the same moving average of absolute values of changes in the nominal exchange
rate as before was greater than 2.5%. Next, to avoid double-counting successful
attacks, I exclude all episodes in which the change in the nominal exchange rate in the
same month or any of the three following months was greater than 10%. I define these
episodes as failed speculative attacks and, as before, I eliminate all failed attacks that
are preceded by a failed attack in any of the twelve previous months, and those
episodes for which indicators of the monetary policy response are not available. This
results in 203 instances of failed speculative attacks.
Relying on reserve losses and increases in market interest rate spreads to
identify failed speculative attacks is problematic, because these indicators potentially
confound speculative pressures and the policy response to these pressures. For
example published data on reserve losses does not permit me to distinguish between
transactions of the monetary authorities to accommodate the increased speculative
demand for their reserves, and direct sales of reserves by the monetary authority in
order to support the currency. Similarly, increases in observed nominal interest rate
differentials may reflect both increases in market participants' devaluation expectations
as well as policy interventions in the money market, as noted in the introduction.
However, to the extent that these considerations are important, the results will be biased
towards finding that tightening monetary policy makes speculative attacks more likely to
fail, simply because the definition of failed attacks in part reflects the presence of tight
monetary policy. It is interesting to note that despite this obvious source of bias in
favour of the conventional wisdom, I find little evidence of this view.
Table 1 lists the full sample of 308 episodes, sorted by success and failure, and
by year. The sample includes a number of familiar episodes, as documented in Table 2.
The recent spate of currency crises in East Asia in 1997 are all represented as
successful speculative attacks, with the exception of Malaysia where the largest monthly
depreciation of the ringgit in August 1997 (6.6 percent) was not large enough to qualify
as a successful attack according to my definition. Table 2 also lists several speculative
attacks associated with the turmoil in the ERM in 1992, and compares the dating of
these attacks with that of Eichengreen, Rose and Wyplosz (1994). My criterion
identifies several of well-known failed attacks during this period, including those on the
7



Danish kroner, French franc, Irish punt and the Spanish peseta, as well as the
successful attacks on the British pound, the Swedish krona and the Finnish markkaa in
the fall of 1992.7 In most cases, the dating of events corresponds fairly closely to thalt of
Eichengreen, Rose and Wyplosz (1994). The only large discrepancy is in the case of
France, where they identify a speculative attack in September 1992 which I do not. T his
is because France's reserve losses of 8 percent in that month are not large enough
according to my definition, while its much larger reserve losses in the fall of 1993 are.
Table 1 nevertheless also includes a number of more questionable episodes.
Upon closer inspection, many of the episodes occur in country-period observations
characterized by underdeveloped financial markets and/or restrictions on capital
movements of various sorts. It is unlikely that the dynamics of speculative attacks in
such distorted environments will be comparable with those occurring in countries with
relatively developed financial markets and free capital mobility. In order to ensure that
the results are not tainted by these questionable episodes, I also define a subsample of
events where domestic credit to the private seclor as a share of GDP (a summary
indicator of financial development) averages more than 20% in the five years prior to the
attack, and the black market premium on foreign exchange (a summary indicator of de
facto currency convertibility) averages less than 10% in the five years prior to the attack.
I refer to this subsample as the financially-developed subsample, and the countries
included in it are indicated with asterisks in Table 1.
In Figure 2, I provide a graphical overview of successful and failed speculative
attacks, plotting the evolution of the nominal exchange rate and reserves during "typical"
attacks. To construct this figure, I compute the median growth rate in the exchange rate
and reserves over all successful and failed attacks for every month in a two-year window
centered on the date of the crisis, and then cumulate these median growth rates on a
base of 100 one year prior to the crisis. By construction, successful attacks are marked
by sharp nominal depreciations preceded by 12 months of very stable exchange rates.
In fact, the median change in the nominal exchange rate prior to these episodes is zero.
Reserves decline steadily over the entire period leading up to the collapse of the
The initial attack on the Swedish krona in the surnmer of 1992 mentioned in the introduction does not
qualify as an unsuccessful attack according to my definition since it was followed by a depreciation within
three months.
8



exchange rate, indicating that speculative pressures emerge in advance of the collapse
in the exchange rate itself, and reserves recover fairly quickly afterwards. Failed attacks
are also by construction preceded by very stable nominal exchange rates, and feature
sharp reserve losses in the month of the attack. As with successful attacks, reserves
recover quickly following failed attacks.
The main question of interest is whether raising interest rates -- or more
generally, tightening monetary policy -- prevents speculative attacks from ending in a
devaluation of the currency. To address this question, I require measures of the stance
of monetary policy around the speculative attack episodes identified above. I primarily
rely on the real central bank discount rate (the nominal discount rate deflated by
contemporaneous annualized monthly inflation) as a measure of the policy instrument
most directly under the control of the monetary authority.8 To the extent that the
monetary authority uses this instrument during a given episode, this variable provides a
good measure of the policy response to the speculative attack. However, as noted in
the introduction, the monetary authorities in these many speculative attack episodes
have a wide variety of instruments at their disposal. Attempting to identify the mix of
instruments actually employed during each of the 308 episodes in the sample, and
hence the appropriate episode-specific measure of the stance of monetary policy, would
be ambitious to say the least.9 1 instead use two other measures as crude "outcome"
indicators of the stance of monetary policy to check the robustness of the results: real
domestic credit growth, and the reserves of deposit money banks held in the central
bank. To the extent that the monetary authorities tighten monetary policy using other
measures (e.g. open market operations, raising reserve requirements, etc.), this will be
reflected in a reduction in real domestic credit and/or increases in bank reserves.'°
8 An unfortunate drawback of this measure is that central bank discount rates are reported by the IMF on an
end-of-period basis only, so that intra-monthly fluctuations in this variable are ignored. Also, there is of
course considerable debate over how to proxy for expected inflation when constructing real interest rates.
The results presented here do not change substantially if I deflate using either past or future inflation rates,
or if I simply consider changes in nominal discount rates.
Even for the United States which has been the subject of decades of intensive research, there is no clear
consensus on how precisely to measure the stance of monetary policy. See for example the discussion in
Bemanke and Mihov (1998). See also Borio (1997) for a description of the bewildering array of instruments
available in a set of developed countries.
10An obvious objection to the domestic credit growth measure is that it does not distinguish between shifts
in the supply and shifts in demand for domestic credit. To alleviate this concern I have also defined tight
9



For each speculative attack episode, it is necessary to determine whether these
measures of monetary policy tightened or not, relative to a suitable benchmark. For
failed attacks, I consider the increase in real discount rates, the decrease in real
domestic credit growth, and the increase in bank reserves in the month of the attack
relative to the month prior to the attack. That is, I ask whether tightening monetary
policy in response to a sudden reserve outflow serves to arrest further reserve losses
and maintain the value of the currency. Given that speculative pressures appear in
advance of the actual devaluation during successful attacks, I consider the change in
each measure of monetary policy in the month prior to the aftack relative to the previous
month. That is, I ask whether tightening monetary policy in response to mounting
speculative pressures serves to prevent attacks from succeeding. I do not include the
month of the attack itself, so as not to capture any post-devaluation policy responses
which may be quite different from those undertaken in defense of the currency prior to
the devaluation.11
(loose) monetary policy as periods where both domestic credit growth fell (increased) and the discount rate
increased (fell), with substantially similar results. A similar objection holds for the bank reserves measure.
The effectiveness of monetary policy in the aftermath of devaluations is studied by Goldfajn and Gupta
(1 999).
10



3. Descriptive Results
In this section I present some simple descriptive statistics on the incidence and
mean value of changes in the stance of monetary policy around successful and failed
speculative attacks. The siimplest possible graphical overview of the evidence is in
Figure 3, which reports the frequency distribution of changes in real discount rates
during successful and failed speculative attacks. The striking feature of this graph is
that there is no apparent difference in the direction or magnitude of changes in discount
rates during successful and failed speculative attacks. More formally, Table 3 uses
contingency tables to summarize the changes in monetary policy during successful and
failed speculative attabks. The three panels of Table 3 correspond to the three different
measures of tighter monetary policy: increases in real discount rates, decreases in real
domestic credit growth, and increases in bank reserves as a share of domestic credit.
Each panel reports a contingency table, with the columns corresponding to successful
and failed attacks, and the rows corresponding to whether monetary policy tightened or
eased. Based on these tables, I report several statistics of interest. I first report the
conditional probability that monetary policy tightens given that a speculative attack fails.
If the conventional wisdom is correct and tightening monetary policy is a necessary
condition to prevent speculative attacks from ending in a devaluation, one would expect
this probability to be near one. In fact, it ranges from 0.33 to 0.57, depending on the
measure of policy. In each case, the upper bound of a 95% confidence interval extends
to no more than 0.65. In fact, for the first two measures of policy, the 95% confidence
interval includes 0.5, so that it is not even possible to reject the null hypothesis that
tighter and looser monetary policy during failed attacks are equally likely. This casts
doubt on the notion that tightening monetary policy is necessary to ensure that
speculative attacks fail.
I also report the conditional probability that a speculative attack fails given that
monetary policy tightens. If the conventional wisdom is correct and tightening monetary
policy is a sufficient condition for speculative attacks to fail, one would expect this
probability also to approach one. In this sample, the estimated probability ranges from
0.60 to 0.72, with a 95% confidence interval extending to at most 0.80. This calls into
1 1



question the idea that raising discount rates is sufficient to ensure that speculative
attacks fail.
More formally, I also report the p.-value lor a chi-squared test of independence
between changes in monetary policy and the success or failure of speculative attacks.
For the first two measures, it is not possible to reject the null hypothesis of
independence at conventional significance levels, suggesting that there is no
relationship whatsoever between changes in the stance of monetary policy and the
success or failure of speculative attacks. For the third measure, the null is (barely)
rejected at the 95% significance level, but this does not constitute evidence in favour of
the conventional view. To see this, note that the probability that an attack fails
conditional on tightened monetary policy is 0.60, while the unconditional probability that
an attack fails is 177/263=0.67 - in other words, speculative attacks are significantly
less likely to fail when monetary policy is tighter than in the sample as a whole. The
rejection of the null of independence tells us that this difference in probabilities is
(barely) statistically significant.
In Table 4, I repeat the analysis, but restricting the sample to the financially-
developed subsample described in the previous section. One might expect that any
evidence on the efficacy of a high interest rate defense would be more apparent in this
smaller set of observations. However, the results in Table 4 show that this is not the
case. Although the conditional probabilities (of tighter monetary policy conditional on
failure, and of failure conditional on tighter monetary policy) are generally a little higher
in this sample, they are still far from one, and in no case can I reject the null hypothesis
that changes in the stance of monetary policy and the outcome of speculative attacks
are independent.
These results have been subjected to a wide variety of robustness checks that
are not reported for brevity. These include: (1) restricting the sample of events to those
for which all measures of monetary policy are available, (2) using a three-month instead
of a one-month window over which to measure changes in the various indicators of
monetary policy around the speculative attack, (3) varying the timing of changes in
monetary policy relative to the date of the attack, and (4) defining tightened monetary
policy as episodes where both real discount rates increased and real domestic credit
12



growth fell. The main conclusion that the stance of monetary policy and the outcome of
the speculative attack are independent is robust to all of these variants.
While Tables 3 and 4 provide a concise summary of the available evidence, they
discard potentially useful information by treating changes in policy as binary events, i.e.
interest rates, domestic credit growth or bank reserves increase or decrease only. I
relax this restriction in Tables 5 and 6, which present two sets of statistics for the full
sample of events and the financially-developed subsample, respectively. In the first
three columns of both tables, I compute the mean change in each measure of monetary
policy during successful and failed speculative attacks and test the null hypothesis that
they are equal. This may be thought of as a weaker version of the tests of necessity in
Tables 3 and 4, in the sense that a rejection of this null hypothesis constitutes evidence
that tightening monetary policy is necessary to prevent speculative attacks from
succeeding.'2  In the next two columns, I estimate the marginal impact of the change in
each measure of monetary policy on the probability that a speculative attack fails,
estimated from a probit regression including a constant term. I report the estimated
marginal effect, and the t-statistic corresponding to the null hypothesis that the
underlying coefficient on the policy measure is zero. This may be thought of as a
weaker version of the earlier tests of sufficiency, in the sense that a positive impact
suggests that tightening monetary policy raises the probability that an attack fails. In the
final column I report the number of observations included in each test. The three
panels of each table again correspond to the three measures of changes in the stance
of monetary policy.13
A further drawback of the previous results is that they do not allow for the
possibility that the effects of monetary policy may depend on fundamentals which vary
across speculative attack episodes. In order to take these possible non-linearities into
account, the rows of Table 5 report results for various subsamples corresponding to
"good" values of such fundamentals. I first distinguish further between financially
12 A stronger test would also require the mean tightening in monetary policy to be positive during failed
attacks and negative during successful attacks.
13Unlike the data description in the previous tables, these summary statistics are not robust to extreme
outliers in the measures of monetary policy. I therefore drop a small number of episodes occurring during
periods of very high inflation where measured changes in real discount rates and real domestic credit
growth are greater than 100 percent in absolute value.
13



developed and less-developed episodes by restricting the sample to the OECD, and to
the 1980s and 1990s. Following the suggestiorn of Goldfajn and Gupta (1999) that
interest rate defenses are only successful when the banking system is strong, I restrict
the sample to those episodes that were not preceded by a banking crisis in any of the
previous five years in the rows labelled "No Banking Crisis". Since one might expect
that tightening monetary policy will only be effective if the exchange rate is not too
overvalued, I construct a crude indicator of real exchange rate overvaluation as the
trend growth rate of the real CPI-weighted exchange rate versus the US in the previous
twelve months. In the rows labelled "No Real Cvervaluation", I restrict the sample to
those episodes where this growth rate is below the median for the entire sample. To
capture the notion that a given defense may be more credible if the monetary authority
can back up its commitment to a fixed exchange rate with a large stock of foreign
currency reserves, I also divide the sample in half according to non-gold reserves
relative to imports, and consider only the high-reserves subsample in the rows labelled
"High Reserves". I also proxy for the overall weakness of the country's external
payments position using the average over the previous twelve months of that country's
borrowing from the International Monetary Fund, expressed as a share of its quota in
the organization. In the rows labelled "Low Quota Drawings", I consider only those
episodes where the country has no obligations to the IMF according to this measure.
Finally, I consider the argument that it is easier to defend against a speculative attack
during a booming economy than during a recession, presumably because the domestic
economy is better able to withstand any of the adverse effects of high interest rates
during the high point in the business cycle. I measure this as the deviation of real per
capita GDP growth in a country from its average, in the five preceding years, and then I
divide countries in two at the median value of this deviation and consider only the
booming economies in the rows labelled 'High FPoint in Cycle".
The results in Tables 5 and 6 are not very supportive of the conventional view
that tightening monetary policy lowers the probability that a speculative attack ends in a
devaluation of the currency. In the vast majority of cases, the mean change in monetary
policy is not significantly different during failed and successful attacks, and changes in
the stance of monetary policy are not statistically significant predictors of the outcome of
the speculative attack. Only in five cases are the estimated effects statistically
significant at the 95% level, notably in the OECD subsample. However, when one
14



considers that there are 96 separate hypothesis tests in Tables 5 and 6, I should expect
around five rejections at the 95% significance level even if the stance of monetary policy
and the outcome of speculative attacks were independent. Moreover, even the few
significant results in the OECD subsample are to a large extent driven by a handful of
successful attacks in Greece, Turkey and Portugal.
At first glance, the descriptive evidence presented in this section is hardly
consistent with the view that tightening monetary policy is effective during speculative
attack episodes. At the same time, it is also hardly consistent with the alternative view
that tighter monetary policy has the perverse effect of weakening the currency under
attack. Rather, this descriptive evidence suggests a striking absence of any systematic
relationship between the stance of monetary policy and the outcome of speculative
attacks.
15



4. The Endogeneity of Policy
Although useful as data description, the evidence in the previous section can
provide only limited information about the effects of policy during speculative attacks.
Since policy is itself likely to respond endogenously to the same fundamentals that drive
speculation, and also to the strength of speculative pressures themselves, it is difficult to
infer any structural relationship frorn the correlatiorns of the previous section. In this
section, I present a simple model which formalizes this issue and illustrates its
ambiguous implications for the evidence of the previcus section.14 I then empirically
address the endogeneity problem by estimating an instrumental variables probit model
that expresses the probability that a given speculative attack ends in a devaluation as a
non-linear function of fundamentals and measures of monetary policy, treating monetary
policy as endogenous. After controlling for endogeneity of policy in this way, I still find
no evidence that raising interest rates either lowers or raises the probability that a
speculative attack ends in a devaluation of the currency.
A Simple Model
I consider a one-period moclel of a srnall open economy that fixes its exchange
rate and comes under speculative attack. The economy is populated by a continuum of
identical atomistic speculators of mass one, and a rnonetary authority. The monetary
authority sets the domestic interest rate, i, at the beginning of the period, and at the end
of the period decides whether or not to devalue the currency by an exogenously-given
and known amount, s. Speculators attack the currency by shorting it, i.e. by taking out
loans in local currency at the interest rate set. by the monetary authority at the beginning
of the period, selling the proceeds to the monetary authority in exchange for US dollars
at the beginning-of-period exchange rate, and then unwinding their positions at the end-
of-period exchange rate.   Speculators determine their demand, S, for the reserves of
14 See Drazen (1999), Lahiri and Vegh (1997, 1999) and Lall (1997) for other models which focus
specifically on the role of interest rates as a defense during speculative attacks.
15 In practice, shorting the domestic currency during speculative attacks is generally done using forward
contracts, rather than domestic currency loans. However, the substance of the analysis is not changed by
this complication. See Goldstein et. al. (1993), Garber and Svensson (1995), and Lall (1997) for details.
16



the monetary authority, R, by maximizing their profits net of borrowing costs, which I
assume for convenience to be quadratic in the volume of speculation:
(1)     max<s> 7*  S -   2
where xt denotes the representative speculator's perception of the probability that the
currency will be devalued.16 Solving this optimization problem and aggregating over all
speculators results in a speculative demand for local currencyS(it,i) =
The monetary authority decides whether or not to devalue the currency by
weighing the costs and benefits of maintaining a fixed exchange rate. There are two
costs to fixing: the monetary authority must spend a fraction   (R'  of its reserves to
R
defend the exchange rate, and in order to maintain a desired level of reserves, it may
need to set domestic interest rates higher than it would otherwise do in the absence of
speculative pressures.'7 These costs are summarized in the following loss function of
the monetary authority:
(2)     L(7r i, *) = 57  i    0 * g
R
where for simplicity I have assumed that the monetary authority's disutility of raising
interest rates is linear in the interest rate, with 0* measuring the strength of its aversion
to high domestic interest rates. The parameter 0* is not known to speculators, who
16 This convenient formulation of speculative behaviour is used by Drazen (1999). In the absence of such
adjustment costs, risk-neutral speculators will take infinite short (long) positions in the currency under attack
if the expected return to shorting is positive (negative). At the cost of complicating the algebra, one can
also motivate a continuous speculative demand for loans by assuming that speculators are risk averse.
17 I follow the conventional (implicit) assumption that the monetary authority dislikes reserve losses and
devalues when these losses are excessive. However, it is natural to ask why this should be the case. One
might also imagine that the monetary authority does not value reserves per se, but rather dislikes the
capital losses it suffers following a devaluation when it restores its target level of reserves by purchasing
them at the depreciated exchange rate. In this case larger reserve losses make devaluations more costly.
Moreover, raising interest rates may have the perverse effect of raising the rationally-expected probability of
a devaluation by making devaluations less costly to the monetary authority.
17



share a common belief that it is equal to i. Let ,B dienote the benefits of maintaining the
fixed exchange rate regime. These benefits are also not known to speculators, who
correctly perceive D to be uniformly distributed on the unit interval. Speculators do know
that if the costs of maintaining a fixed exchange rate exceed the benefits, the monetary
authority will devalue the currency to 1+E.
Speculators rationally form their beliefs regarding the probability that the
monetary authority will devalue, given their perceptions of the "type" of the monetary
authority, 0, and given the interest rate set by the mronetary authority. In particular,
speculators understand that 7t = Prob[L(ir,i,9) > f3], so that the rationally-perceived
devaluation probability is: 18
(3)            7   9 R . 2
R  l-Ģi-F
I plot this probability as a function of the interest rates as a bold line in the top panel of
Figure 4. At low levels of the interest rate, the perceived devaluation probability is
decreasing in i. Over this range, speculation against the currency is intense, and the
marginal benefit of raising interest rates (in terms oi reducing reserve losses S)
outweighs the perceived marginal cost to the domestic economy (as measured by the
parameter i). As a result, raising interest rates lowers the monetary authority's disutility
of maintaining the fixed exchange rate, making a devaluation is less likely. In contrast,
when interest rates are high, the marginal benefit of further increases in interest rates is
smaller than the marginal cost to the domestic economy. Over this range, increases in
the interest rate raise the disutility of the fixed exchange rate regime, and so raise the
probability that the currency will be devalued.
18 To simplify this calculation, I assume that L(7r,i, O) < 1, so that Prob[L(j,i, 0) > P] = L(r,i, i) . It is
straightforward to verify that this holds in equilibrium provided that the following parameter restriction is
satisfied: R     -.  +        < 1. This restriction will hcld provided that the devaluation rate e is
small enough and/or the amount of reserves R is large enough, which together ensure that the speculative
demand for reserves is never too large.
18



The question of interest in this paper is the slope of 7c(i), i.e. whether raising
interest rates raises or lowers the probability that a speculative attack ends in a
devaluation of the currency. However, estimating Tl(i) using the data on speculative
attack episodes described in the previous sections is complicated by two factors. First,
for a given interest rate, the slope of 7r(i) will depend on episode-specific characteristics.
This nonlinearity is illustrated in the lower panel of Figure 4, which considers two
speculative attack episodes that are alike in every respect, except that in the second the
level of reserves is higher than in the first. Not surprisingly, the probability of a
devaluation is everywhere lower in the second episode than in the first, since the
monetary authority has more reserves at its disposal to defend the exchange rate. More
important, at the same level of the interest rate (indicated by the vertical line), a small
increase in interest rates in the first episode will lower the probability of a devaluation,
while in the second episode it raises the probability of a devaluation.
The second difficulty is that the monetary authority's choice of interest rates is
endogenous, and depends on the strength of speculative pressures against the
currency. In order to illustrate this endogeneity within the confines of a very simple
model, I assume that the monetary authority sets interest rates to minimize the costs of
maintaining a fixed exchange rate. In particular, I assume that the monetary authority
chooses i to minimize Equation (2), taking into account the dependence of 7E(i) as given
by Equation (3). The optimal interest rate chosen by the monetary authority is:
( ) R ( ;*
and has a very natural interpretation. Other things equal, the higher is the devaluation
rate e or the lower are reserves R, the greater is the volume of speculation and the
higher is the interest rate set by the monetary authority to deter this speculation. The
greater is the monetary authority's aversion to high interest rates (the higher is G*), the
lower is the optimal interest rate. Finally, the more speculators think the monetary
19



authority dislikes high interest rates (the higher is 0), the higher the monetary authority
needs to raise interest rates to reduce speculation.'9
The important point is of course that the inteirest rate chosen by the monetary
authority in Equation (4) depends on the same fundamentals as speculators' perceived
probability of devaluation in Equation (3). In Figure 5, I illustrate how this endogeneity
problem can either obscure or accentuate the effects of tighter monetary policy during
speculative affacks. In the top panel, I again consider two episodes that are alike in
every respect, except that in the latter the reserves of the monetary authority are higher
than in the former. At the equilibrium in the first episode at A, 7T(i) is decreasing in i, so
that a small increase in interest rates has the conventional effect of lowering the
perceived probability of a devaluation. In the high reserves case, the speculators'
rationally-perceived devaluation probabilities are lower than before (shown as a
downwards shift in n(i)), while the monetary authority reacts to these devaluation
perceptions with a lower interest rate since it has a larger "cushion" of reserves. In this
episode, the equilibrium is at B with a lower interest rate and a lower devaluation
probability. Simply comparing these two episodes, one might easily be led to the
mistaken conclusion that raising interest rates raises the probability of a devaluation,
while precisely the converse is true (since both A and B fall on the downward-sloping
portion of 7rt(i)).
Similarly, the endogeneity problem may also lead to the conclusion that raising
interest rates has the conventional effect of lowering the probability of a devaluation
when in fact the opposite is true. 11 illustrate this possibility in the bottom panel of Figure
5. 1 again consider two identical episodes, which now differ only in the monetary
authority's distaste for interest rates (0*) and speculators' beliefs regarding this
parameter (0). The dashed lines correspond to an episode where both 9* and 0 are
lower than in the episode shown irn solid lines. Not surprisingly, the monetary authority
sets a higher interest rate, and since speculators believe that the monetary authority is
"tough", the devaluation probability is lower for every interest rate i (shown as a
19 assume that the monetary authority knows speculators' perceptions regarding its type, i.e. the monetary
authority knows 0. The main point of the model regarding the endogeneity of policy is unaffected if I
instead assume that the monetary authority does not know 0 Ibut instead takes speculators' perceived
devaluation probabilities as given when minimizing Equation (2).
I0



downwards shift in 7c(i)). Comparing the equilibria A (with a high devaluation probability
and a low interest rate) and B (with a low devaluation probability and a high interest
rate), one might easily conclude that raising interest rates lowers the probability of a
devaluation when the converse is true (since both A and B fall on the upward-sloping
portion of 7c(i)).
This discussion illustrates how the endogeneity of policy can bias the estimated
effects of policy in unknown directions. To the extent that the fundamentals that drive
both speculative pressures and the policy response are not fully observable, partial
correlations between policy and the outcome of speculative attacks will not correctly
identify the effects of policy. To achieve identification, I require an exogenous source of
variation in the interest rate set by the monetary authority that can be used as an
instrument for policy. In this stylized model, the monetary authority's private information
about its "type" (0*) plays this role, since changes in 0* shift the monetary authority's
reaction function without shifting speculators' rationally-perceived devaluation
probabilities. More generally, any private information of the monetary authority which
influences its choice of interest rates can in principle serve to identify the effects of
interest rates on speculators' beliefs that an attack will end in the devaluation of the
currency.
Empirical Specification
I now turn to the empirical specification motivated by this simple model. The
objective is to estimate the impact of monetary policy on probability that a speculative
attack fails. Although this probability is not observable, I do observe a binary indicator of
whether a speculative attack fails or not. I can therefore estimate the marginal effects of
policy on probability that an attack fails using a probit model, with this indicator as the
dependent variable. The first implication of the theory is that this probability will be a
non-linear function of fundamentals and the monetary policy response. Although the
simple model discussed above is too stylized to take the exact functional form implied by
Equation (3) literally, it does suggest that the explanatory variables in the probit equation
should include not only measures of policy and fundamentals, but also interactions
between the two. Accordingly, I consider the following non-linear probit specification:
21



yi =PO + l ii + P2'fj + 3' fj * ij +u
(~ ~       ~       ~~~~~ )  ,if Yj* > 
yi   O, ifyj*<O
where yj* is an unobserved latent variable; yj is an indicator variable taking the value 1 if
speculative attack j ends in a devaluation; ij is a measure of the stance of monetary
policy; fj is a vector of episode-specific fundamentals; and u; is a normally-distributed
disturbance term. I consider the same three measures of the stance of monetary policy
(i1) as in the previous section: increases in real discount rates, decreases in domestic
credit growth, and increases in the reserves of the banking system, and five of the
measures of fundamentals (f1) discussed in the previous section: the presence of
banking crises, the extent of real overvaluation, the adequacy of reserves, indebtedness
to the IMF, and the point in the business cycle prior to the speculative attack.
The second implication of the theory is that ij is endogenous and reacts to the
same fundamentals that drive speculative pressures. To the extent that the observed
fundamentals included in fj do not capture all of these factors, the error term in Equation
(5) will be correlated with policy. It is therefore necessary to instrument for both ij and
f, ij in the above regression. The theory indicates that variables that are the private
information of the monetary authority and influence its choice of monetary policy are
candidate instruments. I rely on two instruments, both of which exploit informational
asymmetries that are likely to exist between speculators and the monetary authorities.
The first is the change in reserves (expressed in months of imports) in the month of the
attack. Since in most countries the monetary authority publishes data on its reserves
only with a lag, it will have information on the extent of aggregate speculative pressures
reflected in reserve losses in advance of market participants who do not have timely
access to this data. To the extent that the monetary authority bases its policy response
on its observed reserve losses, this instrument will be correlated with policy. To the
extent that these reserve losses are unknown to speculators, they will affect speculative
pressures only through their effects on policy, and hence this measure is a valid
instrument.
22



The second instrument is the ex post available information on the country's
borrowing from the International Monetary Fund. If a country comes under speculative
attack, it may seek resources from the IMF for temporary balance of payments support.
To the extent that the IMF places conditions on the stance of monetary policy prior to
agreeing to such support, changes in observed IMF borrowing will be correlated with the
indicators of monetary policy around the speculative attack episode in question. To the
extent that speculators have imperfect information as to the substance of the country's
negotiations with the IMF, the IMF's influence over monetary policy will be known to the
monetary authority, but not to speculators. I therefore proxy for the presence of IMF
involvement by the change in a country's borrowing from the IMF in the three months
following the speculative attack relative to the three months prior to the attack, and use
this variable as an instrument for policy.
Obviously these instruments are imperfect. First, they may not be truly
exogenous. There may be unobserved episode-specific characteristics which both raise
the probability that an attack fails and also accelerate reserve losses or trigger IMF
involvement. Second, despite caricatures to the contrary, the involvement of the IMF
may not be significantly correlated with the subsequent stance of monetary policy. The
first objection is easily addressed if the instruments pass tests of overidentifying
restrictions. The second objection concerns the strength of the instruments. As is well
known (Nelson and Startz (1990), Staiger and Stock (1997)), if the instruments are only
weakly correlated with the endogenous variables, two-stage least squares coefficient
estimates will be biased towards the probability limits of their uninstrumented
counterparts in finite samples. That is, instrumenting with weak instruments will not
correct the problem of endogeneity. This caveat should be kept in mind, since in many
cases the explanatory power of the instruments is not as large as I would like.
I estimate Equation (5) using Amemiya's (1978) generalized least squares
estimator for probit models with endogenous regressors. This is a two-stage procedure,
in which the observed dependent variable yj and the endogenous variables are first all
regressed on the exogenous variables and the instruments. Amemiya's insight is that,
provided that the model is over- identified, the structural parameters in Equation (5) can
be retrieved from a GLS regression of the reduced form parameters of the first-stage
regression involving the dependent variable on the reduced-form parameters from the
23



remaining first-stage regressions. As shown by Newey (1987), this method is
asymptotically equivalent to a minimum chi-squared estimator and is the most efficient
method to extract structural from reduced-form paraimeter estimates. Finally, Lee
(1991) provides a test of overidentifying restrictions for this model.
Results
The results of this instrumental variables probit specification are presented in
Tables 7 and 8 for the full sample and the financially-developed subsample,
respectively. Each table reports the results of 15 probit regressions (three measures of
policy times five measures of fundamentals), with a dummy variable taking on the value
one if the attack fails as the dependent variable. For each regression, I report the
estimated coefficients, their standard errors, and the corresponding marginal effect of
an increase in the right-hand side variable on the probability that a speculative attack
fails. In order to assess the validity of the instruments, I report the p-value associated
with the test of overidentifying restrictions (which tests the null hypothesis that the
instruments affect the outcome of the attack only through their effects on policy) and the
p-value associated with a test of the null hypothesis that the instruments are jointly
significant in the first-stage regression of the policy variable on the instruments.
The results in Tables 7 and 8 provide very little evidence of the efficacy of tight
monetary policy as a defense against speculative attacks. In all but two regressions, the
coefficient on the policy variable is not statistically sig nificantly different from zero, and
the estimated marginal effects are generally tiny. The only exception is when policy is
measured as the change in bank reserves, and the fundamentals are proxied by the
presence of banking crises or indebtedness to the IMF. For these two cases, the
estimated effect of policy is negative -- a tightening of monetary policy lowers the
probability that a speculative attack fails. However, this evidence in favour of "perverse"
effects of tighter monetary policy should not be taken too seriously, given the large
number of other specifications which do not corroborate this finding.
An important caveat regarding Tables 7 and 8 concerns the validity of the
instruments. Although in all cases the instruments comfortably pass tests of
overidentifying restrictions, in many cases they have rather weak explanatory power for
24



policy. This implies that the estimates are likely to be biased towards the probability
limits of their uninstrumented counterparts, and hence may still be tainted by
endogeneity bias. While this is unfortunate, it is not clear a priori whether this will result
in a systematic overstatement or understatement of the effects of monetary policy, given
the previous discussion that the direction of the endogeneity bias is theoretically
ambiguous.
25



5. Conclusions
Do high interest rates help to defend exchange rates that come under
speculative attack? The evidence considered in this paper suggests that the answer is
no. Although proponents of the view that high interest rates can support a currency
under attack can point to episodes such as Sweden in the summer of 1992, while
proponents of the contrarian view that high interest rates weaken currencies can point to
Korea in the fall of 1997 as supportive anecdotes, a systematic examination of interest
rates around a large number of historical speculative attack episodes indicates a striking
lack of evidence that the stance of monetary policy is correlated with the outcome of
speculative attacks. In particular, I find no evidence that interest rates systematically
increase or decrease during failed speculative attacks, nor that raising interest rates
lowers or raises the probability that a speculative attack fails. This basic finding is
robust to alternative measures of the stance of monetary policy, to interactions which
control for differences in fundamentals across speculative attack episodes, and to
controlling for the endogeneity of the policy response to a speculative attack.
Nevertheless, several shortcomings of this paper suggest that it may be
premature to conclude that monetary policy is entirely ineffective in during speculative
attacks. In the interests of covering a sample of spieculative attacks large enough to
include interesting variation in the outcome of speculative attacks, the policy response to
the speculative attack, and the fundamentals that are likely to determine both the
outcome of the attack and the efficacy of the policy response, I have made several
compromises with regards to data and methodology. Three such compromises, and
possible strategies to avoid them in future research, deserve mention.
First, I have relied on readily-available but relatively low-frequency monthly data
to identify speculative attacks and the response of policy. This is unfortunate given that
much of the economically interesting variation during speculative attack episodes is
likely to occur at much higher daily, or even hourly, frequencies. The use of monthly
data also precludes modeling the likely path-dependence in the effects of interest rates
on speculative pressures, a point emphasized by Drazen (1999). Moving to high-
frequency data for the more limited sample of speculative attacks for which such data is
26



available may uncover evidence of the effects of monetary policy that are obscured by
the low frequency and absence of dynamics in the present paper.
Second, in this paper I have relied on the very crude indicators of monetary
policy that can readily be constructed from available monthly data. However, as noted
earlier, monetary authorities have a wide variety of instruments at their disposal,
including open market operations, direct interventions in foreign exchange markets,
imposition of credit ceilings, etc. Disentangling these interventions from the observed
fluctuations in observable high-frequency data, and modeling the choice between
instruments over time and across episodes, is essential to obtaining a better
understanding of the role of monetary policy during speculative attacks.
Third, in this paper I have relied on what turn out to be rather weak instruments
to extract the exogenous component monetary policy. While it is theoretically unclear
how this will systematically bias the results -- given that the direction of the endogeneity
bias is ambiguous -- it is nevertheless unsatisfying if one is interested in understanding
the effects of monetary policy during speculative attacks. One possibility for progress
on this front is that by switching to higher frequency data, the more pronounced
informational asymmetries between speculators and the monetary authority will result in
more robust instruments. Another is to carefully investigate the institutional peculiarities
of the monetary authority during individual speculative attack episodes in the hopes of
identifying changes in these institutions which might serve as valid instruments.
Implementing these improvements for a sufficiently large set of speculative
attack episodes that span the relevant range of country experiences will take time. Until
then, however, it seems that the burden of proof for both the conventional wisdom that
raising interest rates strengthens currencies under speculative attack, and also the
contrarian view that it weakens them, lies with the proponents of these views.
27



References
Amemiya, Takeshi (1978). "The Estimation of A Simultaneous Equation Generalized
Probit Model". Econometrica. 46(5):1193-11205.
Bensaid, B. and 0. Jeanne (1997). "The Instability of Fixed Exchange Rate Systems
When Raising the Nominal Interest Rate is Costly". European Economic
Review. 41:1461-1478.
Bernanke, B. and 1. Mihov (1998). "Measuring Monetary Policy". Quarterly Journal of
Economics. 113(3):869-902.
Borio, C. (1997). "Monetary Policy Operating Procedures in Industrial Countries", in
Bank for International Settlements. Implementation and Tactics of Monetary
Policy. Basle.
Christiano, L, M. Eichenbaum and C. Evans (1998). 'Monetary Policy Shocks: What
Have We Learned and to What End?". National Bureau of Economic Research
Working Paper No. 6400.
Cushman, D. and T. Zha (1997). "Identifying Monetary Policy in a Small Open
Economy Under Flexible Exchange Rates". Journal of Monetary Economics.
39:433-448.
Caprio, Gerard and Daniela Klingebiel (1997). "Bank Insolvency: Bad Luck, Bad Policy
or Bad Banking?". Paper presented at the Annual Bank Conference on
Development Economics 1996, Washington, DC.
Demirguc-Kunt, Asli and Enrica Detragiache (1997). "The Determinants of Banking
Crises: Evidence from Industrial and Developing Countries". World Bank Policy
Research Working Paper No. 1828.
Drazen, A. and P. Masson (1994). "Credibility of Policies versus Credibility of
Policymakers". Quarterly Journal of Econon,ics. 109(3):735-754.
Drazen, A. (1999). "Interest Rate Defense Against Speculative Attack Under
Asymmetric Information". Manuscript. Univiersity of Maryland and NBER.
Edison, H. (1993). "The Effectiveness of Central Bank Intervention: A Survey of the
Literature After 1982". Special Papers in International Economics No. 18,
International Finance Section, Department of Economics, Princeton University.
Eichenbaum, M. and Charles L. Evans (1995). "Some Empirical Evidence on the
Effects of Shocks to Monetary Policy on Exchange Rates". Quarterly Journal of
Economics. 110(4):975-1010.
Eichengreen, B., A. Rose and C. Wyplosz (1994). "Speculative Attacks on Pegged
Exchange Rates: An Empirical Exploration with Special Reference to the
European Monetary System". NBER Working Paper No. 4898.
28



Eichengreen, B., A. Rose and C. Wyplosz (1995). "Exchange Market Mayhem: The
Antecedents and Aftermath of Speculative Attacks". Economic Policy. 21:249-
312.
Eichengreen, B., A. Rose and C. Wyplosz (1996). "Contagious Currency Crises".
CEPR Discussion paper No. 1453.
Furman, Jason and Joseph Stiglitz (1998). "Economic Crises: Evidence and Insights
from East Asia". Paper presented at the Brookings Panel on Economic Activity,
September.
Garber, P. and L. Svensson (1995). "The Operation and Collapse of Fixed Exchange
Rate Regimes", in Grossman, G. and K. Rogoff, eds., Handbook of International
Economics. Amsterdam: North Holland.
Grilli, V. and N. Roubini (1995). "Liquidity and Exchange Rates: Puzzling Evidence
From the G-7 Countries". New York University, Salomon Brothers Working
Paper: S/95131
Goldfajn, Ilan and Poonam Gupta (1999). "Overshooting and Reversals: The Role of
Monetary Policy". International Monetary Fund Working Paper No. 99/42.
Goldstein, M., D. Folkerts-landau, P. Garber, L. Rojas-Suarez and M. Spencer (1993).
International Capital Markets: Exchange Rate Management and International
Capital Flows. Washington: International Monetary Fund.
Lahiri, A. and C. Vegh (1997). "Krugman Balance of Payments Crises: Are They For
Real?". Manuscript, UCLA.
Lahiri, A and C. Vegh (1999). "Output Costs, BOP Crises, and Optimal Interest Rate
Policy". Manuscript, UCLA.
Lail, S. (1997). "Speculative Attacks, Forward Market Intervention and the Classic Bear
Squeeze". IMF Working Paper No. 97/164.
Lee, Lung-Fei (1991). "Amemiya's Generalized Least Squares and Test of
Overidentification in Simultaneous Equations Models with Qualitative or Limited
Dependent Variables." University of Minnesota Center for Economic Research
Discussion Paper No. 262.
Nelson, Charles R. and Richard Startz (1990). "Some Further Results on the Exact
Small Sample Properties of the Instrumental Variable Estimator". Econometrica.
58:4:967-976.
Newey, Whitney K. (1987). "Efficient Estimation of Limited Dependent Variable Models
with Endogenous Explanatory Variables". Journal of Econometrics. 36:231-250.
Radelet, S. and J. Sachs (1998). "The East Asian Financial Crisis: Diagnosis,
Remedies, Prospects". Brookings Papers on Economic Activity. (1):1-90.
29



Sims, C. (1992). "Interpreting the macroeconomic Time Series Facts: The Effects of
Monetary Policy". European Economic Review. 36:975-1000.
Staiger, Douglas and James H. Stock (1997). "Instrumental Variables Regression with
Weak Instruments". Econometrica. 65(3):557-586.
30



Appendix: Data Sources
The monthly data employed in this paper are drawn from the International
Financial Statistics of the International Monetary Fund, as follows:
*  Nominal exchange, local currency units per US dollar, period average (IFS Line rF).
*  Non-gold reserves, US dollars (IFS Line 1I.d)
*  Money market rate, percent (IFS Line 60b)
*  Discount rate, percent per year, end of period (IFS Line 60). For France, Singapore
Sweden and the Netherlands after December 1993, I use repurchase rates (IFS Line
60a)
*  Domestic credit, local currency units (IFS Line 32)
*  Reserves of deposit money banks, local currency units (IFS Line 20)
*  Consumer price index (IFS Line 64). For Australia and Ireland, I use the wholesale
price index (IFS Line 63)
*  Imports, c.i.f., US dollars (IFS Line 71)
*  Total IMF credits and loans outstanding (IFS Line 2tl)
*  IMF quota (IFS Line 2t)
Lower-frequency data corresponding to the speculative attack episodes are
drawn from various sources.
*  Domestic credit to the private sector as a share of GDP is drawn from the World
Bank World Tables (FS.AST.PRVT.GD.ZS)
*  The black market premium is drawn from Easterly and Levine (1997)
* Annual real GDP growth is constant price local currency GDP figures drawn from the
World Bank World Tables (NY.GDP.MKTP.KN)
*  Data on banking crises are based on the IMF's May 1998 edition of the World
Economic Outlook, and augmented using the original sources in Caprio and
Klingebiel (1997) and Demirguc-Kunt and Detragiache (1997).
The sample of countries consists of all countries with per capita GNP greater than $800
in 1995 at Atlas exchange rates and with populations greater than one million, i.e. the
31



World Bank's definition of middle- and upper-income countries in that year. From this
sample, I drop Chile, Algeria, Panama, Romania, Saudi Arabia, El Salvador and
Turkmenistan because requisite data on discount rates was not available.
The daily data in Figure 1 are drawn from Bloomberg: SEK -- Swedish krona/US
dollar, spot, mid-rate, SWBRMRGN -- marginal lending rate, mid-rate, KRW, won/US
dollar, spot, mid-rate, and KWCR1T, overnight call rate, mid-rate.
32



Table 1: Successful and Failed Speculative Attacks
Successful Attacks                                                  Failed Attacks
Large Reserve Losses                      Large Spreads
ISR     62:2    BWA    84:7               BOL      62:1'    PER     76:5    PHL      87:9'            OEU      73:3-
COL     65:9-    PRY     84:3             DEU      62:1-    PRT     75:10    TTO     87:1             ESP      74:5'
ARG     66:11'  THA      84:11'           GTM      62:6    DNK      76:7'    ZAF     87:11            MYS    74:6'
ESP     67:11'   VEN     84:2             PER      62:7'    DOM    76:1     GAS      88:5'            DNK    76:4'
FIN     67:10'  ZAF      84:7'            COL      64:7-    FRA     76:4-    GRC     88:3             ESP      76:7-
GBR     67:12'  AUS      85:2'            MAR    64:6      GBR      76:4'    GTM     88:7             ITA      76:3'
IRL     67:12'   DOM     85:1             SYR      64-8    JAM      76:5    KWr    88:7               NLD      76:8'
ISR     67:11'   ECU     85:12            CHE      65:1'   MAR    76:4'    BOL       89:12'           GBR    781'
JAM     67:12'  TTO      85:12            GBR      65:1'    SWE    76:10'   DOM      89:7             NLD      78:10'
MUS    67:12    GTM      8686             GTM      85:5*    SYR     78:3'    MAR     89:3'            CAN      79:1'
PER     67:9'   IDN      86:9             DOM      68:6    TUR      76:10'   GRC     90:3'            DNK      79:9'
TTO     67:12    PRY     86:12            CHE      67:1'   ZAW      76:2'   MEX      90:3             ARG      80:10'
FRA     69:8'   VEN      86:12            COL      67:1    CHE    77:1-    FIN       91:5'            CAN      80:4'
ARG     70:6'   DOM      87:6'            ESP      67:3'    DNK     77:12'  JOR      918'             BEL      81:4'
ECU     70:8    PER      87:11            SYR      67:12    GAB     771'    TUN      91:4'            MYS    81:10'
PHL     70:3    SYR      88:1             CHE      68:7'    MUS    77:8'    ZAP      91:12'           NOR    81:12
TUR     70:8'    GTM     89:11'           CRI      68:11    NOR    77:11'   CAN      92:11'           ESP      82:5'
ISR     71:8    ISR      89:1             DNK      68:10'   CAN     782'    DEU      92:10'           ESP     842'
KOR    71:7    PRY       89:3             ECU      68:3    CRI      78:10'   ESP     92:9'            MYS    84:10
URY     71:12    VEN     89:3             FRA      68:6'    MUS    78:11    GAB      929'             MEX    85:3
BOL     72:11    DOM     90:4             GAB      68:2    ZAF      78:12'  IRL      92:9'            TUR     86:7'
JAM     731'   HUN       91:1             PHL      68:11    CAN     79:5'    NAM     92:4             TUR      89:11'
AUS     74:10'  POL      91:6             TTo      68:11    DOM     79:7    TTO      92:1             IDN      90:12
ISR     74:11    TUR     91:3'            DEU      69:1'    GAB     79:3    BOL      93:1'            PRT      90:10'
KOR     74:12'   BWA    92:7              FIN      69:5'   JAM      79:8    ONK      931'             POL     91:2'
ARG     75:1    ECU      92:9'            URY      69:10'   SYR     79:8'    FRA     93:11'           ARG      92:7
PER     75:10    FIN     92:9'            CRI      70:7'    URY     79:3'    BOL     94:2'            KWT    92:11
ZAF     75:10'   GBR     92:10'           DOM      70:5    OL      80:4'    DOM     948              NOR    92:9'
AUS    76:12'   LBN      92:3             ECU      70:1    ONK      80:2'    MEX     94.4'            PRT      92:9'
MEX    76:9'   SWE    92:11'              ITA      70:7'    PER     80:1'    SVK     94:7             POL     93:6
ESP     77:7'   TTO      93:4             SYR      70:12    PRT     80:2'   TTO      94:5'            URY     93:6
ISR     77:11'  GAB      94:1'            TWN    70:7    AUS        81:9'    ZAP     94:3'            ARG     94:12
PER     77:11    MEX     94:12'           ARG      71:10    CAN     81:7'    ARG     95:3             LVA      94:5
PRT     77:3'   VEN      95:12'           DNK      71:4'    GTM     81:6    GAB      95:r             MAR    94:1'
IDN     78:11'   BGR     96:5             ECU      71:9    MEX      81:6'   ZAF      95:4'            POL     94:9
JAM     78:5    GTM      97:1'            IDN      71:12'   MUS    81:3'    BGR      96:1             COL    95:4'
TUR     78:3    IDN      97:8'            KOR      71:12    MAR    82:6'    GRC      96:5'            THA      951'
BOL     79:12'   KOR     97:11'           LBN      71:2'    MUS    82:6    NAM       96:9             DOM    96:6
MUS    79:11'   MKO    97:7               MAR    71:10'   PHL       82:1'    SWE    96:11'            LVA     96:2
TUR     79:6    PHL      97:9'            CRI      72:2    URY      82:1'    FIN     97:11'           RUS    98:9
KOR    80:1    THA       97:7'            GBR      72:7'    GAB     83:11'  IRL      97:4'            URY     96:9
ARG     81:2'   BWA    98:7               GAB      73:6    CAN      84:6'    NOR     97:12'           BRA      97:11
CRI     81:1'   MOA    98:10              SYR      73:2    DNK      84:12'   RUS     97:11            EST     97:11
MUS    81:10'   MEX      98:9'            ZAF      73:11'   COL     65:1    BRA      98:9             POL      979
BOL     82:2    NAM      98:7             DOM      74:1    JOR      85:3'   CAN      98:8'            UKR    97:12
ECU     82:5    RUS      98:9             ITA      74:2'    PHL     85:10'                            HKG    98:8'
FIN     82:10'   UKR     98:9             JAM      74:5'    CAN     86:3'                             URY     98:9
MEX     82:2'   ZAP      98:7'            KOR      74:7'    DOM     88:5'
PRT     82:6    BRA      99:1              MUS    74:5    FIN       86:8'
SWE    82:10'  KAZ       99:4             URY      74:4'   JAM      86:10
URY     82:12'                            AUS      75:12'   KOR    86:1'
GRC     83:1                              8OL      75:9'    BOL     87:3
ION     83:4'                             CRI      75:10'   CAN     87:4'
JAM     83:11'                            GAB      75:1'    GAB     87:1'
PHL     83:10                             IDN      75:3'    KOR     87:12
ITA     75:7'
Note: * indicates attacks in the financially-developed subsample.
33



Table 2: Selected Speculative Attack Episodes
Country             Date    Classification    % Growth in:
of Attack   Exchange Rate Reserves
Asia 1997
Indonesia           97:8      Succeed        11.2%       -4.7%
Korea               97:11     Succeed        11.3%       -19.6%
Malaysia 1/         97:8         n/a          6.6%        1.4%
Philippines         97:7      Succeed        10.4%        6.1%
Thailand             97:7     Succeed         17.6%      -6.1%
Europe 1992-93
Belgium 2/          92:11        n/a          7.2%       -14.3%
Denmark 3/          93:1        Fail          1.9%       -28.6%
France 41           93:11       Fail          2.8%       -22.7%
Ireland 5/          92:09       Fail          0.3%       -3.8%
Italy6/             92:10        n/a         11.6%        7.7%
United Kingdom 7/    92:10    Succeed        11.6%       -4.4%
Spain 8/            92:09       Fail          4.5%       -20.0%
Sweden              92:11     Succeed        11.5%       -24.6%
Finland             92:11     Succeed        11.6%        1.9%
11 Nominal devaluation too small to qualify as successful attack.
2/ Nominal devaluation too smail to qualify as successful attack. ERW date is 92:9.
3/ ERW date is 92:9.
41 ERW date is 92:9.
5/ ERW date is 92:11.
61 ERW date is 92:9. Exchange rate too volatile prior to attack to qualify as successful attack.
7/ ERW date is 92:8.
8/ Subsequent devaluation of peseta in 92:10 too smail to prevent classification as failed attack.
ERW: Eichengreen,Rose and Wyplosz (1994).
34



Table 3: Changes In Monetary Policy During Speculative Attacks
(Full Sample)
Discount Rates
Speculative Attack:
Succeeds       Fails        Total
Monetary Policy
Tightens                     37           95           132
Eases                        40           72           112
Total                        77           167          244
Estimate     95% Confidence Interval
P[Tightens i Fails]          0.57         0.49         0.65
P[Fails I Tightens]          0.72         0.64         0.80
P-Value for Independence     0.20
Domestic Credit Growth
Speculative Attack:
Succeeds       Fails        Total
Monetary Policy
Tightens                     41           83           124
Eases                        59          101           160
Total                       100           184          284
Estimate     95% Confidence Interval
P[Tightens I Fails]          0.45         0.38         0.52
P[Fails I Tightens]          0.67         0.58         0.75
P-Value for Independence     0.50
Bank Reserves
Speculative Attack:
Succeeds       Fails        Total
Monetary Policy
Tightens                     39           58           97
Eases                        47          119          166
Total                        86           177         263
Estimate    95% Confidence Interval
P[Tightens I Fails]          0.33         0.26         0.40
P[Fails I Tightens]          0.60         0.50         0.70
P-Value for Independence     0.05
Notes: The contingency tables report the distribution of speculative attacks according to a two-way
classification of whether monetary policy tightened or not using the indicated measure of monetary policy, and
whether the attack succeeded or failed. P(Tightens I Fails] reports the conditional probability that monetary
policy tightens during failed attacks, and P[Fails I Tightens] reports the conditional probability that the attack
fails during episodes where monetary policy tightens. P-Value for Independence reports the p-value
associated with a chi-squared test of independence of the rows and columns of the contingency table.
35



Table 4: Changes In Monetary Polilcy During Speculative Attacks
(Financially-Developed Sample)
Discount Rates
Speculative Attack:
Succeeds       Fails        Total
Monetary Policy
Tightens                      13           47           60
Eases                         15           40           55
Total                         28           87           115
Estimate    95% Confidence Interval
P[Tightens j Fails]           0.54         0.43         0.65
P[Fails I Tightens]           0.78         0.68         0.89
P-Value for Independence      0.48
Domestic Credilt Growth
Speculative Attack:
Succeeds       Fails        Total
Monetary Policy
Tightens                      14           35           49
Eases                         22           49           71
Total                         36           84           120
Estimate    95% Confidence Interval
P[Tightens I Fails]           0.42         0.31         0.52
P[Fails I Tightens]           0.71         0.59         0.84
P-Value for Independence      0.78
Bank Reserves
Speculative Attack:
Succeeds       Fails        Total
Monetary Policy
Tightens                      17           29           46
Eases                         18           53           71
Total                         35           82           117
Estimate    95% Confidence Interval
P[Tightens I Fails)           0.35         0.25         0.46
P[Fails I Tightens]           0.63         0.49         0.77
P-Value for Independence      0.18
Notes: The contingency tables report the distribution of speculative attacks according to a two-way
classification of whether monetary policy tightened or not using the indicated measure of monetary policy, and
whether the attack succeeded or failed. P[Tightens I Failsj reports the conditional probability that monetary
policy tightens during failed attacks, and P[Fails I Tightens] reports the conditional probability that the attack
fails during episodes where monetary policy tightens. P-Value for Independence reports the p-value
associated with a chi-squared test of independence of the rows and columns of the contingency table.
36



Table 5: Weaker Tests of Necessity and Sufficiency
(Full Sample)
Increase in Real Discount Rate
Failed    Successful             Marginal               Number of
Attacks    Attacks    P-Value      Effect    t-Statistic  Observations
Full Sample            1.342      -3.111      0.085      0.003       1.752        239
OECD                   1.173      -8.492      0.040      0.006       2.163        88
1980s and 1990s       -1.378     -2.847       0.644      0.001       0.471        136
No Banking Crises      1.655      -3.127      0.123      0.003       1.601        188
No Real Overvaluation  1.652      -1.279      0.437      0.002       0.861        113
High Reserves          1.067      -4.784      0.123      0.003       1.509        124
Low Quota Drawings     1.411      1.021       0.892      0.000       0.119        112
High Point in Cycle    0.391      -2.443      0.526      0.001       0.677        116
Decrease in Real Domestic Credit Growth
Full Sample           -3.657      0.404       0.482      -0.001      -0.717       230
OECD                  -1.779      13.801      0.056      -0.002      -1.623       75
1980s and 1990s       -2.043     -4.808       0.734      0.000       0.348        131
No Banking Crises     -4.032      4.096       0.204      -0.001      -1.298       184
No Real Overvaluation  0.499      7.143       0.452      -0.001      -0.790       109
High Reserves         -1.515      1.907       0.677      0.000       -0.437       123
Low Quota Drawings     1.628      -3.868      0.501      0.001       0.674        100
High Point in Cycle   -6.527      1.848       0.316      -0.001      -1.001       125
Increase in Bank Reserves
Full Sample           -0.522     -0.140       0.073      -0.033      -1.777       263
OECD                  -0.295      0.038       0.131      -0.071      -1.147       77
1980s and 1990s       -0.399     -0.200       0.450      -0.017      -0.726       154
No Banking Crises     -0.652     -0.328       0.128      -0.033      -1.445       204
No Real Overvaluation    -0.536   0.031       0.017      -0.052      -1.841       126
High Reserves         -0.544     -0.099       0.172      -0.033      -1.403       142
Low Quota Drawings    -0.483     -0.380       0.741      -0.009      -0.321       114
High Point in Cycle   -0.519     -0.208       0.279      -0.021      -0.950       143
Notes: The first two columns report the mean value of the indicated policy variable during failed and
successful attacks, in the indicated subsample of events. The third column reports the p-value associated
with a test of the null hypothesis that the means are equal in the two samples. The fourth column reports the
estimated marginal effect of policy in a probit regression expressing the probability that a speculative attack
fails as a function of a constant and the indicated policy variable, in the indicated subsample of events. The
fifth column reports the t-statistic associated with the estimate of the underlying slope coefricient. The final
column indicates the number of observations for which the policy variables are available in the indicated
subsample of events.
37



Table 6: Weaker Tests of Necessity and Sufficiency
(Financially-Developed Sample)
Increase in Real Discount Rate
Failed   Successful          Marginal              Number of
Attacks   Attacks   PI-Value   Effect   t-Statistic  Observations
Full Sample            -0.155    -4.464     0.177     0.004     1.409        115
OECD                   0.725    -11.236    0.016      0.008     2.540         71
1980s and 1990s        -1.671    -0.972     0.836    -0.001     -0.214        65
No Banking Crises      1.181    -4.713      0.127     0.005     1.535         84
No Real Overvaluation    1.728    -1.163    0.555    0.003      0.640         56
High Reserves          -0.267    -4.331     0.260     0.004      1.130        71
Low Quota Drawings     0.785    -1.590      0.556     0.002     0.515         66
High Point in Cycle    -0.664    -6.405     0.242     0.005      1.258        57
Decrease in Real Domestic Credit Growl.h
Full Sample            -7.339    -3.195     0.561    -0.001     -0.542        108
OECD                   -5.725    6.347      0.103    -0.002    -1.191         63
1980s and 1990s        -8.812    -4.203     0.635    -0.001     -0.438        61
No Banking Crises     -6.166     3.530      0.233    -0.001     -1.083        83
No Real Overvaluation    1.780    -2.841    0.635     0.001     0.458         54
High Reserves          -7.870    -0.305     0.457    -0.001     -0.751        67
Low Quota Drawings     -3.078    -1.327     0.806     0.000     -0.170        62
High Point in Cycle   -11.119    0.831      0.280    -0.001     -0.986        55
Increase in Bank Reserves
Full Sample            -0.247    0.280      0.117    -0.050     -1.654        117
OECD                   -0.340    -0.081     0.246    -0.046    .-0.757        64
1980s and 1990s        -0.099    0.220      0.339    -0.033     -0.837        68
No Banking Crises      -0.380    -0.066     0.149    -0.065     -1.161        87
No Real Overvaluation   -0.246   0.017      0.255    -0.064    -t).808        57
High Reserves          -0.454    0.304      0.141    -0.064    -1.512         65
Low Quota Drawings     -0.391    0.056      0.273    -0.052    -1.107         65
High Point in Cycle    -0.071    0.350      0 224    -0.040    C.0.994        61
Notes: The first two columns report the mean value of the indicated policy variable during failed and
successful attacks, in the indicated subsample of events. The third column reports the p-value associated
with a test of the null hypothesis that the means are equal in the two samples. The fourth column reports the
estimated marginal effect of policy in a probit regression expressing the probability that a speculative attack
fails as a function of a constant and the indicated policy variable, in the indicated subsample of events. The
fifth column reports the t-statistic associated with the estimate of the underlying slope coefficient. The final
column indicates the number of observations for which the policy variables are available in the indicated
subsample of events.
38



Table 7: Controlling for the Endogeneity of Policy
(Full Sample)
Fundamental (f):
Banking    Real    Reserves/  IMF    Growth
Crisis  Overvaluation Imports Borrowing Deviation
Increase in Real Discount Rate
i       01                 0.015       0.009   -0.142    0.037   -0.052
se(13)             0.152      0.462    0.188    0.129    0.220
dP[Fail]/d13      0.005       0.003   -0.017    0.012   -0.012
ixf     13                -0.070       0.000    0.090    0.000    0.013
se(p)              0.244      0.040    0.095    0.001    0.031
dP[Faildtdp       -0.024      0.000    0.011    0.000    0.003
f       1                  -0.221      0.011   -0.192   -0.002   -0.061
se(1)              1.955      0.265    0.295    0.004    0.308
dP[Failydo        -0.074      0.004   -0.023   -0.001   -0.014
p-OlD                      0.830       0.985    0.628    0.702    0.918
p-FSR                      0.031       0.134    0.102    0.015    0.145
Number of Observations       202         199      191      202      180
Decrease in Real Domestic Credit Growth
i  13             -0.105      -0.093    0.118   -0.065    0.001
se(p)              0.134      0.110    0.226    0.084    0.081
dP[Fail]/d13      -0.011      -0.008    0.011   -0.011    0.000
ixf     ,                  0.137       -0.007   -0.038    0.000   -0.002
se(f3)            0.162       0.009    0.052    0.000    0.017
dP[FaillJd        0.014       -0.001   -0.004    0.000   -0.001
f       1                  0.652       0.019    0.215   -0.002    0.051
se(13)             1.538      0.055    0.432    0.003    0.187
dPVFaill/d13      0.066       0.002    0.021    0.000    0.017
p-OlD                      0.716       0.848    0.759    0.559    0.915
p-FSR                      0.329       0.762    0.477    0.682    0.075
Number of Observations       200         197      191      199     182
Increase in Bank Reserves
i       13               -1.339-       -2.266   -4.834  -2.122*   -1.298
se(1)             0.574       5.489    3.419    1.191    1.725
dP[Failydo        -0.254      -0.229   -0.506   -0.263   -0.213
ixf     p                  2.465       -0.065    0.850    0.004   -0.083
se(D)              1.919      0.245    0.764    0.006    0.129
dP[FailIl/dp      0.467       -0.007    0.089    0.001   -0.014
f       .1                 1.345       -0.029    0.451    0.004   -0.022
se(13)             0.841      0.168    0.496    0.003    0.160
dP[Faill/d13      0.255       -0.003    0.047    0.000   -0.004
p-OlD                      0.957       0.952    0.833    0.736    0.730
p-FSR                      0.051       0.155    0.633    0.357    0.087
Number of Observations       229         226      218      228     215
Notes: This table reports the results of a probit regression which expresses the probabiiity that a speculative
attack fails as a function of the monetary policy response (i), a fundamental (f), and the interaction between
the two (ixf), for the indicated measures of policies and fundamentals. All regressions also include a
constant. The rows labelled ,B, se(,B), and dP[Fail]/d,B report the estimated coefficient, its standard error, and
the estimated marginal effect. The policy variable and its interaction with the fundamental are treated as
endogenous, and are instrumented using the contemporaneous change in reserves and the change in IMF
lending, and their interactions with fundamentals. The model is estimated using Amemiya's two-stage GLS
procedure for instrumental variables probit models. p-OlD reports the p-value for the test of overidentifying
restrictions, and p-FSR reports the p-value for the null hypothesis that the instruments are jointly insignificant
in the first-stage regression of the policy variable on the instruments. * (*) denotes significance at the 10%
(5%) level.
39



Table 8: Controlling for the Endogeneity of Policy
(Financially-Developed Sample)
Fundamental (f):
Banking    Real   Reserves/  IMF    Growth
Crisis  Overvaluation Imports Borrowing Deviation
Increase in Real Discount Rate
i        3              -0-039      0.105   -0.004    0.333   -0.005
se(,)           6.700      0.167    0.381    0.395    0.796
dP[Failyd13    -0.010      0.020   -0.001    0.032   -0.001
ixf      0              -0.053     -0.009    0.022   -0.001    0.001
se(p)           9.661      0.010    0.081    0.002    0.097
dP[Failydp     -0.013     -C.002    0.006    0.000    0.000
f        p              -0.907     -0.029    0.024    0.000    0.034
se(p)          28.351      0.069    0.378    0.008    0.757
dP[Failyd3     -0.222     -0.006    0.007    0.000    0.010
p-OlD                    0.993      0.779    0.915    0.886    0.961
p-FSR                    0.366      0.091    0.386    0.479    0.228
Number of Observations     105        104      104      105      104
Decrease in Real Domestic Credit Growth
i       13               0.007     -0.044   -0.010   -0.979    0.055
se(p)           0.628      0.062    0.114   12.762    3.053
dPEFail]/d13    0.002     -0.008   -0.003   -0.013    0.008
ixf      1               0.010      0.004    0.004    0.004   -0.011
se(p)           1.114      0.006    0.025    0.046    0.605
dPtFailyd1      0.003      0.001    0.001    0.000   -0.002
f        p              -0.217      0.018    0.060   -0.006   -0.106
se(p)           8.346      0.060    0:379    0.078    7.385
dP[Fail]1d13   -0.067      0.003    0.019    0.000   -0.016
p-OlD                    0.957      0.705    0.716    1.000    0.999
p-FSR                    0.306      0.591    0.466    0.259    0.228
Number of Observations     98          97       98       98       96
Increase in Bank Reserves
i        1              -2.946     -2.590   -2.811   -0.864   -6.951
se(p)           1.824      9.620    2.661   12.202  162.283
dP[Failydp     -0.320     -0.261   -0.648   -0.193   -0.251
ixf     ,B               2.080      0.116    0.674   -0.001   -0.681
se(13)          5.179      0.921    0.861    0.055   15.232
dP[Failyd13     0.226      0.012    0.155    0.000   -0.025
f        1               1.834      0.0118    0.163    0.002   -0.042
se(p)           1.847      0.420    0.286    0.062    2.365
dP[Faillyd      0.199      0.002    0.038    0.000   -0.002
p-OlD                    0.712      0.903    0.649    0.956    0.996
p-FSR                    0.620      0.965    0.612    0.347    0.897
Number of Observations     105        104      104      105      105
Notes: This table reports the results of a probit regression which expresses the probability that a speculative
attack fails as a function of the monetary policy response (i), a fundamental (f), and the interaction between
the two (ixt), for the indicated measures of policies and fundamentals. All regressions also include a
constant. The rows labelled 1, se(3), and dP[Faillid13 report the estimated coefficient, its standard error, and
the estimated marginal effect. The policy variable and its interaction with the fundamental are treated as
endogenous, and are instrumented using the contemporaneous change in reserves and the change in IMF
lending, and their interactions with fundamentals. The model is estimated using Amemiya's two-stage GLS
procedure for instrumental variables probit models. p-OlD reports the p-value for the test of overidentifying
restrictions, and p-FSR reports the p-value for the null hypothesis that the instruments are jointly insignificant
in the first-stage regression of the policy variable on the instnrments. . * () denotes significance at the 10%
(5%) level.
40



log scale  -            o
Co  cn   o   cn  Co  cA    o   cn  Co                                                   0            CO
C)       C0            C)
02-Jun-97                   -,J F                                                                              0
1 -Jun-92 .
12-Jun-97
m                              11-Jun-92
22-Jun-97          O1- 
23-Jun-92
02-Jul-97                                    og
121-Ju-97                                                    3-Jul-92
22-Jul-97                                                   15-Ju-92
201-Au-97    ~ ^                                             27-Jul-92 -
01-Aug-97                                                    6-Aug-92
210-Au-97 .-                                               18-Aug-92
31-Aug-97              6                                    28-Aug-92-.                                             c
21 -Aec-97                   s                               9 8-Sep-92 :                                                0 
319-Aug-97                                                  2 8-Aug-92 z 
20-Sep-97                                    21~~~~~~~~~~-Sep-92
0.
20-Sep-97                                                     1-Oc-92                     -0
10-oct-97                                         13-b) OcM to > cn (F2ao -92
20-Oct-97                                                   23-Ot-9
o ~~ ~~~~~ ~ ~             ~~~~ oOctn92                                        co
3 0-Oct-97
4-Nov-92
09-Nov-97                                                                                                                 '
19-Nov-97                                   ~~~~~~~~~~~~~16-Nov-92 
29-Nov-97                                                   26Nv9                                          0.
09-Dec-97                                     8~~~~~~~~~~2-Dec-92           c2 
19-Dec-97                                                   18-Dec-92                         :
C: (D~~~~~C
29-Dec-97                                                   183-Dec-92  -        I
0  CJ~~~~  -~~  -~~~  f~~3    ~~~)           0               WJ C  rth.  ui  0)  -4j  0)
0  0      01      ~~      ~~~0 CO
0   0      ~~~0 0)           0                                   Krona/$
0Won/$0        0 



Figure 2: Exchange Rates and Reserves During
Successful and Failed Speculative Attacks
Successful Attacks
120 --
115- /
110  I
0,                              105-
II~ ~ ~~~~~~9
85 -
-12  -10  -8  -6   -4   -2    C    2    4    6    8    10   12
Months Since Attack
Failed Attacks
120 -
110 -
80 -
70 -
,,I   ,                  ,I  ,  , -60       ,      r         ,
-12   -10    -8    -6    -4    -2        0      2      4      6      8     10    12
Months Since Attack
Notes: This figure shows the evolution of the median nominal exchange rate and reserves during
successful and failed speculative attacks. The figures are constructed by cumulating the median (across
all episodes) growth rate of the indicated variables to a base of 100 twelve months prior to the attack.
42



Figure 3: Changes in Real Discount Rates During
Successful and Failed Speculative Attacks
0.25
Mean Change During
Successful Attacks = 0.45
0.2 -                                                  eD Successful Attacks
Mean Change During           l           Failed Attacks
Failed Attacks = 0.30
01
as
0
0
0.1 l
-25   -20   -15   -10    -5       0     5     10    15    20    25    >25
Percent Changes in Real Discount Rates Less Than:
Notes: This figure shows the frequency distribution of percentage changes in real discount rates during
successful and failed speculative attacks. The mean changes during successful and failed attacks are
based on changes in real discount rates less than 25% in absolute value.
43



Figure 4: Devaluation Probabilities as a Function of Interest Rates
Devaluation Probability
0.9
0.8
°0.7 -                                                               2(i)
> 0.6 -
o 0.5
. 0.4
3
2 0.3
0.2
0.1
0-I
0         0.05        0.1        0.15        0.2        0.25        0.3
Interest Rate
Non-Linear Effects of Policy
1
0.9
0.8 -
0
70(i), Low Reserves
o 0.6 -
0.2 -
0
0.4 -
~~~~0.4 ~~~~~~~~~~ir(i), High Reserves
2 0.3-
0.2O-
0.1 
0
0         0.05        0.1        0.15        0.2        0.25        0.3
Interest Rate
44



Figure 5: The Endogeneity of Policy
Case 1: Endogeneity Bias Obscures Conventional View
0.9 
0.8             .
0
~0.7-
> 0.6
Q
0.5
.0
0.4 
0.2-
0.1                       I
0   i
0        0.05       0.1        0.15       0.2        0.25       0.3
Interest Rate
Case 2: Endogeneity Bias Exaggerates Conventional View
1
0.9                    
0.8
0.2
~0.
0.4     -     l          l          l          l   l      l
00.5 -                    0         0           0         0
~0.2 
Interest Rate
45
20.3 ~ ~ ~ ~ ~~4






Policy Research Working Paper BSries
Contact
Titie                             Author                  Date               for paper
WPS2252 Productivity Growth, Capital       Ejaz Ghani               December 1999      N. Mensah
Accumulation, and the Banking    Vivek Suri                                  80546
Sector: Some Lessons from Malaysia
WPS2253 Revenue Recycling and the Welfare  Ian W. H. Parry          December 1999      R. Yazigi
Effects of Road Pricing          Antonio Miguel R. Bento                     37176
WPS2254 Does "Grease Money" Speed Up       Daniel Kaufmann          December 1999      H. Sladovich
the Wheels of Commerce?           Shang-Jin Wei                              37698
WPS2255 Risk and Efficiency in East Asian  Luc Laeven               December 1999      R. Vo
Banks                                                                        33722
WPS2256 Geographical Disadvantage:         Anthony J. Venables      December 1999      L. Tabada
A Heckscher-Ohlin-von Thunen      Nuno Limao                                 36896
Model of International Specialization
WPS2257 Infrastructure, Geographical       Nuno Limao               December 1999      L. Tabada
Disadvantage, and Transport Costs  Anthony J. Venables                       36896
WPS2258 Market Access Bargaining in the    J. Michael Finger        December 1999      L. Tabada
Uruguay Round: Rigid or Relaxed  Ulrich Reincke                              36896
Reciprocity?                     Adriana Castro
WPS2259 Predicting Currency Fluctuations   Daniel Kaufmann          December 1999      E. Khine
and Crises: Do Resident Firms    Gil Mehrez                                  37471
Have an Informational Advantage?    Sergio Schmukler
WPS2260 Regional Integration Agreements:    Anthony J. Venables     December 1999      L. Tabada
A Force for Convergence or                                                   36896
Divergence?
WPS2261 is Knowledge Shared within         Kaushik Basu             December 1999      M. Mason
Households?                      Ambar Narayan                               30809
Martin Ravallion
WPS2262 How Inadequate Provision of Public   Ritva Reinikka         December 1999      H. Sladovich
infrastructure and Services Affects  Jakob Svensson                          37698
Private Investment
WPS2263 When Is Growth Pro-Poor? Evidence  Martin Ravallion         December 1999      J. Israel
from the Diverse Experiences of   Gaurav Datt                                85117
India's States
WPS2264 Do More Unequal Countries          Branko Milanovic         December 1999      P. Sader
Redistribute More? Does the Median                                           33902
Voter Hypothesis Hold?



Policy Research Working Paper Series
Contact
Title                            Author                   Date              for paper
WPS2265 The Political Economy of Distress    Paola Bongini         January 2000       R. Vo
In East Asian Financial Institutions   Stijn Claessens                      33722
Giovanni Ferri
WPS2266 The Impact of Adult Deaths on      Martha Ainsworth        January 2000       S. Fallon
Children's Health in Northwestern    Innocent Semali                        38009
Tanzania