M PRA Munich Personal RePEc Archive
Identification of currency crises in emerging economies: a combined approach Tjeerd M. Boonman Banco de Mexico, General Direction of Financial Stability
1 September 2017
Online at https://mpra.ub.uni-muenchen.de/81445/ MPRA Paper No. 81445, posted UNSPECIFIED
Identification of currency crises in emerging economies: a combined approach1 Tjeerd M. Boonman Banco de México Ave. Cinco de Mayo 1, CP 06000, Mexico City e-mail:
[email protected] September 1, 2017
Abstract The first step for empirical studying currency crises is to identify or date these crises. This is not an easy task – not even the conceptual definition is agreed upon, let alone an operational definition. The choice of a definition has an impact on the identification of crises, and may even provide an explanation to the mixed performance of early warning systems for currency crises. In empirical studies typically a single definition is adopted and the crisis dates are compared to the narrative to determine a set of currency crisis dates. We present a novel way to identify currency crises. We employ combinations of definitions, which provide a wider base than using a single definition. Through combining we can overcome setbacks of the individual definitions. We date currency crises in 35 emerging economies for the period 1990 to 2016, in a monthly frequency. We compare our dataset with the narrative, expert opinions and other studies, and find that a combination of several definitions fits the narrative better than any single crisis definition. We include the narrative in the appendix, which makes this work more transparent and useful for other investigators.
Key words: currency crises, emerging economies, foreign exchange rates JEL classification: G01, F31 1
The opinions expressed here are those of the author and do not represent Banco de Mexico’s or its board of governors’ opinions. We thank Nicholas Bucher for excellent research assistance. All errors or omissions are my own.
1
1. Introduction How to identify and date currency crises has been debated since the mid-1990s. There is no consensus about a formal definition of currency crisis derived from theory, and international organizations do not systematically categorize currency crises (Lestano and Jacobs, 2007). Where country studies often base the crisis episodes on the narrative, for larger data sets it is common to use statistical criteria. The large majority of crisis definitions based on statistical criteria can be categorized in two conceptual groups: a large depreciation of the exchange rate, and a large pressure on the exchange rate (which can result in a depreciation of the currency, or forces the authorities to defend the currency by expending large volumes of international reserves or sharply increasing interest rates). Within each group exists a large number of variations on frequency, thresholds, definitions of indicators, et cetera. Pontines and Siregar (2008) find that there is no single approach that can be clearly considered as far superior to other alternatives. They even consider that the difficulties in identifying crisis dates may provide an explanation to the mixed performance of early warning systems for currency crises. None of the definitions is flawless, which may be overcome by combining definitions. Two often-used databases for currency crises episodes are Reinhart and Rogoff (2011) and Laeven and Valencia (2012). However, both are based on the large depreciation of the nominal exchange rate, and as a consequence do not take into account episodes with high pressure on the exchange rate. Another drawback is that the crisis dating is on an annual base. Given the characteristics of currency crises, a higher frequency is useful to identify crises that reverse (partially) within a year, or crises that develop over two adjacent calendar years. Dating the crises at a higher frequency can also improve the analysis of currency crises. We present a novel way to identify or date currency crises. We date currency crises on a monthly basis by combining several quantitative definitions, which are compared to the narrative, expert opinions, and publicly available databases. Our approach is inspired by Pontines and Siregar (2008) that advise to use various methods and various thresholds to identify crises, and by Lestano and Jacobs (2007) that remark 2
that it is only possible to judge the accuracy of currency crisis dating by comparing the crisis episodes with the narrative and expert opinions. We apply our crisis definition procedure on a sample of 35 emerging economies in the period 1990 to 2016. We first select six definitions that have a different perspective on currency crises – in terms of concept (depreciation and pressure), frequency (monthly, quarterly, annually), components (exchange rate, reserves, and/or interest rate) and definitions (nominal vs real, spread vs change). We find that a combination of definition fits the narrative better than any individual definition. The intuition is that only when an event is picked up in several dimensions the event is large enough to be labeled as a crisis. Our approach is comparable with Babecky et al. (2012) that use annual currency crisis dates in developed economies from five previous studies. Then they date the crises on a quarterly basis, and have experts confirm or adjust these dates. Our work is also related to Kaminsky (2006) that also date the crises with a monthly frequency, although for a smaller number of developed and emerging economies, and by using only one definition. We depart from Babecky et al. (2012) with our focus on emerging economies, our higher frequency and our crisis dating, which is based on calculations rather than adopted from other studies. We differ from Kaminsky (2006) as we use six definitions and include a large number of emerging economies. Our work also differs from these works as the narrative to validate the crisis dates is presented in a transparent manner. Thus, our contributions to the literature are twofold. First, we show that for a large sample of emerging economies a combination of definitions yields better results than relying on a single definition. Second, we offer more transparency in the crisis dating, by providing the narrative for a large number of emerging economies. The database is useful for research projects related to currency crises, such as constructing early warning systems, analyzing the impact from currency crises, and contagion. The remainder of this paper is organized as follows. In Section 2 we describe the two major currency crisis definitions, including advantages and disadvantages of choices related to the operational definitions. In Section 3 we present our approach to date
3
currency crises. In Section 4 we describe the data and Section 5 we present the results. Section 6 concludes.
2. Currency crisis definitions 2.1 The large depreciation approach In this definition a currency crisis is identified when a currency depreciates significantly, where significantly varies among the multiple definitions. The definition is known in the economic literature as the “successful attack approach”. In their seminal paper Frankel and Rose (1996) identify a currency crisis when two conditions are met: (i) the depreciation of the nominal exchange rate of the currency is larger than 25% in a year, and (ii) the rate of the nominal depreciation must be 10 percentage points larger than it was in the previous year. In later research numerous authors have proposed variations. We show an overview in Appendix A. The first variation regards the choice to use the nominal exchange rate or the real exchange rate, with the latter calculated as follows: 𝑃 𝑈𝑆𝐴
𝑅𝐸𝑅𝑡 = 𝑒𝑡 𝑃𝑡𝐷𝑜𝑚 ,
(1)
𝑡
With RERt the real exchange rate; et the nominal exchange rate, expressed as the number of domestic currency units per foreign currency unit (in this case, the US dollar) so that et rises with a depreciation of the domestic currency; 𝑃𝑡𝐷𝑜𝑚 is the price level in the home or domestic country; 𝑃𝑡𝑈𝑆𝐴 is the price level in the reference country, in this case the USA. For the price levels the Consumer Price Index (CPI) is used. The advantage of the real exchange rate is that it eliminates the need to treat high-inflation countries separately. Additionally, changes in the real exchange rate reflect a balance-of-payments crisis better than changes in the nominal exchange rate, because these may be caused by inflation, and such an inflation-driven depreciation of the nominal exchange rate does not necessarily indicate a currency crisis (Kamin and Babson, 1999; Bussiere and Fratzscher, 2006). Esquivel and Larrain (1998) add that the nominal devaluation “has to be 4
meaningful, in the sense that it should affect the purchasing power of the domestic currency”. The second variation is related to how large and rapid the depreciation must be to qualify as a crisis. The choices for frequency (annual, quarterly, monthly and even shorter) and the threshold to classify crises (a fixed percentage, number of times the standard deviation, or quantile) are ultimately ad hoc. The lower the threshold, the more events are selected (with the danger of including events that should not be considered currency crises), and the higher the threshold, the stronger is the focus on the deep crises (with the disadvantages of missing out on currency crises, and having insufficient observations for analysis). In virtually all studies the depreciation is calculated versus the US dollar, but in some the depreciation is calculated versus the relevant anchor currency—US dollar, Japanese yen or the German DM (later euro). Pontines and Siregar (2007) find that crisis dating is sensitive to the choice of anchor currencies. An advantage of the large depreciation approach is that it is an intuitive and straightforward definition. From the viewpoint of an investor or manager of foreign reserves position this (simple) definition is more appropriate (Yu, Wang, Lai and Wen, 2010). A disadvantage is that the definition is focused exclusively on the exchange rate. When a country is able to defend its currency by using its reserves or increasing its interest rates, then there is no crisis according to the definition. However, ignoring this stress situation and labeling the period as tranquil is debatable. A second drawback is the ad hoc choice of the threshold above which a crisis is called is ad hoc. Should this be a 15%, 25% or 30% depreciation, or 2 or 2.5 times the standard deviation? There is no theoretical framework to support any choice; comparing the identified crises with the narrative is the most common argument. Using other studies’ threshold is another common approach.
5
2.2 Exchange Market Pressure approach The Exchange Market Pressure approach is based on the monetary model by Girton and Roper (1977). Eichengreen, Rose and Wyplosz (1995) (from hereon ERW1995) adopted this model to design an index to measure the exchange market pressure: the Exchange Market Pressure Index (EMPI). The index does not only take into account actual devaluation or depreciation of the currency, but also includes periods of great stress of the exchange rate. The latter occurs when the monetary authorities avoid a devaluation or depreciation through the use of its international reserves or by increasing the interest rates. Although the “currency attack” was unsuccessful, one may argue that this should be considered a crisis. In order to capture all attacks (successful and unsuccessful) the EMPI includes a weighted average of exchange rate changes, changes in the international reserve and changes in the interest rates. The weights are the inverse standard deviations of the individual elements, such that none of the components outnumbers the other(s) for having higher variance in general. The EMPI according to ERW1995 is defined as:
𝐸𝑀𝑃𝐼𝑡𝑖 ≡ 𝜎
1 ∆𝑒𝑡𝑖
𝑒𝑖
𝑒𝑡𝑖
−𝜎
1 𝑟𝑚𝑖
∆𝑟𝑚𝑖
( 𝑟𝑚𝑖𝑡 − 𝑡
∆𝑟𝑚𝑡𝐺𝐸𝑅 𝑟𝑚𝑡𝐺𝐸𝑅
)+𝜎
1
𝑟𝑖
∆(𝑟𝑡𝑖 − 𝑟𝑡𝐺𝐸𝑅 ) ,
(2)
where i refers to country i, t refers to time, is the difference operator, e is the nominal exchange rate, rm is the ratio of total non-gold international reserves to M1, r is the nominal short term interest rate (money market rate), e , erm and r are the standard deviations of
∆𝑒𝑡𝑖 𝑒𝑡𝑖
∆𝑟𝑚𝑖
, ( 𝑟𝑚𝑖𝑡 − 𝑡
∆𝑟𝑚𝑡𝐺𝐸𝑅 𝑟𝑚𝑡𝐺𝐸𝑅
) and ∆(𝑟𝑡𝑖 − 𝑟𝑡𝐺𝐸𝑅 ) respectively. The index GER is the
reference country Germany. They use quarterly data. If the index exceeds a predetermined threshold, then a currency crisis is identified. The choice of the threshold is arbitrary. ERW1995 use a threshold of two standard deviations from the mean, while Eichengreen, Rose and Wyplosz (1996) use a threshold of 1.5 standard deviations from the mean. Furthermore, they use an exclusion window of two quarters and one quarter respectively, to rule out measuring the same crisis more than once.
6
The EMPI has been used by the majority of the Early Warning Systems (EWSs) for currency crises. The Exchange Market Pressure approach is known in the economic literature as the “speculative pressure approach”. Many modifications have seen the light, which are shown in Appendix A and are described at continuation. For the exchange rate component either the real or nominal exchange rate is used (see Section 2.1 for arguments). A few studies have used the black market exchange rate instead of the official exchange rate. The black market exchange rate reflects the market pressure better than the official rate. This definition is used in country-specific studies such as Asici (2011) and Cerro and Meloni (2013). A challenge is to find these series for an extended period and with a consistent definition – particularly for a panel dataset this is complicated. The reserves component is simplified to the change in the reserves, instead of the changes in the differential of the reserves to M1 ratio. The interest rate component is originally taken as the change in the differential between the domestic nominal interest rate and the nominal interest rate in a country of reference. Most studies have simplified the component by using directly the change in the domestic nominal interest rate. For consistency, studies that use the real exchange rate also use the real interest rate instead of the nominal interest rate. Kaminsky, Lizondo and Reinhart (1998) do not include the interest rate in the formal expression. Their argument is that in emerging economies interest rate spreads are not always available or useful. The latter is particularly the case for Latin America: the sharp movement of interest rates in Latin American countries is often unrelated to speculative attacks. However, the omission of interest rates in the EMPI is increasingly recognized as a shortcoming in identifying crises (Berg, Borensztein and Patillo, 2005). For example, the 1995 attack on the Argentinian peso is deterred by a rapid increase in domestic interest rates (Hawkins and Klau, 2000). Also the crisis in Hong Kong in 1997 is not picked up if the change in the interest rate differential is not included (Klaassen and Jager, 2011). The components are integrated into a single index. Typically, each component is weighted with the inverse of the standard deviation, to avoid that one component
7
outnumbers the others for having higher variance. Some propose not to weight the components, e.g. Herrera and Garcia (1999), Kroska (2001) and Pontines and Siregar (2008). Angkinand, Li and Willett (2006) argue that when the weights are defined as the inverse standard deviation, then these weights could underestimate or overestimate the size of a change in exchange rate or reserves. As an example, consider a currency in a regime of narrow band fixed exchange rates. Almost all of the effects of incipient payments imbalances will fall on changes in reserves rather than changes in exchange rates. As a consequence, changes in reserves will have a much higher standard deviation, and are thus downplayed. The more attacks on the exchange rate were unsuccessful, the more understated is the change in the reserves. If the country were to abandon the narrow band peg, then the change in the nominal exchange rate would be relatively huge compared to the historical variation. This would overstate the magnitude of exchange rate market pressure. A simple average of the individual components would overcome this problem. Most studies use a monthly or quarterly frequency with the argument that currency crises unfold rapidly. Further, with annual data, one is unable to detect whether a crisis happened in the beginning or in the end of the year. For the threshold, the typical cut-off values are 1.5, 2, 2.5 and 3 standard deviations above the mean. This is either imposed, or a result of a comparison between the crisis dating with the narrative. Andreou, Dufrenot, Sand-Zantman and ZdzienickaDurand (2009) perform this in a more structural way, through a grid search, from 0 to 3 times standard deviation above the mean. The optimal threshold is the one where the identified crises coincide with the actual crises (narrative), which is 0.75 times the standard deviation for their sample. There are alternative ways to determine thresholds, or even avoid to determine these: Extreme Value Theory (a.o. Pontines and Siregar, 2008, Koedijk, Stork and de Vries, 1992, Pozo and Amuedo-Dorantes, 2003 and Lestano and Jacobs, 2007), Markov Switching method (a.o. Vlaar, 1999, Cerra and Saxena, 2000, Pontines and Siregar, 2008, Eichler, Karmann and Maltritz, 2009, Klaassen and Jager, 2011 and Qin and Liu, 2014), or using the continuous index (EMPI) as the dependent variable
8
for the crisis model (Sachs, Tornell and Velasco, 1996, Bussiere and Mulder, 1999, 2000, Nitithanprapas and Willett, 2000 and Kwack, 2000).
An advantage of the exchange market pressure approach is that the method works well under both fixed and floating exchange rate regimes (KLR1998). Frankel and Saravelos (2012) state that the inclusion of reserves is particularly relevant for countries with fixed exchange rate regimes, because capital flight and crisis incidence are present through larger drops in reserves rather than exchange rate weakness. Even in countries that have a de facto freely floating exchange rate, monetary policy makers sometimes intentionally try to influence the volatility of the exchange rates, such as Brazil and Mexico through official forex interventions in the fall of 2008. Also Moreno (1995) notes that policymakers as a rule seek to dampen large fluctuations in the exchange rate, for example to avoid disturbances in consumption due to fluctuations in the real exchange rate. Episodes of very large exchange rate movements may be interpreted as episodes of exchange market pressure that will be viewed with concern by authorities and may trigger a policy response, even in regimes where the exchange rate floats freely, known as “the fear to float” (Calvo and Reinhart, 2000). Therefore, the EMPI is important for monetary policy makers: to know the magnitude of exchange market pressure, to know how effective their instruments are to reduce the pressure, and to know how severe the pressure is on other currencies, which could give a hint on how much pressure they can expect on their own currency as a result of contagion (Klaassen and Jager, 2011). In the economic literature the exchange market pressure approach is the most common definition (Caramazza, Ricci and Salgado, 2000). Despite the widespread use of the EMPI as a crisis indicator, the approach is not without drawbacks. First, the threshold depends on the standard deviation in crisis identification based on the EMPI. When the movements in a particular currency crisis are relatively strong then these may “dwarf” other depreciations that should be marked as currency crises. Zhang (2001) notes that if the sample is large then a period with high volatility will dominate the whole sample, and the threshold will be too high to identify a
9
crisis that happened in the low volatility period. An example is Thailand in May 1997: the EMPI a la ERW1995 does not pick up this crisis, because after high volatility of the exchange rate and reserves in the 1980s, the 1990s were relatively tranquil. The problem can be resolved by using a (shorter) moving time window to calculate the standard deviation, but this may cause underestimation of the volatility. A second disadvantage of the exchange market pressure approach is the ad hoc choice of the threshold above which a crisis is called. A solution is not to convert the index into a binary variable, but to use the continuous exchange market pressure index, or to use Markov Switching.
3. Methodology Since currency crises vary in nature, and crisis dating is controversial, we use a three-step approach in which we combine several definitions, and compare the results with the narrative, expert opinions and publicly available databases. Additionally, we determine which combination of the six definitions that we employ fits our final set with crisis dates the best.
3.1 A three-step approach to determine currency crises
Step 1. Individual crisis definitions Following the recommendation of Pontines and Siregar (2008) we use several indicators to date currency crises. From the wide range of currency crisis definitions that exist we select six. We include definitions which differ in components (one, two or three components), measurements (nominal or real) and frequency (monthly, quarterly or annual). 1. Depreciation of the nominal exchange rate, threshold of 25%, and minimal 10 percentage points greater than the previous year (definition: Frankel & Rose, 1996)2. We calculate the annual depreciation in every month, comparing the
2
Laeven & Valencia (2012) use the same definition with one difference: the threshold is 30% instead of 25%.
10
nominal exchange rate in t with the nominal exchange rate 12 months before (so: t – 12). 2. Depreciation of the real exchange rate, thresholds: (i) quarterly > 20%, or (ii) 2 quarters > 30%, or (iii) 3-6 quarters > 40% (definition: Apoteker & Bartholemy, 2005). 3. Exchange Market Pressure Index with three components: (quarterly) changes in nominal exchange rate, reserves3 and interest rate differential. A quarterly EMPI, where a crisis is called when the EMPI exceeds a threshold of 2 standard deviations above the mean (definition: Eichengreen, Rose and Wyplosz, 1995). 4. Exchange Market Pressure Index (EMPI) with three components: (monthly) changes in real exchange rate, reserves and real interest rates; monthly EMPI, with a threshold 2 standard deviations above the mean (definition: Bussiere & Fratzscher, 2006, 2008). 5. EMPI with two components: (monthly) changes in nominal exchange rate and reserves; monthly EMPI, with a threshold 3 standard deviations above the mean (definition: Kaminsky, Lizondo and Reinhart, 1998). 6. EMPI with two components: (monthly) changes in real exchange rate and reserves; monthly EMPI, with a threshold 2 standard deviations above the mean (definition: Kamin & Babson, 1999).
Step 2: Pre-selection of the currency crises Although the majority of the definitions coincide in calling a crisis, there are differences in crisis dates. Following only one definition could result in incorrect crisis dates. As explained in Section 2 there is no single best indicator and both the large depreciations (definitions 1-2) and the definitions based on the EMPI (definitions 3-6) have drawbacks. The more indicators identify a crisis, the less sensitive the resulting dataset will be for changes in the definition. By demanding that all definitions identify an episode as a crisis
3
Since M1 is missing for nearly half of the countries in our sample, we decided to adjust ERW1995’s original formula: change in reserves instead of using the change in the reserves-to-M1 ratio differential.
11
we have a robust set of crises, but this will only contain the largest, very rare crises. A policy maker or investor will also be interested in less dramatic crises. On the other extreme, not setting a minimum condition would result in a very large number of currency crises (some of them merely hick-ups in the markets), and will vary greatly from one definition to another. We therefore have balanced the condition and found that a minimum of three definitions identifying a currency crisis is optimal4. The intuition for this choice is that a currency crisis is a large, relatively rare event related to the exchange rate. Since we include definitions which differ in components (one, two or three components), measurements (nominal, real) and frequency (monthly, quarterly, annual), we assure that only the more severe events (i.e. crises) are picked up. All definitions include the exchange rate, so a large depreciation in the exchange rate would be picked up by more than one definition. Similar reasoning holds for large changes in the reserves and interest rates, which are typically combined.
Step 3: Narrative and expert opinions With the identification of periods in which there is a significant depreciation and/or pressure on the exchange rate we turn to the narrative and to expert opinions, as recommended by a.o. Lestano and Jacobs (2007) and Babecky et al. (2012). Even with our condition that at least three definitions identify a crisis mistakes can be made. For example, the definitions that include reserves (indicators 3 to 6) could identify a crisis when the reserves drop sharply because a country pays off a large amount of foreign-denominated debt. Obviously this should not be considered a crisis. If interest rates are increased to slow down an overheated economy then this does not necessarily point towards a currency crisis. Next, we compare the crisis dates that we find in step 2 with the crisis dates from two often-used currency crisis dates, and two papers.
4
We compared datasets with the narrative and found that a criteria with one or two definitions yields many episodes which are not considered currency crises, and a more strict criteria in which at least four or more definitions identify a crisis leaves out various episodes that are considered currency crises.
12
-
Reinhart & Rogoff (2011) identify a crisis when depreciation vs. the reference country is 15% or more per year. The reference country is the USA, or the German mark (later euro).
-
Laeven & Valencia (2012) focus on years with an annual depreciation (local currency vs US dollar) greater than 30% annual.
-
Kaminsky (2006) uses the EMPI approach with two components –changes in nominal exchange rate and reserves–, with monthly data, and a threshold of 3 times the standard deviation above the mean.
-
Babecky et al. (2012) determine crisis episodes in two steps. First, they aggregated the “influential papers” (for currency crises: Kaminsky, 2006; Kaminsky and Reinhart, 1999; Laeven and Valencia, 2012; and Reinhart and Rogoff, 2011) into a binary index and assigned value 1 when at least one of them indicated an occurrence. Then they transformed annual data into quarterly. Second, this aggregated file was sent to the country experts for correction (Babecky et al., 2012). The included countries are OECD members only, which means that a large number of countries in our dataset is not included.
3.2 Determination of the best combination of definitions As an alternative for the simple decision rule as explained in step 2 of the previous section we determine which combination of definitions fits our final set of currency crisis dates best. Determining this definition serves three purposes. First, for communication purposes it is easier to define a currency crisis in a more precise way. Second, a combination of two or more definitions is useful to classify currency crises in other countries without having to go through the entire approach that we presented here. Third, we show that for our sample combinations perform better in identifying currency crises than any single definition. We pool the observations of all 35 countries and try different combinations, of one, two, three and four definitions. For each combination we construct a contingency
13
table (see Table 1), and select the combination with the lowest probability of type I and type II errors.
Table 1: Contingency table of crisis identification and realization Actual
Model
Crisis
No crisis
Crisis
A
C
No Crisis
B
D
Notes:
A is the number of observations in which the definition identifies a crisis that actually took place: correct crisis identification;
B is the number of observations in which the definition does not identify a crisis that actually took place: missed crisis, or type I error;
C is the number of observations in which the definition identifies a crisis that in reality did not take place: incorrectly identified crisis, or type II error;
D is the number of observations in which the definition does not identify a crisis that did not take place: correct non-crisis identification.
There are several criteria to determine the best combination. Because of the nature of crises (rare events) the number of non-crisis periods outnumbers the number of crisis periods greatly. As a consequence, the traditional criteria are biased – some towards type I errors, and others towards type II errors. An example of the first is the Total Misspecification Error (TME), which is calculated as the sum of the probabilities of type I and type II errors: 𝐵
𝐶
𝑃𝑟𝑜𝑏(𝑇𝑦𝑝𝑒 𝐼 𝑒𝑟𝑟𝑜𝑟) + 𝑃𝑟𝑜𝑏(𝑇𝑦𝑝𝑒 𝐼𝐼 𝑒𝑟𝑟𝑜𝑟) = 𝐴+𝐵 + 𝐶+𝐷
(3)
Since the number of crises (A+B) is much smaller than the number of non-crisis months (C+D) this measure automatically assigns a higher penalty to missed crises (B) than to incorrectly identified crises (C). The optimal combination according to this criteria will have a low number of missed crisis, but a very high number of incorrectly identified crises.
14
This characteristic can be useful for Early Warning Systems of financial crises, because the costs of missed crises is typically higher than the cost of false alarms, but for crisis identification it is not. Examples of criteria that are biased towards type II errors are the Noise-to-Signal ratio introduced by Kaminsky et al.
(1998)5,
defined as
𝐶⁄ (𝐶+𝐷) 𝐴⁄ (𝐴+𝐵)
, and the
𝐵+𝐶
unconditional probability of misclassification, defined as 𝐴+𝐵+𝐶+𝐷.
An alternative way to determine the best combination is to use the ROC curve, which graphically demonstrates the trade-off between type I and type II errors for each crisis definition, and is presented in Figure 1.
Figure 1: ROC curve, numerical example of combinations
Notes: On the x-axis the probability of incorrectly identified crises (type II error): 𝐶
(𝐶+𝐷), where a low value represents a low number of type II errors; 𝐴
On the y-axis the sensitivity (𝐴+𝐵), or 1 minus the probability of missed crises (type I error), where a high value represents a low number of type I errors. 5
15
The optimal solution is the combination in the top-left corner of Figure 1, marked by the red triangle and letter Z. This combination has zero probability of type I and type II errors, in other words, it perfectly identifies the crises. Unfortunately, usually this combination does not exist. Combinations with different type I and II errors, marked with purple dots, illustrate the criteria. Combination B is superior to combination A, because it has a higher percentage of correctly identified crises for the same probability of an incorrectly identified crisis. Combination B is also superior to combination C, because it has a higher percentage of correctly identified crises and a lower probability of false alarms. Likewise, combination D is superior to combinations C and E, and combination F is superior to combination E. Combination D is not superior to B, nor inferior to F as it depends on the relative costs of type I errors versus type II errors. Summarizing, the combinations B, D and F are superior to the other combinations. The choice among B, D and F depends on the relative costs of incorrectly identifying a crisis (type II error) and missing to identify a crisis (type I error). To give insight in the choice we use a weighted sum of type I and type II errors, where we vary the weights of the type I and type II errors. We denote the 𝐴+𝐵
unconditional probability of a crisis as 𝑃𝐶𝑅𝐼𝑆𝐼𝑆 = 𝐴+𝐵+𝐶+𝐷 . The weighted TME is: 𝐵
𝐶
𝑊 ∙ 𝑃(𝑡𝑦𝑝𝑒 𝐼 𝑒𝑟𝑟𝑜𝑟) + (1 − 𝑊) ∙ 𝑃(𝑡𝑦𝑝𝑒 𝐼𝐼 𝑒𝑟𝑟𝑜𝑟) = 𝑊 ∙ 𝐴+𝐵 + (1 − 𝑊) ∙ 𝐶+𝐷 ,
for 𝑊 = 𝑃𝐶𝑅𝐼𝑆𝐼𝑆 , 2𝑃𝐶𝑅𝐼𝑆𝐼𝑆 , 3𝑃𝐶𝑅𝐼𝑆𝐼𝑆 , 4𝑃𝐶𝑅𝐼𝑆𝐼𝑆 , … With a higher weight we assign more value to type I errors, and with a lower weight we assign more weight to type II errors.
4. Data The choice of the countries is a trade-off. Including more countries implies more observations, and more crisis observations. This leads to more stable thresholds (fewer jumps in type I and type II errors), but on the negative side we have a more heterogeneous group, which can be less representative. Including less countries has the 16
benefit that the sample is more homogenous and thresholds probably more reasonable, but will lead to less (crisis) observations, causing thresholds to be less stable. We select 35 emerging economies with similar features: emerging, open economies, with medium income per capita during part of the time horizon. In these countries capital markets exist, and each country has its own currency. Commodities are important for the majority of these economies. The countries are presented in Table 2.
Table 2: country selection Latin America &
CEE6 and CIS7
Asia
Caribbean
Rest of the world
1
Argentina
Bulgaria
India
Egypt
2
Bolivia
Croatia
Indonesia
South Africa
3
Brazil
Czech Republic
Korea
Turkey
4
Chile
Hungary
Malaysia
5
Colombia
Kazakhstan
Pakistan
6
Costa Rica
Poland
Philippines
7
Dominican Republic
Romania
Sri Lanka
8
Guatemala
Russia
Thailand
9
Jamaica
Ukraine
Vietnam
10
Mexico
11
Paraguay
12
Peru
13
Uruguay
14
Venezuela
We focus on data from the period 19908 to 2016. Starting in 1970 or 1980 would lead to more (crisis) observations, but also more gaps in the data due to hyperinflation (Latin America) and geopolitical structural breaks (Central and Eastern Europe before 1990). Our
6
Central & Eastern Europe Commonwealth of Independent States (former Soviet Union) 8 Some countries start later due to data availability and/or periods with hyperinflation. These are: Argentina (starts in 1992), Brazil (1995), Bulgaria (1998), Croatia (1994), Czech Republic (1994), Kazakhstan (1995), Peru (1992), Poland (1991), Romania (1992), Russia (1995), Ukraine (1996), Vietnam (1992). 7
17
choice to start in 1990 is based on two arguments: (i) data availability, and (ii) most emerging economies started market liberalizations in the late 1980s / early 1990s. For the six Central and Eastern European countries (Bulgaria, Croatia, Czech Republic, Hungary, Poland and Romania) we use the Eurozone (after 1999) and Germany (prior to 1999) as the reference country – both for the exchange rate, the interest rate differential and the inflation (for the real exchange rate calculations). For all other countries we use the USA as the country of reference. We use monthly, quarterly and annual data–all with a monthly frequency (e.g. every month we calculate the annual depreciation in the trailing 12 months and the quarterly depreciation in the trailing 3 months). The variables are: -
Nominal exchange rate versus US dollar (and versus DEM or euro for the CEE countries) at the end of the month.
-
Consumer Price Index for all items9.
-
Real exchange rate is constructed10: 𝑅𝐸𝑅𝑖 = 𝑁𝐸𝑅𝑖 ∙ 𝑅𝐸𝑅𝑖 = 𝑁𝐸𝑅𝑖 ∙
𝐶𝑃𝐼𝐸𝑈𝑅𝑂 𝐶𝑃𝐼𝑖
𝐶𝑃𝐼𝑈𝑆𝐴 𝐶𝑃𝐼𝑖
and
for the CEE countries.
-
International reserves excluding gold (measured in current US dollar).
-
Money market interest rate, or Treasury bill rate if the first is missing. The correlation between these series is on average 0.88 for the 21 countries that contain both series.
-
In the case of missing observations in five or less consecutive months, we used linear interpolation. If six or more consecutive months are missing, then these months remain N/A.
9
For Argentina we took the CPI from the World Bank’s Global Financial Database up to 2013. We converted to annual index into a monthly frequency with the cubic match method. For 2014 and 2015 we took the (monthly) CPI from IFS. For Chile and Venezuela we expanded the series from 2009-2015 resp. 2007-2015 backwards, using the annual inflation percentages. 10 We decided not to use the Real Effective Exchange Rate (REER) as reported by IFS because of large gaps of data unavailability.
18
All variables are taken from the IFS-database. For the standard deviation (used in the weights and the threshold for the EMPI approach) we use the entire period (1990M1 to 2016M1211. We use a window exclusion period of 12 months, which means that if a crisis is called within 12 months from the previous crisis call, then the second crisis is considered a continuation of the first and therefore not considered as a different currency crisis.
5. Empirical results In this section we present the results of the three-step process to determine currency crises (Section 5.1), and we compare the final set with (combinations of) currency crisis definitions to determine a new definition in Section 5.2.
5.1 The three-step approach to determine currency crisis dates
Steps 1 and 2: Crisis dating according to alternative definitions We apply the six crisis definitions to 35 emerging economies (step 1), and identify episodes in which at least three definitions date a crisis (step 2). The results are shown in Table 3. Table 3: Currency crisis dates according to different definitions Country
Crisis
1
2
3
4
5
6
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
Latin America and Caribbean
11
Argentina
1995M3
Argentina
2002M1-6
V
V
Argentina
2014M1
V
V
Argentina
2015M12-2016M2
V
V
Brazil
1998M9
Brazil
1999M1-2
Brazil
2000M4
V
V V
Our results are robust for the use of a smaller sample period.
19
V
V
V
Brazil
2002M7-9
V
V
Brazil
2008M11
V
V
V
Chile
2008M10
V
V
V
Colombia
1995M5-8
Colombia
1998M6-9
Colombia
2008M9-10
Colombia
2015M7-8
V
V
Dominican Republic
1990M8
V
V
Dominican Republic
2003M5-7, 10, 2004M1
V
V
Guatemala
1990M8-10
V
V
Jamaica
1991M5-12
V
V
V
Mexico
1994M12-1995M3
V
V
V
Mexico
1995M10-11
V
V
V
Mexico
2008M10, 2009M2
V
V
Paraguay
2002M5-7
V
V
Paraguay
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
2008M10
V
V
V
V
Peru
2008M10
V
V
Uruguay
2002M2, 6-8
V
V
V
V
Venezuela
1994M5-6
V
Venezuela
1995M12
Venezuela
V
V
V V
V
V
V
V
V
V
V
V
1996M4
V
V
V
V
V
V
Venezuela
2002M2-5
V
V
V
V
V
V
Venezuela
2003M1
V
V
V
V
V
V
Venezuela
2009M1
V
V
V
Venezuela
2010M1
V
V
V
Venezuela
2011M1
V
V
V
V
V
V
Venezuela
2013M2
V
V
V
V
V
V
Venezuela
2016M2
V
V
V
Central and Eastern Europe, and Commonwealth of Independent States Bulgaria
1998M8
V
V
Croatia
1999M2-3
V
V
Croatia
2001M8
Croatia
2009M1
V
Czech Republic
1997M5
V
V
20
V
V
V V
V
V
V
V
V
V
Czech Republic
1998M8
V
V
V
Hungary
1995M3
V
V
V
Kazakhstan
1999M4-6
V
V
V
V
Kazakhstan
2009M2-3
V
V
V
V
Kazakhstan
2015M8-2016M1
V
V
V
V
Poland
1991M4-5
Poland
2008M10-2009M2
V
V
Romania
1992M3-6
V
V
Romania
1997M1-2
V
V
Romania
1999M11
V
Russia
1998M8-1999M1
V
V
Russia
2009M1-2
V
V
V
V
Russia
2014M12-2015M1
V
V
V
V
Ukraine
1998M9-10, 1999M3
V
V
V
V
V
V
Ukraine
2008M10-12
V
V
V
V
V
V
Ukraine
2014M2-2015M2
V
V
V
V
V
V
India
1991M4, 7
V
V
V
V
V
V
India
1993M3
V
V
V
India
2008M9-11
V
Indonesia
1997M8-1998M6
V
V
Indonesia
2008M10-11
V
V
Korea
1997M11-1998M2
V
V
Korea
2008M10-11
V
Malaysia
1997M7-1998M1
V
Malaysia
2015M8
V
Pakistan
1993M7-9
Pakistan
1995M10
Pakistan
1996M10
Pakistan
1999M5
Pakistan
2000M9-10
Philippines
1990M11-1991M1
Philippines
1997M7-12
V
Philippines
2000M10
V
V V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
Asia
V V V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
21
V
V
V
V
V V
V V
Sri Lanka
2012M2-4
V
Thailand
1997M7-1998M1
Vietnam Vietnam
V
V
V
V
V
V
V
V
1998M2, 8
V
V
V
V
2009M11
V
V
V
V
Rest of the world Egypt
1990M7
V
V
V
V
Egypt
1991M3
V
V
V
V
Egypt
2003M1-4
V
V
V
V
Egypt
2016M11
V
V
V
V
V
V
South Africa
1996M4
V
V
V
V
South Africa
1998M6-8
V
V
V
V
V
South Africa
2001M12
V
V
V
V
V
South Africa
2008M10, 2009M1
V
V
V
V
V
V
Turkey
1994M1-4
V
V
V
V
V
V
Turkey
2001M2-4
V
V
V
V
V
V
Turkey
2008M10
V
V
V
V
V
Notes: Condition: Minimal three definitions indicate a crisis episode. The length of the crisis can be extended to various months in case the crisis unfolds or lasts more than one adjacent month.
Step 3: Comparison with narrative and expert opinions We compare the list of potential crisis dates from Table 3 with the narrative (Appendix B). For most crises (68 out of 88) the narrative confirms the crisis dates that were determined with the quantitative definition. We discard 20 (out of 88) crises. We exclude 4 (out of 36) crises from Latin America: Brazil 2000, Colombia 1995, Peru 2008, and Venezuela 2009; 8 (out of 21) from CEE and CIS: Bulgaria 1998, Croatia 1999, 2001 and 2009, Czech Republic 1998, Hungary 1995, Poland 1991 and Romania 1992; 8 (out of 21) from Asia12: India 1993, Pakistan 1993, 1995, 1996, 1999 and 2000, and Vietnam 1998 and 2009; and 0 (out of 10) from the
12
Excluding Pakistan and Vietnam, only 1 (out of 14) crisis would be discarded
22
rest of the world. The reasons for discarding are diverse. In most cases the change in the exchange rate, reserves and/or interest rates is technically sufficient to call a crisis, but according to the narrative it is either part of a long-lasting, consistent policy to use small devaluations to maintain international competitiveness (Pakistan’s five crises, Vietnam’s two crises, Hungary 1995) or the threshold is so low (due to low historical volatility) that even a minor depreciation leads to a crisis identification (India 1993, Peru 2008, Bulgaria 1998, Croatia 1999, 2001 and 2009). Another reason to discard is a large drop in reserves that is not used to defend the currency but to pay off debt (Brazil 2000) or to deposit in a development fund (Venezuela 2009). In some cases the events should be considered as deep recessions (Colombia 1995, Czech Republic 1998) instead of currency crises. Lastly, there are some cases where the devaluation occurs at the end of the hyperinflation period (Poland 1991, Romania 1992). We note that discarded crises are dominated by EMPIdefinitions, 16 out of 20 discarded crises are identified exclusively with three or four EMPI definitions. We include two crises that our approach does not pick up13: Brazil 2015 and Colombia 2002. These events should be considered currency crises according to the narrative. Finally, we apply the window exclusion period of 12 months. For example, Egypt’s 1990M7 and 1991M3 crises are considered as a single crisis dated 1990M7 up to and including 1991M3. The resulting set of data is shown in Table 4.
Table 4: Resulting crisis dates Country
Crisis episode
Latin America and Caribbean: 29 crises in 11 countries (2.6 average) and 3 countries without crises Argentina
1995M3, 2002M1-6, 2014M1, 2015M12-2016M2
Bolivia
None
Brazil
1998M9-1999M2, 2002M7-9, 2008M11, 2015M7-9
13
In both cases only two out of six definitions identify these as crises. The currency crisis in Colombia 2002 is also identified by Reinhart and Rogoff, 2011.
23
Chile
2008M10
Colombia
1998M6-9, 2002M12-2003M2, 2008M9-10, 2015M7-8
Costa Rica
None
Dominican
1990M8, 2003M5-2004M1
Republic Guatemala
1990M8-10
Jamaica
1991M9-12
Mexico
1994M12-1995M10, 2008M10-2009M2
Paraguay
2002M6, 2008M9-12
Peru
None
Uruguay
2002M6-7
Venezuela
1994M5-6, 1995M12–1996M4, 2002M2–2003M1, 2010M1, 2011M1, 2013M2, 2016M3
Central and Eastern Europe and Commonwealth of Independent States: 13 crises in 6 countries (2.2 per country) and 3 countries without crises Bulgaria
None
Croatia
None
Czech
1997M5-7
Republic Hungary
None
Kazakhstan
1999M4-5, 2009M2-3, 2015M8-2016M1
Poland
2008M10-2009M2
Romania
1997M1-2, 1999M11
Russia
1998M8-11, 2009M1-2, 2014M11-2015M1
Ukraine
1998M9-1999M3, 2008M10-12, 2014M2-2015M2
Asia: 13 crises in 8 countries (1.6 per country) and 2 countries without crises India
1991M4-7, 2008M9-11
Indonesia
1997M8-1998M6, 2008M10-11
Korea
1997M11-1998M2, 2008M10-11
Malaysia
1997M7-1998M1, 2015 M8
Pakistan
None
Philippines
1990M11-1991M1, 1997M7-12, 2000M10
Sri Lanka
2012M2-3
24
Thailand
1997M7-1998M1
Vietnam
None
Rest of the world: 10 crises in 3 countries Egypt
1990M7-1991M3, 2003M1-M4, 2016M11
South Africa
1996M4, 1998M6-8, 2001M12, 2008M10-2009M1
Turkey
1994M1-4, 2001M2-4, 2008M10
We identify 65 episodes with currency crises in 27 countries. There are eight countries without currency crises in the period 1990-2016: Bolivia, Costa Rica, Peru, Bulgaria, Croatia, Hungary, Pakistan and Vietnam.
Lastly, we compare our crisis dates with the crisis dates from Reinhart and Rogoff (2011), Laeven and Valencia (2012) and two other papers, shown in Table 5. Notice that these external databases do not coincide in the crisis dating amongst themselves. In the narrative in Appendix B we also include crises that are identified by these external data sources.
Table 5: External data bases with currency crisis dating; starting in 1990 Reinhart &
Laeven &
Babecky et
Kaminsky
Rogoff (2011)
Valencia
al. (2012)
(2006)
Boonman (2017)
(2012) Frequency
Annual
Annual
Quarterly
Monthly
Monthly
Period
1800-2010
1970-2011
1970Q1-
1970M1-
1990M1 - 2016M12
2011Q4
2002M12
covered Latin America and Caribbean Argentina
2002
2002
2002M2
(1992-)
1995M3, 2002M1-6, 2014M1, 2015M12-2016M2
Bolivia
None
None
None
None
Brazil
1999, 2001-
1999
1999M1
1998M9-1999M2, 2002M7-9,
(1995-)
02, 2008
Chile
2008
2008M11, 2015M7-9 None
None
25
None
2008M10
Colombia
1980-91,
None
1995M8,
1998M6-9, 2002M12-2003M2,
1995, 1997-
1997M9,
2008M9-10, 2015M7-8
2000, 2002
1998M9, 1999M8
Costa Rica
1991
1991
None
Dominican
1990, 2002-03
1990, 2003
1990M8, 2003M5-2004M1
None
None
1990M8-10
1991
1991M5-12
Republic Guatemala Jamaica Mexico
1994-95,
1995
1998,
1994Q1-
1994M12
1995Q4
1994M12-1995M10, 2008M102009M2
2008 Paraguay
1992-93,
2002
2002M6, 2008M9-12
1998-99, 2001-02 Peru (1992-)
1987-93, 1998
None
None
None
Uruguay
1974-97,
1990, 2002
None
2002M2-8
1989-96,
1994,
1994M5,
1994M5-6, 1995M12-1996M4,
2002, 2004,
2002, 2010
1995M12
2002M2-2003M1, 2010M1,
2001-02 Venezuela
2010
2011M1, 2013M2, 2016M3
Central and Eastern Europe, and Commonwealth of Independent States Bulgaria
None
None
None
(1998-) Croatia
None
None
(1994-) Czech Rep
None
1997Q2-3
1997M5-7
None
None
None
(1994-) Hungary
1991, 1993, 1995-97, 1999
Kazakhstan
1999
1999M4-6, 2009M2-3, 2015M8-2016M1
26
Poland
1984-93,
(1991-)
1995-97,
None
None
2008M10-2009M2
1996
1996Q1-
1997M1, 1999M11
1999, 2008 Romania
1990-2001,
(1992-)
2008, 2010
Russia
1987-96,
(1995-)
1998-99, 2008
Ukraine
1997Q1 1998
1998M8-11, 2009M1-2, 2014M12-2015M1
1998, 2009
1998M9-1999M3, 2008M10-12,
(1995-)
2014M2-2015M2
Asia India
1991, 1993,
None
1991M4-7, 2008M9-11
2008 Indonesia
1997-98,
1998
1997M12,
2000, 2008 Korea
1997, 2008
1997M8-1998M6, 2008M10-11
1998M1 1998
1998Q1-4
1997M11-1998M2, 2008M10-11
Malaysia
1997
1998
1997M8-
1997M7-1998M1, 2015M8
1998M6 Pakistan Philippines
None 1990, 1997,
None
1998
1997M12
2000
1990M11-1991M1, 1997M7-12, 2000M10
Sri Lanka
None
None
Thailand
1997, 2000
1998
2012M2-3 1997M7-
1997M7-1998M1
1998M1, 1999M92000M7 Vietnam
None
None
1990
1990M7- 1991M3, 2003M1-4,
(1992-) Rest of the world Egypt
1989-91, 2001, 2003
South Africa
1996, 1998,
2016M11 None
1996M4, 1998M6-8, 2001M12,
2000-01, 2008
2008M10-2009M1
27
Turkey
1977-2001,
1991,
1991Q1-4,
1994M3,
1994M1-4, 2001M2-4,
2008
1996, 2001
1994Q1-4,
2001M2
2008M10
1996Q1-4, 2001Q1-4
Notes: Dates in red & itallic identify crises that we have not identified. In the narrative these dates are included, as to explain why these should not be considered currency crises. Dates in blue & bold identify crises that we identify but none of the peer studies does. Also for these we refer to the narrative for further explanations. None means that the country is included in the sample, but no crisis is identified. The last column is taken from Table 4.
Although a direct comparison is complicated because the country selection and time periods are distinct14, we mention some stylized facts of the comparison: -
The large crises are picked up by all studies. E.g. Argentina 2002, Brazil 1999, Mexico 1994-95, Russia 1998, Asia 1997, Turkey 2001.
-
Reinhart and Rogoff (2010) have a relatively low threshold to define crises and as a consequence identify a large number of crises.
-
Babecky et al. (2012) include few emerging countries and identify a very low number of crises. Remarkable is that no crisis in 2008-09 is identified.
-
Kaminsky (2006) coincides relatively most with our crisis dates for the fourteen countries in both samples: 15 crises in common, 9 different.
5.2 Determination of the best combination of definitions There are 170 crisis months and 11,170 non-crisis months, or 1.52% of all observations is a crisis month. The average length of each of the 65 crisis episodes is 2.6 months. We compare the quantitative results according to the six definitions with our final dataset of currency crises, shown in Table 6. For each definition we calculate the
14
Particularly the end date of the alternative databases is far shorter than our sample. Only two databases include crises up to an including 2011.
28
probability of type I and II errors, the total misspecification error, the noise-to-signal, the total probability of misspecification and the weighted TME, for different weights of type I and type II errors. Definition #5 is optimal when type I errors have a relatively low cost, definition #1 is optimal when type I errors have a relatively high cost –and are thus less common, although at the cost of more type II errors.
Table 6: Measures of fit. Individual definitions #A #B #C #D %A %B %C %D Prob (type I error) Prob (type II error) TME (unweighted) Noise-to-Signal Total Misspecif. Prob. Weighted TME, W=1*UPM Weighted TME, W=2*UPM Weighted TME, W=3*UPM Weighted TME, W=4*UPM Weighted TME, W=5*UPM Weighted TME, W=6*UPM Weighted TME, W=7*UPM Weighted TME, W=8*UPM Weighted TME, W=9*UPM Weighted TME, W=10*UPM
#1 132 38 350 10,820 1.2% 0.3% 3.1% 95.4% 22.4% 3.1% 25.49% 4.04% 3.42% 3.43% 3.72% 4.01% 4.30% 4.60% 4.89% 5.18% 5.47% 5.77% 6.06%
#2 129 41 353 10,817 1.1% 0.4% 3.1% 95.4% 24.1% 3.2% 27.28% 4.16% 3.47% 3.48% 3.80% 4.12% 4.44% 4.76% 5.07% 5.39% 5.71% 6.03% 6.35%
#3 108 62 194 10,976 1.0% 0.5% 1.7% 96.8% 36.5% 1.7% 38.21% 2.73% 2.26% 2.27% 2.79% 3.32% 3.85% 4.38% 4.91% 5.44% 5.97% 6.49% 7.02%
#4 85 85 139 11,031 0.7% 0.7% 1.2% 97.3% 50.0% 1.2% 51.24% 2.49% 1.98% 1.99% 2.73% 3.47% 4.21% 4.95% 5.70% 6.44% 7.18% 7.92% 8.66%
#5 76 94 48 11,122 0.7% 0.8% 0.4% 98.1% 55.3% 0.4% 55.72% 0.96% 1.25% 1.26% 2.10% 2.93% 3.77% 4.60% 5.44% 6.27% 7.11% 7.94% 8.78%
#6 107 63 148 11,022 0.9% 0.6% 1.3% 97.2% 37.1% 1.3% 38.38% 2.11% 1.86% 1.87% 2.41% 2.96% 3.50% 4.04% 4.59% 5.13% 5.68% 6.22% 6.76%
Notes:
#A is the number of correctly identified crises; %A is the percentage of correctly identified crises over the total number of observations: %A = A / (A + B + C + D) ;
#B is the number of missed crises (type I error);
#C is the number of incorrectly identified crises (type II error);
#D is the number of correctly identified non-crises;
Probability of a type I error is calculated by B/(A+B);
Probability of a type II error is calculated by C/(C+D);
TME = Total Misspecification Error = Probability of a type I error + Probability of a type II error = 𝐵
𝐶
𝐵
𝐶
= 𝐴+𝐵 + 𝐶+𝐷 = 170 + 11,170 = 0.005882 ∙ 𝐵 + 0.000089 ∙ 𝐶
29
𝐶⁄ (𝐶+𝐷)
Noise-to-Signal ratio = 𝐴 ⁄
(𝐴+𝐵)
=
𝐶⁄ 11,170 𝐴⁄ 170
𝐶
= 0.0152 ∙ 𝐴 𝐵+𝐶
Unconditional probability of misclassification=
Weighted Total Misspecification Error, 𝑊𝑇𝑀𝐸 = 0.0152 ∙ 𝐵
𝐴+𝐵+𝐶+𝐷
𝐵+𝐶
= 11,340 = 0.000088 ∙ (𝐵 + 𝐶). 𝐵 𝐶 + 0.9848 ∙ 𝐶+𝐷 𝐴+𝐵
𝑓𝑜𝑟 𝑊 =
𝐶
0.0152; 0.0304 ∙ 𝐴+𝐵 + 0.9696 ∙ 𝐶+𝐷 𝑓𝑜𝑟 𝑊 = 2 ∙ 0.0152, 𝑒𝑡𝑐.
Next, we proceed with combining definitions, to see whether it is possible to improve the fit. We try all possible combinations consisting of 2, 3, or 4 definitions, and apply different conditions. Figures 2.A and 2.B illustrate the results. The combinations in Figure 2.A do not have any condition, while the combinations in Figure 2.B do. In Figure 2.A we note that using one indicator (group A; blue triangles) yields low type II error probabilities, but also a low probability of correctly identified crises. Using more than one definition increases the probability of correctly identified crises, but at the cost of higher probability of a type II error.
Figure 2.A: Probability of type I and type II errors for combinations without conditions
Notes:
30
The probability of a type II error (on the x-axis) and the probability of correctly identifying a crisis (1 – probability of a type I error; on the y-axis). The observations in the figure are divided in four categories. A crisis is dated if: A = one definition identifies a crisis (as in Table 6) B = using two definitions: at least one definition identifies a crisis C = using three definitions: at least one definition identifies a crisis D = using four definitions: at least one definition identifies a crisis
Figure 2.B: Probability of type I and type II errors for combinations with conditions
Notes: A crisis is dated if: E = using two definitions: both definitions identify a crisis F = using three definitions: all three definitions identify a crisis G = minimal one definition from group 1 (definitions 1 and 2) identifies a crisis, and minimal one from group 2 (definitions 3 to 4) identifies a crisis H = minimal one definition from group 1 identifies a crisis, and minimal two from group 2 identify a crisis 31
I = both definitions from group 1 identify a crisis, or minimal two from group 2 identify a crisis Note that A-D is shown to demonstrate that these combinations are inferior to the combinations with the conditions.
In Figure 2.B we mark the two combinations that yield the best results. Table 7 presents details on the performance for these combinations. Compared to the performance of the individual definitions (Table 6), we conclude that these selected combinations yield better results. Selecting the overall best combination depends on the relative importance that one puts on type I and type II errors. For completeness we also show the performance with the simple decision rule as used in step 2 of the approach to identify currency crises.
Table 7: Measures of fit. Selection of combinations Combination Description #A #B #C #D A% B% C% D% Prob. (error type I) Prob. (error type II) TME (unweighted) Noise-to-Signal Total Misspecif. Prob. Weighted TME, W=1*UPM Weighted TME, W=2*UPM Weighted TME, W=3*UPM Weighted TME, W=4*UPM Weighted TME, W=5*UPM Weighted TME, W=6*UPM Weighted TME, W=7*UPM Weighted TME, W=8*UPM Weighted TME, W=9*UPM Weighted TME, W=10*UPM
(i) (ii) (#1 or 2) and (#3, (#1 and #2) or 4, 5 or 6) (minimal two from #3, 4, 5 or 6) 131 39 60 11110 1.16% 0.34% 0.53% 97.97% 22.94% 0.54% 23.48% 0.70% 0.87% 0.88% 1.22% 1.56% 1.90% 2.24% 2.58% 2.92% 3.26% 3.61% 3.95%
152 18 203 10967 1.34% 0.16% 1.79% 96.71% 10.59% 1.82% 12.41% 2.03% 1.95% 1.95% 2.08% 2.22% 2.35% 2.48% 2.62% 2.75% 2.89% 3.02% 3.15%
signal: minimal 3 136 34 50 11120 1.20% 0.30% 0.44% 98.06% 20.00% 0.45% 20.45% 0.56% 0.74% 0.75% 1.04% 1.34% 1.64% 1.94% 2.23% 2.53% 2.83% 3.13% 3.42%
Based on our findings we propose two new definitions, one with a preference for a low probability of type II errors (at the cost of a higher probability of type I errors), and one 32
with a preference for a low probability of type I errors (at the cost of a higher probability of type II errors). Definition (i): (i)
A large depreciation, in real or nominal terms;
and (ii)
at least one EMPI exceeding the threshold, where the EMPI can be with quarterly or monthly frequency, consist of exchange rate and reserves, or exchange rate, reserves and interest rate, and can use real or nominal values.
Definition (ii): (i)
A large depreciation, both in real and in nominal terms;
or (ii)
at least two EMPI´s exceeding the threshold, where the EMPI can be with quarterly or monthly frequency, consist of exchange rate and reserves, or exchange rate, reserves and interest rate, and can use real or nominal values.
6. Conclusion Identifying currency crises is the first step in empirical work related to currency crises, whether it is for Early Warning Systems, analyzing the impact of currency crises, or contagion. Unfortunately, there is no consensus on the conceptual definition, nor on the operational definition. We propose the use of a combination of quantitative definitions. Since there is no official dataset with currency crisis dates we compare our outcomes with the narrative and expert opinions, as suggested by Lestano and Jacobs (2007). We have identified and dated currency crises for 35 emerging economies, for the period 1990-2016. Compared to other studies we have included more emerging economies with high frequency, we have updated (up to and including 2016) the dataset, and we have shown the narrative in a transparent way. We find that a simple decision rule (a crisis is dated when at least three out of six definitions call it a crisis) yields good results: 33
the large majority of the crises identified by the quantitative definitions is confirmed by the narrative and expert opinions. No single definition yields similar good results. This provides support for the suggestion from Pontines and Siregar (2008) that the use of a combination of definitions is recommendable. Additionally we propose two new definitions that consist of a combination of existing definitions. These fit the narrative better than any of the individual definitions do. The first definition is a large depreciation, in real or nominal terms, and at least one of the four exchange market pressure indices identifying a crisis. This definition yields low type II errors. The second definition is a large depreciation, both in real and in nominal terms, or at least two of the four exchange market pressure indices identifying a crisis. This definition yields low type I errors. We contribute to the literature in two ways. First, we show that a combination of definitions yields better results than relying on a single definition. Second, we offer more transparency in the crisis dating, by showing the narrative for a large number of emerging economies.
References
Alvarez-Plata, Patricia and Mechtild Schrooten, 2004. “Misleading indicators? The Argentinean currency crisis”, Journal of Policy Modeling, Vol. 26(5), 587–603. Andreou, Irène, Gilles Dufrénot, Alain Sand-Zantman, and Aleksandra ZdzienickaDurand, 2009. “A Forewarning Indicator System for Financial Crises: the Case of Six Central and Eastern European Countries”, Journal of Economic Integration, Vol. 24(1), 87-115, March. Angkinand, Apanard, Jie Li, and Thomas Willett, 2006. “Measures of Currency Crises: A Survey”, mimeo Claremont Graduate University. Apoteker, Thierry and Sylvain Barthelemy, 2005. “Predicting financial crises in emerging markets using a composite non-parametric model”, Emerging Markets Review, Vol. 6, 363 – 375. Ari, Ali, 2012. “Early warning systems for currency crises: The Turkish case”, Economic Systems, Vol. 36, 391–410. Ari, Ali and Raif Cergibozan, 2016. “The Twin Crises: Determinants of Banking and Currency Crises in the Turkish Economy”, Emerging Markets Finance and Trade, Vol. 52 (1), 123-135.
34
Asici, Ahmet Atil, 2011. “Exchange rate regime choice and currency crises”, Economic Systems, vol. 35, 419–436. Aziz, Jahangir, Francesco Caramazza and Ranil Salgado, 2000. “Currency crises: in search of common elements”, IMF Working Papers, 00/67, International Monetary Fund. Babecký, Jan, Tomáš Havránek, Jakub Matějů, Marek Rusnák, Kateřina Šmídková and Bořek Vašíček, 2012. “Banking, debt and currency crises early warning indicators for developed countries”, ECB Working Papers No. 1485. Berg, Andrew, and Catherine Pattillo, 1999. "Are Currency Crises Predictable? A Test", IMF Staff Papers, Palgrave Macmillan Journals, vol. 46(2). Berg, Andrew, Eduardo Borensztein and Catherine Pattillo, 2005. “Assessing Early Warning Systems: How Have They Worked in Practice?”, IMF Staff Papers, Palgrave Macmillan Journals, Vol. 52 (3), 462-502. Bauer, Christian, Bernhard Herz, Volker Karb, 2007. “Are twin currency and debt crises special?”, Journal of Financial Stability, Vol. 3, 59–84. Brüggemann, Axel and Thomas Linne, 2002. “Are the Central and Eastern European Transition Countries still vulnerable to a Financial Crisis? Results from the Signals Approach”, IWH-Discussion Papers No.157. Bucevska, Vesna, 2015. “Currency Crises in EU Candidate Countries: An Early Warning System Approach”, Panoeconomicus, Vol. 62(4), 493-510. Bussiere, Matthieu, and Christian Mulder, 1999. "External Vulnerability in Emerging Market Economies - How High Liquidity Can Offset Weak Fundamentals and the Effects of Contagion", IMF Working Papers 99/88, International Monetary Fund. Bussiere, Matthieu, and Christian Mulder, 2000. "Political Instability and Economic Vulnerability," International Journal of Finance & Economics, vol. 5(4), 309-30. Bussiere, Matthieu and Marcel Fratzscher, 2006. “Towards a New Early Warning System of Financial Crises”, Journal of International Money and Finance, 25(6), 953–973. Bussiere, Matthieu and Marcel Fratzscher, 2008. “Low probability, high impact: Policy making and extreme events”, Journal of Policy Modeling, Vol. 30, 111–121. Bussiere, Matthieu, 2013a. “Balance of Payment Crises in Emerging Markets: How Early Were the ‘Early’ Warning Signals?”, Applied Economics, Vol. 45(12), 1601– 1623. Bussiere, Matthieu, 2013b. ”In Defense of Early Warning Signals”, Banque de France Working Paper, No. 420. Bussiere, Matthieu, Sweta C. Saxena and Camilo E. Tovar, 2012. “Chronicle of currency collapses re-examining the effects on output”, Journal of International Money and Finance, Vol. 31, 680-708. Calvo, Guilllermo A. and Carmen M. Reinhart, 2000. “Fear of floating”, NBER Working Paper, No. 7993, November 2000.
35
Candelon, Bertrand, Elena-Ivona Dumitrescu and Christophe Hurlin, 2014. “Currency crisis early warning systems: Why they should be dynamic”, International Journal of Forecasting, vol. 30, 1016–1029. Caramazza, Francesco, Luca Ricci and Ranil Salgado, 2000. “Trade and financial contagion in currency crises”, IMF working paper, 00/55. Catao, Luis A.V., 2006. “Sudden Stops and Currency Drops: A Historical Look”, IMF Working Paper, 06/133. Cerra, Valeria and Sweta Chaman Saxena, 2000. “Contagion, monsoons, and domestic turmoil in Indonesia: a case study in the Asian currency crisis”, IMF Working Paper 00/60. Cerro, Ana María and Osvaldo Meloni, 2013. “Crises and Crashes: Argentina 18252002”, Revista de Historia Económica, vol. 31(2), 219 – 252. Christofides, Charis, Theo S. Eicher and Chris Papageorgiou, 2016. “Did established Early Warning Signals predict the 2008 crises?”, European Economic Review, Vol. 81, 103–114. Collins, Susan, 2003. “Probabilities, Probits and the Timing of Currency Crises”, mimeo Georgetown University. Comelli, Fabio, 2013. “Comparing Parametric and Non-parametric Early Warning Systems for Currency Crises in Emerging Market Economies”, IMF Working Papers 13/134. International Monetary Fund. Comelli, Fabio, 2014. “Comparing the Performance of Logit and Probit Early Warning Systems for Currency Crises in Emerging Market Economies”, IMF Working Papers 14/65. International Monetary Fund. Cumperayot, P and R. Kouwenberg (2013). “Early warning systems for currency crises: A multivariate extreme value approach”. Journal of International Money and Finance, Vol. 36, pp. 151–171. Edison, Hali J., 2003."Do Indicators of Financial Crises Work? An Evaluation of an Early Warning System," International Journal of Finance and Economics, John Wiley & Sons, Ltd., vol. 8(1), pp. 11-53. Eichengreen, Barry, Andrew Rose and Charles Wyplosz, 1995. "Exchange Market Mayhem: The Antecedents and Aftermath of Speculative Attacks", Economic Policy, Blackwell Publishing, vol. 10(21), 249-312. Eichengreen, Barry, Andrew Rose and Charles Wyplosz, 1996. “Contagious currency crises”, NBER Working Paper, No 5681. Eichler, Stefan, Alexander Karmann and Dominik Maltritz, 2009. “The ADR shadow exchange rate as an early warning indicator for currency crises”, Journal of Banking & Finance, vol. 33, 1983–1995. Eijffinger, Sylvester C.W. and Bilge Karataş (2013). “Three sisters: the interlinkage between sovereign debt, currency and banking crises”, CEPR Working Paper, No. 9369. Esquivel, Gerardo and Felipe Larrain, 1998. “Explaining currency crises”, Harvard Institute for International Development, June.
36
Frankel, Jeffrey, and Andrew Rose, 1996. "Currency Crashes in Emerging Markets: An Empirical Treatment," Journal of International Economics, Vol. 41, 351-366. Frankel, Jeffrey A. and George Saravelos, 2012. “Can Leading Indicators Assess Country Vulnerability? Evidence from the 2008–09 Global Financial Crisis”, Journal of International Economics, Vol. 87(2), 216–231. Frost, Jon and Ayako Saiki, 2013. “Early warning for currency crises: what is the role of financial openness?”, DNB Working Paper, No. 373, De Nederlandse Bank. Ghosh, Swati R., and Atish R. Ghosh, 2002. "Structural Vulnerabilities and Currency Crises", IMF Staff Papers, Palgrave Macmillan Journals, vol. 50(3). Girton, Lance and Don Roper, 1977. “A Monetary Model of Exchange Market Pressure Applied to the Postwar Canadian Experience”, American Economic Review, Vol. 67, 537-548. Glick, Reuven and Michael Hutchison, 1999. “Banking and currency crises: how common are twins?”, Centre for Pacific Basin Monetary & Economic Studies working paper PB99-07, Federal Reserve Bank of San Francisco. Goldstein, Morris, Graciela Kaminsky and Carmen M. Reinhart, 2000. “Assessing Financial Vulnerability: an Early Warning System for Emerging Markets”, Peterson Institute for International Economics, Washington DC, June. Haile, Fasika and Susan Pozo, 2008. “Currency crisis contagion and the identification of transmission channels”, International Review of Economics and Finance, Vol. 17, 572–588. Hawkins, John and Marc Klau, 2000. "Measuring potential vulnerabilities in emerging market economies," BIS Working Papers 91, Bank for International Settlements. Herrera, Santiago, and Conrado Garcia, 1999. "User's Guide to an Early Warning System for Macroeconomic Vulnerability in Latin American Countries," Policy Research Working Paper Series 2233, The World Bank. Inoue, Atsushi and Barbara Rossi, 2008.”Monitoring and Forecasting Currency Crises”, Journal of Money, Credit and Banking, Vol. 40, 523-534. Kamin, Steven B. and Oliver D. Babson, 1999. “The contributions of domestic and external factors to Latin American devaluation crisis: an early warning systems approach”, US Federal Reserve System International Finance Discussion Paper, No 645, September. Kamin, Steven B., John Schindler and Shawna Samuel, 2007. “The contribution of domestic and external factors to emerging market currency crises: an Early Warning Systems approach”, International Journal of Finance and Economics, Vol. 12, 317–336. Kaminsky, Graciela 2006. “Currency Crises: Are They All the Same?”, Journal of International Money and Finance, Vol. 25(3), 503–527. Kaminsky, Graciela, Saul Lizondo and Carmen M. Reinhart, 1998. "Leading Indicators of Currency Crisis", IMF Staff Papers, Palgrave Macmillan Journals, Vol. 45(1).
37
Kaminsky, Graciela L., and Carmen M. Reinhart, 1999, “The Twin Crises: Causes of Banking and Balance of Payments Problems”, American Economic Review, Vol. 89(3), 473-500. Klaassen, Franc and Henk Jager, 2011. “Definition-consistent measurement of exchange market pressure”, Journal of International Money and Finance, Vol. 30(1), 74–95. Knedlik, Tobias, 2006. “Signaling currency crises in South Africa”, IWH-Discussion Papers, No. 19, October. Koedijk, Kees G., Philip A. Stork and Casper G. de Vries, 1992. “Differences between foreign exchange rate regimes: the view from the tails”, Journal of International Money and Finance, Vol. 11, 462–473. Kraay, Aart, 2003. “Do high interest rates defend currencies during speculative attacks?”, Journal of International Economics, Vol. 59, 297–321. Krkoska, Libor, 2001. “Assessing macroeconomic vulnerability in central Europe”, EBRD Working Paper No. 52, European Bank for Reconstruction and Development. Kumar, Mohan, Uma Moorthy and William Perraudin, 2003. "Predicting Emerging Market Currency Crashes", Journal of Empirical Finance, Vol. 10(4), 427-454. Kwack, Sung Yeung, 2000. “An empirical analysis of the factors determining the financial crisis in Asia”, Journal of Asian Economics, Vol. 11, 195–206 Laeven, Luc and Fabian Valencia, 2012. “Systemic Banking Crises Database: An Update”, IMF Working Paper No. 12/163. Lestano and Jan P.A.M. Jacobs, 2007. “Dating currency crisis with ad hoc and extreme value-based thresholds: East Asia 1970–2002”, International Journal of Finance and Economics, Vol. 12, 371–388. Lin, Chin-Shien, Haider A. Khan, Ruei-Yuan Chang and Ying-Chieh Wang, 2008. “A new approach to modeling early warning systems for currency crises: Can a machine-learning fuzzy expert system predict the currency crises effectively?”, Journal of International Money and Finance, Vol. 27, 1098–1121 Lomeli, Jacques, 2005. “Un ejercicio empírico para Latino América. ¿Es posible predecir las crisis cambiarias?”, Problemas de desarrollo. Revista Latinoamericana de Economía, Vol. 36 (143), 213-227. Milesi-Ferretti, Gian Maria and Assaf Razin, 1998. “Current account reversals and currency crises: empirical regularities”, IMF working paper 98/89, June. Moreno, Ramon, 1995. "Macroeconomic Behavior During Periods of Speculative Pressure or Realignment: Evidence from Pacific Basin Economies", Pacific Basin Working Paper Series 95-05, Federal Reserve Bank of San Francisco. Moreno, Ramon and Bahrat Trehan, 2000. "Common Shocks and Currency Crises", FRBSF Working Paper, Federal Reserve Bank of San Francisco. Nag, A., and A. Mitra, 1999. "Neural Networks and Early Warning Indicators of Currency Crisis", Reserve Bank of India Occasional Papers, No. 20(2), 183-222. Nitithanprapas, Ekniti, and Thomas Willett, 2000. "A Currency Crises Model That Works: A Payments Disequilibrium Approach", Claremont Colleges Working Papers 2000-25. 38
Osband, Kent, and Caroline van Rijckeghem, 2000. "Safety from Currency Crashes", IMF Staff Papers, Palgrave Macmillan Journals, Vol. 47(2). Peng, Duan and Claustre Bajona, 2008. “China's vulnerability to currency crisis: A KLR signals approach”, China Economic Review, Vol. 19, 138–151. Pontines, Victor and Reza Siregar, 2007. “The Yen, the US dollar, and the trade weighted basket of currencies: Does the choice of anchor currencies matter in identifying incidences of speculative attacks?”, Japan and the World Economy, Vol. 19, 214–235. Pontines, Victor and Reza Siregar, 2008. “Fundamental pitfalls of exchange market pressure-based approaches to identification of currency crises”, International Review of Economics and Finance, Vol. 17, 345–365. Pozo, Susan and Catalina Amuedo-Dorantes, 2003. “Statistical Distributions and the Identification of Currency Crises”, Journal of International Money and Finance, Vol. 22; 591- 609. Qin, Xiao and Liya Liu, 2014. “Extremes, return level and identification of currency crises”, Economic Modelling, Vol. 37, 439–450 Reinhart, Carmen M. and Kenneth S. Rogoff, 2011. “From Financial Crash to Debt Crisis”, American Economic Review, Vol. 101(5), 1676–1706. Sachs, Jeffrey, Aaron Tornell, and Andres Velasco, 1996, "Financial Crises in Emerging Markets: The Lessons from 1995", Brookings Papers on Economic Activity vol. 27 (1), pages 147-216. NBER Working Paper, No. 5576. Sevim, Cuneyt, Asil Oztekin, Ozkan Bali, Serkan Gumus, and Erkam Guresen, 2014. “Developing an early warning system to predict currency crises”, European Journal of Operational Research, Vol. 237, 1095–1104. Shimpalee, Pattama L., and Janice Boucher Breuer, 2007. "An event study of institutions and currency crises", Review of Financial Economics, vol. 16, 274–290. Tovar, Camilo, 2010. “Currency collapses and output dynamics: a long-run perspective”, BIS Quarterly Review, June 2010. Vlaar, Peter J.G., 1999. “Currency Crisis Models for Emerging Markets”, Research Memorandum WO&E, Nr. 595/9928, De Nederlandsche Bank. Yu, Lean, Shouyang Wang, Kin Keung Lai and Fenghua Wen, 2010. “Multiscale neural network learning paradigm for financial crisis forecasting”, Neurocomputing, Vol. 73, 716–725 Zhang, Zhiwei, 2001. "Speculative Attacks in the Asian Crisis", IMF Working Papers 1/189, International Monetary Fund.
39
Appendix A. Overview of currency crisis definitions Table A.1: Overview of definitions which follow the large depreciation approach Authors Frankel and Rose (1996), Bauer, Herz and Karb (2007) Milesi-Ferretti and Razin (1998)
Esquivel and Larrain (1998)
Exchange rate Frequency Nominal Real Annual
X
Monthly
X
Monthly
X X
Monthly
X
Monthly
X
Moreno and Trehan (2000)
Quarterly
X
Bruggemann and Linne (2002)
Daily
X
Condition 2 > 10 percentage points higher than previous year
Condition 3
Depreciation > 25% per year, or
Depreciation > 2x depreciation previous year > 10 percentage points higher than previous year
Depreciation in previous year < 40% Depreciation in previous year < 10%
Depreciation > 15% per year
Monthly
Osband and Van Rijckeghem (2000)
Condition 1 Depreciation > 25% per year
Depreciation > 15% in 3 months Depreciation > 2.54 x standard deviation in one month Depreciation > 10% per month
> 4 percentage points higher than previous year Depreciation > 2 standard deviations above mean Depreciation in one quarter > m + 2 of The value of the currency at the end of the crisis quarter is below the last 20 quarters where it was a year ago. Depreciation > 20% in ten trading days (i) unanticipated currency crash = large (annual) devaluation, adjusted for interest rate differential:
Monthly
X with e t is nominal exchange rate in t, i f and i d foreign, resp. domestic interest rate (with maturity ), and y1 is 5% or 10%. (ii) total depreciation crash = large (annual) devaluations which exceed the devaluation in the previous year by some multiple:
Kumar, Moorthy and Perraudin (2003)
Monthly
X
with y2 is 5% or 10%. Quarterly
X
Depreciation > 20% in 1 quarter, or
Apoteker and Barthelemy Quarterly (2005) Quarterly
X
Depreciation > 30% in 2 quarters, or
X
Lomeli (2005)
Monthly
X
Catao (2006)
Annual
X
Innoue and Rossi (2008) Yu, Wang, Lai and Wen (2010)
Monthly
X
Daily
X
Tovar (2010)
Monthly
X
Reinhart and Rogoff (2011)
Annual
X
with y3 is 100%.
Depreciation > 40% between 3 to 6 quarters Depreciation > 2 x standard deviation above the mean in one month, with standard deviation on a rolling window base (last 24 months) Depreciation > 2 x standard deviation Depreciation is not fully reversed above the mean in one year, with the within three years standard deviation from the entire period (1870-1913) Depreciation > 20% per year Depreciation > 2.5 x standard deviation above the mean in one day Annual depreciation in the top quartile of all depreciation episodes in the sample Depreciation > 15% per year
40
Authors
Bussiere, Saxena and Tovar (2012)
Laeven and Valencia (2012)
Exchange rate Frequency Nominal Real Monthly
X
Monthly
X
Monthly
X
Annual
X
Kraay (2003) and Eijffinger Monthly and Karataş (2012)
X
Cumperayot & Kouwenberg (2013)
X
Monthly
Condition 1 Depreciation > 25% per year, or Depreciation > 15% per year
Condition 2 Depreciation > 2x depreciation previous year > 10 percentage points higher than previous year
Annual depreciation in the top quartile of all depreciation episodes in the sample Depreciation > 30% per year > 10 percentage points higher than previous year Advanced countries: Depreciation Emerging economies: exceeds 5% in a given month following Depreciation exceeds 10% in a an episode of stable exchange rates given month following an (i.e. the average absolute percentage episode of stable exchange rates change should be lower than 1% for the (i.e. the average absolute 12 month-period prior to the large percentage change should be depreciation). lower than 2.5% for the 12 monthperiod prior to the large depreciation). Depreciation greater than 9.6% monthly
41
Condition 3 Depreciation in previous year < 40% Depreciation in previous year < 10%
Table A.2: Overview alternative EMPI definitions Data frequency
Paper
Change in exchange rate nominal real
black market
Threshold
Nominal interest rate Change in differen- change in reserves Change tial differential
Eichengreen, Rose and Wyplosz (1995)
Quarterly
X
X1
X
m + 2
Eichengreen, Rose and Wyplosz (1996)
Quarterly
X
X1
X
m + 1.5
Moreno (1995)
Monthly
X
X
Sachs, Tornell and Velasco (1996), Bussiere and Mulder (1999, 2000), Kwack (2000), Nitithanprapas and Willett (2000)
Monthly
X
X
No threshold: continuous index
Kaminsky, Lizondo and Reinhart (1998), Kaminsky and Reinhart (1999), Berg and Patillo (1999), Goldstein, Kaminksy and Reinhart (2000), Collins (2003), AlvarezMonthly Plata and Schrooten (2004), Kaminsky (2006), Shimpalee and Boucher (2007), Peng and Bajona (2008), Lin, Khan, Chang and Wang (2008), Comelli (2013)
X
X
m + 3
X
X
Herrera and Garcia (1999) 2 Kamin and Babson (1999), Glick and Hutchison (1999), Ari and Cerigibozan (2016) Nag and Mitra (1999), Christofides, Eicher and Papageorgiou (2016)
Monthly
Annual
Vlaar (1999)3
> m + 1.5, for any component (separately)
X
m + 1.5
X
X
m + 2
X
X
m + 2
Monthly
X
X
Aziz, Caramazza and Salgado (2000)
Monthly & annual
X
X
Caramazza, Ricci and Salgado (2000)
Monthly
X
X
Monthly
X
X
Monthly
X
X
Krkoska (2001)
Monthly
X
Extreme Value Theory (EVT) > m + 1.5, with m and from entire pool > m + 1.645, with m and from entire pool m + t(n, 5%) , and m + t(n, 0.5%) > m + 3, for each component (separately)
2
Zhang (2001) Ghosh and Ghosh (2002)
4
Monthly
X
Pozo and Amuedo-Dorantes (2003)
Quarterly
X
Edison (2003)
Monthly
X
Monthly
X
Monthly
X
Monthly
X
Knedlik (2006)
X
X
m + 1.5, and with EVT m + 2.5 > m + X , with X: 1.5, 1.645, 1.75, 2, 2.5 and 3
X X
Idem Idem
X
Bussiere and Fratzscher (2006, 2008), Bussiere (2013a)
Monthly
X
X
Kamin, Schindler and Samuel (2007)
Two months
X
X
Haile and Pozo (2008)
m + 2
X X1 X
Monthly
Lestano and Jacobs (2007)
X
1
Quarterly
X
X
Quarterly Quarterly
X X
X1 X
42
Idem m + 2
X5
m + 1.75 X
m + X , with X = 1.5, 2, 2.5 and 3; and EVT
X
EVT
Paper
Data frequency
Change in exchange rate nominal real
black market
Andreou, Dufrenot, Sand-Zantman and Zdzienicka-Durand (2009)
Quarterly
Klaassen and Jager (2011)
Monthly
Asici (2011)
Monthly
Ari (2012)
Monthly
Frost and Saiki (2013)
Monthly
X
Monthly
X7
X7
X
X
Annual
X7
X7
X
X
X
Grid search (0 to 3), with optimal: 0.75 No threshold: Markov switching method m + 3
X
X
X6
X X X
X X
m + 3
X
m + 1.5 and m + 2
X
Cerro and Meloni (2013)
Bussiere (2013b)
Monthly
X
mild: EMPI > m + 1.5, deep: > m + 2, very deep: > m + 3 mild: EMPI > m + , deep: > m + 1.5, very deep: > m + 2 No threshold: continuous index
X
X
X
m + 2
Monthly
X
X
X
m + 3
Monthly Quarterly
X X
X X
Candelon, Dumitrescu and Hurlin (2014) Monthly Sevim, Oztekin, Bali, Gumus and Guresen (2014) Comelli (2014) Bucevska (2015)
X
Threshold
Nominal interest rate Change in differen- change in reserves Change tial differential
X
m + 2 and m + 3 m + 1.5
Notes: 1 Uses the change in the differential (vs reference country) of the ratio Reserves to M1. 2 Do not use weights to integrate the three components: the simple average is used 3 Weights for all countries the same (instead of country-specific weights) 4 Additional condition: real GDP growth decreases minimal 3% compared to average real GDP growth in the previous five years 5 Use the change in the real interest rate 6 Use the difference between the actual level and the interest rate chosen if the country had no exchange rate objective. For practical reasons they use the interest rate differential with the reference country 7 Uses black market exchange rate for periods in which Argentina intervened heavily in the forex markets.
43
Appendix B. Narrative In this section we briefly comment on the crisis episodes by using the narrative. We focus on the episodes that are identified as crises (minimal three definitions identify a crisis), on episodes that were identified in other databases (see Table 5), and on episodes after 201115 in which the exchange rate is known to have been under severe pressure. In Appendix B.1 we analyze the episodes with currency crises for Latin American and Caribbean countries, Appendix B.2 for Central and Eastern European countries, Appendix B.3 for Asian countries, and Appendix B.4 for the rest of the world.
Appendix B.1 Latin America and Caribbean Argentina (1994-1995) In 1994, the tight monetary policy in the United States leads to worldwide interest rate increases, contributing to banking fragilities and a credit crunch amid a severe recession in Argentina. Following the Mexican crisis in December 1994, Argentina's banking system suffers from large deposit withdrawals, as rumors spread that systemic banks may fall due to their exposure to sovereign debt. The lending volume drops by more than 50% and banks suffer from liquidity problems. Deposits are fleeing from wholesale banks - taken out by large and informed investors. In March 1995 the peg is successfully defended through the use of reserves and interest rate. This event is considered the first currency crisis of the Convertibility Plan that was launched in April 1991 (Kaminksy, Mati and Choueri, 2009). In several academic papers this crisis is used as an example to emphasize why the EMPI should include the change in interest rate. The crisis is profound, but lasts only shortly, and is successfully contained by the policy response (Guidotti and Nicolini, 2016). (2001-2002) In 2001 the international capital markets are still wary of emerging economies (Asian, Russian, LTCM, Brazilian and Turkey crises). Domestically the political uncertainty and the financial turmoil following Brazil’s crisis in January 1999 has severely affected private investment and consumption in Argentina. The country is stuck with twin deficits, a trade gap and a budget gap, that foreigners are less and less willing to finance (Economist, 2005). As the situation continues to deteriorate in 2000 and 2001, the government seeks more financing, and resorts to compulsory placing of government bonds at banks, with banks becoming increasingly more exposed to government default (Kaminksy, Mati and Choueri, 2009). The Argentinian government receives a USD 40 billion IMF-led emergency assistance package in December 2000. By June 2001 a bank run starts. The withdrawal of bank deposits, which continues through the whole year, ends up in 15
Since the external databases only cover up to and including 2011, we include also episodes that may be considered currency crises although the quantitative procedure does not identify these.
44
the "corralito" (imposition of restrictions on deposit withdrawals) in December 2001 (Guidetti and Nicolini, 2016). The peg is maintained at the cost of decreasing reserves and higher interest rates. But the situation becomes unsustainable. In December 2001, the government defaults on its debt, and a few weeks later it abandons the currency peg to the dollar. In January 2002 Argentina has no other choice than to abandon the peg and float the currency. The exchange rate rapidly lost value, from 1.00 : 1.00 in December 2001 to 1.40 : 1.00 in January 2002 and 3.75 : 1.00 in June 2002 (Weisbrot et al., 2011). The banking sector collapses, due to its high exposure on sovereign debt and the high default rate of the private sector in the face of the crisis, making it a triple crisis. (2014) In January 2014 the exchange rate depreciates 21% in a month, which implies an accumulated depreciation of 59% compared to 12 months before. From the Economist (2014): "The peso has looked overvalued since at least 2011, when the government clamped down on all foreign-currency transactions in an attempt to stem capital flight. The black-market exchange rate has at times exceeded the government’s artificial rate by as much as 70%. To close this gap, the government has been allowing the peso to devalue: 12% in 2012 and 33% in 2013." In 2014 Argentina defaults again, but this time due to legal issues related to New York based judge Griesa, who orders that Argentina cannot pay anything to creditors who had agreed to the earlier restructuring until Argentina would pay the bond holdouts fully. (2015) The new president Mauricio Macri removes limits on exchange transactions. This causes the official value of the Argentinian peso to crash overnight, dropping 30% in less than 24 hours (Al-Shawwa, 2015) , and 36% in the month of December. The depreciation is inevitable: due to the high inflation the exchange rate is overvalued, and not sustainable. Reserves are at a seven-year low, exports not competitive. "Macri's government eliminated or reduced taxes on commodity exports and abolished exchange controls, resulting in a de facto devaluation of around 35% against the dollar" (Stiglitz and Guzman, 2016). With Macri’s arrival, the long-lasting settlement of the 2001-default has come to an end, when the New York judge Griesa changed the verdict and a final agreement was reached (Moyer, 2016). We conclude that Argentina has experienced currency crises in: 1995M3, 2002M17, 2014M1, and 2015M12. The narrative confirms the quantitative determination of crises.
Brazil (1998) With the Russia crisis (started in August 1998) and to a lesser extent the LTCM bailout (September 1998) capital flows to emerging economies, including Brazil, come to a halt (Fraga, 2000). The pressure on the real is relieved by increase in the interest rate (19% to 34%), and the use of reserves (drop of 33%) in September 1998. The pressure on the exchange rate does not terminate into a devaluation, although it can be considered the prelude for the deep crisis in January 1999. The currency is considered 15-25% overvalued in real terms, and as 45
a consequence industrial production has decreased due to low competitiveness (Gruben and Welch, 2001), and the current account deficit exceeds 5% of GDP (Fraga, 2000). The country repeats the same recipe as in 1995 and 1997 (increasing the interest rate), but this time the policy fails as the fiscal reform package lacks credibility. The policy of higher interest rates turns the situation more complicated because the risk of a debt default increases. Capital outflows increase, various states declare to suspend debt payments to the federal government (Gruben and Welch, 2001). In November 1998 the Brazilian government signs an emergency loan package from the IMF to avoid a financial meltdown. (1999) In January 1999 the currency is floated: the exchange rate peg is released and the currency depreciates 64% (from R1.21 to R1.98). The reserves drop 21% in January 1999, and the interest rate remain stable at a high level (30%). The banking sector is healthy and well-regulated, but the government's solvency is very low, with a high proportion of short-term debt, and not enough resources to service the debt obligations (Gruben and Welch, 2001). With the IMF assistance (emergency loan that was signed in November 1998) a default is avoided. (2000) The 27% drop in reserves in April 2000 is caused by the early repayment of part ($10 billion) of the debt from the 1998-rescue package (Buckman, 2013), and is therefore not considered a currency crisis. (2002) In the run-up to the presidential election in October 2002, Brazil's economy comes under renewed pressure as financial markets are worried about lingering fiscal and current account problems, the crisis in neighboring Argentina, and the prospect of a left-leaning candidate winning the elections (IMF, 2007). The Brazilian real depreciates strongly in July and September 2002. A large IMF standby arrangement helps to restore confidence and to stabilize the economy. (2008) The fall of Lehman Brothers in September 2008 leads to worldwide depreciations of emerging economies’ currencies. The Brazilian real is one of the hardest hit currencies. From July 2008 to February 2009 the Real depreciates by 51% (R1.57 to R2.38). On an annual base the depreciation is –partly– offset by the appreciations that occur before August 2008 and after March 2009, which explains why Laeven and Valencia (2012) do not label 2008 as a currency crisis year. Contrary to past crises, Brazil’s real economy is not strongly affected, and there is no banking or debt crisis. The initial conditions (including a regulated banking sector), the commodity price cycle and abundant capital inflows play a role in this improved economic performance (ECB, 2016). (2014-201516) Brazil’s economic situation has deteriorated significantly in 2014 and 2015. The economy enters into recession in 2014 and the situation worsens in 2015, with real GDP declining by 3%, while inflation has remained close to 10% (ECB, 2016). From January to September 2015 the real depreciates by 58% vis-à-vis the US dollar, with acceleration in July-September 2015 (more than 30% in three
16
We include Brazil 2015 since there has been wide attention on the pressure on the exchange rate. Since peer studies (as shown in Table 5) do not reach out further than 2011, we decided to include the events in 2015 in the narrative.
46
months). The depreciation is partially due to external factors, such as China's slowdown, the fall in commodity prices, the US's imminent decision to raise its interest rates and the consequent tendency of withdrawing capital from riskier emerging countries for safer gains in the US. But there also domestic factors, in particular the unsustainable debt level, which forces the government to cut expenses and increase taxes, thus affecting economic growth. In September 2015 Standard and Poor’s downgrades Brazil’s debt to speculative grade. Worries about the credibility of the country’s policies to keep inflation down and the necessary structural reforms affect confidence and put pressure on the exchange rate (Gallas, 2015). In later months the depreciation is partly offset by the appreciation of the currency. Two out of our six definitions identify a crisis, because of the large nominal and real depreciation. However, none of the EMPI definitions identifies a crisis, which is due to the (very) high volatility of the indicators in the 1990s that make the threshold relatively very high. We conclude that Brazil experienced currency crises in: 1998M9-1999M3, 2002M7-9, 2008M11, and 2015M7-9. The narrative confirms the crises identified by the quantitative criteria, except for 2000M4 (excluded), and 2015M2-9 (included).
Chile (2007-2008) During 2007 the reserves fluctuate on a monthly base, but remain at a relatively high level, mainly due to the high dollar revenues from commodity exports. By the beginning of 2008 the Chilean central bank starts to intervene in the financial markets, to depreciate the overvalued Chilean peso and make exports more competitive. The peso loses 11% in value from January to August 2008 (from $465 to $516), while reserves increase by almost 6 billion US dollar in the same period (Briones, 2010). With the fall of Lehmann Brothers this program is suspended and the peso rapidly decreases in value, 29% from August to October (from $516 to $665). After October 2008 the exchange rate starts to appreciate against the dollar. During 2008 the interest rate is increased to curb inflationary pressures, caused by the high commodity prices (in particular oil). In 2009 the country uses the Social and Economic Stability Fund with windfalls from the commodity boom for countercyclical policy, and lowers the interest rates drastically, from more than 8% to less than 1%. The healthy banking sector, the recovery of the commodity prices and the large fiscal stimulation package help Chile to overcome the crisis fast: in the last quarter of 2009 the economy starts a new growth path (Agosin and Montecinos, 2011; Briones, 2010). We conclude that Chile has experienced one currency crisis, dated in 2008M10. The narrative confirms the strong impact on the exchange rate, although part of the depreciation was on a voluntary base prior to September 2008. Colombia (1995) Government spending increases strongly in the 1990s (from 10% of GDP in 1990 to 18.8% of GDP in 1999); up to 1995 it is financed with higher revenues – 47
mostly due to the high economic growth. The trade balance deficit increases, caused by a continuous appreciation of the real exchange rate. Large capital inflows enter the country to finance these imbalances, and (private) foreign debt increases strongly. Colombia installs non-remunerated reserve requirements on capital inflows in 1993 (Villar and Rincon, 2000). After 1995 the government runs fiscal deficits, which are financed by privatizations of state-owned companies and increasing debt levels, from 12% of GDP in 1994 to 28% of GDP in 1999. In 1995 economic activity is weak, partly due to the tequila crisis in Mexico. The exchange rate weakens approximately 10%, while reserves and interest rate are volatile during 1995. Galindo and Maloney (1998) mention that Colombia is one of the Latin American countries that is largely unaffected by the tequila crisis. We conclude that the narrative does not confirm that Colombia has experienced a currency crisis in 1995. (1997-1999) in the midst of the Asian and Russian economic crises in 1997-1998, the Colombian peso comes under severe pressure, and the central bank devalues the exchange-rate band twice. The sudden stop in international lending leads to an abrupt adjustment in the current account, which means a large contraction in aggregate demand. Increases in international interest rates together with expectations of devaluation of the peso cause interest rates to rise: the average interest rate on loans increased almost 1000 basis points in the summer of 1998, thus
contributing to the contraction of GDP. The recession and the bursting of a realestate bubble also result in a major banking crisis. Public banks and private mortgage banks are hard hit, and the subsequent government intervention to aid some of the distressed financial institutions adds pressures on public expenditure (Steiner and Vallejo, 2010). The economy enters the deepest and one of the longest recessions since the 1930s (Vargas, 2005). The 1998 banking crisis in Colombia is accompanied by a currency crisis. In 1999 its exchange rate regime (a target zone) collapses and the exchange rate is allowed to float freely (Arias, 2000). (2002-200317) In 2002 the uncertainty about Brazil’s economic and political outlook leads to a fast increase in Colombia’s sovereign spread. Additionally, revenues from the two main commodities are under pressure: oil production decreases, and coffee prices are low (Steiner and Vallejo, 2010). The crisis in Venezuela at the end of 2002 leads to a reduction in Colombia’s exports revenues, adding more pressure on the peso. The peso depreciates by 30 percent between April 2002 and February 2003. The higher exchange rate causes high inflation, because of the pass-through of higher of import prices. The situation induces the authorities to intervene for the first time since the exchange rate was floated in 1999. The central bank increase interest rates in December 2002 and April 2003 and use foreign reserves to reverse the depreciation trend (Vargas, 2005). The latter is picked up by only two definitions, and is also identified by Reinhart & Rogoff (2011).
17
For Colombia we also include 2002-2003 in the narrative, because these years are identified by Reinhart and Rogoff (2010) as currency crisis years.
48
(2008) Prior to September 2008 the Central Bank has increased interest rates steadily (from 6% in April 2006 to 10% in the summer of 2008), as it is confronted with increased governmental expending, high foreign capital inflows, and an expansionary credit market. When the global crisis breaks out, the Central Bank does not use reserves nor increase the interest rates to defend the exchange rate. The Colombian peso loses 40% of its value (from June 2008 to February 2009) – very similar to other currencies in the region. To revitalize the economy, interest rate are lowered in 2009, capital controls are cut and investments are stimulated through a temporary increase of the maximum level of the fiscal deficit and a stand-by credit from the IMF (Arguello, 2011). (2015) The (36%) depreciation of the peso from the summer of 2014 to the summer of 2015 is caused by both external and internal factors. External: lower prices and demand for commodities -in particular oil and coffee-, and internal: high government debt (which has increased during the boom years of the commodity markets) combined with fiscal imbalance (low government revenues) (Lemaitre, 2015). We conclude that Colombia has experienced currency crisis in: 1998M6-9, 2002M12-2003M2, 2008M9-2009M2 and 2015M7-8. Compared to the quantitative selection of crises, the narrative discards a crisis in 1995M5-8 and dates a crisis in 2002M12-2003M2.
Costa Rica (199118) Costa Rica’s colon depreciates in a gradual manner, according to crawling peg exchange rate regime. In real terms the annual depreciation is single-digit in 1991. As Carstens (2004) mentions in a speech, Costa Rica has not suffered a major financial crisis since the late 1980s. We conclude that Costa Rica has not experienced currency crises in the period 1990 to 2016. The narrative and the quantitative selection procedure coincide. Dominican Republic (1990) In 1990 the Dominican Republic faces deteriorating macroeconomic performance and policies. After three years with a growing fiscal deficit financed by inflationary domestic policies and heavy borrowing abroad, in August 1989 the government can no longer service its outstanding debts (Boughton, 2012). The ruling president is reelected in May 1990, in the midst of great social tensions (Fearon and Laitin, 2006). In the summer of 1990, the authorities embark on a serious effort to get their fiscal accounts on a more sustainable level, then enter into long discussions with the IMF staff aimed at settling arrears and obtaining new financial and other assistance (Boughton, 2012). The exchange rate is devalued by 42% (from 7.45 to 10.35) in August 1990, and further, smaller devaluations follow in the next months. Reserves drop to low levels by the end of 1990. 18
For Costa Rica we also include 1991 in the narrative, because this year is identified by Reinhart and Rogoff (2010) as a currency crisis.
49
(2003) The 1990s are a period of macroeconomic stability and economic growth for the Dominican Republic. In this decade the country experiences a de facto dollarization of its banking system (Sánchez-Fung, 2005). In 2003 Dominican Republic suffers from a large banking crisis, centered around the country’s secondlargest private bank, Baninter. When the bank collapses in April 2003, a large rescue package by the government is undertaken, with an estimated cost of 20% of GDP. In total three large banks fail in 2003 due to accounting malpractices and mismanagement (Jaramillo and Sancak, 2007). The banking crisis triggers a currency crisis in May 2003 (NY Times, 2003). Particular factors that affect the economy at the time are the slowdown in tourism after the 11th September attacks on the USA, the reduced growth rate of the USA’s economy, and higher oil import prices. This leads to exchange rate instability, high inflation, and weakened the effectiveness of monetary and fiscal (stabilization) policies during 2003. The nominal exchange rate loses more than 100% of its value from April 2003 to January 2014, the reserves drop by 50% and the interest rate surges by more than 25 percentage points. As a result of these problems the Dominican Republic signs a stand-by agreement with the IMF that is formally approved in August 2003 (Sanchez-Fung, 2005). We conclude that Dominican Republic has experienced two currency crises, in 1990M8 and 2003M5-2004M1. The narrative confirms the quantitative determination of crises.
Guatemala (1990) Guatemala is not included in most studies on currency crises. This can be attributed to the low number of currency crises that the country has experienced since 1990. The last crisis is dated in August 1990 by Esquivel and Larrain (1999). The Guatemalan quetzal is floated on November 1989, and the consequent period of depreciations lasts until September 1990 (Morán Samayoa, 1999), with the largest depreciations in the last months (July-October 1990). We conclude that Guatemala has experienced one currency crises, in 1990M8-10. The narrative confirms the quantitative determination of crises. Jamaica (1991) In September 1991 the financial system of Jamaica is liberalized, including the removal of foreign exchange, trade and capital controls (Harriott, 2000). With the removal of exchange rate controls, the Jamaican dollar depreciates by more than 50% (from 13.67 to 20.91 per US dollar) from September to December 1991. In the same period interest rates rise sharply and reserves drop (Nelson-Douglas, 2001).
We conclude that Jamaica has experienced one currency crises, in 1991M9-12. The narrative confirms the quantitative determination of crises.
Mexico (1994) Political unrest in Chiapas is followed by the assassination of Colosio, PRIcandidate for presidency in March in Tijuana. The rising social and political 50
uncertainty increases the pressure on the (pegged) exchange rate (Musacchio, 2012). To defend the currency the interest rate is increased and reserves are used. In April 1994 reserves drop massively (32% in one month, and 12% in the previous month), and the interest rate almost doubles (from 9.5% to 18%). The peg is maintained, although at a high cost. The pressure on the exchange rate increases again in the last months of 1994. Reserves and increases in the interest rate are not sufficient to fence off the pressure on the peso. Newly installed president Zedillo sees no other way out then to let the currency float rate (Musacchio, 2012). The peso loses 54% of its value in December 1994, and continues to depreciate in the months after, in particular in March and October-November 1995. As a consequence of the devaluation (currency mismatch in the funding) and the widespread defaults in the corporate sector the poorly regulated banking sector collapses. Only with IMF assistance a sovereign default can be avoided –Manasse et al. (2003) consider this a debt crisis. The country falls in a deep economic crisis (Musacchio, 2012). (199819) The Russia crisis (started in August 1998) and the LTCM bailout (September 1998) affect Mexico as international risk aversion increases (Dungey et al., 2002). Many emerging economies are affected. In Mexico the exchange rate depreciates (12%), reserves drop (6%) and interest rate differential increases (4%), but we do not find evidence in the narrative that this should be considered a currency crisis. The country is still recovering from the 1994-1995 financial crisis and the increased volatility in 1998 is not perceived as a new currency crisis. (2008) The fall of Lehmann Brothers in September 2008 triggers a period of depreciation of currencies from emerging economies as global capital is withdrawn from risk-perceived assets, which include emerging economies. Mexico is hit relatively hard due to its high exposure to the US economy, and the sudden stops in capital flows and in international trade (WEO, 2009). Between August 2008 and February 2009 the peso loses 47% of its value. Contrary to past financial crises, Mexico does not have a banking crisis (prudent regulations and supervision) nor a debt crisis (low debt levels). The economy slips into a very deep recession, with negative growth in 2009 (Rojas-Suarez, 2011). We conclude that Mexico has experienced two currency crises, in 1994M121995M10 and 2008M10-2009M2. The narrative confirms the quantitative determination of crises.
Paraguay (1990-199220) Reforms are implemented after a change in the political regime in 1989. These reforms include the (managed) floating of the domestic currency
19
For Mexico we also include 1998 in the narrative, because this year is identified by Reinhart and Rogoff (2010), Eijffinger and Karata (2013) and Cumperayot et al. (2013) as a currency crisis. 20
For Paraguay we also include 1992 and 1998 in the narrative, because these years are identified by Reinhart and Rogoff (2010) as currency crisis years.
51
Peru
(guaraní) and the liberalization of interest rates in 1990-1991 (Garcia-Herrera, 1997). As a consequence these events are not to be considered a currency crisis. (1998-1999) The Brazilian crisis has a contagious effect on Paraguay. The country experiences a political crisis after the assassination of the vice president in March 1999. Six months after the start of the Brazilian crisis, Paraguay benefits from a large loan from Taiwan, allowing it to revert its contractionary interest rate defense (IMF, 2009). We do not consider these events as a currency crisis, but more like a political crisis. Although there is some pressure on the exchange rate, the actual depreciation, drop in reserves and increase in interest rates is rather modest. (2001-2002) The 2001-2002 crisis in Argentina has a larger impact on Paraguay than the 1998-1999 crisis in Brazil. The crisis spills over into the domestic banking system through a local subsidiary of an Argentinean bank (IMF, 2009). The country experiences rising inflation, high fiscal deficit, high external debt service, and a deep economic recession. This period of stagnation and financial instability leads to a financial crisis in 2002. S&P downgrades Paraguay’s credit rating to SD (Selective Default) in February 2003 (Franks, Mercer-Blackman, Sab and Benelli, 2005). (2008) As most Latin American currencies, the guaraní appreciates in the year up to September 2008. From September to December 2008 the currency depreciates by 25% to reach the same level at the beginning of the year. The policy of limited interventions helps to avoid a disorderly adjustment that could jeopardize the stability of the dollarized banking system, while preserving the reserve cushion (IMF, 2009). Reserves drop by almost 15% from August to November 2008, while at the beginning of 2009 the interest rate increases sharply. We conclude that Paraguay has experienced currency crises in 2002M6 and 2008M9-12. The narrative confirms the quantitative determination of crises. We consider the events in 1992 and 1998-1999 to be associated with structural reforms and a political crisis respectively.
(1990-199421) After being elected as an independent candidate, president Fujimori initiates an ambitious stabilization program in 1990–1992, fully in line with the Washington consensus –with a focus on liberalization and a small government. The structural reforms set the basis for high economic growth in the future (Santos, IMF, 2015). The multiple exchange rates are abolished and only the market rate is used. Inflation decreases spectacularly, from over 7,000% in 1990 to 50% in 1992, then decreasing to single digits from 1997 on. A managed float for the exchange rate is implemented. In April 1992 president Fujimori carries out a military coup, to overcome the political deadlock by the opposition parties. The coup is condemned
21
For Peru we also include 1992-1994 and 1998 in the narrative, because these years are identified by Reinhart and Rogoff (2010) as currency crisis years.
52
by the international community, but opens up the road for reforms. In March 1993 the debt from the 1980s is renegotiated and Peru gets access to foreign funding (McClintock, 2000). From April 1993 on the currency stabilizes: monthly depreciation is on average 1.3% monthly, and in later years even less. The economy grows at record levels. We do not consider the events in 1992 as a currency crisis, for two reasons. First, because the country comes out of a period of hyperinflation, with high yet sharply lower inflation and nominal exchange rate depreciations. Second, because the crisis is associated more with political issues than with a currency crisis. (1998-1999) After years of high growth Peru suffers a setback in 1998. The Asian and particularly Russian crises have a huge impact on capital flows to and from emerging economies, and Peru is no exception. Additionally, prices of raw materials decrease, and El Niño has a devastating effect on agriculture and fishing (IMF, 1999). The economy suffers a recession. Since the 1980s Peru’s financial system holds an important share of its assets and liabilities in foreign currency. A depreciation and the decline of foreign funds paralyzes the flow of local credit, worsening the country’s financial conditions (Rossini, Quispe and Serrano, 2013). The banking sector is in deep trouble and requires bailouts from the government in November and December 1998 (Hidalgo, 1999). In June 1999 Peru completes negotiations for an Extended Fund Facility (IMF, 1999). The exchange rate drops 6% in January 1999, 21% below the value one year earlier. We consider this a banking crisis rather than a currency crisis, because the depreciation is relatively modest – although it has a huge impact on the banking sector due to its high external debt. (2008) The impact from the GFC has been modest for Peru. The policy reaction in Peru (and many other emerging economies) is in a reversed sequence compared to developed economies. Liquidity easing measures are implemented first, followed by policy rate reductions when the forex and money markets have stabilized (Quispe and Rossini, 2010). The exchange rate depreciates 15% from July 2008 to February 2009, reserves drop marginally and interest rates remain virtually constant. We conclude that Peru has not experienced currency crises in the period 1992 to 2016. According to the narrative there is no currency crisis in 2008 while according to the quantitative determination there is. The latter can be explained by turning to the volatility. Humala and Rodriguez (2009) present empirical evidence that official intervention in the foreign exchange market in Peru has reduced excess volatility in the foreign exchange rate. As a consequence of this low volatility, the threshold to call a crisis is low. So a relatively small adverse movement of the components is sufficient to call a crisis.
Uruguay (2002) The government runs fiscal deficits since the start of the recession in 1999 and finances these with public debt (large majority foreign currency denominated debt). Partially responsible for the weak performance of the Uruguayan economy 53
since 1999 are the successive devaluations of the Brazilian real and the Argentinean peso in 1999 and 2001 respectively, which have made Uruguay less competitive. The two largest banks are considered weak and in general the highly dollarized banking sector is vulnerable for external shocks. When the crisis in Argentina starts in December 2001 foreign depositors withdraw (US dollar) funds from their Uruguayan banking accounts. In February 2002 Uruguay intervenes for the first time in the banking sector, and in the same month loses investment grade rating. The run on dollars and bank deposits accelerates and by July 2002 the country’s reserves are exhausted (an 80 percent decline since December 2001), and the government is forced to float the exchange rate, which immediately falls 27%. By the end of July 2002, a cumulative 38 percent of total deposits has been withdrawn from the system, and the peso has depreciated by 57 percent since December 2001 (de la Plaza and Sirtaine, 2005). We conclude that Uruguay has experienced a currency crisis in 2002M6-7. The narrative confirms the quantitative determination of crises.
Venezuela (1994-1996) Venezuela continues the 1990s similar to the 1980s. Inflation remains high, economic growth continues to be volatile and oil-dependent, growth per capita stagnates, and public sector deficits endure. Between 1988 and 1999 the country suffers some sort of economic emergency, whether it is a critical fiscal deficit, a banking crisis, a currency crisis, an economic recession or a combination of these (Corrales, 1999). The macroeconomic reforms that start in 1989 have to be accompanied by financial sector reforms, but these are never carried out. The highly concentrated banking sector lacks supervision and efficiency. The rise in real interest rates and the 1993-recession increases non-performing loans. Banco Latino, the second largest bank, collapses in January 1994 after rumors spread that the bank is in financial distress. Bank runs on other banks follow and capital flight leads to a large loss in reserves (almost 40% from December 1993 to May 1994). The exchange rate drops by 40% in May 1994 and 20% more in the next month, and is also high in real terms (almost 50% in May-June 1994), turning the banking crisis into a twin banking-currency crisis. Banks are supported and later nationalized. The banking crisis continues until mid-1995. The costs for the rescue operation weakens the fiscal balance and increases the domestic public debt. Two years after the banking crisis starts, the participation of private banks has dropped from 90% to 58%, with foreign-owned banks and nationalized banks increasing their market participation (Garcia-Herrero, 1997). In December 1995 and April 1996 the fixed exchange rate of the bolivar is devalued by 70% and 60% respectively (in real terms the percentages are 60% and 50% respectively), accompanied by sharp increases in the interest rates, but no changes in the (still low) reserves. (2002-2003) The economy is affected by political instability in the first four years of Chavez’s presidency that started in 1998. Through a military coup in April 2002 the president increases his power. This culminates in a devastating oil strike from 54
December 2002 to February 2003, which sends the economy into a severe recession, during which Venezuela loses 24 percent of GDP. From the second quarter of 2003 the political situation begins to stabilize (Weisbrot and Sandoval, 2007). The exchange rate shows great volatility in 2002, but with a clear depreciating trend, with large depreciations in February 2002, May-June 2002 and January 2003. The bolivar depreciates in 12 months by 140%, going from 0.76 to 1.85 bolivar per USD. The interest rate increases sharply. In February 2003 the bolivar is pegged to the US dollar at 1.60 bolivar per USD (the Guardian, 2003). Since 2003 the country has devalued the main official exchange rate four times up to 2015 and alternates between single and multiple exchange rate systems. In 2015 there are four exchange rates, including the (illegal) black market rate. The two lowest rates are used for imports of government-authorized priority goods such as food, medicine and car parts (Crooks, 2015). (2009) Reserves drop sharply in January 2009. However, this drop (from 33 to 19 billion US dollars) is not caused by an attempt to defend the currency, but part of the government’s periodic transfers reserves into an opaque “development fund” (Economist, 2010b). We do not consider this a currency crisis. (2010) As of January 8 2010 the multiple exchange rates of the bolivar vs the US dollar are devalued by 17% (for imports of medicine and food) and 100% (all other imports) (Economist, 2010a). (2011) As of January 1 2011, the bolivar is devalued. Dollars available for essential imports such as medicines and some foods (accounting for roughly a third of all forex transactions) are exchanged at 4.30 bolivars per dollar from 2.60 (MercoPress, 2011). (2013) President Chavez orders his government to weaken the exchange rate by 32 percent to 6.3 bolivars per dollar starting February 13 2013. The devaluation is meant to help narrow the budget deficit (the fiscal deficit has tripled in election year 2012) by increasing the amount of bolivars the government receives from oil exports (Devereux and Pons, 2013). (2016) Maduro announces that the primary exchange rate used for essential imports, such as food and medicine, will weaken to 10 bolivars per dollar from 6.3 in March 2016 (Rosati and Pitts, 2016). We conclude that Venezuela has experienced currency crisis in 1994M5-6, 1995M12–1996M4, 2002M2–2003M1, 2010M1, 2011M1, 2013M2 and 2016M2. The narrative confirms the quantitative determination of crises, except for the events in 2009M1.
55
Appendix B.2 Central and Eastern Europe (CEE) and Commonwealth of Independent States (CIS) Bulgaria (1996-199722) A severe twin banking and currency crisis occurs in 1996-1997. Fourteen banks, equivalent to almost a third of the banking sector, go under in a little over a year (Economist, 2014; Roussenova, 2002). In January 1997 Bulgaria experiences a currency crisis (Bruggemann and Linne, 1999). The currency crisis is a consequence of expansionary monetary policy, which leads to irreversible depletion of international reserves and causes the eventual collapse of the exchange rate (Kovatchevska, 2004). The underlying causes are a deteriorating macroeconomic environment (characterized by declining GDP and accelerating inflation) combined with external balance vulnerabilities (high balance of payment deficit, high external debt service obligations and short-term capital outflows). The country has to sign an agreement with the IMF to avoid a default on its debt. On 1 July 1997 Bulgaria moves from a free or managed-float system of the lev toward a currency board arrangement (first based on the German Deutsche mark, then on the euro) (Crespo-Cuaresma, Fidrmuc and Silgoner, 2005). We conclude that Bulgaria has not experienced any currency crises in the period 1998 to 2016. The narrative does not confirm the crisis identified by the quantitative criteria. Croatia (1999) Since January 1994 Croatia has officially a managed float exchange rate regime, but Reinhart & Rogoff (2004) consider it a de facto fixed peg, with a de facto band of the kuna around the Deutsche mark and the euro with a band width of +/- 2% (Crespo-Cuaresma, Fidrmuc and Silgoner, 2005). Per January 1st 1999 the kuna is pegged to the euro instead of to the German Deutsche mark. The Croatian kuna comes under strong speculative pressure in the first months of 1999 at the end of a domestic banking crisis that starts in March 1998. The exchange rate depreciates and the central bank intervenes through selling reserves (Tatomir, 2009). However, Ahec-Šonje (1999) do not consider these events as a currency crisis, but rather a “mild currency disturbance culminating at the beginning of 1999”. (2001) There is no evidence from the narrative that Croatia suffered a currency crisis. (2009) Croatia’s currency depreciates vis-à-vis the euro, but by approximately 5% from August 2008 to March 2009. Since the exchange rate volatility is so low due to the de facto fixed peg, the EMPI indicators have picked up the depreciation as a crisis. However, there is no evidence from the narrative that Croatia suffered a currency crisis in this year. 22
The 1996-1997 currency crisis in Bulgaria is not included in our sample. Due to data unavailability Bulgaria is analyzed from 1998 on.
56
We conclude that Croatia has not experienced any currency crisis in the period 1991 to 2016. The narrative confirms none of the three crises identified by the quantitative criteria.
Czech Republic (1997) The Czech currency crisis in May 1997 is the first typical currency crisis in the group of more advanced transition economies (Horvath, 1999). Problems start in 1996 in the financial sector (Bruggemann and Linne, 1999). Banks and investment funds that have been involved with privatization fall due to incompetence or fraud. Investors –both domestic and foreign- become wary of the Czech economy. The situation further deteriorates with a large trade deficit, tight monetary policy and austerity to curb the fiscal deficit. After using large amounts of reserves the national currency is allowed to float freely on May 26 1997, and drops immediately (Stroehlein et al., 1999). (1999) Since the exchange rate floats freely, the koruna has appreciated slowly over time, with some small, temporary depreciations, such as in 1999. These are not considered currency crises. We conclude that Czech Republic has experienced one currency crisis, in 1997M5. The narrative confirms one out of two crises identified by the quantitative criteria. Hungary (1990-1995) From 1990 until March 1995, the value of the forint is pegged to a currency basket, but adjusted through 23 –small– devaluations between January 1990 and March 1995, to support export competitiveness (Szikszai et al., 2012). In this respect the multiple devaluations are expected and small, and should not be considered currency crises. The country suffers a banking crisis when the banking system collapses in 1993 after a transitional recession in 1991-1993 (Banai et al., 2009; Szikszai et al., 2012). In March 1995 the government adopts a stabilization package, which gives new momentum to the process of institutional reform, reinstates fiscal discipline, introduces drastic measures of wage restraint and an import surcharge, devalues the forint by 10% and introduces a pre-announced crawling peg regime for the exchange rate (Golinelli and Rovelli, 2002). These events are not considered a currency crisis as the stabilization package is in the same order as the multiple devaluations that take place in the previous five years. We conclude that Hungary has not experienced any currency crisis in the period 1990 to 2016. The narrative confirms none of the two crises identified by the quantitative criteria. Kazakhstan (1999) The sharp devaluation of the Russian ruble in August 1998 and the consequent uncertainties for domestic and international investors become a major concern for Kazakhstan. Just before the outbreak of the Russian crisis the economy of Kazakhstan shows major improvements, in terms of higher economic growth, low inflation and a smaller current account deficit. When the ruble devalues, 57
Kazakhstan is affected through the trade channel (lower exports to Russia and to common trade partners) and the finance channel (lower capital flows due to increased uncertainty). To reduce the impact of the Russian crisis, Kazakhstan follows an expansionary fiscal policy (worsening the external debt position) and intensifies trade and exchange controls. Interest rates are initially not increased, to avoid a reversal in the finally recovered economic growth. Kazakhstan uses large amounts of reserves and by the end of 1998 increases interest rates to support the de facto exchange rate peg. It becomes clear by the beginning of 1999 that this is not sustainable and in April 1999 the exchange rate is floated, and loses more than 30% of its value in a month (Pastor and Damjanovic, 2001). (2007-2009) Since 2000, as the prices of oil and gas skyrocket, Kazakhstan’s banks borrow heavily on international markets to finance a massive construction, investment and real estate boom (Bohr, 2009). When international investors’ risk aversion increases in the 2007-2009 Global Financial Crisis, Kazakhstan suffers a banking crisis. Thanks to the initial favorable conditions in the public sector (low public debt and a savings fund from windfalls in oil revenues during the boom years 2000-2007) the troubled bank sector can be rescued with a large-scale policy package. Fiscal and monetary policies are supportive and the Kazakh tenge is devalued by 25 percent in February 2009, easing pressures on reserves and restoring competitiveness with Russia, its large neighbor and key trading partner (IMF, 2010). (2015) In August 2015 the tenge plunges by more than 30% after the authorities stop to intervene (Euler Hermes Economic Research, 2016). The reasons for the policy change are the falling oil prices and the earlier devaluations by its main trade partners, Russia and China (Kramer, 2015). We conclude that Kazakhstan has experienced three currency crises, in 1999M4-5, 2009M2-3 and 2015M8-2016M1. The narrative confirms all three crises identified by the quantitative criteria.
Poland (1989-1999) The “Balcerowicz Plan” in Poland is an ambitious reform plan that stabilizes the economy after a period of default, foreign exchange rationing and hyperinflation, and is considered to have laid the fundaments for the marketeconomy. The “shock therapy” is later followed in a.o. Czechoslovakia, the Baltic countries and Bulgaria (Roaf, Atoyan, Joshi and Krogulski, 2014). The exchange rate regime is focused on stabilization and bringing down hyperinflation. The exchange rate is fixed to the USD until May 1991, when it is devalued by 17% against the USD and fixed to a basket of currencies until October 1991 (Kokoszczyñski, 2001). By 1992 Poland has renegotiated its external debt under the Paris Club framework (Roaf, Atoyan, Joshi and Krogulski, 2014). From October 1991 to mid-1995 the country adopts a crawling peg policy with daily devaluation against the basket. Since the rate of inflation is higher than the rate of parity devaluation the real exchange rate is still appreciating. This leads to two devaluations, in February 1992 (12% against the basket) and August 1993 (8%) (Pruski and Szpunar, 2005). In May 58
1995 Poland switches to a crawling band policy to let the Polish zloty appreciate from its undervalued level. Since 2000 the currency is free floating (Kokoszczyñski, 2001). The country becomes an OECD member in 1996, and EU member in 2004 (the currency has been floating freely even that Poland is part of the EU). There is no evidence in the narrative that the episodes of relatively small devaluations and switches in exchange rate regimes should be considered currency crises. (2008) In the run-up to the crisis in 2008 Poland is one of few CEE countries to run fiscal deficits and to increase the external debt (the other countries are Czech Republic and Hungary) (Roaf, Atoyan, Joshi and Krogulski, 2014). The free floating zloty loses almost half its value (from 3.20 zloty per euro in July 2008 to 4.65 in February 2009). This depreciation is believed to be part of the reason why Poland has not experienced a recession in 2008–200923, because it helps to sustain Polish exports (Adam et al., 2012). The currency crisis does not spread around: the country does not experience a banking crisis due to the sector’s sound financial condition, nor does it fall into deep fiscal problems (Adam et al., 2012). We conclude that Poland has experienced one currency crisis, in 2008M102009M2. The narrative confirms one out of three crises identified by the quantitative criteria (1991 is excluded as it follows directly on a period of hyperinflation).
Romania (1992) Romania starts its transition in 1990. In the first years the country suffers hyperinflation and the currency depreciates accordingly. After 1993 a macrostabilization program is applied, with positive consequences for economic growth and inflation rate (Constantin et al., 2011). The sharp depreciations up to and including 1992 are associated with the hyperinflation period that last until 1992. (1997) Romania suffers high inflation in early 1997 caused by price liberalization, a rapidly depreciating exchange rate and a deep recession, following years of delayed reforms (Roaf et al., 2014). Additionally, the populist policies in 1996 result in the widening fiscal deficit which has to be financed through money creation (Arefieva et al., 1999). The exchange rate drops by more than 75% between December 1996 and February 1997. (1999) Early 1999 problems related to the signing of the stand-by arrangement with the IMF (IMF, 1999), the start of the war in Yugoslavia and an overvalued currency contribute to the deterioration of inflation expectations and currency depreciation (Racaru et al., 2006). The authorities manage to temporarily reduce inflation by managing the exchange rate, but the strategy leads to considerable real exchange rate appreciation and a ballooning current account deficit – subsequently leading to a renewed currency crisis in late 1999 (Roaf et al., 2014). 23
In fact, Poland is the only country in Europe that maintained positive growth figures (Adam et al., 2012). It is important to note that some do not consider this as a currency crisis, because there was no negative impact on the real economy. We do not share this view, because there may be a negative impact for other agents in the economy, such as corporations with foreign-denominated debt or derivatives traders.
59
We conclude that Romania has experienced two currency crises, in 1997M1-2 and 1999M1-12. The narrative confirms two out of three crises identified by the quantitative criteria (1992 is excluded because of hyperinflation).
Russia (1997-1998) In 1997, Russia’s economic growth is positive for the first time since 1991. Inflation has decreased sharply, the World Bank and the IMF support the Russian economy by giving financial aid, and oil prices increase. However, in the fourth quarter of 1997, market sentiment deteriorates drastically as a result of the Asian crisis. In November 1997 the ruble comes under attack and reserves have to be used to maintain the exchange rate. World commodity prices start to drop as a result of the turmoil. The current account deficit increases and interest rates skyrocket (with negative impact on the already delicate fiscal position) to defend the exchange rate and reduce capital flight. In August 1998 the outbreak of a severe banking, currency and sovereign debt crisis cannot be prevented. By the beginning of September 1998 the ruble is floated freely and loses two thirds of its value in three weeks (Van der Wiel, 2013; Chiodo and Owyang, 2002). The Russia crisis is a major shock to emerging markets, with spreads widening sharply in Latin America and Asia as well as Central and Eastern Europe, contributing to Brazil’s currency crisis, and the collapse of the prominent US hedge fund, Long Term Capital Management (Roaf et al., 2014). (2008-2009) In the boom-years (2000-2007) Russia can reduce its debt-to-GDP ratio and accumulate reserves through fiscal surpluses, made possible by high economic growth and high oil revenues. With the shock caused by the unexpected bankruptcy of Lehmann Brothers in September 2008 Russia makes substantial use of its very high reserves to avoid sharp depreciations that can damage corporate, household, and bank balance sheets (Roaf et al., 2014). The Russian government adopts a policy of "controlled devaluation". Reserves drop by more than 60% (from 583 to 371 billion USD) between July 2008 and January 2009. The Russian ruble loses 25% of its value between July and December 2008. In January 2009 the ruble depreciates by more than 20% in one month, partly as a consequence of the political-economic conflicts between Russia and Ukraine related to gas deliveries to Europe (Gow, 2009). From February 2009 on reserves stabilize and the exchange rate even appreciates to its December 2008 level, but the country enters a deep recession. (2014) The exchange rate depreciates by more than 100%, from 33.60 in June 2014 to 68.90 in January 2015, and with the biggest change in the period November 2014 to January 2015. In the same period reserves drop by 24% and interest rates are doubled from 8.2% to 17%. The main causes of the currency crisis in Russia in 2014-2015 are the economic slowdown, the decreasing oil prices, and the Western economic sanctions as a consequence of the Krim annexation and Russia’s meddling in Ukraine (Rodionov et al., 2015). An additional cause for the speculative attacks is the fear for a financial meltdown because the corporate 60
sector is heavily indebted with foreign currency denominated external debt (Economist, 2014; Matthews, 2014). We conclude that Russia has experienced three currency crises, in 1998M8-11, 2009M1-2 and 2014M11-2015M1. The narrative confirms all three crises identified by the quantitative criteria.
Ukraine (1998) The shock from the Russia crisis in August 1998 spreads around the region, including Ukraine. It is felt strongest through the effect of the collapse of Russian imports, which halve in the months following the devaluation. Ukraine is forced to follow Russia with a large exchange rate devaluation: in September 1998 the currency decreases 50% in value. (Roaf et al., 2014). (2008) One of the most severely hit countries in CEE and CIS is Ukraine. The currency loses more than 50% of its value in the last quarter of 2008. Ukraine receives an emergency loan from the IMF in October 2008 for US$16.5 billion to avoid a banking and sovereign debt crisis (Gordon, 2009). (2014-2015) In the early months of 2014 Ukraine suffers strong political instability due to the conflict with Russia, and related internal political turmoil. The country has relatively high external debt (both public and private), which makes it more vulnerable for adverse currency movements. The economy is struggling to recover from a recession and has been shaken by capital flight, as worried investors move their money abroad (Walker, 2014). The Ukrainian hryvnia is devalued by more than 35% in January and February 2014, despite the depletion of 30% of the reserves in the same period. The IMF approves a $17.1 billion bailout for Ukraine to avoid a default, and other donors (including the World Bank, EU, Canada and Japan) add funds worth $15 billion (BBC, 2014). In the beginning of 2015 the situation deteriorates further. The conflict in East-Ukraine continues and reserves used to defend the currency drop to 5 billion USD, down from 10 billion USD in September 2014. In February 2015 the central bank announces that it can no longer support the currency with regular interventions and will allow greater fluctuations. The hryvnia falls by 70% against the USD in February 2015 (RT, 2015). Ukraine suffers from a widening current-account deficit and the IMF expects the economy to contract by 9 per cent in 2015 due to gas payments, debt servicing and the war in the east (Moore, 2015; Economist, 2015). We conclude that Ukraine has experienced three currency crises, in 1998M91999M3, 2008M10-12, 2014M2-2015M2. The narrative confirms all three crises identified by the quantitative criteria.
61
Appendix B.3 Asia Asia crisis (1997-1998) On July 2, 1997, Thailand devalues its currency relative to the US dollar. This development, which follow months of speculative pressures that have substantially depleted Thailand’s official foreign exchange reserves, mark the beginning of a deep financial crisis across much of East Asia. In subsequent months, Thailand’s currency, equity, and property markets weaken further as its difficulties evolve into a twin balance-of-payments and banking crisis. Malaysia, the Philippines, and Indonesia also allow their currencies to weaken substantially in the face of market pressures, with Indonesia gradually falling into a multifaceted financial and political crisis. Severe balance-of-payments pressures in South Korea bring the country to the brink of default. Across East Asia, capital inflows slow or reverse direction, and growth slows sharply. Banks come under significant pressures, investment rates plunge, and some Asian countries enter deep recessions, producing important spillovers to trading partners across the globe. Years of rapid domestic credit growth and inadequate supervision have resulted in a significant build-up of financial leverage and doubtful loans. Overheating domestic economies and real estate markets add to the risks and lead to increased reliance on foreign savings, reflected in mounting current account deficits and a build-up in external debt. Heavy foreign borrowing, often at short maturities, also expose corporations and banks to significant exchange rate and funding risks—risks that have been masked by longstanding currency pegs. When the pegs prove unsustainable, firms see sharp increases in the local currency value of their external debts, leading many into distress and even insolvency (Carson and Clark, 1998). Corsetti, Pesenti and Roubini (1998) argue that fundamental imbalances trigger the currency and banking crisis in 1997. Once the crisis starts, market overreaction and herding cause exchange rate, asset prices and economic activity to drop more than warranted by the initial weak economic conditions. India (1991) In 1991 the Indian rupee starts to depreciate. The authorities at the Reserve Bank of India slow the decline in value by expending international reserves. With reserves nearly depleted, however, the exchange rate is devalued sharply on July 1 and July 3 against major foreign currencies. Additionally, interest rates are increased sharply. The crisis has been attributed to continued current account deficits leading up to the crisis, triggered in 1990-1991 through Iraq’s invasion of Kuwait (higher oil prices and lower remittances) and slow economic growth in India’s export markets (Ghosh, 2006). Additionally, external debt has risen and foreign exchange reserves have declined steadily since the beginning of the 1980s (Cerra and Saxena, 2000). By June 1991 reserves cover only two weeks of imports, and India is close to defaulting on its sovereign debt for the first time in its history (Ghosh, 2006). (1993) In March 1993 India adopts a managed floating exchange rate regime (Dua and Ranjan, 2010; Prakash, 2012). The currency drops almost 20% in value versus 62
the US dollar. There is no evidence in the narrative that this is a currency crisis, only a change in the exchange rate regime. (2008) In the aftermath of Lehman Brothers the rupee depreciates by 25% vis-à-vis the US dollar between April and December 2008. The central bank of India intervenes heavily in the foreign-exchange market to stop the rupee from falling too sharply, using more than 20% of its reserves (Economist, 2008). The pressure on the rupee diminishes by early November 2008, and later the currency appreciates against the US dollar (Bajpai, 2011). We conclude that India has experienced two currency crises, in 1991M4-7, and in 2008M9-11. The narrative confirms two out of three crises identified by the quantitative criteria (1993M3 is excluded).
Indonesia (1997-1998) Indonesia is one of the hardest hit countries in the Asia crisis. The country asks for IMF assistance, which is granted in October 1997. A large number of banks is closed or bailed out (Nasution, 2000). The Indonesian rupiah loses most of its value, going from 2,450 rupiah per US dollar in June 1997 to 10,375 in January 1998, and to 14,900 in June 1998. (2000-200124) In the aftermath of the Asia crisis Indonesia establishes new policies, to reconstruct and recapitalize the banking sector. The movements of exchange rate are largely driven by changes in political and economic policy perceptions, even in terms of the announcement of new economic policies or new agreements with the IMF. The political situation becomes more stable with Abdurrahman Wahid winning the presidential election in 2000. Exchange rate fluctuations become relatively small and stability of macroeconomic conditions increase significantly (Kasri, 2011). No evidence of a new currency crisis in these years. (2008-2009) As most emerging economies’ currencies, the Indonesian rupiah comes under deep depreciation pressures. The exchange rate is influenced by developments in the GFC, turbulent commodity prices and slowing global economy which leads to deteriorating investor perceptions and market expectations. Risk perceptions soar, prompting investors to move their funds away from emerging markets. The current account also sustain pressure from plunging commodity prices and low economic growth in trading partner nations. These developments combine to put pressure on the rupiah (Bank of Indonesia, 2009). The international reserves deplete by around 7 billion US dollar (equivalent to 13% of reserves) in September-October owing to market interventions to prevent rupiah depreciation. The central bank abandons market interventions in October 2008 and allows the currency to be relatively free-floating. The rupiah loses almost 30% of its value in October and November 2008 (Titiheruw et al., 2009).
24
For Indonesia we also include 2000-2001 in the narrative, because these years are identified by Reinhart and Rogoff (2010) and Cumperayot et al. (2013) as a currency crisis.
63
We conclude that Indonesia has experienced two currency crises, in 1997M81998M3 and in 2008M10-11. The narrative confirms both crises identified by the quantitative criteria.
Korea (1997-1998) Korea develops strongly in the 1980s and 1990s, and in 1995 the country becomes member of the World Trade Organization, in 1996 of the Organization for Economic Cooperation and Development and the Bank for International Settlements. Korea is reclassified by the IMF as an “advanced economy”. With the bankruptcy of Hanbo Steel Industry in January 1997 and the investigations that follow, the banking sector turns out far less solid than officially stated – with non-performing loans at an estimated 20% (instead of the official 1%). At the same time market participants believe that the won and the Korean equities market are overvalued. When the crisis breaks out in Thailand, more pressure comes on the Korean won. Korea’s financial problems worsen in the second half of October, after Taiwan and Hong Kong are affected by Thailand’s crisis. In Korea six of the 30 largest conglomerates (chaebol) declare bankruptcy, causing more pressure on the banking system and on the potential bailout costs to the government. On October 23 Standard & Poor’s downgrades its assessment of Korea’s sovereign debt. Since August 1997 the won is falling against the U.S. dollar, complicating debt service on the large external debt. Although foreign reserves are high, a large portion of the central bank’s reserves consists of deposits in overseas branches of Korean banks, and those banks have committed the money to cover their own external debts. In November and December 1997 the won falls 20% resp. 45% versus the US dollar. In the same period the reserves drop from 30 to 20 billion US dollar. By November 1997 the country has no alternative than to ask for IMF assistance (Boughton, 2012). (2008-2009) Following the collapse of Lehman Brothers in September 2008, the financial and foreign exchange markets in Korea are thrown into turmoil. The Korean won plummets against major currencies because of the outflow of foreigners’ investment funds and the deterioration of foreign currency borrowing conditions for domestic banks. As worries about credit risk mount, credit crunches emerge in the financial markets (Chung, 2012; Lee and Rhee, 2012). The exchange rate drops from 1,200 won per US dollar in September 2008 to 1,470 in November 2008. Reserves drop from 240 to 200 billion US dollar between September and November 2008. We conclude that Korea has experienced two currency crises, in 1997M11-1998M2 and in 2008M10-11. The narrative confirms both crises identified by the quantitative criteria. Malaysia (1997-1998) As its neighboring countries, Malaysia is hit hard by the Asian crisis. The Malaysian ringgit depreciates by 80% between June 1997 and January 1998. 64
(2015) In 2015 political scandals, economic slowdown, uncertainty on China’s economy and a possible US Fed interest rate increase trigger fears of a crisis in Malaysia. Compared to the situation prior to the 1997-1998 Asia crisis Malaysia faces higher leverage –household, quasi-public and external–, making the country vulnerable for external shocks (Riley, 2015). However, compared to 1997 the currency floats in 2015, there is no current-account deficit (with fixed exchange rates), and the banking system is stronger (Economist, 2015b). The devaluation of the Chinese yuan in August 2015 affects countries in the region as the competitiveness diminishes. The ringgit depreciates by 26% against the US dollar from its peak in August 2014, reaching its lowest level since the 1997-1998 crisis, and reserves drop below the $100bn psychological threshold (Durden, 2015). Malaysia's currency crash has many causes: on one hand factors that affect many emerging economies, such as the worsening global outlook and the plunging commodity prices, and on the other hand country-specific factors, such as the political scandal centered around Prime Minister Najib Razak, the country’s relative high indebtedness (household, quasi-public and external debt were higher than in 1997), the high dependence on commodities (including oil), and the lack of structural reforms that other countries in the region implemented after the 19971998 crisis (Pesek, 2015; Economist, 2015a). We conclude that Malaysia has experienced two currency crises, in 1997M71998M1 and in 2015 M8. The narrative confirms both crises identified by the quantitative criteria.
Pakistan (1990-1996) The period 1988–1996 is characterized by repeated attempts to stabilize the economy amid weak efforts at structural reforms that push the economy into a series of vicious cycles: fiscal shortages, low public and private investments, low productivity and profitability, low employment in the private sector, and –to compensate– overstaffing in the public sector. The exchange rate arrangements of Pakistan in the 1990s consist of a managed float. The Central Bank seeks to manage the rupee-dollar exchange rate to balance between containing inflationary pressures and maintaining a competitive tradable sector. This policy results in long phases during which the Central Bank implements minor “technical” adjustments vis-a-vis the dollar, interspersed with more substantial step devaluations that allow the real effective exchange rate to recover eroded competitiveness (Zubair Khan, 1999). The devaluations are in the range of 5% to 10%, and are not considered currency crises. (1999-2000) The devaluations in May 1999 and September-October 2000 are similar to the devaluations in 1993, 1995 and 1996: a devaluation of approximately 10% in one month to restore competitiveness after countries in the region have suffered large depreciations in the Asia crisis. We conclude that Pakistan has not experienced currency crises in the period 1990 to 2016. Only the EMPI measures identify crises, which can be attributed to the 65
low volatility of the indicators; the (small) devaluations are policy and the narrative does not provide evidence that these should be considered crises. Philippines (1990) The peso depreciates sharply by the end of 1990. This balance of payment crisis is a consequence of the rise in oil prices following the Iraq invasion of Kuwait (Yap, 1996). Although some authors consider balance of payment crises as a different type of financial crisis (for instance Ishihara, 2005), it is more common to consider these as -first generation- currency crises. (1997) The initial manifestations of the crisis are similar to those in Indonesia, Korea and Thailand, including a loss in investor confidence resulting in large capital outflows, a decline in reserves, stock market collapses and large currency depreciations (IMF, 2000). When the Asia crisis starts in Thailand in early July, Philippines increases its interest rate and uses reserves to defend the Philippine peso. It is not sufficient and step-wise the peso is devalued in the following months, from 26.40 peso per US dollar in June 1997 to 42.40 in January 1998. (2000) A political crisis during president Joseph Estrada’s term leads to uncertainty and a 10% drop in the value of the peso in October 2000, and cumulative drop of 25% in the year (Genato Rebullida, 2006). The country responds to the pressure on the currency by allowing the exchange rate to move, by intervening in foreign exchange markets, as well as by using domestic monetary policies to defend the value of the exchange rate (Hernandez and Montiel, 2001). We conclude that Philippines has experienced three currency crises, in 1990M111991M1, 1997M7-12 and 2000M10. The narrative confirms all three crises identified by the quantitative criteria. Sri Lanka (2012) By 2011 the Sri Lanka rupee is considered overvalued. As a consequence the current account deficit increases. To defend the exchange rate of the rupee the Central Bank heavily intervenes, releasing some $2.5 billion dollars, or 25% of its reserves in a very short period of time. Later the government allows for a gradual depreciation of the rupee. A small devaluation in November 2011 is not enough to reverse the sharply increasing trade and current account deficits. The International Monetary Fund at the time believes that “the rupee is not less than 20% overvalued”. The Central Bank decides to float the rupee in February 2012. The exchange rate depreciates from 113 to 133 rupees in the following two months (Transparency International Sri Lanka, 2012; IMF, 2012). We conclude that Sri Lanka has experienced one currency crisis, in 2012M2-4. The narrative confirms this crisis identified by the quantitative criteria. Thailand (1997-1998) See separate section on Asia crisis. Thailand’s currency crisis in the summer of 1997 is the trigger for the Asian crisis, which spreads around the region (and beyond) in the second half of 1997. 66
(200025) In 2000 the Thai financial system is still very fragile, and a new central bank law is under discussion that would increase the independence of the bank. In 2000 the exchange rate stabilizes (Jansen, 2001). The country has moved to greater exchange rate flexibility compared to the pre-Asian crisis exchange rate regime; the post-crisis period is considered a relatively tranquil period (Hernandez and Montiel, 2011). There is no evidence in the narrative that this should be considered a currency crisis. We conclude that Thailand has experienced one currency crisis, in 1997M71998M1. The narrative confirms this crisis identified by the quantitative criteria.
Vietnam (1997-1998) The collapse of financial markets and exchange rates across Southeast and East Asia in 1997 and 1998 affects Vietnam much less than those of most other Asian countries. Part of the explanation is that the banking sector reforms are incomplete and financial markets are not developed fully: the Vietnamese dong is non-convertible and therefore not traded internationally. In order to maintain export competitiveness the Vietnamese dong is devalued twice since the outbreak of the Asia financial crisis, in February 1998 (5%) and in August 1998 (7%) (Jeffreys, 2006; Tranh, 2000). (2009) Prior to 2009 Vietnam experiences a twin deficit: a large current account deficit and a large fiscal deficit. The rigid exchange rate policy plays an important part in the chronic trade deficit of Vietnam. The government of Vietnam prefers a strong national currency and maintains a fixed exchange rate regime, pegged to the US dollar, with irregular adjustments. The rationale for a strong currency even though it is not supportive for the export-led growth strategy is that the country imports machinery and most of its intermediate inputs for exporting. More importantly, a strong currency is an advantage for its foreign currency denominated public debt. In the last quarter of 2008 and the first quarter of 2009 a sudden stop in capital flows and a steep decline in exports put pressure on Vietnam to devaluate the dong (Nguyen et al., 2011). The central bank intervenes in the exchange rate market to stabilize the currency (Thanh et al., 2000), but the pressure remains. By late November 2009, the central bank devalues the local currency by approximately 5% against the US dollar (Nguyen et al., 2011). We conclude that Vietnam has not experienced a currency crisis in the period 1990-2016. The narrative does not confirm any of two crises identified by the quantitative criteria; the periods 1998M2-8 and 2009M11 are characterized by high pressure on the exchange rate, but are not to be considered currency crises.
25
For Thailand we also include 2000 in the narrative, because this year is identified by Reinhart and Rogoff (2010) and Kaminsky (2006) as a currency crisis.
67
Appendix B.4 Rest of the world Egypt (1990-1991) Egypt faces a severe foreign exchange shortage before the Gulf war breaks out. Its debt arrears are piling up and it becomes more difficult to obtain new loans. The Gulf crisis war makes this situation even worse, as it affects remittances sent home by Egyptian workers in the Gulf, Suez Canal tolls and revenues from tourism, as well as merchandise exports to Iraq and Kuwait. Support from the US and IMF and a Paris Club deal bring relief (Pripstein, 1991). In this period the exchange rate is devalued twice, from 1.10 to 2.00 Egyptian pounds (EGP) per US dollar in July 1990 and then to 3.17 in March 1991. These devaluations coincide with reforms towards a more efficient functioning of the economy (El-Shazly, 2011). (1999-200126) In the second part of the 1990s the economy experiences balanceof-payments and banking problems. The central bank uses administrative measures in 1999 that aim at rationalizing the demand for foreign currency. At the same time, the central bank releases slowly part of its international reserves and keeps a high interest rate. In January 2001, the authorities devalue the currency and effectively allow a crawling peg regime. In August 2001 the authorities allow a wider band for exchange rate movements (within 3% compared to an earlier 1.5% of the market-based rate of the central bank) and further devalue the Egyptian pound. In the year 2001 the currency drops from 3.69 to 4.49 EGP per US dollar, approximately 20%. Although both events (1999 and 2001) are classified by ElShazly (2011) as currency crises we disagree: the pressure on the exchange rate is not strong enough, nor is there a large (or fast) enough depreciation to call these events currency crises in our definition. The definition of El-Shazly (2011) of a currency crisis is “lighter” than common with the purpose to capture all episodes with increased pressure on the exchange rate – not necessarily equal to currency crises. The events seem to be based more on regime and policy changes. (2003) In January 2003, the authorities adopt a floating exchange-rate regime and thus effectively cancel the central bank rate that has been used since 2001 as a reference market rate (El-Shazly, 2011). This decision is preceded by several steps in 2001 and 2002 to decrease the value of the Egyptian pound against the dollar, and leads to an increase in the exchange rate, from EGP 3.85 to EGP 4.51 to the US dollar. The decision to float the pound leads to a depreciation of the pound, from EGP 4.51 to EGP6.14 by July 2003 (InfoMineo, 2016). (2016) Egypt announces on November 3 2016 that it will free float the Egyptian pound. The currency is devaluated 48%, from 8.88 to 13.00 per US dollar. The new rate is intended to serve as jumpstart the market, from where the central bank will float the pound. At the same time the central bank increases its main policy rate by 300 basis points (FocusEconomics, 2016). A free-floating currency is key to 26
For Egypt we also include 2001 in the narrative, because this year is identified by Reinhart and Rogoff (2010) as a currency crisis.
68
securing a $12 billion IMF emergency loan that is agreed in August (Alkhalisi, 2016; Feteha et al., 2016). By the end of November the exchange rate has depreciated further, ending at 17.77 per US dollar, equivalent to a 100% depreciation since the beginning of the month. We conclude that Egypt has experienced three currency crises, in 1990M71991M3, 2003M1-M4, and 2016M11. The narrative confirms all three crises identified by the quantitative criteria.
South Africa (1996) South Africa is confronted with large capital inflows after the successful elections in 1994. The Reserve Bank is following a dual strategy: an interest rate policy based on monetary targets and a nominal exchange rate stabilization policy. The unsterilized interventions to avoid appreciation in 1994 result in monetary expansion. This leads to inflationary pressure in 1995. Additionally, the liberalization policy (also with regard to exchange controls) along with the end of international sanctions result in concerns about the increase of the current account deficit, rising inflation and increasing public debt at the end of 1995. The depreciation episode starts in mid-February 1996, accumulating to 20 percent by April 1996. The Reserve Bank intervenes heavily in the spot and forward markets to counteract the depreciation (Knedlik, 2006). Also Bhundia and Ricci (2005) and Aron and Muellbauer (2000) identify these months as a currency crisis episode. (1998) During 1997 South Africa sees an increases in capital inflows due to a large interest rate differential to the US and a slow appreciation process. A currency crisis starts in April 1998 when funds are moving out of emerging markets in reaction to currency crises in Russia and Asia. Heavy interventions and interest rate increases cannot prevent a depreciation of 34 percent by July 1998 (Knedlik, 2006). Also Bhundia and Ricci (2005) and Aron and Muellbauer (2000) identify these months as a currency crisis episode. (2001) Between September and December 2001 the South African rand depreciates by 26 percent. The Reserve Bank does not intervene in favor of the rand. Potential explanations of the depreciation are the acceleration in money growth in 2001, the delay of the privatization of the telecom service provider, the continued policy of drawing down the net open forward position of the Reserve Bank (intervention against the domestic currency), Reserve Banks announcements to tighten the enforcement of exchange controls in October, contagion from political and social trouble in Zimbabwe, and a crisis in Argentina (Knedlik, 2006). Also Bhundia and Ricci (2005) identify these months as a currency crisis episode. (2008-2009) A sudden stop in June-July 2008 of international capital flows produces a collapse of the exchange rate. Prior to the crisis South Africa runs current account deficits, which are financed by inflows of private capital (African Development Bank, 2009). South Africa enters the crisis with a greater degree of vulnerability than other BRICKS (Brazil, Russia, India, China, Korea and South Africa), having a large current account deficit, high interest rates, and high inflation 69
(Lin et al., 2013). The collapse of commodity prices hits the mining and commodity industries particularly hard (Viegi, 2008). We conclude that South Africa has experienced four currency crises, in 1996M4, 1998M6-8, 2001M12 and 2008M10-2009M1. The narrative confirms all four crises identified by the quantitative criteria.
Turkey (199127) The first Gulf War (August 1990 – February 1991) affects Turkey’s economy. The nominal exchange rate depreciates some 30% in the first four months of 1991, but in real terms the depreciation is barely 10%. Reserves drop in the same period and the interest rate increases. For other countries these movements would point towards a crisis, but since Turkey has a very volatile past, the event is identified as a crisis by only two definitions. We have not been able to find any evidence that this episode is to be considered a currency crisis –other than the external databases based on large nominal exchange rate depreciations. (1990-1994) After the large structural reforms in the 1980s Turkey is an example of a ‘‘success story’’ for other developing countries. In the early 1990s inflation has reduced sharply, export earnings have increased, economic growth is high, and the financial system is liberalized. However, a deep financial crisis in the beginning of 1994 overshadows this success. A sharp downgrading in Turkey’s credit rating in the beginning of 1994 plays a triggering role in the occurrence of a severe currency crisis, which soon spreads to the banking sector (Ari, 2012). Concern on the viability of the economy and the debt situation leads to a sudden capital outflow (Rodrik, 2012), causing a sharp depreciation (the Turkish lira loses half its value from January to April 1994) – despite the use of high interest rates and reserves to defend the currency (Celasun, 1998). The crisis is controlled with the announcement of a stabilization program supported by a standby agreement with the IMF on April 5, 1994 (Ari, 2012). (199628) The nominal depreciation is large, but in real terms the exchange rate is almost constant. Reserves and interest rates are not affected. Like the depreciation in 1991 this would be an inflation-driven depreciation of the nominal exchange rate, which does not necessarily indicate a speculative attack (Bussiere and Fratzscher, 2006). A good example why a too-mechanical approach (focus on large nominal depreciation) is not always a good idea. (2001) In February 2001 Turkey faces the second large currency crisis since the 1994 crisis. The country has structural problems with the high and persistent inflation, a troubled banking sector (poorly supervised, currency mismatch), a structural high fiscal deficit and consequent high public debt, and a current account deficit financed by foreign capital. With the Asia, Russia and Brazil crises in 27
For Turkey we also include 1991 in the narrative, because this year is identified by Reinhart and Rogoff (2010), Laeven and Valencia (2012) and Babecky et al. (2012) as a currency crisis. 28 For Turkey we also include 1996 in the narrative, because this year is identified by Reinhart and Rogoff (2010), Laeven and Valencia (2012) and Babecky et al. (2012) as a currency crisis.
70
the late 1990s the sentiment of the international capital markets turns negative. In November 2000 a banking crisis breaks out, later followed by political instability and speculative attacks on the lira. The authorities defend the currency by using reserves and increasing the interest rates, but are forced to let the currency float on February 22 2001 (Ari, 2012). (2008-2009). In the run-up to the GFC Turkey faces very high unemployment rates, a highly indebted private sector, and difficulties to attract direct investments from abroad (Ari and Cergibozan, 2016). Its economic growth strategy (focus on export earnings and short-term capital inflows to finance the increasing current account deficit) make the country vulnerable for to external shocks. The Global Financial Crisis spreads after September 2008 via financial markets and trade. The direct impact is net capital outflows, currency depreciation, a fall in stock prices, rising risk premia and tightening liquidity in the banking sector (Rawdanowicz, 2010). We conclude that Turkey has experienced three currency crises, in 1994M1-4, 2001M2-3, 2008M10. The narrative confirms all three crises identified by the quantitative criteria.
References for Appendix B: Narrative References for Appendix B.1: Latin America and Caribbean Argentina Al-Shawwa, Yasmeen (2015). Argentina's Black Market for Pesos Is Going Dark. Fortune, 28 December 2015. Retrieved from: http://fortune.com/2015/12/28/argentina-peso-black-market/ Economist (2005). A victory by default? Retrieved from: http://www.economist.com/node/3715779 March 3, 2005. Economist (2014): " First decline, now fall. January 23 2014. Retrieved from: http://www.economist.com/blogs/americasview/2014/01/argentinas-peso Guidotti, Pablo and Juan Pablo Nicolini (2016). The Argentine banking crises of 1995 and 2001: An exploration into the role of macro-prudential regulations. Federal Reserve Bank of Minneapolis. First draft: May 2016. Kaminsky, Graciela, Amine Mati and Nada Choueiri (2009). “Thirty years of currency crises in Argentina: external shocks or domestic fragility?, NBER Working Paper, No. 15478. Moyer, Liz (2016). Argentina’s debt settlement ends 15 year battle. New York Times. February 29 2016. Retrieved from: http://www.nytimes.com/2016/03/01/business/dealbook/argentinas-debtsettlement-ends-15-year-battle.html Stiglitz, Joseph E. and Martin Guzman (2016). What's on Argentina's balance sheet? WEF Forum. February 5 2016. Retrieved from: 71
https://www.weforum.org/agenda/2016/02/what-s-on-argentina-s-balancesheet/ Weisbrot, Mark, Rebecca Ray, Juan A. Montecino, and Sara Kozameh (2011). The Argentine Success Story and its Implications. CEPR. October 2011
Brazil Buckman, Robert T. (2013). Latin America 2013. The World Today series. Rowman and Littlefield. ECB (2016). Update on economic and monetary developments. European Central Bank Eurosystem Economic Bulletin, Issue 1 / 2016. Retrieved from: https://www.ecb.europa.eu/pub/pdf/ecbu/eb201601.en.pdf Fraga (2000). Monetary Policy During the Transition to a Floating Exchange Rate: Brazil's Recent Experience. Finance and Development. IMF, Volume 37, Number 1, March 2000. Retrieved from: http://www.imf.org/external/pubs/ft/fandd/2000/03/fraga.htm Gallas, Daniel (2015). “Real worries: Why is Brazil's currency now so weak?”. BBC News. October 7 2015. Retrieved from: http://www.bbc.com/news/business34455980 Gruben, William C. and John H. Welch (2001) Banking and currency Crisis Recovery: Brazil’s Turnaround of 1999. Federal Reserve Bank of Dallas. IMF (2007). Brazil: Helping Calm Financial Markets. November 21, 2007. Retrieved from: https://www.imf.org/external/np/exr/articles/2007/112107.htm Chile Agosin, Manuel R and Alexis Montecinos (2011). Chile en los años 2000: evolución macroeconómica y financiera. Mimeo Universidad de Chile. March 2011. Briones, Ignacio (2010). Chile Country Report. In: Bertelsmann Stiftung (ed.), Managing the Crisis. A Comparative Assessment of Economic Governance in 14 Economies. Gütersloh: Bertelsmann Stiftung, 2010. Colombia Arguello, Ricardo (2011). The international economic crisis and the Colombian crisis. Universidad del Rosario Working Paper No. 98. May 2011. Arias, Andres F. (2000). “The Colombia banking crisis: Macroeconomic Consequences and What to Expect”. Mimeo. August, 2000 Galindo, Arturo J. and William F. Maloney (1998). “Second Thoughts on Second Moments. Panel Evidence on Asset-Based Models of Currency Crises”. Policy Research Working Paper 1939. World Bank. Lemaitre (2015). Collapsing currency signals need for reform in Colombia. Global Risks Insights. July 29 2015. Retrieved from: http://globalriskinsights.com/2015/07/collapsing-currency-signals-need-forreform-in-colombia/
72
Steiner, Roberto and Hernan Vallejo (2010). Economic history. In: Colombia: a country study. Federal Research Division, Library of Congress ; edited by Rex A. Hudson. 5th edition. Vargas, Hernando (2005). Public Debt Market Risk: The Effects on the Financial System and on Monetary Policy - The Case of Colombia. BIS Working Papers No. 28. Retrieved from: http://www.bis.org/publ/bppdf/bispap28j.pdf Villar, Leonardo and Hernán Rincon (2000). The Colombian economy in the nineties: capital flows and foreign exchange regimes. Paper presented at the Conference on “Critical Issues in Financial Reform: Latin American-Caribbean and Canadian Perspectives”, University of Toronto, June 1-2, 2000.
Costa Rica Carstens, Augustin (2004). “Twenty Years Without a Crisis in Costa Rica: The IMF's View”. Speech by Agustín Carstens, Deputy Managing Director, IMF. July 12, 2004. Retrieved from: http://www.imf.org/en/News/Articles/2015/09/28/04/53/sp071204 Dominican Republic Boughton, James M. (2012). Chapter 9: “Five Fat Years: Recovery from the Debt Crisis, 1990–94 “. Taken from the book: “Tearing Down Walls. The international Monetary Fund 1990-1999”. International Monetary Fund. Retrieved from: https://www.imf.org/external/pubs/ft/history/2012/ Fearon, James and David Laitin (2006). “Dominican Republic”. Mimeo Stanford University. Retrieved from: http://web.stanford.edu/group/ethnic/Random%20Narratives/Dominican%20Rep ublicRN1.2.pdf Jaramillo, Laura and Cemile Sancak (2007). “Growth in the Dominican Republic and Haiti: Why has the Grass Been Greener on One Side of Hispaniola?”. IMF Working Paper WP/07/63. NY Times (2003). Dominican Republic in crisis. December 29 2003. Retrieved from: http://www.nytimes.com/2003/12/29/opinion/dominican-republic-in-crisis.html Sánchez-Fung, José R. (2005). “Exchange Rates, Monetary Policy, and Interest Rates in the Dominican Republic during the 1990s Boom and New Millennium Crisis”. Journal of Latin American Studies Volume 37, Issue 4 November 2005, pp. 727-738 Guatemala Esquivel, Gerardo and Felipe Larrain (1999). Currency Crises: Is Central America Different? Center for International Development at Harvard University. CID Working Paper No. 26. September 1999. Retrieved from: https://www.hks.harvard.edu/index.php/content/download/69202/1249618/versi on/1/file/026.pdf
73
Morán Samayoa, Hilcías Estuardo (1999). “Dinámica macromonetaria de una crisis cambiaria para Guatemala”. Retrieved from: http://www.banguat.gob.gt/inveco/notas/articulos/envolver.asp?karchivo=1401& kdisc=si
Jamaica Harriott, Kevin (2000). “Essential Facts about Inflation”. October 2000. Mimeo Bank of Jamaica. Nelson-Douglas, Bosede (2001). “Estimation of speculative attack models and the implications for macroeconomic policy: the implications for macroeconomic policy: Jamaica 1991—00”. Mimeo Research & Economic Programming Division, Bank of Jamaica. January 2001. Retrieved from: http://www.boj.org.jm/uploads/pdf/papers_pamphlets/papers_pamphlets_Estima tion_of_Speculative_Attack_Models_and_the_Implication_for_Macroeconomic_P olicy.pdf Mexico Dungey, Mardi, Renee Fry, Brenda Gonzalez-Hermosillo and Vance Martin (2002). International Contagion Effects from the Russian crisis and the LTCM near-collapse. IMF Working Paper 02/74. April 2002. Manasse, Paolo, Nouriel Roubini and Axel Schimmelpfennig (2003). Predicting Sovereign Debt Crises. IMF Working paper 0 3/221, November 2003. Musacchio, Aldo. "Mexico's Financial Crisis of 1994-1995." Harvard Business School Working Paper, No. 12–101, May 2012. Retrieved from: https://dash.harvard.edu/bitstream/handle/1/9056792/12-101.pdf?sequence=1 Rojas-Suarez, L. (2011). The international financial crisis: Eight lessons for and from Latin America. In F. Fukuyama and N. Birdsall (Eds.), New Ideas in Development after the Financial Crisis. Baltimore, Maryland: The John Hopkins University Press. WEO (2009). World Economic Outlook: Crisis and Recovery. April 2009. International Monetary Fund. Paraguay Franks, Jeffrey, Valerie Mercer-Blackman, Randa Sab, and Roberto Benelli (2005). Paraguay: Corruption, Reform, and the Financial System. IMF. Retrieved from: https://www.imf.org/external/pubs/nft/2005/paraguay/reform.pdf Garcia-Herrera, A. (1997). Banking crises in Latin America in the 1990s: lessons from Argentina, Paraguay and Venezuela. IMF Working Papers 97/140. October 1997. IMF (2009). Paraguay: 2009 Article IV Consultation—Staff Report; Public Information Notice on the Executive Board Discussion; and Statement by the Executive Director for Paraguay. June 2009. Retrieved from: https://www.imf.org/external/pubs/ft/scr/2009/cr09182.pdf
74
Peru
Hidalgo, Manuel (1999). Peru’s banking crisis: just the beginning. Executive Intelligence Review, Volume 26, Number 2, January 8, 1999. Humala, Alberto and Gabriel Rodriguez (2009). Foreign Exchange Intervention and Exchange Rate Volatility in Peru. Banco Central de Reserva del Peru. Working Papers 2009-008. April 2009. IMF (1999). IMF Concludes Article IV Consultation with Peru. Public Information Notice (PIN) No. 99/56. July 6, 1999. Retrieved from: https://www.imf.org/external/np/sec/pn/1999/pn9956.htm McClintock, Cynthia (2000). The United States and Peru in the 1990s: cooperation with a critical caveat on democratic standards. Mimeo George Washington University. June 2000. Retrieved from: https://www2.gwu.edu/~clai/working_papers/McClintock_Cynthia_06-00.pdf Quispe, Zenon and Renzo Rossini (2010). Monetary policy during the global financial crisis of 2007–09: the case of Peru. BIS Papers No 54. Retrieved from: http://www.bis.org/publ/bppdf/bispap54r.pdf Rossini, Renzo and Zenon Quispe and Enrique Serrano (2013). Foreign exchange intervention in Peru. BIS Papers No 73. Retrieved from: http://www.bis.org/publ/bppdf/bispap73r.pdf Santos, Alejandro (2015). Peru’s road to economic success. IMF and Diálogo a fondo, October 1 2015. Retrieved from: http://www.imf.org/external/np/blog/dialogo/100115.pdf
Uruguay de la Plaza, Luis and Sophie Sirtaine (2005). “An Analysis of the 2002 Uruguayan Banking Crisis”. World Bank Policy Research Working Paper 3780, December 2005 Venezuela Corrales, Javier (1999). “Venezuela in the 1980s, the 1990s and beyond”. Revista: Harvard Review of Latin America. Fall 1999. Retrieved from: http://revista.drclas.harvard.edu/book/venezuela-1980s-1990s-and-beyond Crooks, Nathan (2015). “Venezuela, the Country With Four Exchange Rates”. Bloomberg. February 19 2015. Retrieved from: https://www.bloomberg.com/news/articles/2015-02-19/venezuela-the-countrywith-four-exchange-rates Devereux, Charlie and Corina Pons (2013). “Chavez Devaluation Puts Venezuelans to Queue on Price Raise”. Bloomberg. February 11 2013. Retrieved from: https://www.bloomberg.com/news/articles/2013-02-11/chavez-devaluation-putsvenezuelans-to-queue-before-price-raise Economist (2010a). “The weakening of the “strong bolívar”. January 14 2010. Retrieved from: http://www.economist.com/node/15287355 Economist (2010b). “Disappearing dollars”. September 16 2010. Retrieved from: http://www.economist.com/node/17043041 75
Garcia-Herrero, Alicia (1997). “Banking crises in Latin America in the 1990s: lessons from Argentina, Paraguay and Venezuela”. IMF Working Papers WP/97/140. Guardian (2003). “Venezuela pegs bolivar to dollar”. February 6, 2003. Retrieved from: https://www.theguardian.com/business/2003/feb/06/venezuela.internationalnew s MercoPress (2011). Venezuelan currency devalued again. January 3, 2011. MercoPress: South Atlantic News Agency. Retrieved from: http://en.mercopress.com/2011/01/03/venezuelan-currency-devalued-again Rosati, Andrew and Pietro Pitts (2016). "Venezuela Hikes Gas Price, Devalues Bolivar as Economy Tanks”. Bloomberg. February 17 2016. Retrieved from: https://www.bloomberg.com/news/articles/2016-02-17/venezuela-raisesgasoline-prices-for-first-time-since-1996-ikrf0ppn Weisbrot, Mark and Luis Sandoval (2007). “The Venezuelan Economy in the Chávez Years”. CEPR Working Paper, July 2007. Retrieved from: http://cepr.net/documents/publications/venezuela_2007_07.pdf
References for Appendix B.2: Central and Eastern Europe, and Commonwealth of Independent States Bulgaria Brüggemann, Axel and Linne, Thomas (1999). “How Good are Leading Indicators for Currency and Banking Crises in Central and Eastern Europe? An Empirical Test”. IWH Discussion Papers, No. 95. Halle Institute for Economic Research (IWH). Crespo-Cuaresma, Jesús, Jarko Fidrmuc and Maria Antoinette Silgoner (2005). “On the Road: The Path of Bulgaria, Croatia and Romania to the EU and the Euro” Europe-Asia Studies, Vol. 57 (6), 843-858. Economist, The (2014). “Why the run on banks?” Retrieved from: http://www.economist.com/blogs/easternapproaches/2014/07/bulgaria July 1, 2014. Kovatchevska, Preslava (2004). The Banking and Currency Crises in Bulgaria: 19961997. Center for Social and Economic Research. Working Paper series, No. 204. Roussenova, Lena (2002). The 1996-1997 financial crisis in Bulgaria. Paper prepared for the International Seminar on Comparative Experiences in Confronting Banking Sector Problems in Central/Eastern Europe and Central Asia, April 22-24, 2002. Croatia Ahec-Šonje, Amina (1999). “Leading indicators of currency and banking crises: Croatia and the world”. Economic Trends and Economic Policy (Privredna kretanja i ekonomska politika), 1999, No. 75, 31-85. 76
Crespo-Cuaresma, Jesús, Jarko Fidrmuc and Maria Antoinette Silgoner (2005). “On the Road: The Path of Bulgaria, Croatia and Romania to the EU and the Euro” Europe-Asia Studies, Vol. 57 (6), 843-858. Tatomir, S. (2009). “Exchange Market Pressure on the Croatian Kuna”. Financial Theory and Practice 33 (2), 187-199. Reinhart, Carmen M. and Kenneth S. Rogoff (2004). “The Modern History Of Exchange Rate Arrangements: A Reinterpretation”. NBER Working Paper No. 8963.
Czech Republic Brüggemann, Axel and Thomas Linne, 2002. “Are the Central and Eastern European Transition Countries still vulnerable to a Financial Crisis? Results from the Signals Approach”, IWH-Discussion Papers No.157. Horvath, Julius (1999). “The May 1997 Currency Crisis in the Czech Republic”. PostCommunist Economies,Volume 11, Issue 3, Pages 277-298 Stroehlein, Andrew, Jan Culik, Steven Saxonberg and Kazi Stastna (1999). “The Czech Republic 1992 to 1999: From unintentional political birth to prolonged political crisis”. Central Europe review. Vol 1, No 12, September 1999. Retrieved from: http://www.ce-review.org/99/12/stroehlein12.html Hungary Banai, Ádám, Júlia Király and Márton Nagy. “The demise of the halcyon days in Hungary: “foreign” and “local” banks – before and after the crisis”. BIS Papers No 54. Retrieved from: http://www.bis.org/publ/bppdf/bispap54l.pdf Golinelli, Roberto and Riccardo Rovelli (2002). “Painless Disinflation? Monetary Policy Rules In Hungary, 1991-1999“. Economics of Transition, Volume 10 (1), 55–91 Szikszai, Szabolcs, Csilla Raffai and Andras Tóthmihály (2012). “The Hungarian Financial System”. FESSUD, Studies in Financial Systems No 8. Retrieved from: fessud.eu/wp-content/uploads/2012/08/Hungary-studies1.pdf Kazakhstan Bohr, Annette (2009). “Kazakhstan: End of the Banking Boom”. Chatham House, REP Programme Paper: REP PP 2009/01. May 2009. Retrieved from: https://www.chathamhouse.org/sites/files/chathamhouse/public/Research/Russia %20and%20Eurasia/0509_bohr.pdf Euler Hermes Economic Research (2016). “Kazachstan: Hard hit by persistently low oil prices and impact of Russia crisis”. September 2016. Retrieved from: http://www.eulerhermes.com/economic-research/countryreports/Pages/Kazakhstan.aspx IMF (2010). “Republic of Kazakhstan: 2010 Article IV Consultation—Staff Report; IMF Country Report No. 10/241”. July 2010. Retrieved from: http://www.imf.org/external/pubs/ft/scr/2010/cr10241.pdf
77
Kramer, Andrew E. (2015). “Kazakhstan’s currency plunges”. New York Times. August 20 2015. Retrieved from: http://www.nytimes.com/2015/08/21/business/international/kazakhstanscurrency-plunges.html Pastor, Gonzalo and Tatiana Damjanovic (2001). “The Russian financial crisis and its consequences for Central Asia”. IMF Working papers 01/169. Retrieved from: https://www.imf.org/external/pubs/ft/wp/2001/wp01169.pdf
Poland Adam, Michał, Witold Koziński and Janusz Zieliński (2012). To what extent can central banks influence exchange rates with foreign exchange interventions? The case of Poland. BIS Papers No 73. Retrieved from: http://www.bis.org/publ/bppdf/bispap73t.pdf Kokoszczyñski, Ryszard (2001). From Fixed to Floating: Other Country Experiences: The Case of Poland. Paper presented at the IMF seminar “Exchange Rate Regimes: Hard Peg or Free Floating?”. Washington, DC, March 19-20, 2001. Retrieved from: https://www.imf.org/external/pubs/ft/seminar/2001/err/eng/kokos.pdf Pruski, Jerzy and Piotr Szpunar (2005). Exchange rate policy and foreign exchange interventions in Poland. BIS Papers No. 24. Retrieved from: http://www.bis.org/publ/bppdf/bispap24u.pdf Roaf, James, Ruben Atoyan, Bikas Joshi, Krzysztof Krogulski and an IMF Staff Team (2014). 25 Years of Transition Post-Communist Europe and the IMF. Regional Economic Issues Special Report. Washington, D.C.: International Monetary Fund, October 2014. Retrieved from: https://www.imf.org/external/pubs/ft/reo/2014/eur/eng/pdf/erei_sr_102414.pdf Romania Arefieva, Maria, Melissa E. Dodge and Erin Webster (1999). “Romania: A MultiCountry Evaluation of Trade Imbalances”. Elliott School of International Affairs, The George Washington University. April 1999. Retrieved from: http://internationalecon.com/tradeimbalance/romania.html Constantin D.L., Goschin Z., Danciu A.R. (2011). “The Romanian Economy from Transition to Crisis. Retrospects and Prospects”. World Business Institute Australia World Journal of Social Sciences 1(3): 155–171. Retrieved from: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.470.8646&rep=rep1&t ype=pdf IMF (1999). “Letter of Intent and Memorandum on Economic Policies of the government of Romania”. July 26, 1999. Retrieved from: https://www.imf.org/external/np/loi/1999/072699.htm Racaru, Irina, Mihai Copaciu and Ion Lapteacru (2006). “Early Warning Systems on Currency Crises”. National Bank of Romania. Occasional Papers No. 5. July 2006. Retrieved from: www.bnr.ro/files/d/Pubs_en/OP/op0506.pdf
78
Roaf, James, Ruben Atoyan, Bikas Joshi, Krzysztof Krogulski and an IMF Staff Team (2014). 25 Years of Transition Post-Communist Europe and the IMF. Regional Economic Issues Special Report. Washington, D.C.: International Monetary Fund, October 2014. Retrieved from: https://www.imf.org/external/pubs/ft/reo/2014/eur/eng/pdf/erei_sr_102414.pdf
Russia Chiodo, Abbigail J. and Michael T. Owyang (2002). “A Case Study of a Currency Crisis: The Russian Default of 1998”. Working Paper The Federal Reserve Bank of St. Louis, November-December 2002. Retrieved from: https://research.stlouisfed.org/publications/review/02/11/ChiodoOwyang.pdf Economist (2014). “What’s gone wrong with Russia’s economy”. December 16 2014. Retrieved from: http://www.economist.com/blogs/economistexplains/2014/12/economist-explains-16 Gow, David (2009). “Russia-Ukraine gas crisis intensifies as all European supplies are cut off”. The Guardian. January 7 2009. Retrieved from: https://www.theguardian.com/business/2009/jan/07/gas-ukraine Matthews, Chris (2014). Russian ruble’s fall: A classic ‘currency collapse’. Fortune. December 16 2014. Retrieved from: http://fortune.com/2014/12/16/russianruble-currency-collapse/ Roaf, James, Ruben Atoyan, Bikas Joshi, Krzysztof Krogulski and an IMF Staff Team (2014). 25 Years of Transition Post-Communist Europe and the IMF. Regional Economic Issues Special Report. Washington, D.C.: International Monetary Fund, October 2014. Retrieved from: https://www.imf.org/external/pubs/ft/reo/2014/eur/eng/pdf/erei_sr_102414.pdf Rodionov, C.G., V. V. Pshenichnikov and E. D. Zherebov (2015). Currency crisis in Russia on the spun of 2014 and 2015: causes and consequences. Procedia - Social and Behavioral Sciences 207 (2015 ) 850 – 857. Van der Wiel, Iris (2013). “The Russian Crisis 1998”. RaboResearch - Economic Research, Rabobank. September 16, 2013. Retrieved from: https://economics.rabobank.com/publications/2013/september/the-russian-crisis1998/ Ukraine BBC News (2014). “Ukraine economy: How bad is the mess and can it be fixed?”. May 1 2014. Retrieved from: http://www.bbc.com/news/world-europe-26767864 Economist (2015). “Crisis in Ukraine”. June 4 2015. Retrieved from: http://www.economist.com/blogs/graphicdetail/2015/06/ukraine-graphics Gordon, Lydia (2009). “The global financial crisis: Eastern Europe and CIS hit hard”. Euromonitor International. Retrieved from: http://blog.euromonitor.com/2009/01/the-global-financial-crisis-eastern-europeand-cis-hit-hard.html
79
Moore, Elaine (2015). “Ukraine’s debt crisis: Default fears rise as repayments are due and reserves shrink”. Financial Times. July 23 2015. Retrieved from: https://www.ft.com/content/2d058776-312b-11e5-8873-775ba7c2ea3d Roaf, James, Ruben Atoyan, Bikas Joshi, Krzysztof Krogulski and an IMF Staff Team (2014). 25 Years of Transition Post-Communist Europe and the IMF. Regional Economic Issues Special Report. Washington, D.C.: International Monetary Fund, October 2014. Retrieved from: https://www.imf.org/external/pubs/ft/reo/2014/eur/eng/pdf/erei_sr_102414.pdf RT (2015). Ukrainian hryvnia in free fall after Central Bank scraps currency support. 5 Feb, 2015. Retrieved from: https://www.rt.com/business/229607-ukrainehryvnia-free-fall/ Walker, Andrew (2014). “Ukrainian currency hits 10-year low” BBC World Service. 26 February 2014. Retrieved from: http://www.bbc.com/news/business-26353505
References for Appendix B.3: Asia Asia 1997-1998 crisis Carson, Michael and John Clark (1998). “Asian Financial Crisis, July 1997 December 1998”. Federal Reserve Bank of New York. Retrieved from: http://www.federalreservehistory.org/Events/DetailView/51 Corsetti, Giancarlo, Paolo Pesenti, and Nouriel Roubini (1998). “What Caused the Asian Currency and Financial Crisis? Part I: A Macroeconomic Overview,” NBER Working Paper 6833, National Bureau of Economic Research, Cambridge, MA, 1998. India Bajpai, Nirupam (2011). “Global Financial Crisis, its Impact on India and the Policy Response”. Working Paper No. 5. July 2011. Columbia Global Centers South Asia, Columbia University, Mumbai, India. Retrieved from: http://globalcenters.columbia.edu/mumbai/files/globalcenters_mumbai/Global_Fi nancial_Crisis_its_Impact_on_India_and_the_Policy_Response_CGCSA_Working_P aper_5.pdf Cerra, Valerie and Sweta Chaman Saxena (2002). What Caused the 1991 Currency Crisis in India? IMF Staff Papers, Vol. 49, No. 3. Dua, Pami and Rajiv Ranjan (2010). “Exchange Rate Policy and modeling in India”. Development Research Papers No. 33, Reserve Bank of India. Economist (2008). “India and the credit crisis“. October 14 2008. Retrieved from: http://www.economist.com/node/12411151 Ghosh, Arunabha (2006). “Pathways through financial crises: India”. Global Governance, Vol. 12, No. 4, Understanding Pathways Through Financial Crises and the Impact of the IMF (October–December 2006), 413-429. 80
Prakash, Anand (2012). “Major Episodes of Volatility in the Indian Foreign Exchange Market in the Last Two Decades (1993-2013): Central Bank’s Response”. Reserve Bank of India Occasional Papers, Vol. 33, No. 1 & 2. Retrieved from: https://rbidocs.rbi.org.in/rdocs/Content/PDFs/8MEVIF270614.pdf
Indonesia Bank of Indonesia (2009). 2008 Economic report on Indonesia. Retrieved from: http://www.bi.go.id/en/publikasi/laporantahunan/perekonomian/Documents/bc444121546e41ec92bae5fc6dfdcc76LPI2008 ingg.pdf Kasri, Rahmatina A. (2011). “Explaining the Twin Crises in Indonesia”. Working Paper in Economics and Business, Volume I No. 2. Department of Economics, Faculty of Economics, University of Indonesia. Retrieved from: ftp://ftp.repec.org/opt/ReDIF/RePEc/lpe/papers/201102.pdf Nasution, Anwar (2000). “The Meltdown of the Indonesian Economy: Causes, Responses and Lessons”. ASEAN Economic Bulletin, Vol. 17, No. 2, The Asian financial crisis: Hindsight, Insight and Foresight (August 2000), pp. 148-162 Titiheruw, Ira S., Hadi Soesastro and Raymond Atje (2009). “Global Financial Crisis Discussion Series. Paper 6: Indonesia”. Retrieved from: https://www.odi.org/sites/odi.org.uk/files/odi-assets/publications-opinionfiles/4327.pdf Korea Boughton, James M. (2012). Chapter 11: “Asian Flu: Financial Crisis in the Pacific“. Taken from the book: “Tearing Down Walls. The international Monetary Fund 1990-1999”. International Monetary Fund. Retrieved from: https://www.imf.org/external/pubs/ft/history/2012/ Chung, Hee Chun (2010). “The Bank of Korea’s policy response to the global financial crisis”. BIS Working papers No. 54. Retrieved from: http://www.bis.org/publ/bppdf/bispap54o.pdf Lee, Hangyong and Changyong Rhee (2012). “Lessons from the 1997 and the 2008 Crises in the Republic of Korea”. ADB Economics Working Paper Series No. 298. Retrieved from: https://www.adb.org/sites/default/files/publication/29659/economics-wp-298.pdf Malaysia Durden, Tyler (2015). “This Is What Happened The Last Time Malaysia Faced A Currency Crisis”. Zero Hedge. August 29, 2015. Retrieved from: http://www.zerohedge.com/news/2015-08-29/what-happened-last-timemalaysia-faced-currency-crisis Economist (2015a). “Plunging like it’s 1998”. Economist, August 8 2015. Retrieved from: http://www.economist.com/news/finance-and-economics/21660557rupiah-and-ringgit-plumb-depths-unseen-asian-financial 81
Economist (2015b). “Why emerging-market weakness is not a replay of the 1997 financial crisis”. Economist, August 23 2015. Retrieved from: http://www.economist.com/blogs/economist-explains/2015/08/economistexplains-17 Pesek, William (2015).”1990s Come Back to Haunt Malaysia”. Bloomberg News. August 4 2015. Retrieved from: https://www.bloomberg.com/view/articles/201508-10/1990s-come-back-to-haunt-malaysia Riley, Charles (2015). Ghosts of 1997 financial crisis return to haunt Asia”. CNN Money. August 18 2015. Retrieved from: http://money.cnn.com/2015/08/18/investing/malaysia-indonesia-economycurrency/
Pakistan Zubair Khan, Mohammad (1999). “Liberalization and Economic Crisis in Pakistan“. Asia Regional Integration Center, Asian Development Bank (ADB). Retrieved from: https://aric.adb.org/pdf/aem/external/financial_market/Pakistan/pak_mac.pdf Philippines Genato Rebullida, M.L.G. (2006). “Chapter 9: The Philippine executive and redemocratization”. From: Philippine Politics and Governance: An Introduction. Eds. Teresa S. Encarnacion Tadem, Noel M. Morada. University of the Philippines & Philippine Commission on Higher Education, 2006. Hernandez, Leonardo and Peter Montiel (2001). “Post-crisis exchange rate policy in five Asian countries: filling in the “Hollow Middle”?”. IMF Working Paper 01/170, November 2001. IMF (2000). Recovery from the Asian Crisis and the Role of the IMF. IMF Staff. June 2000. Retrieved from: https://www.imf.org/external/np/exr/ib/2000/062300.htm#box4 Ishihara, Yoichiro (2005). “Quantitative Analysis of Crisis: Crisis Identification and Causality”. World Bank Policy Research Working Paper 3598, May 2005. Yap, Josef T. (1996). Inflation and Economic Growth in the Philippines. Philippine Institute for Development Studies. PIDS Discussion Paper Series No. 96-11. Retrieved from: http://dirp4.pids.gov.ph/ris/dps/pidsdps9611.pdf Sri Lanka IMF (2012). “Sri Lanka: Seventh Review under the Stand-By Arrangement”. April 2012. Retrieved from: https://www.imf.org/external/pubs/ft/scr/2012/cr12198.pdf Transparency International Sri Lanka (2012). “The rupee crisis”. Report on the discussion on the Rupee Crisis on 25 April 2012. July 2012. Retrieved from: http://www.tisrilanka.org/pub/pp/pdf/Rupee_Eng_Long.pdf
82
Thailand Jansen, Karel (2001). “Thailand, financial crises and monetary policy”. Journal of the Asia Pacific Economy 6 (1), 124–152. Hernandez, Leonardo and Peter Montiel (2001). “Post-crisis exchange rate policy in five Asian countries: filling in the “Hollow Middle”?”. IMF Working Paper 01/170, November 2001. Vietnam Jeffreys, Ian (2006). “Vietnam. A guide to economic and political developments“. Routledge. Taylor & Francis Group. Nguyen, N. A., D. C. Nguyen, C. N. Nguyen, and T. H. Bui (2011). ‘Fiscal Issues in Vietnam Economy: Assessment on the Impact of Stimulus, Fiscal Tranparency and Fiscal Risk’, in Ito, T. and F. Parulian (eds.), Assessment on the Impact of Stimulus, Fiscal Transparency and Fiscal Risk. ERIA Research Project Report 2010-01, pp.249282. ERIA [online]. Available at: www.eria.org/publications/research_project_reports/images/pdf/y2010/no1/ch8F iscal_Is sue_Vietnam.pdf Thanh, Vo Tri, Dinh Hien Minh, Do Xuan Truong, Hoang Van Thanh and Pham Chi Quang (2000). “Exchange rate arrangement in Vietnam: information content and policy options”. East Asian Development Network. Individual Research Project. Retrieved from: http://eadn.org/files/Working%20Papers/WP_18_Vo_Tri_Thanh_Exhange_Rate_A rrangement_in_Vietnam_Information_Content_and_Policy_Options.pdf
References for Appendix B.4: Rest of the world Egypt Alkhalisi, Zahraa (2016). “Why Egypt just let its currency crash by 48%”. CNNMoney. November 3, 2016. Retrieved from: http://money.cnn.com/2016/11/03/news/economy/egypt-pound-devaluationbailout/ El-Shazly, Alaa (2011). “Designing an early warning system for currency crises: an empirical treatment”. Applied Economics, 43:14, 1817-1828. Feteha, Ahmed, Alaa Shahine , Tarek El-Tablawy and Tamim Elyan (2016). “Egypt Free Floats Pound, Raises Lending Rates to Spur Economy”. Bloomberg. November 3 2016. Retrieved from: https://www.bloomberg.com/news/articles/2016-1103/egypt-free-floats-pound-raises-lending-rates-to-spur-economy FocusEconomics (2016). “Egypt: Central Bank free floats the pound”. FocusEconomics, November 4 2016. Retrieved from: http://www.focuseconomics.com/countries/egypt/news/exchange-rate/central-bank-free-floatsthe-pound 83
InfoMineo (2016). “Egyptian Pound Devaluation and its Impacts”. June 20 2016. Retrieved from: http://www.infomineo.com/egyptian-pound-devaluation-and-itsimpacts-on-the-egyptian-economy/ Pripstein Posusney, Marsha. “Economic Impact of the Crisis in Egypt”. Middle East Report, Vol. 21, January/February 1991. Middle East Research and Information Project. Retrieved from: http://www.merip.org/mer/mer168/economic-impactcrisis-egypt
South Africa African Development Bank (2009). “Impact of the Global Financial and Economic Crisis on Africa”. African Development Bank – African Development Fund: Office of the Chief Economist. February 2009. Retrieved from: https://www.afdb.org/fileadmin/uploads/afdb/Documents/Knowledge/Financial% 20crisis_Impacts%20on%20Africa.pdf Aron, J. and Muellbauer, J. (2000): Estimating Monetary Policy Rules for South Africa. Central Bank of Chile. Working Papers, 89. Bhundia, A. J.; Ricci, L. A. (2005): The Rand Crises of 1998 and 2001: What have we learned? in: Nowak, M. & Ricci, L.A. (eds.). Post-Apartheid South Africa: The first ten years. 156-173. Washington D.C.: IMF. Knedlik, Tobias, 2006. “Signaling currency crises in South Africa”, IWH-Discussion Papers 19, October. Lin, C.Y.Y., L. Edvinsson, J. Chen and T. Beding (2013). Chapter 2: Impact of the 2008 Global Financial crisis. In: “National intellectual capital and the financial crisis in Brazil, Russia, India, China, Korea and South Africa”. Springer Verlag. Retrieved from: file:///C:/Users/K15077/Downloads/9781461460886-c1%20(1).pdf Viegi, Nicola (2008). “The impact of the financial crisis in South Africa”. Institute of Development Studies, Sussex. Turkey Ari, Ali (2012). “Early warning systems for currency crises: The Turkish case”. Economic Systems, 36, 391–410. Ari, Ali and Raif Cergibozan (2016). “The Twin Crises: Determinants of Banking and Currency Crises in the Turkish Economy”. Emerging Markets Finance and Trade, 52:1, 123-135 Celasun, Oya (1998). “The 1994 Currency Crisis in Turkey”. World Bank. Policy Research Working Paper. April 1998. Rawdanowicz, Łukasz. (2010). “The 2008-09 Crisis in Turkey: Performance, Policy Responses and Challenges for Sustaining the Recovery”, OECD Economics Department Working Papers, No. 819, OECD Publishing, Paris. Rodrik, Dani (2012). “The Turkish Economy after the Global Financial Crisis”. Ekonomi-tek Volume 1, No 1, January 2012, 41-61.
84