2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
Asian Currency Crisis: Adding Trigger to Vulnerability
June 22-24, 2008
Oxford, UK
Ismail Ait-Saadi
Curtin University of Technology, Sarawak, Malaysia
Mansor Jusoh
National University of Malaysia
Norghani Md.Nor
National University of Malaysia
1

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
Asian Currency Crisis: Adding Trigger to Vulnerability
ABSTRACT
The Asian crisis which erupted in 1997-98 has not only been one of the most damaging in decades, but also one of the most controversial. Now and after ten years; there is still no agreement among economists on what really caused it to happen. Some analysts have claimed that it is caused by weak fundamentals as it is argued by first generation models, while others have claimed exactly the opposite saying that the five Asian countries had in fact strong fundamentals and the only explanation is panic and herd behavior as suggested by second generation Models. Recently and as a reaction to this hot debate, many analysts started getting more convinced that Asian crisis had in fact elements of both first and second generation models.
Advocates of this view argue that the depth and the severity of the Asian crises cannot be explained solely by weak fundamentals. To make sense of what happened in Asia we introduce a new understanding of the crisis in terms of "Trigger" and "Vulnerability". We argue that crises can happen only if both trigger and vulnerability are present. The main contribution of this study is to use for the first time an empirical model using trigger and vulnerability as the only factors that explain the crisis. To that end we combine Classification And Regression Tree (CART) methodology with Logit model. The results support the argument that in addition to weak fundamentals (vulnerability), self-fulfilling prophecies (trigger) have played a major role in the collapse of the Asian economies
JEL: F31, F32, F41
Keywords: Currency Crises, Asia, Trigger, Vulnerability, CART
June 22-24, 2008
Oxford, UK
2

2008 Oxford Business &Economics Conference Program
I.
INTRODUCTION
ISBN : 978-0-9742114-7-3
The Asian crisis which erupted in 1997-98 has not only been one of the most damaging in decades, but also one of the most controversial. Now and after ten years; there is still no agreement among economists on what really caused it to happen. Some analysts have claimed that it is caused by weak fundamentals as it is argued by first generation models, while others have claimed exactly the opposite saying that the five Asian countries had in fact strong fundamentals and the only explanation is panic and herd behavior as suggested by second generation models. These conflicting explanations have convinced others that the standard models, namely first and second generation models are inadequate for explaining what really happened in Asia stressing the need to look for new models that can explain these new waves of crises(See, Berg & Pattillo, 1998; Krugman, 1999; Radelet & Sachs, 1998a; Stiglitz, 1998).
Krugman (1999) suggests a third generation model “
We badly need a “third-generation” crisis model both to make sense of the recent crises and to help warn of crises to come
” 1
. This pressing need for new models has lead to the birth and emergence of few alternative frameworks among which two deserve to be mentioned. The first one suggested by Radelet & Sachs (1998b) where they identify five main types of financial crises: Macroeconomic policy-induced crisis,
Financial panic, Bubble collapse, Moral hazard and Disorderly workout. The second interesting classification is the one suggested by Kaminsky (2003); she identifies six major types of crises:
First generation models, Second generation models, Third generation models, Sovereign Debt
Models, Sudden stop models and Self-fulfilling Models.
Despite the existence of these models which classify and explain currency crises, economists still differ substantially over the real causes behind them (Kaminsky & Schmukler,
1
Interestingly Paul Krugman (2001) explores ways to have a fourth generation model in which asset prices are at the source of the crisis. It is more of general financial crisis rather than just a currency crisis.
June 22-24, 2008 3
Oxford, UK

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
1999). The early models developed in the 1980s stress the role of weak fundamentals as being the main cause of the crises. These are often referred to as “first generation models” where monetization of the large budget deficit often leads to a massive loss of international reserves that ultimately forces the authorities to abandon the peg by either devaluing or floating the domestic currency. However, models of the 1990s called “second generation models” argue that even countries with immaculate economic fundamentals are also at risk of seeing their currency under attack. This is usually attributed to sudden shifts in investors’ sentiments which make the crisis self-fulfilling. This disagreement among economists became very apparent during the
Asian crisis where the question of whether Asian crisis was caused by weak fundamentals or by self-fulfilling prophecies is still a hot debate.
Some economists (Fundamentalists)
2
consider Asian crises as being mainly caused by weak fundamentals such as Budget deficit, Real exchange rate misalignment, and the ratio of M2 to international Reserves. They believe that misguided macroeconomic policies are at the heart of the causes of the Asian crisis. According to Fischer (1998), “ financial and corporate inefficiencies were at the epicenter of the economic crisis and have to be dealt with to restore durable growth
”. Corsetti, Pesenti & Roubini (1999) claim that “ the crisis reflected structural and policy distortions in the countries of the region. Fundamental imbalances triggered the currency and financial crisis in 1997”
.
On the other hand “the self-fulfilling view” held by some leading economists, suggests that Asian crisis was self-fulfilling and has nothing to do with bad economic policies or weak fundamentals. Bhagwati (1998) notes that “ none of the Asian economies that were hit had any serious fundamental problems that justified the panic that set in to reverse the entire huge
2
While the proponents of this view consider the deterioration in macroeconomic conditions as the main cause of the crisis, they do however accept that the panic created and the sudden shift in sentiments among investors may well have played a role in making the collapse more severe and damaging than what is warranted by fundamentals.
June 22-24, 2008 4
Oxford, UK

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 capital inflow…and the only explanation that accounts for the massive net outflows is panic and herd behavior ”. While referring to Asian crisis, Tobin (1998) argues that “ countries can suffer liquidity crises through no fault of their own”
. Radelet and Sachs (1998a) believe that the crisis
“ was triggered by dramatic swing in creditors’ expectations about the behavior of other creditors, thereby creating a self-fulfilling -although possibly individually rational- financial panic”
. Hill (1999) emphasizes the fact that “ the majority of pre-crisis economic and financial indicators in the crisis economies looked quite healthy”.
Recently and as a reaction to this hot debate, many analysts started getting more convinced that Asian crisis had in fact elements of both first and second generation models.
Meaning to say that what caused Asian crisis were neither weak fundamentals alone nor selffulfilling prophecies alone, but a combination of both. Advocates of this view argue that the depth and the severity of the Asian crises cannot be explained solely by weak fundamentals as most of the common indicators were not that bad and in fact some were even reasonably good. In their explanation of the twin crises in Asia Burnside et al (2000) made this point very clear when they chose a title “On the Fundamentals of Self-fulfilling Speculative Attacks” for their paper in which they argue, “ while self-fulfilling beliefs have an important role to play, twin crises do not happen just anywhere, they happen in countries where there are fundamental problems”
.
Radelet and Sachs (1998a; , 1998b) also highlight the self-fulfilling nature of the crisis as well as the role of fundamentals, such as growing short-term debt and expanding bank credit.
The proliferation of new models mirrors the complexity of such crisis and the theoretical ambiguity on this issue. In this study we adhere to the view that Asian crisis had elements of
June 22-24, 2008
Oxford, UK
5

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 both weak fundamentals and self-fulfilling prophecies. We suggest an alternative framework that can handle these differences and make sense of the competing explanations of currency crises in general and Asian crisis in particular. To make sense of what happened in Asia we introduce a new understanding of the crisis in terms of "Trigger" and "Vulnerability". We argue that crises can happen only if both trigger and vulnerability are present. Vulnerability means that the country is predisposed to suffer a currency crisis because of the weak "fundamentals" as suggested by first generation models. And "trigger" is the element -such as panic - that would send a vulnerable situation into a real collapse, and this is the argument of second generation models. The main contribution of this study is to use for the first time an empirical model using trigger and vulnerability as the only factors that explain the crisis. To that end we combine
Classification And Regression Tree (CART) methodology with Logit model. The results support the argument that in addition to weak fundamentals (vulnerability), self-fulfilling prophecies
(trigger) have played a major role in the collapse of the Asian economies.
The rest of the paper is organized as follows. Section two presents some puzzles that have surfaced after the Asian crisis and the way to deal with them. Section three introduces CART methodology and how it is used to identify elements of trigger and vulnerability and lastly section four concludes.
II.
NEW PUZZLES: THE MORE WE KNOW THE MORE WE
DON’T KNOW
Can we say that our understanding of currency crises is better than what it was a few decades ago? The answer is yes and no. Because of their devastating effects, researchers and policymakers have emphasized the necessity to further explore how these crises come about, how to prevent them from happening and how to deal with them when they hit, especially in
June 22-24, 2008
Oxford, UK
6

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 emerging markets. And this has without doubt deepened our understanding of currency crises over time. However, each time a crisis erupts we tend to look at it as something new and different from the previous ones which necessitates new understanding and new models. Asian crisis was no exception; it is seen by many as very unique “ it hit the most rapidly growing economies in the world, and prompted the largest financial bailouts in history. It is the sharpest financial crisis to hit the developing world since 1982 debt crisis. It is the least anticipated financial crisis in years ” (Radelet & Sachs, 1998b). That is why First and second generation models which once were considered standard models are no longer sufficient to explain the new waves of crises which seem to elicit new style of models (Eichengreen et. al, 1995).This view may not be shared by others who believe that Asian crisis is a classical financial crisis similar in many ways to the previous crises like the ones in Chile in 1982 and Mexico in 1994 (Chang &
Velasco, 1998; Honohan, Daniela, & Andrea, 1999).
While this debate may not end anytime soon, one thing for sure which made Asian crisis unique, is that it has in many ways challenged our views on how currency crises come about, and left us with many puzzles that are yet to be resolved. For example contagion theories suggest that contagion effects are usually channeled from large to small economies, but what happened in
Asia was exactly the opposite. Another puzzle is the defense of the currency, the common belief is that high interest rates may well defend the currency during a speculative attack, but empirical research stimulated by Asian crises tends to refute it
3
. The puzzle we would like to focus on in this study is maybe the most remarkable one: what are the economic and financial indicators that can cause currency crises?
The literature has identified a large number of economic and financial indicators that reflect weak fundamentals which lead to the collapse of the currency and ultimately to the crisis. However Asian experience shows that this was not always true, because
3 See Kraay (2000)
June 22-24, 2008
Oxford, UK
7

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 during the Asian crisis in 1997 -98, there were many other countries that had relatively worse economic fundamentals and yet did not suffer any crisis. How can this puzzle be explained?
Causes
One important term that needs to be clearly defined when talking about currency crises is the use of the term “Cause”. Part of the confusion over what causes crises to happen is due to the vague meaning and the ambiguous understanding of what is meant by causality. Furman and
Stiglitz (1998, p. 5) define causality as “ a factor that inevitably leads to a given consequence ” and interestingly come to the conclusion that based on this definition “ none of the alleged causes of the East Asian crisis satisfy
”. This is because Standard models of currency crises have not been able to answer the following question: why some countries that had weak economic fundamentals have not experienced crises, while others with relatively better fundamentals had?
Before we answer this question let’s present some facts about five leading and commonly used indicators of currency crises
4
. Real exchange rate appreciation, the ratio of M2 to foreign reserves, the ratio of current account to GDP, the ratio of short-term debts to foreign reserves and the ratio of domestic bank credit to GDP. Table 1 shows that the ratio of short-term debts to reserves was quite high in Indonesia, Korea and Thailand which suffered a crisis in July 1997, however Mexico and Argentina which also had a high ratio did not suffer any. The ratio of M2 to foreign reserves was in excess of one in all countries with the lowest in Chile, but none of the
Latin American countries were hit by the crisis. Real exchange rate overvaluation which is one of the most important and leading indicator of currency crises had appreciated by more than 22% percent in four Asian countries. But the appreciation was even higher in Latin American
4 See Bustelo (2000)
June 22-24, 2008
Oxford, UK
8

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 countries with the highest in Brazil (67%) and no pressure was put on the Brazilian currency. An astonishing example is South Africa which had a bad record in almost all indicators and yet was spared from the crisis. How could this be explained?
To answer the above question and make sense of the data we should try and understand the crises not in terms of causality but rather in relation to two important concepts: vulnerability and trigger. The fact is that weak fundamentals reflected by these economic and financial indicators do not guarantee the onset of crisis; there need to be some triggering elements that push the situation into a full blown crisis. Let's look at this more closely.
Vulnerability & Trigger
Vulnerability means that the country is predisposed to suffer a currency crisis because of the weak fundamentals reflected in the unsustainable deterioration in macroeconomic conditions.
Dornbusch (1997, p. 21) defines the state of vulnerability as a situation where
“if something goes wrong, then suddenly a lot goes wrong” . This is what First Generation models put emphasis on.
However as mentioned by Athukorala and Warr (2002, p. 37), “a state of vulnerability by itself does not give rise to a currency crisis; there need to be a certain disturbance (a trigger) that will push a vulnerable situation into an actual collapse .” Trigger is similar to what second generation models call panic or pure contagion. So theoretically if two countries with similar economic conditions are believed to be in the state of vulnerability, therefore making them exposed to an attack on their currencies, may not both suffer a currency crisis if the element of trigger is present only in one and not in the other. It is possible to represent this as follows:
Crisis = Vulnerability + Trigger
June 22-24, 2008
Oxford, UK
9

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
Understanding currency crises in relation to the concept of Trigger and Vulnerability tells us that currency Crises are the result of both weak fundamentals (Vulnerability) and Panic
(Trigger). Furthermore this approach helps us gain new insights into understanding how currency crises come about and why in certain countries not in others with similar fundamentals. For example, many studies have tried to understand and explain how the crisis has spread from
Thailand to other neighboring countries using contagion or herd behavior theories, but what theses studies can’t explain is how it started in the first place and why in Thailand not in other countries which had relatively weaker fundamentals? The answer is that Weak fundamentals make the country vulnerable to crisis but do not necessarily “Cause” one.
While the identification of vulnerability element may be quite easy, trigger element is not as tangible as vulnerability factors. Although many observers like to mention events such as the collapse of Hanbo Steel in Korea and the problems faced by Samprason Land in Thailand to pay its foreign debts and the political uncertainties in Indonesia, the Philippines and Malaysia, to be triggering events that ignited the fire, however we believe that triggering events can be very random and can be anything that would make market participants expect a devaluation of the currency. The study of crises in terms of vulnerability and trigger, suggests that the state of vulnerability is more likely to lead to a currency crisis because of the randomness of the trigger element. However trigger element alone is less likely to cause a currency crisis because the healthy economic conditions can easily absorb the shock and prevent the collapse of the currency.
June 22-24, 2008
Oxford, UK
10

2008 Oxford Business &Economics Conference Program
III.
METHODOLOGY
ISBN : 978-0-9742114-7-3
Our innovation in this study is the combination of two methodologies, Classification And
Regression (CART) methodology and Logistic model.
CART is a non-parametric methodology developed for analyzing classification problems.
If the dependent variable is categorical, CART produces a classification tree, and if the dependent variable is continuous, it produces a regression tree. CART methodology is built through a process known as binary recursive partitioning. This is an iterative process of splitting the data into partitions, and then splitting it up further on each of the branches. The algorithm divides the sample data according to a “splitting rule” and “goodness of split criteria”. Splitting rule is a question of the form, is X > D, where X is a variable and D is a constant within the range of that variable. Depending on the answer given to the question the algorithm splits the sample into two, one on the right and one on the left side called child nodes, The process continues until no more useful splits can be found. The goodness of split criteria compares different split and chooses the most homogeneous sub samples.
One major advantage of using CART is its ability to capture interaction effects among variables especially when this interaction has a significant effect on the outcome, as is the case in currency crises. Probit and Logit models using Ceteris Paribus assumption fail to identify such interactions, and give the marginal effect of each of the explanatory variables on the probability of the crisis holding all other variables constant at their mean. This assumption is not always the best especially when complementarities between factors play an important role. Ignoring these complementarities that exist among variables has resulted in dropping factors that may have been statistically insignificant when taken in isolation, but can trigger a currency crises when accompanied by other small fragilities (Kaminsky, 2003). To remedy to this problem Kaminsky
June 22-24, 2008
Oxford, UK
11

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
(2003) suggested the use of Classification And Regression Tree (CART) methodology. She examined crises episodes for twenty industrial and developing countries and used 18 economic variables out of which only nine had contributed to the outcome.
Recently CART has started gaining more popularity and credibility in the field of currency crises. To my knowledge two more studies have also applied this methodology. The first is an IMF working paper entitled “Rules of Thumb for Sovereign Debt Crises” by Manasse and Roubini 5 (2005). Their work was confined to “Debt Crises”. They used fifty variables, out of which ten appeared to have contributed to the classification and prediction of such crises. The second is also an IMF working paper entitled “Structural Vulnerabilities and Currency Crises”, by Ghosh and Ghosh (2005), They used CART methodology to focus on structural factors such as the rule of law, corporate governance, and shareholders’ rights in generating currency crises.
Sample and Data
Our model is estimated for the five Asian countries: Indonesia, Malaysia, Korea, The
Philippines and Thailand using monthly data covering the period 1981–1998. Monthly observations are more widely used in the literature because of their ability to capture the sudden nature of currency crises. Where only annual or quarterly data is available; monthly data was generated by linear interpolation. Most data is taken from International Financial Statistics (IFS), external debt and short-term debts are taken from the World Development Indicators (WDI).
The number of economic variables that can be included as indicators of vulnerability is very large and our choice is mainly determined by the availability of data. The list of variables used consists of variables reflecting fiscal problems such as fiscal deficit. Current account
5 Nouriel Roubini is one of the leading economists in the area of currency crises; his website: http://www.rgemonitor.com/ contains almost everything that relates to currency crises, making it the most popular site in currency crises topic.
June 22-24, 2008
Oxford, UK
12

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 problems are measured using real exchange rate
6
, value of imports, exports and interest rate differential between domestic and US real interest rates. M2 multiplier, domestic credit, bank deposits, bank foreign liabilities and the ratio of claims on private sector to GDP, represent financial indicators. Lastly to account for debt problems especially short-term debts we use the ratio of short-term debt to reserves and the ratio of short-term debts to total external debts, making the total number of variables included equals 14. These macroeconomic and financial variables are widely used in the empirical literature 7 .
Most variables were defined as the percentage change in their level with respect to their level 12 months ago. Furthermore, all variables were transformed to percentiles of the distribution; this is to standardize all indicators and to allow a better comparison between countries
8
. It is important to note that we do not consider structural problems such as corporate governance, banking and corruption; and political factors such as government victory (loss) and type of ruling party.
CART methodology
After running CART algorithm on the data, a classification tree was generated. Figure 1 represents a visual illustration of the full-grown tree, which consists of 16 nodes. 7 non-terminal nodes and 9 terminal nodes (4, 5, 7, 10, 11, 13, 14, 15 and16).
6 This was calculated as a deviation from trend using HP filter.
7 Because of data availability we used only 14 variables; our next project is to include as many variables as possible since CART can handle that easily.
8 “For example a monthly 20-percent fall in stock prices can be as usual in emerging markets but is a strong signal of crisis in a mature economy” (Kaminsky, 2003, p. 9).
June 22-24, 2008 13
Oxford, UK

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
The first observation is the number of variables included in the model. Out of 14 variables used in this study, only 8 were included: real exchange rate, reserves, M2 multiplier, short-term debts/total debts, private claims, exports, imports and M2multiplier. One major advantage, which is unique to CART methodology, is its ability to identify the best predictors out of a large number of variables, and this is particularly useful when trying to identify indicators of vulnerability to currency crises.
Another important point that needs to be motioned is that the first split of the data is based on the real exchange rate; meaning that exchange rate overvaluation is the most important predictor of the East Asian currency crises. This confirms the findings by many previous studies that the currencies of the Asian countries were overvalued.
The way the tree is interpreted is very simple. For example looking at node number 5, which is predicted to be a “No-Crisis” node, we can see that it is the result of the interaction of two factors: a currency which is not overvalued and low Claims on private sector, making the probability of crises occurrence extremely small (0.4%). However, node 15, which is predicted to be a “Crisis” category, is the result of the interaction of an overvalued real exchange rate, low foreign reserves, high exports, high M2/reserves and low imports. Out of these five factors, exports and imports seem not to be causing any problem while others appear to be causing trouble reflecting high crisis probability (66.7%).
June 22-24, 2008
Oxford, UK
--INSERT FIGURE 1 ABOUT HERE--
14

2008 Oxford Business &Economics Conference Program
Vulnerability Factors 9
ISBN : 978-0-9742114-7-3
This section presents major indicators of vulnerability identified by CART methodology that have jointly contributed to the East-Asian crisis in 1997-98. The findings also point to the significance and practical importance of understanding currency crises as a result of the interaction of two elements, trigger and vulnerability.
Real exchange rate overvaluation: the first split of the tree is based on the real exchange rate, which makes it the most important indicator of vulnerability during the Asian crisis. Many studies have considered the deviation of the exchange rate from the trend as a leading indicator of currency crises especially in Asia, Edison (2003) has ranked the deviation of real exchange rate from trend as top performer indicator of currency crises. Kaminsky (2003) who used CART methodology to identify leading indicators of currency crises in twenty countries has also generated a tree with the first split based on real exchange overvaluation.
Claims on Private Sector /GDP: in addition to foreign debt accumulation, domestic bank lending to private sector known as “over borrowing syndrome 10 ” or “a lending boom” was a norm in may Asian countries. Banking claims on the private sector in Thailand, Korea and
Malaysia reached more than 140% of GDP in 1996. What made the situation even more fragile is the fact that a bigger proportion of domestic bank lending was used by the private sector in over investment, especially in real estate, property and stock market speculation, this has led to lower efficiency of capital. In addition to that a rising share of the Banks’ claims on the private sector relative to GDP is also a sign of an unhealthy domestic banking system as it increases the ratio of
9 This part is mainly drawn from Ismail and Mansor (2007)
10 This was introduced by McKinnon and Pill (1995).
June 22-24, 2008
Oxford, UK
15

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 bad loans to total assets. In such situations an attack on the currency may not be evitable, as investors know that the cost of defending the currency when it is under pressure will be very high because it would result in many bankruptcies and corporate failures.
Reserves adequacy: one important element of first generation models ala Krugman, is the attempt of the authorities to defend the currency by intervening in the foreign exchange market until foreign reserves are totally exhausted. A central bank with enough reserves may resist an attack on the currency by defending it in the foreign exchange market; however there is no guarantee that the attack will not be successful. Now the Asian countries hit by the crisis provide a regular update of the amount of reserves available, this is to give investors and market participants a sense of confidence in the performance of their economies and also to show their capability to defend their currencies against any attack. The literature on currency crises uses a variety of variables to capture the role of foreign reserves; we chose to use “reserves” and the ratio of M2 to reserves “M2/Reserves”, these two variables are found to be among top leading indicators of currency crises (Edison, 2003; Kaminsky, 2003). While “reserves” measure the stock of foreign exchange available, “M2/reserves” however is a measure of the capacity of the central bank to meet the demand by agents who may want to convert their liquid monetary assets into foreign exchange. In the era of easy movement of capital, a sudden stop or even a reversal of capital can easily put a country under extreme pressure and subsequently be forced to devalue its currency. Reserves are very important during such critical periods. Stanley Fischer, the former deputy-managing director of the International Monetary Fund (IMF), has made this clear when he says, “ reserves matter because they are a key determinant of a country’s ability to avoid
June 22-24, 2008
Oxford, UK
16

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 economic and financial crisis. This is true of all countries but especially of emerging markets open to volatile international capital flows ” Fischer (2001, p. 2).
Bank foreign Liabilities:
There is a well established view that financial liberalization in emerging economies was a leading factor to banking crises. The ease of entry barriers has resulted in a proliferation of private banks and the emergence of foreign financial intermediaries representing a sizeable share of the banking system. What made the situation even worse are the deficiencies in the financial institutions such as weak regulations, poor bank supervision, implicit and explicit government guarantees which lead to excessive risk taking by depositors, borrowers and banks. The immediate effect was a lending boom which was financed through foreign financial intermediaries, therefore increasing bank liabilities with foreigners. This in turn has made these countries more vulnerable to external liquidity shocks.
Trigger elements: Another interesting finding of this study, is that the definition of currency crises in relation to “Trigger” and “Vulnerability” and the use of CART methodology has made the interpretation of the results not only possible but also reasonable. A closer look at and a simple comparison between node 15 and 16 (See Table 2) shows that node 15 has three elements of vulnerability: real exchange rate, reserves and m2/reserves, however node 16 which has on the top of the three variables mentioned in node 15 another indicator of vulnerability which is imports, has been classified as “No-crisis” node. Cases like this one are not easy to interpret, because all signs point to an inevitable crisis. If node 15 with three indicators of vulnerability has 66.7 % probability of crisis occurrence, obviously node 16 with an added
June 22-24, 2008
Oxford, UK
17

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 vulnerability should definitely be a crisis node, but CART algorithm classifies it as a” No-crisis” node. Why? Here comes the importance of our definition of a crisis and the convenience of using
CART methodology.
Let’s explain this further. The interpretation of the result similar to those in node 16 using standard method such as probit or logit model would mean that the signs of the coefficients were not as expected, therefore concluding that those variables in fact reduce the probability of the crisis, while all theories and empirical research says the opposite. This conclusion simply defies theories related to how currency crises come about and has to be rejected. So how is node 16 interpreted? The only answer that is consistent with the assumptions made and the methodology used is that the “Trigger” element was not present. This explains why a crisis may hit a country with relatively good fundamentals, while another with worse fundamentals may avoid the crisis all together.
Was Asian crisis self-fulfilling?
For the node to be classified as self-fulfilling it has to meet two criteria. First, it must be a crisis node; second it must not have any indicator of vulnerability. Table 2 shows that there is no such node. Therefore Asian crisis cannot be attributed solely to panic or herd behavior. This confirms the view that self-fulfilling prophecies are the characteristics of developed countries and not emerging economies ( Kaminsky, 2003).
Capturing the elements of “Trigger” and “Vulnerability”
Very few authors use trigger and vulnerability in the sense introduced in this paper, in order to explain currency crises. The exception is Radelet and Sachs (1998b), and Athukorala and Warr (2002). However these studies could not be translated into an empirical model that would explain crises using only these two elements. This is due to the difficulty of capturing all
June 22-24, 2008
Oxford, UK
18

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 factors of vulnerability using one variable called “Vulnerability”, and capturing all elements of trigger using one variable called “Trigger”.
Since crises are the result of trigger and vulnerability, it is important to further explore this new approach in order to comprehend currency crises. One major contribution of this paper is the use of CART methodology which allows us to capture these two elements quite easily.
Looking at Table 2 which contains final nodes and using our understanding of crises as a combination of “Trigger” and “Vulnerability”, we can say that “Crisis” nodes (7, 11, 14 and 15) must contain both elements: trigger and vulnerability. Node number 5 has no vulnerability element, but since it is a "No-Crisis" node and following the argument by Burnside et al (2000) that self-fulfilling prophecies (trigger) usually happen where there are fundamental problems
(vulnerability), it is more sensible to assume that it has no trigger element as well. All the remaining nodes (4, 10, 13 and 16) have indicators of vulnerability but classified as a "No-
Crisis" node; therefore the element of trigger is not present.
To run a logistic model using Trigger and Vulnerability as the only predictors, we need two “catch- all” terms: "Vulnerability" term which reflects all vulnerability factors. This would include all indicators identified by CART methodology. And "Trigger" term reflects all factors and events that would turn a vulnerable situation into a full scale crisis. While it is very difficult to identify such triggering factors because of their randomness, however it is possible to capture
Trigger element by reference to the final nodes of the tree generated by CART (See Table 2).
Having two categories for each term, the estimated logistic model takes the form:
June 22-24, 2008
Oxford, UK
19

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 ln
1

   
1
(
)
 
2
(
)
Where P is the probability of crisis occurrence, “TRIGGER” and “VULNERABILITY” are two dummy variables that represent "catch-all" terms for trigger and vulnerability factors respectively.
The results are summarized in Table 3. The model as a whole looks highly significant with both elements trigger and vulnerability having the expected sign and also significant at 1% significance level. Modeling currency crises in such a way is very promising as it helps identify the relative importance of each element and their contribution to the crisis, this in turn will help answer the question of whether Asian crisis was mainly caused by the deterioration of economic fundamentals as it is suggested by first generation models or the result of panic as it is the case in self-fulfilling models or the combination of both.
To examine the role and the impact of trigger and vulnerability on the crisis, we estimate the predictive probability for the crisis, when both elements of trigger and vulnerability are present; when both elements are non-existent and when only one element is present. These four scenarios are reported in TtTable 4. The results show that the highest probability of crisis (86%) is the outcome of the presence of both trigger and vulnerability (Part I), while the probability of the crisis when both elements are nonexistence is only 2% (Part II). This explains pretty well our assumption that crises are the outcome of the combination of trigger and vulnerability. To address the issue of how much damage was caused by vulnerability (Fundamentals) and how much damage was caused by trigger (Self-fulfilling), we calculated the probability of the crisis when only one factor is available. Part IV in Table 4 shows that the impact of vulnerability alone on the crisis is not very high (10.7%), while the impact of trigger alone in part III is almost
June 22-24, 2008
Oxford, UK
20

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
55%.The very high impact of trigger element confirms the view that while weak economic fundamentals (vulnerability) may have played a role in the lead up to the crisis; however these weaknesses were not large enough to cause a crisis with such depth and severity (International
Monetary Fund, 1998; Radelet & Sachs, 1998b; Summers, 2000).
To further explore how the relationship between Trigger and Vulnerability affects the predicted probability of the crisis, we created a visual interpretation of the same predicted probabilities shown in Table 4 11 . While Figure 2 displays exactly the same information contained in Table 4, however it has the advantage of revealing some insights into how Trigger and
Vulnerability affect the probability of crisis occurrence. Figure 2 clearly shows that the predicted probability is very sensitive to the Trigger element. In a vulnerable situation the predicted probability increases from 11% (point B) when there is no "Trigger" to 86% (point D) where there is "Trigger".
This suggests that trigger (panic) has played an important role in causing the huge reversal in capital which has led to an ultimate collapse of domestic currencies of the affected countries and made Asian crisis one of the most devastating in the last few decades.
Another important point that may not be clear without this visualization is the effect that Trigger has on Vulnerability. When there is no Vulnerability the difference between the predicted probabilities of a situation where there is Trigger element (Point C) and where there is not (Point
A) is 53%. However the same difference in a vulnerable situation (D-B) becomes much bigger
75%. This is a very interesting result because it strongly supports the argument that weak economic fundamentals (Vulnerability) not only create but also intensify panic (Trigger). This means self-fulfilling crises are more likely to happen where there are fundamental problems.
This should send a strong signal to emerging market economies: before they blame external players for such crises; first they have to get there home in order.
11 This is done using STATA program (See, Michael & Xiao, 2005)
June 22-24, 2008
Oxford, UK
21

2008 Oxford Business &Economics Conference Program
IV.
CONCLUSION
ISBN : 978-0-9742114-7-3
In this study we tried to understand and model currency crises in relation to trigger and vulnerability, focusing on the five Asian countries hit by the crisis in 1997-98. First, we introduced a new methodology called CART (Classification And Regression Tree Analysis), to help identify indicators of vulnerability to crises. Second, the tree has made it easy for us to capture the elements of trigger and vulnerability causing the crisis. Third, we use a logistic model that can explain and predict crises using trigger and vulnerability as the only variables in the regression.
Our findings show strong evidence that Asian crisis is the result of vulnerability (weak fundamentals) as well as trigger (panic and herd behavior). The overvaluation of the currency seems to be a major factor together with the domestic lending boom and foreign reserves. This, points to the necessity of carefully addressing the issue of macroeconomic weaknesses and inconsistencies especially where local currencies are pegged against the Dollar.
With regard to triggering elements that have contributed to the crisis it is difficult to ascertain that particular events have triggered the crisis, however in the case of Asia, triggering elements were not only strong but also obvious. The collapse of Hanbo Steel in Korea, the problems faced by Samprason Land in Thailand and the political uncertainties in Indonesia, the
Philippines and Malaysia were all making the news. And because of the geographical and financial links among these countries, this has lead to a kind of multiplier effect which has worsened the situation.
In conclusion, the results confirm the view that even though fundamentals may have played a role in the lead up to the crisis, however panic and self-fulfilling prophecies have made the situation even worse and intensified the severity of the crisis. No one can guarantee that a
June 22-24, 2008
Oxford, UK
22

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3 currency will not be attacked, especially in an environment where ease of capital mobility and financial integration is becoming a predominant feature of the global economy. Asian crisis has taught us a lesson: currency crises can happen anywhere anytime and without warning. Although second generation models put the blame on irrational behavior of the investors, and their selffulfilling expectations, it is important to note that good fundamentals and strong financial sector can play a big role in reducing vulnerability to currency crises and preventing a collapse of the economy. We may see new types of currency crises in the future, which may call for new models with different diagnosis and remedies, but one way to defend the currency is to have the “house in order”: sound fundamentals, strong financial sector and credibility.
REFERENCES
Athukorala, P.-c., & Warr, P. G. (2002). Vulnerability to a Currency Crisis: Lessons from the
Asian Experience The World Economy, 25 (1), 33-57.
Berg, A., & Pattillo, C. A. (1998). Are currency crises predictable? : A test. IMF working paper ;
WP/98/154 , 61 p.
Bhagwati, J. (1998, 2-3 November). Asian Financial Crisis Debate: Why? How Severe?
Paper presented at the International conference Managing the Asian Financial Crisis: Lessons and Challenges, Kuala Lumpur.
Burnside, C., Eichenbaum, M. S., & Rebelo, S. (2000). On the fundamentals of self-fulfilling speculative attacks. NBER working paper series ; wp 7554 , 43, [43] p.
Bustelo, P. (2000). Novelties of Financial Crises in the 1990s and the Search for New Indicators.
Emerging Markets Review, 1 (2000), 229-251.
Chang, R., & Velasco, A. (1998). The Asian Liquidity Crisis. Federal Reserve Bank of Atlanta,
Working Paper 98-11 .
Corsetti, G., Pesenti, P. A., & Roubini, N. (1999). What caused the Asian currency and financial crisis? . Japan and the World Economy, 11 (3), 305-373.
Dornbusch, R. (1997). A Thai Mexico Primer:Lessons for Outmaneuvring a Financial
Meltdown. The International Economy (September / October), 20-23.
Edison, H. J. (2003). Do indicators of financial crises work? An evaluation of an early warning system. International Journal of Finance & Economics, 8 (1), 11-35.
Eichengreen, B. J., Rose, A., Wyplosz, C., Dumas, B., & Weber, A. (1995). Exchange Market
Mayhem: The Antecedents and Aftermath of Speculative Attacks. Economic Policy,
10 (21), 249-312.
Fischer, S. (1998, 10/03/98). Lessons from a crisis. The Economist, 349, p23--27.
June 22-24, 2008
Oxford, UK
23

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
Fischer, S. (2001, April 28). Opening Remarks.
Paper presented at the IMF/ World Bank
International Reserves: Policy Issues, Washington DC.
Furman, J., & Stiglitz, J. (1998). Economic Crises: Evidence and Insights from East Asia.
Brooking Papers on Economic Activity, 2 , 1-114.
Ghosh, S. R., & Ghosh, A. R. (2005). Structural Vulnerabilities and Currency Crises. IMF Staff
Papers, 50 (3).
Hill, H. (1999). An overview of the Issues. In H. W. A. a. H. Hill (Ed.), East Asia Economic
Crisis: Origins, Lessons and the Way Forward (pp. 1-15). Singapore: Institute of
Southeast Asian Studies.
Honohan, P., Daniela, K., & Andrea, M. (1999). Is the Fiscal Cost of Bank Crises Increased by
Poor Resolution Policy? World Bank, Washington, D.C.
International Monetary Fund. (1998). The Asian Crisis: Causes and Cures. Finance &
Development, 35 (2).
Ismail, A.-S., & Mansor, J. (2007). Vulnerability to Currency Crises: the Asian Experience. In
M. L. Mackenzie & S. L. Rosenberg (Eds.), Emerging Business Theories for Educators and Practitioners . Newcastle, UK: Cambridge Scholars Publishing.
Kaminsky, G. L. (2003). Varieties of currency crises. NBER Working Paper (10193).
Kaminsky, G. L., & Schmukler, S. L. (1999). What triggers market jitters? : a chronicle of the
Asian crisis . Washington, DC: World Bank Development Research Group
Macroeconomics and Growth.
Kraay, A. (2000). Do high interest rates defend currencies during speculative attacks ? Policy
Research Working Paper Series 2267, The World Bank .
Krugman, P. R. (1999). Balance Sheets, The Transfer Problem, and Financial Crises.
Preliminary Draft prepared for the festschrift volume in honor of Robert Flood .
Krugman, P. R. (2001, March 25-26). Crises: the Next Generation?
Paper presented at the Razin
Conference, Tel Aviv University.
Manasse, P., Roubini, N., & International Monetary Fund. Fiscal Affairs Dept. (2005). "Rules of thumb" for sovereign debt crises . Washington, D.C.: International Monetary Fund Fiscal
Affairs Dept.
McKinnon, R. I., & Pill, H. (1995). Credible liberalizations & international capital flows : the
"over-borrowing syndrome". CEPR publication ; no. 437 , 35, [12].
Michael, N. M., & Xiao, C. (2005). Visualizing Main Effects and Interactions for Binary Logit
Models. The Stata Journal, 5 (1), 64-82.
Radelet, S. C., & Sachs, J. (1998a). The East Asian Financial Crisis: Diagnosis, Remedies,
Prospects. Brooking Papers on Economic Activity, 2 , 1-89.
Radelet, S. C., & Sachs, J. (1998b). The onset of the East Asian financial crisis. NBER Working
Paper (6680).
Stiglitz, J. (1998). The Role of International Financial Institutions in the Current Global
Economy (Adress to Chicago Council on Foreign Relations).
Summers, L. H. (2000). International Financial Crises: Causes, Prevention, and Cures. AEA
Papers and Proceedings, 90 (2).
Tobin, J. (1998). Asian Financial Crisis. Japan and the World Economy, 10 (3), 351-353.
June 22-24, 2008
Oxford, UK
24

2008 Oxford Business &Economics Conference Program
List of Tables
ISBN : 978-0-9742114-7-3
Table 1: Common Indicators Before the Crisis
Current
Account
Balance
(%of GDP)
Domestic
Credit
(%of GDP)
M2/
Reserves
Shortterm
Debts/
Reserves
Real
Exchange
Rate (a)
Country
Name
Indonesia
Philippines
Thailand
Malaysia
Korea, Rep
Mexico
Chile
Argentina
Brazil
South Africa
1996
-3.37
-4.77
-8.08
-4.42
-4.42
-0.7
-5.12
-2.51
-3
-1.31
1996
54.23
73.95
142.63
142.91
69.5
27.09
64.55
28.22
40.91
140.16
1996
6.11
3.91
3.81
3.34
6.49
4.52
1.89
3.13
3.59
31.13
1996
1.77
0.79
1.26
0.41
2.09
1.54
0.18
1.3
0.62
11.5
Source: World Bank Development Data
(a)(1990 =100). An increase means depreciation. (From Radelet and Sachs 1998b).
Jun-97
78
54
76
75
89
79
55
42
33
N/A
June 22-24, 2008
Oxford, UK
25

2008 Oxford Business &Economics Conference Program ISBN : 978-0-9742114-7-3
Table 2: Final Nodes & Indicators of Vulnerability
Final
Nodes
Factors
Predicted
Category
4
5
7
10
11
13
14
15
16
Real Exchange Rate < 15.6
Reserves >87
Real Exchange Rate > 15.6
Private Claims /GDP < 68
Real Exchange Rate < 15.6
Reserves < 87
Exports < 40.7
Real Exchange Rate > 15.6
Private Claims /GDP > 68
M2 Multiplier > 47
Real Exchange Rate < 15.6
Reserves < 87
Exports > 40.7
M2 Reserves < 38
Real Exchange Rate > 15.6
Private Claims /GDP > 68
M2 Multiplier < 47
Short-term Debts < 67
Real Exchange Rate > 15.6
Private Claims /GDP > 69
M2 Multiplier < 47
Short-term Debts > 67
Real Exchange Rate < 15.6
Reserves < 87
Exports > 40.7
M2 Reserves > 38
Imports < 28.1
Real Exchange Rate < 15.6
Reserves < 87
Exports > 40.8
M2 Reserves > 38
Imports > 28.2
No Crisis
No Crisis
Crisis
No Crisis
Crisis
No Crisis
Crisis
Crisis
No Crisis
Crsis
Prob%
Vulnerability Trigger
7.4
2.3
87.7
7.1
85.7
26
92.6
66.7
4
Factors in Bold represent Vulnerability
YES
NO
YES
YES
YES
YES
YES
YES
YES
NO
NO
YES
NO
YES
NO
YES
YES
NO
June 22-24, 2008
Oxford, UK
26

2008 Oxford Business &Economics Conference Program
Table 3: Logit Model
ISBN : 978-0-9742114-7-3
Log likelihood = -176.05218 Pseudo R2 = 0.5001
LR chi2(2) = 352.30 Prob > chi2 = 0.0000
Crisis | Coef. Std. Err. z P>|z|
Trigger 3.96 .355 11.16 0.000
Vulnerability 1.65 .369 4.47 0.000
Table 4: Currency Crisis: Predicted Probabilities
I
95% Conf. Interval
Pr(y=1|x): 0.86 [0.80, 0.93]
Pr(y=0|x): 0.14 [0.07, 0.20]
Trigger Vulnerability
x= 1 1
III
95% Conf. Interval
Pr(y=1|x): 0.5475 [ 0.3203, 0.7746]
Pr(y=0|x): 0.4525 [ 0.2254, 0.6797]
Trigger Vulnerability x= 1 0
Tt
II
95% Conf. Interval
Pr(y=1|x): 0.022 [0.0094, 0.0357]
Pr(y=0|x): 0.978 [ 0.9643, .9906]
Trigger Vulnerability x= 0 0
IV
95% Conf. Interval
Pr(y=1|x): 0.1070 [ 0.0681, 0.1459]
Pr(y=0|x): 0.8930 [ 0.8482, 0.9319]
Trigger vulnerability x= 0 1
June 22-24, 2008
Oxford, UK
27

2008 Oxford Business &Economics Conference Program
List of Figures
Figure 1: Classification Tree
ISBN : 978-0-9742114-7-3
<=0.156
Node1
Reserves
<=0.87
Node3
Exports
>0.87
Node4
No Crisis
Node0
Exchange Rate
>0.156
Node2
Claims Private
<=0.68
Node5
No Crisis
>0.68
Node6
M2 Multiplier
<=0.407
Node7
Crisis
<=0.38
Node11
Crisis
>0.407
Node8
M2/Reserves
>0.38
Node12
Imports
<=0.47 >0.47
Node9
Short-term Debts
Node10
No Crisis
<=0.67
Node13
No Crisis
>0.67
Node14
Crisis
<=0.281
Node15
Crisis
>0.281
Node16
No Crisis
June 22-24, 2008
Oxford, UK
28

2008 Oxford Business &Economics Conference Program
.7
.6
.5
.4
.3
.2
.1
0
1
.9
.8
C
A
0
Vulnerability
Trigger=0 Trigger=1
ISBN : 978-0-9742114-7-3
D
B
1
Figure 2. Currency Crisis: Predicted Probabilities (Visualization)
June 22-24, 2008
Oxford, UK
29