The Backgrounds of Banking and Currency Crises

(1)

The Backgrounds of Banking and

Currency Crises

An update of Kaminsky and Reinhart (1999) and an

analysis of selected crises

by

Eda Aral

August 2010

UNIVERSITY OF GRONINGEN

(2)

THE BACKGROUNDS OF BANKING AND CURRENCY CRISES: AN UPDATE OF KAMINSKY AND REINHART (1999) AND AN ANALYSIS OF SELECTED CRISES

Master Thesis for Master of Science in International Business & Management

University of Groningen

Faculty of Economics and Business Duisenberg Building (H)

Nettelbosje 2

9747 AE Groningen The Netherlands http://www.rug.nl

Supervisor: Prof. Dr. Harry Garretsen

Written by: Eda Aral (1940384)

(3)

Management Summary

Last decades of the 1990s were scene to numerous financial crises. Latin American, European, and Asian economies experienced devastating economic problems. With improved economic policies and technological developments the 2000s went well for most of the countries in the world. Then in 2007, a seemingly atypical financial crisis unfolded in the United States (US) and impacted the rest of the world. A majority of countries have faced with recession, sectors have shrunk, and unemployment rates have risen. The crisis started in 2007 has been recognised as the second largest economic crash since the Great Depression of the 1930s.

The huge cost created by the global crisis ignited the researchers‟ interest in the formation of economic crises. The causes and initial indicators of the economic crises are widely discussed. The majority of economists (i.e. Borio, 2008; Reinhart and Rogoff, 2009) have concluded that the global crisis of 21st century is similar to other cases in its formation. So it must have been predictable. If it was predictable, it could have been prevented. Therefore, understanding the causes of financial turmoils, and having insight to foresee the future crises became more important than ever; for policy makers, strategists, and managers of financial corporations. Understanding the formation of the crises, and being able to interpret the precedents of financial crises is essential in making effective policies that would lessen the risk of future crises. However, focusing only on similarities between the causes of crises can mislead policy makers to make wrong decisions. Although the crises are quite similar to each other in terms of formation, case-specific factors can have major impacts in the backgrounds of crises. In this research, the focus is on the backgrounds of banking and currency crises. The investigation starts with the update of a previous research by Kaminsky and Reinhart (1999). The research sample is composed of twenty-three countries with advanced, developed, developing, large and small economies. The observation period extends from 1990 to 2009. Thus, many Latin American, European, East Asian crises and the current global crisis are covered. The analysis spanned forty-five currency, twenty-five banking and twenty twin crises. A set of seven macroeconomic variables regarding financial, external and real sectors is examined. It is concluded that the banking and currency crises are linked, and their macroeconomic backgrounds are similar.

(6)

predictable. As a second step, the focus is narrowed on five selected cases. Apart from the macroeconomic indicators in first step of analysis, country-specific characteristics and events are also reviewed. In the selection of these crises, the heterogeneity of the sample is considered to ensure the generalizability. 2008 twin crises of Denmark and Iceland, and the twin crises of Argentina and Turkey in the early 2000s are investigated in an effort to have both banking and currency problems in the consideration. Furthermore, the US subprime crisis of the 2007 is also examined. It is found that the economic indicators considered in the first step of analysis behave similarly in the selected cases. However, there are also other factors contributing to the formation of the crises. Among the common causes of crises, there are real estate booms, inept management policies, political instabilities, moral hazard problems, and contagion of problems from other economies. Based on the observed causes of the selected five crises, alternative indicators are proposed for predicting the future crises. This research has some limitations. First of all, the causality between the selected indicators and crises are not subject to regression analysis. There are also other problems with the measurement of the fragility of economies on the eve of crises. Alternative approaches are recommended to overcome these limitations.

(7)

1. INTRODUCTION AND RESEARCH DESIGN

1.1. Introduction

Financial crises have been around the world since the birth of money and markets. The earliest crises were mostly driven by currency debasements when the rulers of a country reduced the gold or silver amount of coins to relieve budget problems. As financial systems have advanced, governments do not need to shrink coins to overcome budget deficits anymore. But financial crises have continued to hinder countries‟ economic prosperity without differentiating the advanced from emerging (Reinhart and Rogoff, 2009).

Latin American economies have long been notorious for their crises in 1970s, 1980s, 1990s, and in the 2000s. Devastating crises also hit Asian economies and Russia in the end of 1900s and affected other economies in the region. Many European economies, especially Finland (Mayes, 2009), had their share of big crises in the early 1990s, while Spain, Greece, Italy, and Ireland experience problems nowadays. Among the advanced economies, Japan has been a very problematic one, still suffering from inadequate growth since its crisis in the 1990s. Then the subprime mortgage crisis unfolded in the United States (US) and had global impact. Indeed the global crisis, or the “Second Great Contraction” as Reinhart and Rogoff (2009) refer, has been recognised as the most devastating crisis since the Great Depression of the 1930s.

A crisis has multi-dimensional impacts on economy. Alongside the burden for the government, the economy loses pace, production activities slow down and unemployment rises. Therefore, crisis-preventive policies are crucial for governments‟ since they are responsible for maintaining the welfare of people. Furthermore, the corporations also need to understand the mechanisms that create crises. Markets‟ expectations can lead an economy into crisis. So the interpretation of the economic conditions is important in companies‟ strategy-making processes. Companies today can take advice from consulting firms or from their financial consultants about the future of the economy. A substantial rise in inflation, interest rates, and/or the value of domestic or foreign currencies can be a sign of future economic problems. Hence, corporations should consider the economic developments and adjust their investment plans and recapitalisation structures accordingly.

(8)

attracting utmost interest. Understanding the formation of the crises, and having insight into the precedents and analogies of financial crises are essential in making effective policies that would lessen the probability of future crises.

However, focusing only on analogies of the crises can bias the policy makers. Although the crises are quite similar to each other in terms of formation, case-specific factors can have major impacts in the backgrounds of crises. A housing bubble led both the US and Denmark to banking crises, but the complexity and over-estimated reliability of the American finance system made the American crisis global. The inept management of fixed exchange rate regimes led Turkey and Argentina to twin crises in the early 2000s, but the inability of government in controlling provincial expenditures was one of the biggest problems in Argentina while it was completely irrelevant in Turkey case. Hence, the peculiarities of economies must be included in the analysis of the precedents of crises. Studying the maximum number of possible factors that can lead an economy to crisis can help in improving crisis-preventive strategies.

In this research, the backgrounds of banking and currency crises are investigated by first updating a previous research by Kaminsky and Reinhart (1999). For a sample of twenty-three countries observed during 1990-2009 period, it is concluded that the macroeconomic backgrounds of crises are analogous. Then, the predictability of crises is tested by observing the behaviours of macroeconomic indicators and it is inferred that crises are mostly predictable. As a second step, the focus is narrowed on five selected cases. Other incidents in countries‟ financial systems, commercial sectors, politics and international relationships are observed, and fragilities are pointed out. Based on the causes of the selected five crises, alternative indicators are proposed for predicting the future crises.

1.2. Research Objective and Scope

(9)

are more costly to economies since, with twin crises, the financial resources of the state and private sectors melt down together. Considering these, the scope of this research is narrowed to banking and currency crises, and the joint occurrences of them „the twin crises‟.

Believing in the importance of crisis-preventive strategies, I focus on the causes and preceding economic conditions of banking and currency crises. I structure my research according to a previous research by Kaminsky and Reinhart (1999) which is mainly a study of common behavioural patterns of economic variables around crises. After updating their research by applying the same methodology, I review a set of selected crises1 in an endeavour to find non-common behaviours of economic variables around crises. This research therefore has primary interest in the causes of banking and currency crises.

Briefly, this research is aiming to contribute to the area of pre-crisis analyses, by; - examining the behaviours of macroeconomic indicators around the crises,

- suggesting generally applicable thresholds levels for economic variables to be interpreted as signals for upcoming crises,

- investigating a set of selected crises to control whether the behavioural patterns of economic indicators show similarity to the general,

- and by trying to find out which case-specific factors take place in the formation of the crises.

1.3. Research Questions

The research questions of this investigation are listed below in order of importance:

1. What are the common and specific causes of banking, currency and twin crises, separately and together?

2. What happens in the macroeconomic backgrounds of banking, currency and twin crises?

3. Which macroeconomic variables signal for an upcoming banking or currency crisis? The sub questions that will be operationalised in the analysis related to the third research question above are listed below:

1

(10)

i. Do macroeconomic variables signal for an upcoming banking or currency crisis?

ii. If so, when do macroeconomic variables signal for an upcoming banking or currency crisis?

1.4. Outline

The structure of this research is as follows: First, a review of the literatures on financial, banking, currency and twin crises will be introduced. Then, in the second section, the research methodology will be explained, the data set will be introduced and the results of the quantitative analysis, which is based on the research of Kaminsky and Reinhart (1999), will be presented and interpreted. In the third section, a rather qualitative method of analysis will be made. Five crises will be reviewed with a focus on the causes of the crises and the macroeconomic backgrounds. In the forth section, the five crises2 will be reviewed for the selected macroeconomic indicators, and alternative indicators that can signal for upcoming crises will be proposed. The fifth section will conclude, reveal the research limitations and suggest further implications.

1.5. Literature Review

1.5.1. Financial Crises

The economics‟ literature has been frequented by the analyses of financial crises. For example, European, Latin American and Asian crises of 1990s have been commonly discussed from every perspective, though the global crisis takes most of the attention nowadays. The causes of and linkages between different types of financial crises, the recovery from crises, the preventive and ameliorating policies have been reviewed. Krugman (1979), Velasco (1987), Minsky (1992), Obstfeld (1994), Eichengreen et al. (1995), Mishkin (1996, 2001), Demirgüç-Kunt and Detragiache (1998b), Eichengreen and Rose (1998), Kaminsky and Reinhart (1999), Glick and Hutchison (2000), Levine (2002), Caprio and Klingebiel (2003), Komulainen and Lukkarila (2003), Claessens et al. (2008), Laeven and Valencia (2008), Mayes (2009), and Reinhart and Rogoff (2009) are remarkable examples from the literature.

Mishkin (2001) mentions that a financial system performs the essential function of canalising funds to those who has promising investment opportunities, reminding that accurate

2

(11)

judgement of the investments‟ feasibility, productivity and creditworthy is critical for the lenders. Here, the problem of asymmetric information is inevitable, in which “one party to a financial contract has much less accurate information than the other party” (p.1). The problem lies in borrower‟s advantage to the lender in having better information about the potential risks and returns and of the investment projects. Leading to the so-called adverse selection and moral hazard problems, asymmetric information leads has been recognised as an important issue for the financial systems. After this explanation, Mishkin (1996, p.17 and 2001, p.2) describes a financial crisis as “a disruption to financial markets in which adverse selection and moral hazard problems become much worse, so that financial markets are unable to efficiently channel funds to those who have the most productive investment opportunities.” Seemingly tender in this definition, a financial crisis renders the financial markets inefficient and sometimes even disabled, resulting in a sharp contraction of economic activity.

In his seminal hypothesis of financial instability, Minsky (1992) approaches the issue of financial crises from an exceptional perspective. He first defines the speculative financial units as units who can fulfil their payment obligations partly and need to roll over their liabilities, and the Ponzi units as the ones who have inadequate cash flows from operations to pay the debts back. Ponzi units sell assets and/or borrow again, eventually they exceed the safety level of indebtedness, and thus they decrease the asset value of the loan for the creditor. Minsky asserts that over a protracted period of good times, capitalist economies tend to move from financial relations that make for a stable system to those that make for an unstable system. Moreover, in inflationary economies with considerable number of speculative units the contractionary monetary policy, which is aimed at stabilising the inflation, can turn speculative units to Ponzi units. Since the Ponzi borrowers are insolvent, the loans lent to Ponzi units are worthless. The collapse of asset values can translate into the turmoil of the financial system.

(12)

crises (sovereign defaults), and (iii) domestic debt crises. This paper focuses on currency crises, banking crises and joint occurrences of the two. Financial crises are pernicious, and yes, almost inevitable but worse, they can also be contagious. Some crises spread over the countries, even to ones without obvious vulnerabilities. Unsound macroeconomic policies, unsustainable indebtedness, capital bonanzas, balance sheet weaknesses, credit booms, and political shocks have been the causes underlying the financial crises (Laeven and Valencia, 2008). In many financial crises currency and maturity mismatches have been fundamental like in the case of 2000-2001 turmoils of Turkey, while off-balance sheet operations of the banking sector were obtrusive like in the background of the crisis in the US in the end of 2000s.

1.5.2. Banking Crises

Are banks good for economies or are they only a menace as they indebt households and companies, ask for charges for the any kind of transactions that people has to do inevitably? As if the dullness of the banking stuff is not enough, banks seem to fall in crises in every now and then. Levine (2002) presents an interesting cross-country examination where he compares bank-based and market-based systems and asks which one is better for promoting the economic growth. In an earlier study, Levine (1997) explains that banks have a positive role in mobilising capital, differentiating the good projects from the bad ones, monitoring the actors in the industries and managing risk. Since banks do not disclose the information of their customers, well-developed financial markets can make better conditions for innovative projects that foster economic growth. Gerschenkron (1962) advocates that stock markets are less effective in financing industrial growth in comparison to banks, especially in under-developed economies. Moreover, Reinhart and Rogoff (2009) mention that banks transform short-term deposit funding into long-term loans, thus effect the maturity transformation in the system3. On the other hand, the market-based view puts a special emphasis on the problems with banks. Banks that have close ties fit firms can hinder competition by protecting particular firms and therefore thwart innovative projects of others (Levine, 2002 and the references therein). The main finding of the cross-country examination of Levine is however, although overall financial improvement is linked with economic growth, neither the bank-based nor market-based view is supported.

(13)

Still not knowing whether banking system is an effective supporter of economic growth, we need to discuss the problems of banking systems. Caprio and Klingebiel (2003) describe the situation where the aggregate value of the banking system liabilities exceeds the aggregate value of the assets as a systemic banking crisis. The economic history has been scene to numerous banking crises.

According to Reinhart and Rogoff (2009, p.141), the incidence of banking crises proves to be similar everywhere, like high, middle and low income countries. Poor, under-developed countries of Africa and time to time emerging markets from Latin America have had banking crises in the form of domestic default. The making of those crises has a simple formula: High financial suppression rules, entire savings and payments are monopolised by domestic banks so that government takes tax from every last cent, banks are indebted via tight reserve requirements to provide finance to government at low cost, restrictions and regulations force banks to lend to fund public debt, but then government has to stop paying its debts, and domestic default. Another and more typical variety of banking crises has been experienced by emerging and developed economies as well. As explained above, banks bear the risk of being vulnerable to bank runs. In tranquil times, banks maintain a certain amount of liquidity to cope with unexceptional rushes in deposit withdrawals. However, any panic created by the deterioration of confidence in the banks multiplies the withdrawals, banks then become obliged to liquidate their assets in order to fulfil the clients‟ requirements, at fire sale prices. Normally liquid assets turn into illiquid ones in periods of mass panic, banks become insolvent, and failures follow one another. Falling asset prices, not without other problems of course, led Scandinavian countries to crises and decelerated the economic growth (Drees and Pazarbasioglu), while the bursting asset price bubble in Japan left the banking sector insolvent (Demirgüç-Kunt and Detragiache, 2005).

(14)

that President Obama had to approve a large-scale bailout that would support the recovery of the financial system in the US4, within the first weeks of his duty in the Office.

Wrecking havoc on the development of economies, the making of banking crises is subject to numerous academic studies. Eichengreen and Rose (1998) explain that, developing in the early twentieth century, banking systems have been perceived as both a benefit and a menace to financial stability, especially in emerging economies. Even the Great Depression of the 1930s has been attributed to the problems within the banking systems. With the strict regulations of the banking systems in the aftermath of the Great Depression and World War II, the world has been more or less immune from banking crises. However, as the financial liberalisation movements ignited after 1970s, the banking crises gained frequency and placed on the top lines of the academic agenda of economies.

Sundararajan and Balino (1991) report that periods of rapid growth and increased imbalances in the external accounts precede the banking crises in developing economies. Supporting this, Kaminsky and Reinhart (1999) and Gavin and Hausmann (1996) add the unsustainable credit booms to the causes of crises list. However, Caprio and Klingebiel (1997) conduct a broader analysis that covers both developing and industrial economies where developing economies stand for crisis-prone components of the sample and the OECD countries as the control group. They conclude that banking crises are not related to excessive credit growth stories.

Another seminal analysis among the banking crisis studies is by Demirgüç-Kunt and Detragiache (1998b, updated in 2005). Their main findings concentrate on unsound growth and high inflation, while they present high real interest rates and deteriorating terms of trade among the problem-making factors. They stress the risk in sudden capital outflows and insufficient liquidity in the banking sector. They find robust relation between explicit deposit insurances and crises, suggesting that moral hazard distorts the intentions. They also criticise countries with weak institutions which they measure by a „law and order‟ index. Demirgüç-Kunt and Detragiache (2005) investigate the determinants of banking crises by „multivariate logit approach‟, which is a different technique from Kaminsky and Reinhart‟s signals approach (1999). With this approach, “the probability that a crisis occurs is assumed to be a function of a vector of explanatory variables. A logit econometric model is fitted to the data and an estimate of the crisis probability is obtained by maximising the likelihood function.

(15)

Thus, the model produces a summary measure of fragility (the estimated probability of crisis).” (Demirgüç-Kunt and Detragiache, 2005, p.70).

Eichengreen and Rose (1998) specify their analysis on only developing countries and present their core finding as there is a significant and robust correlation between interest rate changes in industrial countries and banking crises in emerging markets. An example is the turmoil in Argentina in 1995. They remark that not only the fragilities in domestic financial system could provide explanation for banking crises, but also unfavourable external conditions have a stake in the groundwork for banking crises. This is consistent with Mishkins‟s emphasis (1996).

Another threat to banking system as emphasized by Mishkin is financial dollarization. Denominating liabilities5 in foreign currencies has been common in developing economies according to Mishkin (1996), however it has important adverse results: It makes a financial system more vulnerable to crises and makes financial crises more severe since it limits the policy options available. The foreign exchange rate risk that is borne by the banks turn into an unmanageable problem as unexpected devaluation make borrowers insolvent, and loans nonperforming. Foreign currency loans were indeed among underlying problems in Mexico (Mishkin, 1996), in the Nordic countries in the beginning of last decade of 20th century (Drees and Pazarbasioglu, 1998) and in Turkey in 1994 (Demirgüç-Kunt and Detragiache, 2005).

1.5.3. Currency (Balance-of-Payments) Crises

Though the notion of money is as old as history, currency crises studies were not popular before Krugman‟s seminal balance-of-payment crises model (1979). Krugman constructs his model on the existence of a pegged (fixed) exchange-rate regime. Countries who have well-developed financial markets can defend an exchange parity via regulations such as bank reserve requirements, open market operations, intervening operations in forward exchange market and direct operations in foreign assets, yet all these policy instruments are subject to limits. The policymakers‟ endeavours to keep the value of the domestic currency above a limit can deplete the foreign reserves, as well as it hampers the country‟s borrowing possibilities. On the other hand, avoiding the upward pressure on the currency, striving against the appreciation of the domestic currency creates an inflationary impact on the economy. When the government throw in the towel, in other words when the fixed parity cannot be defended anymore, there is a balance-of-payments (BOP) crisis, explains Krugman.

(16)

Still classified in the earlier models of BOP problems, Lahiri and Végh (2003) hold the same assumption of fixed exchange-rate regime and limited foreign reserves. However, they object to the assumption of central banks‟ staying passive as reserves shrink. They assert that, in real fact, central banks defend pegs by raising short-term interest rates. They analyse the feasibility and optimality of using interest rates as a tool to postpone a potential currency crisis. They have two noteworthy findings. First, delaying is possible but eventually the crisis will always occur. Second, defending the peg by large increases in interest rates can create an inflationary pressure on the economy.

Later crisis events, such as Nordic crises of early 1990s, Mexico crises of 1994 and Asian crises of 1997 (Kaminsky, 2000) as covered in the sample of this paper, inspired different approaches to the currency crises issue, mostly aimed to explore the linkages between banking and balance-of-payments problems. Liquidity problems due to sudden reversals in capital flows are recognised among factors that create balance-of-payments instability as expressed by Kaminsky (2000) and references therein. From a differentiated perspective, Velasco (1987) works on small open economies, particularly on Chile, and discusses the currency crashes rooted from excessive money creation which is a result of printing money to bailout a troubled banking system.

Remind that Reinhart and Rogoff (2009) classify currency crashes in crises which are defined by quantitative thresholds. Indeed, currency crises are typically described as large changes in some indicator of currency value (Glick and Hutchison, 2000). The issue of speculative attacks to currencies have been approached differently in the literature. Some studies (Eichengreen et al, 1995; Kaminsky and Reinhart, 1999) give importance to such speculative attacks where exchange-rate does not adjust in all the cases; thanks to successful defend of the currency by the governments. The defending endeavour is reflected by raised interest rates and / or the loss of foreign reserves in such analyses, as a result of authorities‟ intervention to the foreign exchange markets. However, in their analysis of the similarity of currency crashes in developing countries Frankel and Rose (1996) pay attention to substantial depreciation episodes only.

(17)

1.5.4. Twin Crises

Though it has a respectively smaller share in the academic literature, the joint occurrence of financial crises has been among the most important issues of crisis studies due to enormous damage that they make. Leaving alone triple crises which are fortunately rare events, I focus on entwined banking and currency crises, and briefly present the theoretical studies of the linkages between banking problems and currency crashes.

The channel of causation can be both ways or even joint. The causality going from banking system problems to currency crises has been widely accepted in the literature. In the study of “Logic of Currency Crises”, Obstfeld (1994) reminds the rational expectations idea and suggests that if rational speculators predict the rise of inflation in the expense of exchange rate stability to prevent possible bank bankruptcies, then speculative attacks to currency can cause a crash. Thus present bank problems indirectly result in future currency crises. Financial dollarization is also attributed to currency problems which take root from banking system troubles (Mishkin, 1996). The exposure to foreign exchange risk can reach dangerous levels if dollarization is common in a country. When future expectations turn bad, this can throw banks into illiquidity problems that create upward pressure on the foreign exchange-rate. Depending on the exchange-rate regime and international reserves of the government, continuous upward pressure on the foreign exchange-rate can eventually result in the devaluation of the domestic currency. Velasco (1987) presents another feature of exchange-rate regimes and bank panics. In fixed exchange-exchange-rate regimes, the government‟s bailout to the banking system, which is aimed at the relief of illiquidity problems, can damage the currency if the bailout is inconsistent with exchange-rate stability.

(18)

unhedged foreign currency liabilities (Glick and Hutchison, 2000). Overborrowing is likely in systems of fresh financial liberalisation and deposit insurance. McKinnon and Pill (1997) report that a credit boom of foreign and domestic currencies with moral hazard incentives can lead the economy to the joint collapse of the currency and the banking system.

1.5.5. The Empirical Studies of Twin Crises

1.5.5.1. The Article by Kaminsky and Reinhart (1999)

Constructing my paper on the influential article of Kaminsky and Reinhart (1999), I would like to present their analysis first among the empirical studies of twin crises. Inspiring from Kindelberger‟s study of financial crises (1978), Kaminsky and Reinhart assume that economic variables behave in repetitive, generalizable ways and they seek common economic patterns underlying the financial crises. Indeed, by examining events in the banking and external sectors, they surmise that the balance-of-payments problems, banking turmoils and financial liberalisation have similar backgrounds. Once acknowledged the causes of financial complications they suggest that when macroeconomic indicators pass certain „threshold‟ levels, a crisis may be possible in near future. In other words the development of economic variables can be used as a signalling system where the frequency of the signals coming from macroeconomic indicators point to the fragility of economy. Therefore, Kaminsky and Reinhart (1999) propose a simple framework for predicting the crises.

To examine the link between banking and balance-of-payments problems, and macroeconomic variables Kaminsky and Reinhart (1999) first identify the dates of banking and currency and banking crises. The years that they analyse make a time period that many countries applied fixed exchange-rate systems, had high and even in some cases hyper inflation periods, tried exchange-rate based stabilisation programmes, liberalised financial accounts, and had several financial crashes. Taking such developments of the economies into account, they focus on events with regard to banking crises, while they themselves specify the dates of currency crises by an index of turbulence.

(19)

label to beginning of banking crises by “(1) bank runs that lead to the closure, merging, or takeover by the public sector of one or more financial institutions, and (2) if there are no runs, the closure, merging, takeover, or large-scale government assistance of an important financial institution (or group of institutions) that marks the start of a string of similar outcomes for other financial institutions.” Another decision to be made about marking the dates of banking crises is whether taking the beginning or the peak of the crisis into account, where the peak refers to the time with the most bank closures and/or heaviest government intervention. Doing the analysis twice, once with the beginning dates and once with the peaks, they find that using the peak dates provides no valuable information. Note that the core idea behind Kaminsky and Reinhart‟s analysis is that usually banking crises precede the currency crises, and that the peak of the banking crises comes after the currency crisis. Hence, using the beginning of the banking crises makes more sense as their findings provide proof.

As to the dates of currency crises, Kaminsky and Reinhart (1999) construct their own index of market turbulence inspiring from Eichengreen et al (1995, 1996). While Eichengreen et al (1996, p.475) take exchange-rate, international reserve and interest rate changes into the account of „exchange market pressure‟, Kaminsky and Reinhart ignore the interest rate component. Their index measures the significance of the change in exchange-rate and the loss of foreign exchange reserves that government use in maintaining the value of domestic currency. The dates of critical change are marked as crisis. Finally, “the episodes in which a currency crisis occurs within the forty-eight months of the beginning of a banking crisis are marked as twin crises.” (Kaminsky and Reinhart, 1999, p.477)

The sample of Kaminsky and Reinhart (1999), therefore, consists of seventy-six currency and twenty-six banking crises from 1970 until 19956. Their selection of countries is based on the criteria of being small open economies, applying fixed or crawling peg regime or band through portions of the sample and data availability, making a set the same with mine minus the US, Germany and Iceland. The US has mostly been excluded from the late empirical studies of financial crises but not after the global crisis of 2007 was born in the US.

The economic variables they focus on are the ones popularly emphasized in the theoretical and empirical literature of financial crises, especially parallel the studies of Eichengreen et al (1995). The economic variables that Kaminsky and Reinhart (1999) focus on are the ones

6_{Kaminsky and Reinhart (1999) also take other cases from out of the sample, which are Indonesia‟s Aug.1997}

(20)

popularly stressed in the theoretical and empirical literature of financial crises, especially parallel the studies of Eichengreen et al (1995). Foreign exchange reserves, measurement of excess money balances, interest rates and external shocks to foreign trade are the ones often related to the twin crises in the literature. Imports, capital flows, bank credits, equity prices, and production are commonly studied factors in relation to inflation stabilisation programmes and financial liberalisation stories. Among these economic variables that Kaminsky and Reinhart pay attention, I focus on a rather limited set, which still covers the most relevant issues.

(21)

finding is derived from the third one. Since the banking and currency crises are mostly predated by bad economic conditions, before twin crises the economic fundamentals are expected to be rather worse. Indeed they are, as proved by Kaminsky and Reinhart (1999).

1.5.5.2. Other Empirical Studies of Twin Crises

In March 2000, Glick and Hutchison present an analysis that is strongly parallel Kaminsky and Reinhart‟s analysis of 1999. They also study on the individual and intersected incidences of banking and currency crises for almost the same period of 1995-1997, but with larger sample of ninety industrial and developing countries. While Kaminsky and Reinhart (1999) focus on the countries which experienced crises in the analysis period, Glick and Hutchison also include countries that did not have either a banking crisis or currency crash during the sample period. They emphasise this approach‟s controlling and generalizability benefit. Glick and Hutchison too assess the linkage between crises, by testing the causality from banking sector distress to currency crashes, vice versa, and joint. Applying the noise-to-signal ratio methodology that was first used by Kaminsky and Reinhart (1999), they conclude that banking crises are indicators of future currency crises, specifically in emerging market economies where thirty percent of the banking crises entwine with currency crashes. However, currency crises do not seem to be a significant signal of upcoming banking crises. An interesting feature to be pointed out from Glick and Hutchison‟s study is that they come up with parallel findings although they start with different definitions of crises. The currency crises in their analysis are marked by using an index of currency pressure, which is constructed as the weighted average of real exchange rate changes and reserve losses and records „large‟ changes at crisis dates. The approach of large changes differentiates from Kaminsky and Reinhart (1999) by first taking the purchasing power of domestic currencies into account, second in the issue of handling high and/or hyper inflationary periods, and third in the reading of index7. As for banking crises, Glick and Hutchison posit the commonly shared data limitation problems about the measurability of the scale of bank runs, government interventions and the degree of deterioration of asset quality in the financial system. Like Kaminsky and Reinhart (1999) and I do, they also refer to other studies and well-known sources of information like International Monetary Fund‟s WebPage. The basic reference that

7_{Glick and Hutchison define „large changes in exchange rate pressure as changes in pressure index when it}

(22)

constructs their list of banking crises is first Demirgüç-Kunt and Detragiache‟s study (1998a) and second Caprio and Klingebiel‟s study (1996), which is also a main source for Kaminsky and Reinhart (1999).

Glick and Hutchison come up with a sample of 90 banking and 202 currency crises of where there are 37 twin occurrences. As a nuance, they use annual crisis observations in their analysis instead of months-approach which they find arbitrary. They assert that banking crises cannot be dated in such precision, and moreover monthly presumes would hinder data availability and thus make the scope of the sample smaller.

2. DATA, METHODOLOGY AND RESULTS

Four sets of decisions should be made to examine the causes of crises and ascertain the fragility of economies on the eve of crises. These decisions are based on the research of Kaminsky and Reinhart (1999). The first necessary definition is the notion of crisis. Second, it should be decided upon which economic variables can properly signal an upcoming crisis. Third, the behaviours of the economic indicators should be classified as either a signal of a crisis or normal (not a signal). Forth, a period of time should be defined. When an indicator signals, a crisis should happen within this proper time period if it is a true signal. When no crisis happens, the signal is to be classified as false alarm.

2.1. Definitions, Dates, and Incidence of Crises

In the search of a connection between the banking and currency crises and what economic developments can lead countries to such crises, I need to define the balance-of-payments (currency) crisis, the banking crisis and the twin crises where currency and banking crashes intersect. I applied the same approach with Kaminsky and Reinhart (1999) to define the dates and incidence of the crises.

Currency (Balance-of-Payments) Crises: I use the „index of currency market turbulence‟,

which is constructed by Kaminsky and Reinhart (1999, p.498) based on previous studies of Eichengreen et al. (1996).

(23)

𝐼 =

∆

𝑒

−

σ

_e

σ

_R

×

∆

_R

R

(1)

σe is the standard deviation of the rate of change of exchange rate and σR is the standard

deviation of the rate of change of reserves.

The most basic translation of a currency crash is the depreciation of the currency, hence the changes of the exchange rate has a positive weight in the index. On the contrary, the changes of reserves has a negative weight attached reflecting the loss of reserves in a country.

The month where the index is three standard deviations or more above the mean is marked as the beginning of a currency crisis. In the sample, there are countries with turbulent periods that last long. There, exchange rates and reserves have shocks over and over again during consecutive months. So the index labels consecutive months as crises. However, following the exclusion principle of Eichengreen et al (1996), the list of found crises is shortened so that each country contributes no more than two crises in a year. In other words, to avoid counting the same turmoil more than once, the later observation(s) is (are) excluded when two (or more) crises occur in consecutive quarters8.

A modification is made for the countries that had hyperinflation since substantial devaluation of the currency is not unlikely in the countries with hyperinflation, while 100 percent depreciation may be an ordeal for low-to-moderate inflation economies. To eliminate the possible biasing effect of hyperinflation periods over the whole sample and to avoid missing sizable losses of reserves and devaluations, the sample is partitioned according to whether the inflation was higher than 150 percent in the previous six months. Then separate indices are constructed for each subsample.

The resulting list of currency crises, which is composed of forty-five incidences, is presented in Table 1. It fairly accords with the historical chronology of currency market turmoils of the real world.

Banking Crises: Relying on the existing studies of banking crises9 and on the financial press, I constructed a list of crises (Table 1). Parallel to Kaminsky and Reinhart (1999, p.476), the

8_{Glick and Hutchison (2000), on the other hand, discard any later observation within the 24-month window of}

the first crisis.

9_{The order of reference is Kaminsky and Reinhart (1999), Laeven and Valencia (2008, "Systemic Banking}

(24)

incidents of “bank runs that lead to the closure, merging, or takeover by the public sector of one or more financial institutions, and/or large-scale government assistance of an important financial institution (or group of institutions) that marks the start of a string of similar outcomes for other financial institutions” are labelled as banking crises. Other contemporary studies are also referred. The second reference source is the research of Laeven and Valencia (2008) where systemic banking crises are defined according to the incidents when;

- the frequency of bankruptcies substantially increase among corporate and financial sectors,

- the corporations experience severe illiquidity problems, - the number of non-performing loans surge,

- asset prices fall sharply,

- interest rates significantly increase in real terms, - capital inflows contract, capital outflows increase, - and bank panics occur.

The third reference for defining the banking crises, Caprio and Klingebiel (2003, p.1), have a simpler definition for systemic banking crises, which is „incidents where much or all of bank capital being exhausted‟. The approach of Reinhart and Rogoff (2009) is explained in the first section as a part of the literature review for the banking crises. Jacome (2008) who examines the banking crises in Latin America labels the events where at least one institution is intervened and/or closed, or becomes subject to resolution as banking crises. By referring to multiple sources where banking crises are approached from various perspectives, I try to ensure the reliability of this research. The final list of banking crises is composed of twenty-five crises.

I focused on the beginnings of the banking crises, not the peaks, following the approach of Kaminsky and Reinhart (1999). The banking crises whose beginning month is not clarified in the literature and press are assumed as starting in the sixth month of the relevant year.

Twin Crises: “Episodes in which the beginning of a banking crisis is followed by a

(25)

Table 1 The Timing of the Banking, Currency, and Twin Crises

COUNTRIES Banking Crises Currency Crises Twin Crises

Argentina January 1995 March 2001

January 2002 January 2002

Bolivia November 1994 October 2008 -

Brazil February 1990 July 1994 January 1999 - Chile - June 1999 October 2008 -

Colombia April 1998 August 1995 -

Denmark August 2008 October 2008 October 2008 Finland September 1991 October 1991 October 1991

Germany - February 1991

October 1992

- Iceland 1993

September 2008 September 2008 September 2008 Indonesia November 1992

November 1997 May 2002

December 1997 and June 1998 October 2008

December 1997 and June 1998

Israel - October 1998

October 2008

-

Malaysia September 1997 July and December 1997 December 1997 Mexico October 1992 December1994

October 2008 December1994 Norway - November 1992 December 1997 October 2008 - Peru 1999 December 1990

April and September 1991 September 1992

-

Philippines July 1997 July and December 1997 October 2000

July and December 1997 October 2000

Spain - September 1992 -

Sweden November 1991 November 1992 October 2008

November 1992

Thailand May 1996 July and November 1997 July and November 1997 Turkey January 1991 April 1994 November 2000 March 1994 February 2001 October 2008 March 1994 February 2001

United States August 2007 - -

Uruguay January 2002 June 2002 June 2002

Venezuela October 1993 May 1994 December 1995 April 1996 February 2002 January 2003 January 2009 May 1994 December 1995 April 1996

(26)

As listed in Table 1, I construct a set of forty-five currency and twenty-five banking crises for the period of 1990-2009. Twenty crises out of forty-five currency crises are twin crises. The frequency of crises over time is depicted below in Table 2 and graphically in Figure 1.

Table 2 Frequency of Crises over Time

Number of Crises

1970-1995 1980-1995 1970-2009 1990-1999 2000-2009

Type of

Crisis Total

Average

per year Total

Average

Average per year Currency 76 2.92 50 3.13 106 2.65 28 2.8 17 1.7 Twin 19 0.73 18 1.13 33 0.83 14 1.4 6 0.6 Single 57 2.19 32 2.00 73 1.83 14 1.4 11 1.1 Banking 26 1.00 23 1.44 39 0.98 18 1.8 7 0.7

Figure 1 - Frequency of Crises over Time

To understand the links between banking and currency crises, the unconditional probability of banking and currency crises should be clarified first to have a preliminary and objective insight about the likelihood of crises at any time. For this, crisis windows are defined as definite time periods before and/or after the beginning of the crises. Same as the definitions of

(27)

Kaminsky and Reinhart (1999, p.479), the currency crisis window is the 24 months preceding the currency crisis, while the banking crisis window is the 12 months before and after the beginning of the banking crisis. This approach of shifting the window 12 months forward for the banking crises can mitigate the drawbacks of focusing on the beginnings of banking crises (not the peaks) since later periods where the worst of crises may come are also covered. The

unconditional probabilities of currency and banking crises are calculated as the proportion of

total number of months in the respective crisis windows to the total number of months in the whole sample. The currency crisis probabilities conditional on the beginning of a banking crisis are calculated as the number of months in the currency crisis windows that occur within 24 months of the beginning of banking crisis divided by the total number of months in the banking crisis windows. The probabilities of banking crisis conditional on the beginning of a currency crisis are calculated as the number of months in the banking crisis windows that occur within 24 months of a currency crisis divided by the total number of months in the currency crisis windows. The results are shown in Table 3, where conditional probabilities on both ways are higher than the unconditional probabilities, as expected10.

Table 3 Probabilities of Crises

Probabilities of currency crises

Unconditional 16 %

Conditional on the beginning of a banking crisis 21 %

Probabilities of banking crises

Unconditional 11 %

Conditional on a currency crisis 14 %

2.2. The Macroeconomic Background of Crises

To see whether banking and currency crises both have similar backgrounds, a set of macroeconomic and financial indicators around the time of the crises is analysed. The indicator associated with financial liberalisation is the ratio of domestic credit to gross domestic product (GDP). Another financial indicator that can be related to the level of liberalisation is real commercial bank deposits. The indicators linked to the current account include the percent deviation of the real exchange rate from trend, and the values of exports

10

(28)

and imports. The indicator associated with capital account is foreign exchange reserves. Finally, the indicator of real sector is an index of stock prices.

Domestic Credit/GDP: IFS11 line 32 (not line 52 as in Kaminsky and Reinhart (1999), that one is no longer available) divided by IFS line 64 to obtain domestic credit in real terms, which was then divided by monthly real GDP amounts. Real GDP data is interpolated from annual data obtained from World Economic Outlook of IMF (2010) since the original data source of Kaminsky and Reinhart (1999) which was IFS line 99b.p is no longer available. Base month for real term series is December 2000.

Deposits: IFS line 24 plus 25 deflated by consumer prices (IFS line 64). The base month is

December 2000 for real term series.

Real Exchange Rates: Monthly real exchange rates are broad for 58 countries from the BIS12

database. Unlike the data of Kaminsky and Reinhart (1999), an increase of the real exchange rate index denotes an appreciation of the domestic currency in my data set. Instead of annual growth, I focus on deviations of the real exchange rate from trend as a measure of misalignment. The trend is specified as linear.

Exports and Imports: IFS lines 70 and 71 respectively. All values are converted to US

dollars with the exchange rate data from the IFS database.

Reserves: Data gathered from IFS line 1L.d. and converted to US dollars for all countries.

Stock Prices: The quotes from the main boards are gathered from the Datastream database.

All stock prices are in US dollars.

I focus on the 12-month percent changes of all variables, except from the real exchange rates where deviations from trend are examined. The behaviours of variables before and after the crises are compared against the average behaviours during the „tranquil periods‟ which are all the remaining observations in my sample. For currency crises (total, single and twin), the 18 months before and after the beginning months of the crises are marked as non-tranquil months. For banking crises, the 27 months before and after the beginning month of the crises are marked as non-tranquil months. Banks try hard to hide the weaknesses in their balance sheets; it usually takes time for banks to reveal their problems, and for governments to take

11

International Financial Statistics (IFS) is a database maintained by the International Monatary Fund.

12

(29)

supporting actions. Bank panics, government bailouts and take-overs are the events that specify the beginning dates of the banking crises in my analysis – as it is in Kaminsky and Reinhart (1999) – that may cause the late marking of crisis dates. Having the crisis window wider for banking crises enable to mitigate this complication and to conduct a more sensitive investigation of the evolution of those crises.

The steps of analysis are as follows:

- The average 12-month change of each variable (deviation from trend for real exchange rate) during the tranquil periods are calculated for each indicator and for each country. - The percent difference between the tranquil and crisis periods are calculated for each

month in the non-tranquil (crisis) periods, for each variable and for each country. - To reach a general conclusion, the monthly percent differences between the tranquil

and crisis periods is aligned together for all the crises, separately for banking and currency crises (total, single and twin). Then the average differences are calculated for each month in the non-tranquil (crisis) periods, for each variable, for all crises.

The graphs under the Financial, External and Real Sector headings under section 2.2 show the average behaviours of economic indicators relative to tranquil times, for banking crises, currency crises totally, and twin crises in comparison to single crises separately. The percent values of differences between the crises and tranquil times are reported on the vertical axes (1 = 100 percent). The horizontal axes depict the numbers of months before (with a negative sign) and after the crises. The months with missing data are deducted from the calculations of averages.

(30)

Furthermore, there is another problem with comparing the annual growth of a variable (or deviation from trend for the real exchange rate indicator specifically) against its average annual growth (deviation from trend for the real exchange rate indicator) during tranquil times, in terms of percentages. Let 𝑥𝑡 be the annual growth of a variable (or deviation from

trend for the real exchange rate indicator) at time 𝑡, and 𝑥_𝐴𝑉𝑇 be its average during the tranquil times. The difference between those in percentages is as follows:

𝑥

_𝑡

− 𝑥

_𝐴𝑉𝑇

𝑥

_𝐴𝑉𝑇 (2)

Note that when 𝑥_𝐴𝑉𝑇 of an indicator has a negative value, the result of (2) changes sign. For example, let the country „c‟ has a negative deviation from trend for its real exchange rate on average during non-crisis (tranquil) periods – a value like -0.003. The real exchange rate decreases when there is a devaluation of the domestic currency. Normally, the crises have a depreciating impact on the value of the currency such that in a crisis event at time 𝑡, the value of 𝑥_𝑡 has a smaller negative value – say -0.30 – for any given month in the crisis window. The percent difference between be 𝑥_𝑡 and 𝑥_𝐴𝑉𝑇 becomes 99 percent, positive. The graphical presentation of this would then be as if the real exchange appreciates so much during the crisis period when compared to its trend in non-crisis times. Indeed, some of the countries in my sample have negative averages for tranquil times and they do experience substantial devaluations with the crises. This is why the end graphics that cover all countries and crises show the real exchange rate evolution as if the rates appreciate by the crises.

To eliminate this effect of taking the percent difference between the behaviours of variables during the crisis times and the tranquil times, I recalculate the differences without taking the percentage by the formula below:

𝑥

_𝑡

− 𝑥

_𝐴𝑉𝑇 (3)

The graphs regarding this alternative comparison are provided in the Appendix13.

The following interpretations can be made regarding the behavioural patterns of economic variables on the onset and aftermath of the crises by having a closer look at the indicators related to financial, external and real sectors.

13_{Note that, the interpretations here are based on the figures which are obtained by applying the method of}

(31)

2.2.1. The Financial Sector

In the research of Kaminsky and Reinhart (1999), the financial sector indicators are originally aimed at studying the effects of financial liberalisation experiences of countries since financial liberalisation seems to precede banking crises in many cases. However, in the period that is subject my research the countries in the sample have already had years in the liberal system. Therefore, rather than directly connecting the development of crises to the liberalisation movement, I study the post and indirect effects of liberalisation in this research.

0 2 4 6 8 10 12 14 16 18 20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Domestic Credit / GDP - Currency Crises

Domestic Credit / GDP -20 -10 0 10 20 30 40 50 -27 -24 -21 -18 -15 -12 Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis 12 15 18 21 24 27

Domestic Credit / GDP - Banking Crises

(32)

The domestic credit / GDP ratio grows markedly higher than normal as the crises approach.

The highest difference between pre-crisis periods and tranquil times is observed in the cases of twin crises, while almost no rise is observed in the cases of single currency crises. However when currency crises are examined without any distinction of single or twin crises, the ratio soars above the average of tranquil times, as expected in credit boom stories that usually take place prior to crises. On the other hand, the annual growth of domestic credit / GDP records the highest decline (from the high levels that it reaches on the eve of the crises) as the banking crises unfold. Note that the crisis period for banking crises is considered longer in respect to currency crises, a more sensitive analysis of the evolution of this ratio is possible here. The ratio seems to get in a diminishing trend around the first months of the crisis (see the levels before and after the first months named as „crisis‟ on the horizontal axes). A surge follows this decline of the ratio, which is probably due to the increase of money in the system that is provided by central bank to help banks get out of financial troubles. However, as the problems become conspicuous and the economy sinks into the depths of the crisis, the ratio falls below the average of tranquil periods. For all types of crises, the ratio is highly volatile.

-20 -10 0 10 20 30 40 50 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Domestic Credit / GDP - Twin & Single Crises

(33)

During the months preceding the crises in Kaminsky and Reinhart‟s analysis (1999), the growth rate of real bank deposits seems to keep close to normal times. However in my

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Deposits - Currency Crises

Deposits -1.5 -1 -0.5 0 0.5 1 1.5 2 -27 -24 -21 -18 -15 -12 Cri sis Cri sis Cri sis Cri sis C ris is Cri sis Cri sis 12 15 18 21 24 27

Deposits - Banking Crises

Deposits -2 -1.5 -1 -0.5 0 0.5 1 1.5 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Deposits - Twin & Single Crises

(34)

research, the indicator has a better performance which gradually worsens until the crisis. This pattern is similar in both banking and currency crises. For the cases of banking crises, it records a higher fall in comparison to currency crises such that the loss of deposits (which is a source for banks) may indeed be one of the causes that lead the banks into trouble. The time needed to catch up the pre-crises growth rate is lengthy for both types of crises, which is longer than eighteen months. This may be a result of bank runs or if not that much, change in the saving and investment behaviours of the households, companies, and investors. Once the crisis starts, the parties having money tend to prefer the safest ways of investing and/or saving. Saving in terms of gold as a safety heaven is a known as a common behaviour in the crisis times. Companies tend to strengthen their capital stock as a safety buffer from bankruptcy. Banks prefer to call in outstanding receivables from borrowers. So, the deposits of saving can be spent for recapitalisation and/or credit-payment purposes. The lack of deposits or even a slow growth would make it harder for economies to recover from crises, since banks need a robust base of deposits to be able to lend and companies usually need bank credits to conduct their activities.

2.2.2. The External Sector

-2 -1.5 -1 -0.5 0 0.5 1 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Exports - Currency Crises

(35)

The annual growth of exports records a worse than normal performance in the eighteen months preceding both types of crises. This poor performance makes an exception in both types of crises again, seen as a modest and short-lived rise as early as one year prior to the crises. The decline in the midst of the crises is worst and sharpest during the twin crises events, as expected. The difference of exports growth between crisis times and normal times is around 100 percent during banking crises, and worse, around 150 percent during currency crises; possible due to the high fluctuation in the exchange rates, and ponderous trade activities. The appreciation of exports growth, so that it catches up the tranquil times‟ pace, is only after one year from the crises.

-1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 -27 -24 -21 -18 -15 -12 Cri sis Cri sis Cri sis Cri sis C ris is Cri sis Cri sis 12 15 18 21 24 27

Exports - Banking Crises

Exports -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Exports - Twin & Single Crises

(36)

The growth of imports, on the other hand, also records a bad performance prior to the crises; it remains below the growth average of tranquil times. The adverse development of imports is

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Imports - Currency Crises

Imports -1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 -27 -24 -21 -18 -15 -12 Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis 12 15 18 21 24 27

Imports - Banking Crises

Imports -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Imports - Twin & Single Crises

(37)

worse in the case of currency crises, probably due to the negative impacts of exchange rate conditions. Imports growth makes the sharpest decline in the cases of twin crises. Recovery can be achieved only after sixteen months from the beginning of the crises, even later with the banking crises possibly because of the tightened credit lending by the banks to the sectors and due slowed economic activity.

-15 -10 -5 0 5 10 15 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Real Exchange Rates - Currency Crises

Real Exchange Rate

-15 -10 -5 0 5 10 15 -27 -24 -21 -18 -15 -12 Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis 12 15 18 21 24 27

Real Exchange Rates - Banking Crises

(38)

Real exchange rate index14 indicator is way above zero on the eve of the crises, which can be translated as being overvalued relative to its average during non-crisis periods. The appreciation of the index is especially higher prior to the twin crises events, and not surprisingly the vastest devaluation of the currency occurs when the currency clashes and financial system meltdown intersect. Another notable act of the index is around the banking crises where the trough is reached with a lag. In other words, the declining trend of the index lasts longer in the cases of banking crisis. The sharp fall of the real exchange rate indices make sense since the Latin American, Asian and Turkish currency catastrophes of 1994-2001 periods, and the recent global crises are included in the sample. Note that half of the countries in the sample have balance-of-payments problems in October 2008. The failures of central banks in defending fixed exchange rates have resulted in the abandonment of the pegs. Moreover, the exchange-rate based inflation stabilisation plans today are no longer as popular as they were once.

14

Note that the interpretation related to the real exchange rate index is based on the figure where the difference between the behaviour of the index during crisis times and non-crisis times is not in terms of percentages.

-25 -20 -15 -10 -5 0 5 10 15 20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Real Exchange Rates - Twin & Single Crises

(39)

Reserves, as a proxy for capital account developments, show the typical trend of falling on

the eve of the crises, as an indicator of central banks‟ efforts to maintain the exchange rates in

-1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

Reserves - Currency Crises

Reserves -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 -27 -24 -21 -18 -15 -12 Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis Cri sis 12 15 18 21 24 27

Reserves - Banking Crises

Reserves -3 -2 -1 0 1 2 3 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18

The Backgrounds of Banking and Currency Crises