Is There a Relationship between Financial Crises and the Type of Financial System?

Is There a Relationship between Financial Crises

and the Type of Financial System?

A Cross-Country Analysis

By

Cristel Stegge

University of Groningen

&

University of Uppsala

Faculty of Economics and Business

MSc. International Financial Management

April 2008

Abstract

This research examines to which extent the type of financial systems has an impact on the occurrence of financial crisis. Two types of systems were examined (bank-based and market-based) and three types of financial crises (currency, banking and debt crisis). It was assumed that market-based systems are more vulnerable to the occurrence of a financial crisis due to the theory of asymmetric information which in literature is the main cause of financial crisis. It is argued that bank-based systems are better able to control for asymmetric information. By means of a logistic regression analysis it was proved that financial systems have a significant impact on the probability of the occurrence of currency and banking crises and that the market-based countries are more vulnerable for this. There was no significance found for the impact of financial system on debt crisis.

Keywords

Market-based system, bank-based system, currency crisis, banking crisis, debt crisis Supervisor

Dhr. Niels Hermes Referent

Table of Content

1.5 Organization of the paper 6

Chapter 2 Theoretical Background 7

2.1 Financial Systems 7

2.1.1 Bank-based financial systems 7

2.1.2 Market-based financial systems 8

2.1.3 The bank-based or market-based system? 10

2.2 Financial Crises 11

2.2.1 Currency Crises 11

2.2.2 Banking Crises 13

2.2.3 Debt Crises 14

2.2.4 Contagion 16

2.3 Connection between Financial System and Financial Crises 16 2.3.1 Asymmetric Information, Moral Hazard and the Comparative

Advantage of Banks 17

2.3.2 Firm’s Relationships with Banks 18

2.3.3 Globalization, Liberalization, International Capital flows and

Information Asymmetries 19 2.3.4 Bank Concentration 22 Chapter 3 Methodology 23 3.1 Methodology 23 3.1.1 Financial Systems 23 3.1.2 Financial Crises 24 3.1.3 Control variables 25 3.2 Research design 28 3.2.1 Sample 28

3.2.2 Dependent and Independent variables 28

3.2.3 Logistic Regression Analysis 28

Chapter 4 Analysis 34

4.1 Currency Crises 34

4.2 Banking Crises 37

4.3 Debt Crises 39

4.4 The effect of financial system on financial crisis 42

Chapter 5 Discussion and Conclusion 44

References 48

Chapter 1 Introduction

Financial systems are a crucial mechanism in economies since they allocate resources. They channel household savings to the corporate sector and allocate investment funds among firms, they allow inter-temporal smoothing of consumption by households and expenditures among firms. So they allow both households and firms to share risk (Allen and Gale, 2000). When looking at the different countries around the world we can see that financial systems differ extremely, triggered by different fundamentals, history and development of policy making. In literature a major attempt has been done in understanding the different financial systems and which work best. In general two broad categories of systems are being distinguished: market-based financial systems and bank-market-based financial systems. Authors like Levine, Demirguc-Kunt, Gale, Allen, and Beck have dedicated lots of their work to this topic.

In literature there have been many discussions what the better system is, or whether there actually is a perfect system, particularly in the relationship with economic growth. Several researchers found that there is no perfect and fully efficient system for economic growth (like Levine). Imperfections and inefficiencies are also occurring in the different systems and sometimes they get so severe that it can result in a financial crisis (Allen and Gale, 2000). So, the question arises: does financial crises occur more in one particular system than the other? Or does financial system not matter much in this case, and why?

In the past much research has been done on the relationship between the financial system and economic growth in order to test the successfulness of one of the two systems.

In my research I am going to examine the relationship between financial system and the likelihood whether a financial crisis is more likely to occur in one system or the other, in this way we can find out whether economies in one system are more susceptible to it than another. The type of crisis will be examined as well. In order to find out I will make a cross-country comparison between the type of system and type of financial crises occurring in these countries and systems.

Because of my research it would become clear whether a particular system would be more sensitive to a crisis, and from the other side which one is the most stable towards a financial crisis. There have been much research in the field of financial crises, but still it is surrounded by mystery of why it occurs and why one country is more susceptible to it than the other. Even though my research might be only a little part of the puzzle, it could help in understanding this complex matter a little more. This brings me to the following research question:

1.2 Research Question

Is there a relationship between the type of financial system in a country (market-based or bank-based) and types of financial crises occurring in these systems?

1.3 Sub questions

1) What are the characteristics of the market-based and bank-based financial systems and what are the main differences in functioning?

2) What are the characteristics of the financial crises occurred in the sample countries? 3) What is the link between financial crises and financial systems?

4a) Is there a significant relationship between the financial crises and the financial systems available in the countries

1.4 Conceptual model

1.5 Organization of the paper

The paper will be organized as follows. In chapter 2 I will first discuss the theories on the different types of financial systems and financial crises. At the end of that chapter those two streams of literature will be linked together. Chapter 3 will explain the methodology used in this research. In chapter 4 the results of the logistic regression analysis will be discussed followed by the discussion and conclusion in chapter 5.

Financial crisis

Currency

crisis Banking crisis

Chapter 2 Theoretical Background

In this chapter I will use the available literature on both the characteristics of the different financial systems and the literature of why the different financial crises occur. In the end these two different streams of literature will be linked and expectations of relationships will be made, which will be tested in the analysis phase of my research.

2.1 Financial systems

The financial system of a country is considered to be the mixture of financial instruments, markets and intermediaries (Demirguc-Kunt and Levine, 1999). They are crucial for the allocation of resources in a modern economy, but not all systems in the world are doing this in the same way. Many experts view stock markets as an ideal mechanism in allocating resources. However, there are many economies in which stock markets are unimportant. In these countries funds are obtained through banks. From this view financial systems can then be divided in market-based systems and bank-based systems. But, there is more. Throughout the world no financial system is the same. Much of the research on financial systems has been done focusing on four countries. Germany and Japan were considered to be bank-based countries while the United States and the United Kingdom are market-based. Considering the fact that only four countries are researched here and they all are well developed, it is hard to make generalizations about it. It is also particularly hard to make generalizations within the same category of financial systems. For example, Germany and France are both considered to be bank-based countries, however when we compare these countries with each other, we can see that within the both countries there are considerable differences in how systems work (i.e. in France the importance of stock markets is slightly higher than in Germany). With keeping this in mind, the next section will explain the two different systems more in depth.

2.1.1 Bank-based systems

appoint one information producer in acquiring information. This way duplication costs are avoided and the process will be more efficient (Ramakrishan and Thakor, 1984); 2) Banks are able to manage cross-sectional, inter-temporal and liquidity risk and thereby enhancing investment efficiency and economic growth. According to the proponents, banks are able to examine and manage risk better than markets due to the fact that they have a better insight in the situation of a firm because of close bank-firm relationships, in markets this might be harder due to the fact that the information that is revealed in markets is more incomplete; 3) Mobilizing capital to exploit economies of scale (Levine, 2002). Because banks operate as an intermediary for a coalition of investors they are argued to be more effective and efficient than markets to exploit economies of scale in information processing, ameliorate moral hazard through effective monitoring and form long-term relationships with firms to ease information distortions.

Proponents of the bank-based view state that, traditionally the features of a bank-based system is that they are better in mobilizing savings, identifying good investments, and exerting sound corporate control, particularly during the early stages of economic development (Levine, 2002). This is because of the relation-based nature of the banks, companies and sectors that are in development can get the investment they need from banks. In these early development periods, the financial markets are still very volatile and may not have the capacity to work well in this situation (Rajan and Zingales, 2001). They also claim that banks, because of their intermediary role, are better at monitoring firms and reducing moral hazard than uncoordinated markets (Boot and Thakor, 1997; Levine, 2002). This is because players in a market operate alone and do not merge these monitoring activities, which make the cost lower and quality higher for banks on these monitoring activities. The proponents of this view also stress that liquid markets create a short-sighted investor climate (Bhide, 1993), with the consequence that in liquid markets investors can inexpensively sell their shares, so that they have fewer incentives to exert rigorous corporate control. According to Sabani (1992), market-dominated economies will restructure financially distressed borrowers less than bank-dominated economies. This is also due to the long-term relationship banks are having with their borrowers. Banks can collect the cost they make on these restructurings in a later stage when the borrower is back in a sound financial state.

2.1.2 Market-based systems

the market-based view will reduce inefficiencies associated with banks and enhance economic growth.

2.1.3 The bank-based or market-based system?

So, can we say that one system is better? Among researchers and countries the opinions differ significantly and researchers look at the same problems from a different view. Allen and Gale (2000) state that in recent years countries are moving more towards the market-based system. They take as example France and Japan that are developing policies and reforms towards the increase of the importance of financial markets (Allen and Gale, 2000) and that there is a tendency towards the convergence of systems. Boot and Thakor (1997) conclude in their article on financial system architecture that as when a financial system evolves in an economy, even when it starts as a bank-dominated system, will move towards a system where banks lose market share to the financial market (Boot and Thakor, 1997). This same view is supported by Stulz, who claims that when a financial system develops, the stock market will play a more important role (Stultz, 2001).

On the other hand Chakraborty and Ray state in their 2006 article that neither system is superior, however the bank-based system outperforms the market-based one on several dimensions like a higher investment and income per capita in bank-based countries, and the existence of a lower income inequality in these systems. So, different opinions rule in this field.

A comparison of different financial systems is complex, each type of system has its own advantages and disadvantages. For that reason it is important to understand the different trade-offs between intermediaries and markets in financial systems. A conclusion on this matter can then be captured by a quote of Allen and Gale (2000): “In the end, it is not a question of markets versus intermediaries but rather of markets and intermediaries.”

2.2 Financial Crisis

The last decades the world has been hit by a large amount of financial crises. They can be distinguished into three broad categories: currency crises, banking crises, and debt crises (both sovereign and private sector debt crises) (Aziz, Caramazza, Salgado, 2000; Bussiere and Fratzscher 2002). In literature the term currency crisis is used the most, while when discussed the specific definition is not necessarily a currency crisis alone. This is because a financial crisis often involves a collapse of the currency, which significantly loses its value against another currency (Hausmann and Velasco, 2004). This is the reason why economists often speak of currency crises to describe all kind of meltdowns. But of course, there is more. Sometimes the crisis comes with a default of the government, which gets unable to pay its external debt (sovereign debt crisis). Other times it is the private firms that cannot pay their debts (private sector debt crisis). In other cases, it are banks that are getting into trouble because their own position gets weakened, or when their two major customers, government and corporations, are weakened, and the public is getting scared and bank runs occur (banking crisis) (Hausmann and Velasco, 2004). In the next sections the three different types of crises will explained in more depth.

2.2.1 Currency crises

Reinhart, 1998). More recent papers have claimed that authorities abandon parities on other reasons than depletion of international reserves. They just let the currency devalue because they may be concerned about the adverse consequences of policies to maintain parity on other key economic variables. For example, they are afraid that higher interest rates might negatively influence the level of employment (Kaminsky, Lizondo and Reinhart, 1998). The concern is that the problem will be moved rather than solved and will even have more negative effects on the long term.

Other recent models have also shown that crises may also develop without a change in fundamentals and that it is self-fulfilling. In these models economic policy is not predetermined, but responds to the changes in the economy. Agents take these relationships into account in forming their expectations and act accordingly. At the same time these expectations and actions of agents affect variables that may influence economic policy and then a self-fulfilling crisis may exist without a change in fundamentals (Kaminsky et al. 1998). Another theory on the occurrence of currency crises is contagion. This concept will be explained in paragraph 2.2.4.

Aziz et al. (2000) found some common characteristics for currency crises. They found that typically, before a currency crisis occurs 1) the economy was overheated. This is shown by a relatively high inflation and overvalued currency, which in turn affects the export sector and the current account balance; 2) Monetary policy was significantly expansionary. Authorities are trying to stimulate the economy because they see that problems are arising and trying to prevent downturn by for example trying to stimulate consumption. This can be seen by an increase of domestic credit in an economy; 3) Financial vulnerability of the economy was increasing. This is seen by rising liabilities of the banking system, unbacked by foreign reserves and falling asset prices. This influences the degree of liquidity of a country in the sense that the lower the liquidity, the lower the ability of a country to anticipate a speculative attack on the currency.

in the economy also respond to this stabilization policy that might negatively influence the economy in the long-term. They also found that crises have not grown more severe with time. 2.2.2 Banking crisis

A banking crisis exists when a panic occurs among bank customers about the soundness of the banks they have deposited their funds in. As a result they withdraw their money and hold it in currency (Allen and Gale, 2007). This can happen when customers are getting signs that a bank might be in trouble or when economic circumstances deteriorate. The definition of systemic banking crises usually refers to situations where many banks simultaneously come under pressure and may be forced to default since much of the bank capital is exhausted (Allen and Gale, 2007).

The theory of asymmetric information plays a large role in the occurrence of a banking crisis. According to Mishkin’s theory (1996), financial crises occur because of the asymmetric information theory. According to Mishkin the impediment to the efficient functioning of the financial system is asymmetric information, which is the situation in which one party to a financial contract has much less complete information than the other party (Mishkin, 1996). This leads to two basic problems: adverse selection and moral hazard. Adverse selection occurs before the transaction and occurs when ‘potential bad credit risks’ are most likely to seek out a loan. This is because they have most to gain from a loan since the expected return on their risky project is very high, while their cost is only the borrowed amount (Mishkin, 1996). Moral hazard occurs after the transaction because the lender is subjected to the risk that the borrower has incentives to perform activities that are undesirable from the lender’s point of view (activities that make it less likely that the loan will be repaid, like taking an excessive risk) (Mishkin, 1996). This theory of asymmetric information also explains why banks are such important players in the financial systems. Banks are well-suited to reduce adverse selection and moral hazard problems in financial markets, because they have good monitoring tools and long-term relationships with firms. This issue will come back in great detail later in this paper.

Detragiache, 1998). A number of authors have researched the causes of banking crises (among them Demirguc-Kunt and Detragiache, 1998; Caprio, 1998; Hardy and Pazarbasioglu, 1998). The main reasons mentioned when countries are more sensitive to systemic banking crises in these researches are: 1) A low GDP-growth, indicating a weak macroeconomic environment. In literature a significant correlation between a low growth rate and an increased risk to the banking sector is proved by Demirguc-Kunt and Detragiache (1998) and Hardy and Pazarbasioglu (1998). This shows that a deterioration in macroeconomic circumstances increase the probability of banking crises; 2) Excessive high interest rates and a high rate of inflation might also increase the risk of banking sector problems. According to Demirguc-Kunt and Detragiache (1998) this is because the high and volatile interest rates associated with high inflation, makes it difficult for banks to perform maturity transformation, which increases the risk of a crisis; 3) The presence of an explicit deposit insurance scheme tends to increase the probability of systemic banking sector problems. This is because these schemes cause increased moral hazard. When agents know that the deposits are insured against bank insolvencies, they are more likely to take an excessive risk since they know the losses will be compensated anyway.

Other aspects that play a role in systemic banking sector problems are: a speculative attack on the currency. When a devaluation is expected, depositors withdraw their funds and convert them in other currencies abroad, causing illiquidity for domestic banks. The degree of capitalization of banks also plays a role. Banking systems that are less capitalized are more sensitive to shocks because of their lower capital buffer (Demirguc-Kunt and Detragiache, 1998).

2.2.3 Debt Crisis

The debt crises can be divided into two different categories: the sovereign debt crisis, when the debt instrument is guaranteed by the government and the private debt crisis. When a traditional debt crisis occurs, the government debt has risen to a certain amount on which the government is unable to pay the debt and has to default on it.

moment, the arrears in payment does not vanish from one day to the other. The debt needs to be rescheduled and therefore the time before a country is debt-free can take some time.

Possible structural factors which may raise or lower the probability of debt crisis in a particular country have been mentioned by a number of researchers. These variables include: 1) High levels of foreign debt to GDP. This is an indicator of the solvency of a country, so whether it is able to pay its foreign debt. As this amount is increasing, it gets more difficult for a country to service this debt and the chance of sovereign default increases; 2) High levels of short-term debt, making the economy vulnerable to a debt crisis because external liquidity is decreasing, and this indicates a higher probability of default on debt. An increase in the level of short-term debt also indicates capital reversals from a country which is a sign that investors fear that economic circumstances deteriorate. Short-term funds are easiest to withdraw for investors; 3) Current account imbalances, which is also a measure of insolvency. Persistent imbalances may lead to an accumulation of debt and at some point surpluses are needed to avoid insolvency. So these imbalances may affect the ability to pay; 4) Exchange rate depreciation can also affect a country’s ability to pay. When there is a high level of foreign debt and the currency is depreciating, this can lead to severe balance sheet effects because the stock of debt in real terms can sharply increase after a devaluation; 5) Low GDP-growth and low trade openness. These variables affect the willingness of a country to pay its foreign debt, so here the chance of default is larger. The relative costs of paying their external debt might be higher for a country than when it decides to default on their debt. The main cost of defaulting is a loss of access to international capital markets and the output and trade cost due to the default. When growth is low, this loss of access to international capital markets is less costly and more open economies will lose more from the disruptions of international trade triggered by the default than less open economies; 6) Macroeconomic policy instability also influences a country’s risk to default. This is reflected in variables like a high inflation and high money growth. These variables reflect a country’s policy credibility and predictability and this influences the investors risk attitudes towards a country and their perceptions of the country’s willingness to pay its foreign debt (Manasse et al., 2003 and 2005).

2.2.4 Contagion

In all the above discussed crises, contagion may play a significant role. It occurs when a crisis occurs in one country and then spreads to neighbours who did not encounter the negative shock which hit the first country. This happened practically at all the large crises in the past, like the Mexican crisis, where it spread from Mexico to other Latin American countries like Brazil and Argentina and in the Asian Crisis. Contagion occurs because of the linkages that are existent between countries. The process is observed through co-movements in exchange rates, stock prices, sovereign spreads, and capital flows (Dornbusch, Yung and Claessens, 2000). The causes can be divided into two categories. The first category is called “Fundamentals-Based contagion” this theory emphasizes spillovers that result from the normal interdependence among market economies. This interdependence means that shocks, whether of a global or local nature, can be transmitted across countries because of their real and financial linkages. These forms of co-movements would not normally constitute contagion, but if they occur during a period of crisis, they may be expressed as contagion (Dornbusch, Yung and Claessens, 2000). The second category can be defined as contagion caused by investors sentiment. This involves a financial crisis that is not linked to observed changes in macroeconomic or other fundamentals but is solely the result of the behaviour of investors or other financial agents. Under this definition, contagion arises when a co-movement occurs, even when there are no global shocks and interdependence and fundamentals are not factors. An example is that a crisis in one country may, lead investors to withdraw their investments from many markets without taking into account differences in economic fundamentals (Dornbusch, Yung and Claessens, 2000).

2.3 Connection between financial system and financial crisis

period. This is proved by Bordo, Eichengreen, Klingebiel and Martinez-Peria (2001) on currency crises, Caprio and Klingebiel (1996) on banking crises and Sylla and Wallis (1998) and Standard and Poor’s (2002) on debt crises. In the next section a theoretical discussion will be given on these issues and what implications it has on the different financial systems.

2.3.1 Asymmetric information, Moral Hazard and the Comparative Advantage of Banks In developing a theory of why a financial crisis occurs more in one system than in another, the asymmetric information theory plays a central role. In the discussion of why a financial crisis exists, many authors go back to the fact that this happens because of weak macroeconomic fundamentals and asymmetric information that leads to moral hazard (Stiglitz and Weiss (1981), Mishkin (1996), Boot (2000)). A financial crisis can occur in a system because markets are not perfect and because of this reason resource allocation cannot happen in the most efficient way. Asymmetric information is a very large symptom of these imperfect markets and mentioned as one of the major causes of financial crises the past years (Mishkin, 1996). As explained in paragraph 2.2.2, adverse selection and moral hazard is a consequence of asymmetric information. Because markets are imperfect, intermediaries are inevitable players in partly solving the frictions, since monitoring activities are very important in managing asymmetric information (Mishkin, 1996).

Banks are also having a comparative advantage over markets in controlling for asymmetric information due to relationship banking between the banks and the firms. There are some reasons mentioned in literature that suggests this advantage. The first is concerned with information exchange. There is information that a borrower would reveal to its bank that it would never disseminate to the financial markets (They would not like to make information public that would benefit its competitors). The second reason is that banks have better incentives to invest in information production about borrowers. It is true that information production is costly, however, bank’s incentives are high because they are an enduring lender. Eventually, because of the long-term relationship between bank and borrower, there is a better insight into the situation of the borrower and can generate an improved information flow between bank and borrower (Bhattacharya and Chiesa, 1995 and Boot, 2000).

So, in sum, banks are better at monitoring activities at a lower cost than financial markets and they are more advantageous in information collection activities because they have the ability to engage in long-term customer relationships. So it can be assumed from these theories that it is more likely that financial crises will occur easier in market-based systems than in bank-based systems due to the information asymmetry theory.

2.3.2 Firm’s relationships with banks

The firms’ perspective can also be taken into account when looking at the financial systems situation in a country. Rajan and Zingales proof in their 2001 article, that banks are more cooperating with firms in bank-based countries than in market-based countries. With this they mean that they are more willing to help out financially distressed borrowers. They also suggest that firms are less likely to be bailed out in market-based economies (Rajan and Zingales, 2001).

the competitive market-based system the focus is more on the short-run. It works as follows in a bank-based system: in the short-run, when the company is in trouble, the lender can offer a below market rate and later, when the company is healthy again, recoup its losses by charging above market rates over the long-run when the company can afford it again (Rajan and Zingales, 2001). Because of this suggested help to firms it can be assumed that the probability of a private debt crisis is more likely to happen in market-based systems than in bank-based systems, because bailout and provision of liquidity will happen quicker in bank-based systems.

2.3.3 Globalization, Liberalization, International Capital Flows and Information Asymmetries

The latest trends in the global financial systems cannot be ignored either in this discussion. The past two decades the world has been globalized and liberalized in a major extent. Schinasi, 2006 and Oosterloo, 2007 have claimed that since the 1980s financial globalization and liberalization caused:

1) Major expansion of cross-border financial activities;

2) New interdependencies among market participants, markets and financial systems; 3) Greater international mobility of capital;

4) Greater complexity of financial instruments and trading strategies.

There are major gains because of globalization and liberalization: sources of financing are cheaper, there are new opportunities for risk sharing and there is a more efficient allocation of capital. The downside is that financial stability is harder to maintain, because of greater complexities in financial systems and financial risks and vulnerabilities are harder to assess. Financial disturbances can quickly move from one party to the other and because of international capital flows it goes in large volumes, with a result that a small shock can quickly develop into a serious financial crisis (Oosterloo, 2007).

Liberalization and information asymmetries

financial liberalization. This is proved by Kaminsky and Reinhart (1999) and Demirguc-Kunt and Detragiache (1998). The reason why crises increased simultaneously with globalization and liberalization, can be brought back to the asymmetric information theory. This view is supported by Erturk (2005). According to him liberalization destabilizes the economy because of asymmetric information and herding behaviour. This is because following liberalization, the trend of developing asset prices tend to become fairly predictable, because of the pro-cyclical nature of capital flows, which is stimulating speculation on part of foreign investors. Aziz et al. (2000) claim that a reason of why liberalization leads to more instability is that there is also a decrease in the level of financial intermediation because of the increased importance of stock markets, and as motivated by Mishkin (1996), Boot (2000), and Rajan and Zingales (2001) intermediaries are very important in a system for controlling the frictions that exists due to asymmetric information. So there is an increase of moral hazard and adverse selection in the market-based systems because the role of intermediaries are decreasing in these systems. The theories above provided by Erturk (2005) and Aziz et al. (2000) provide support for the assumption that financial crises occur more often in market-based systems than in bank-based systems.

International capital flows and information asymmetries

a country the exchange rate may also come under pressure, which could cause a devaluation and there will be an inevitable loss of international reserves when the currency is defended. If this loss is not sterilized, it might lead to a credit crunch, increased bankruptcies with as consequence a banking crisis (Kaminsky, 1999). Furthermore, the capital flowing in after liberalization is mostly short-term and specifically, this short-term capital is also reversed easier by investors. When this happens, illiquidity can happen relatively quick due to the quick rise of short-term debt, which improves the probability of a debt crisis. So, all three types of crises can happen when capital reversals occur.

In sum, due to the increased international capital flows when systems are liberalizing, it becomes clear that because of this process, instability and complexity in financial markets is increasing due to asymmetric information that increases uncertainty for international investors. Market-based countries are influenced more by this process since stock markets play a larger role in these countries, where asymmetric information is harder to control for than in a bank-based system. All three types of crises are occurring because of this process. This is proved for currency crises by Aziz et al. (2000), banking crises by Kaminsky (2005) and debt crisis by Erturk (2005).

Countries are not always better off in a market-based system

been moving towards a market-based system in the recent years without a clear consensus that such systems are necessarily better for their economies.

2.3.4 Bank concentration

In the research by Beck, Demirguc-Kunt and Levine (2003) where they study the impact of bank concentration on the likelihood of suffering from banking crises, they found that bank concentration has a stabilizing effect on the economy and that systemic banking crises occur less in countries where bank concentration is higher. It is also proved by Allen and Gale (2000) that a less concentrated banking sector with many small banks is more sensitive to financial crises than a concentrated banking sector with a few large banks. Reasons that crises occur less in concentrated systems are: 1) Large banks can diversify better so that banking systems characterized by a few large banks will be less fragile than banking systems with many small banks. 2) Concentrated banking systems may enhance profits and therefore lower bank fragility. High profits provide a “buffer” against adverse shocks and increase the franchise value of the bank, reducing incentives for bank owners to take excessive risk (Hellmann, Murdoch, and Stiglitz, 2000). 3) Large banks are easier to monitor than many small banks, so that corporate control of banks will be more effective and the risks of contagion less pronounced in a concentrated banking system (Beck et al., 2003). According to Rajan and Zingales (2001) and Allen and Gale (2000) bank-based systems are more likely to be concentrated than market-based systems. This is due to the fact that in market-based systems competition among banks is more promoted (Allen and Gale, 2000). So, from these theories it can be assumed that bank-based systems, due to the high concentration of banks, are more robust against financial crises than market-based systems are.

When the literature in this paragraph is linked together it cannot be avoided to come to the conclusion that asymmetric information plays a key-role in the occurrence of financial crises and that a bank-based economy is better at managing this imperfection. The following assumption then arises:

Chapter 3 Methodology

It is difficult to divide the financial systems around the world into two different categories since no country is the same and all have different histories, institutions and development. The same goes for financial crises. However, in order to understand the relationship between financial crises and types of financial systems, circumstances need to be simplified in order to be able to do a cross-country comparison.

In order to do the analysis I will make a logistic regression analysis to explore the relationship between crisis and system. The data that is collected on financial systems and crises will come from secondary data sources, which is publicly available. The databases that will be primarily used are IMF databases (the World Economic Outlook and International Financial Statistics Database) and the United Nations Statistics database. If possible, the retrieved data will be compared to data from other databases to guarantee reliability of the data. Given the large amount of data that needs to be collected (54 countries, from the period 1980 until 1999) there is an inevitable amount of missing data for some countries on some variables. The research will not go further back then 1980 due to data restraints. The analysis will go until 1999 because during data collection it became clear that there is still a large amount of missing and incomplete data for the years following 1999, especially on the financial system variable, and therefore it has been decided that it was best to keep the years after 1999 out of the analysis. 3.1 Methodology

3.1.1 Financial system

In order to make a categorization whether a country is market-based or bank-based, I will use the data used by Demirguc-Kunt and Levine (1999) from their World Bank research, which they describe as follows:

development is below the mean are classified as bank-based. Thus, this grouping system produces two categories of countries: market-based and bank-based”

Demirguc-Kunt and Levine found in their research that banks and stock markets tend to be larger, more active and more efficient in countries with higher levels of GDP per capita. However, it should be emphasised that in this research the focus is on banks relative to stock markets.

As mentioned before, when calculating whether a country will be classified as a bank-based or a market-based country, the authors are using three measures of financial structure to come to the index number to make the classification, which are size, activity and efficiency.

For the size, they look at the ‘domestic stock capitalization’ relative to the ‘domestic assets of deposit money’. For activity at the ‘total value of stock transactions on the domestic stock exchanges’ relative to the ‘ratio of private credit by deposit money banks’. And to calculate the efficiency for the markets they look at the ‘total value of the stock market as a share of GDP’ and for banks, at the ‘total value traded relative to GDP versus the overhead costs’. A conglomerate index is created after the results from the three different measures were retrieved; this is because the different indices generated slightly different results. The conglomerate index is calculated by removing the means of each series, and then the average is taken of size, activity and efficiency (Demirguc-Kunt and Levine, 1999).

The data to come to this index number for financial system goes until 1999. Given the complex methodology and data that is needed to calculate the index numbers, there was a lack of resources and time to get to the data and the accompanying index numbers for the years following 1999. Therefore it has been decided to make the analysis until 1999.

3.1.2 Financial Crisis

Currency crises

In literature the definition of currency crises is most complicated because different authors use different methods and definitions to date a currency crisis. In my research the data by Bordo, Eichengreen, Klingebiel and Martinez-Peria (2001) will be used to identify currency crises. They consider it to be a currency crisis when they observe “a forced change in parity, abandonment of a pegged exchange rate or an international rescue”. On top of these observations they constructed an exchange market pressure index, which is calculated as the weighted average of exchange rate change, short-term interest rate change and reserve change, relative to the same variables of the centre country which is the United States. When this index exceeds a certain threshold, it is considered to be a currency crisis.

Banking crisis

For banking crises, the data and methods of Caprio and Klingebiel (2003) were used to date and identify banking crises in the different countries. In line with their methodology, the data used in this research will be divided into two different categories: systemic banking crises and non-systemic banking crises. With systemic crises the authors mean severe crises where all or much of the bank capital is exhausted. Non-systemic crises are the smaller banking crises. In this research only systemic banking crises will be considered as banking crises, in order not to complicate this research any further. The dates they attached to the crises are those that are generally accepted by finance experts that are familiar with the countries that are in the sample (Caprio and Klingebiel, 2003).

Debt crisis

For the debt crises, data from Manasse, Roubini and Schimmelpfennig (2003) is used. They define a country to be in a debt crisis when it is classified by Standard and Poor’s to be in a debt default or if it receives a large non-concessional IMF loan defined as “access in excess of 100 percent of quota”. Standard and Poor’s is rating a country in default when a government fails to meet principal or interest payment on external obligations on the due date (Manasse et al., 2003).

3.1.3 Control variables

In this way a more complete picture can be estimated, which variables are of influence on the occurrence of a specific financial crisis. At the same time, the importance of the financial system variable can be compared to the other variables that already proved significance in former research. Each equation will have around nine control variables. All are selected from literature only when they have proved in previous research to be of significant influence in causing a crisis. Due to time and data availability constraints, it has been decided to limit the amount of included variables to around eight.

First, a description will be given of what influence the control variables will have on financial crises, the sources that found these variables significant can be found in appendix 3, 4 and 5. GDP-growth is considered to be an important explanatory variable in the occurrence of all three types of crises. This is because a low GDP-growth indicates that the macroeconomic environment is weakened, which makes the economy more vulnerable to a crisis. This is proved by a great number of authors, among them Kaminsky et al. (1998) on currency crises, Demirguc-Kunt and Detragiache (1998) on banking crises and Manasse et al. (2003) for debt crises. A low GDP-growth is also related to an increased risk of the banking sector, because macroeconomic shocks hurt banks since the share of non-performing loans are increasing, which is another sign for the increased vulnerability of crises.

Interest Rate is an explanatory variable in the currency crises and banking crises models. This variable also gives an indication of the situation a country’s macroeconomic environment. In the case of currency crises, when the interest rate is sharply raised, it is a sign that the authorities are trying to defend the currency. In the case of the possible occurrence of banking crises, high interest rates may lead to banking sector problems because high lending rates result in a larger fraction of non-performing loans. It also serves as a proxy for financial liberalization. This is because the liberalization process tends to lead to high rates, this was proved in the research of Demirguc-Kunt and Detragiache (1998). This financial liberalization process may lead to increased banking sector fragility because of the increased opportunities for risk taking. This is proved by Kaminsky and Reinhart (1996).

a sign of an instable macroeconomic environment and are increasing the probability of financial crises.

International Reserves is an important explanatory variable in currency crises. When the currency gets under pressure, authorities will try to defend the currency by using international reserves in an attempt to maintain parity. When the international reserves get depleted, authorities are no longer able to defend the currency and has to devalue the currency and this increases the chance of the occurrence of currency crises. So a decrease in international reserves increases probability of currency crises

The real exchange rate is an explanatory variable in all three types of crises. For currency crises a sharp depreciation of the exchange rate increases the probability of a crisis, the sharp depreciation of the currency is the consequence of the speculative attack. The depreciation increases the probability of a banking crisis because when depositors are expecting a devaluation of the currency, they may withdraw their bank deposits and convert them into foreign currency deposits abroad and leaving the domestic banks illiquid (Demirguc-Kunt and Detragiache, 1998). It has an effect on the occurrence of a debt crisis because when the currency depreciates, it becomes more difficult for countries to repay their external debt, because in real terms their debt increases.

The M2/Reserves variable is much used in currency crises and in the banking crises models. It measures the degree of financial vulnerability of a country and is taken as a measure of liquidity. The ratio indicates how good the central bank will be able to handle the substitution of local currency to foreign currency. The larger the ratio, the lower the ability of a country to anticipate a speculative attack on the currency, which increases the probability of a currency crisis. The ratio reflects the probability of a banking crisis. When agents fear devaluation, they may substitute local currency for foreign currency which may cause banking sector problems as well.

Domestic credit to private sector is included in all three models on financial crises and is an indicator of the macroeconomic situation. The increase of this ratio points to an expansionary monetary policy, that authorities are using to stimulate the economy and try to protect it from an economic downturn or crisis. When this ratio is increasing, the stock of money in the economy increases. According to Demirguc-Kunt and Detragiache (1998) this variable also serves as a proxy for the progress of financial liberalization and this can also lead to an increased vulnerability to crises.

a sign of expansionary fiscal and credit policies in the sense that it shows a higher demand of traded goods, which shows weak exports and an excessive import growth. This could lead to deterioration in the current account, and a greater sensitivity to financial crises.

Presence of an explicit deposit insurance scheme is a variable used in the banking crisis model and it tends to increase the probability of systemic banking sector problems. It serves as a proxy for increased moral hazard because when agents know that the deposits are insured against bank insolvencies, they are more likely to take an excessive risk than in the absence of such a scheme, because they know the losses will compensated.

The External Debt to GDP ratio is used in the debt crisis model. High levels of external debt are increasing the probability of a debt crisis. It is a measure of solvency which shows the ability of a country to pay its external debt. The higher this level, the larger the chance that a country is not able to repay its debt and has to default.

Short-term debt to international reserves is also used in the debt crisis model. This variable shows the liquidity of a country. High levels of short-term debt increases the probability of a debt crisis since the external liquidity is decreasing, which increases the chance that the country has to default. This variable is also a proxy for large capital reversals from a country, since short-term funds are easier to withdraw for investors. Large capital reversals are a sign that investors fear that economic circumstances in a country are deteriorating.

3.2 Research Design 3.2.1 Sample

A sample of 54 countries will be taken to perform the analysis. In the sample there will be a diverse range of countries, both developed as developing, from different continents. A list of countries included in the research can be found in appendix 1 and in appendix 2 there will be a list of crises episodes. The analysis will cover a period 19 years, from 1980 until 1999. 3.2.2 Dependent and Independent variables

In this research, financial crisis is considered to be the dependent variable and financial system an independent variable. The control variables are included as the other independent variables in the equation, which are also influencing the probability of a crisis.

3.2.3 Logistic Regression Analysis

non-metric dependent variable, which is the case in this research, where the dependent variable is categorized in two groups, the occurrence of financial crisis (1) or no occurrence of financial crisis (0). An advantage of the use of a logistic regression model is that it is quite robust in handling the fact that the number of observations in the no-crisis group will be much larger than the number of observations in the crisis group, since financial crisis is a relatively rare event. It also is robust in not meeting the statistical assumptions that are required to be met with other methods, like multivariate normality and the equal variance among groups.

In the analysis it will become clear within the groups whether the financial system variable will explain a significant part of the equation of when a specific crisis occurs or not. The question “does financial system influences the occurrence of financial crisis?” can then be answered. It will also become clear how the financial system variable explains the occurrence of crises in relation to the other independent variables. Between groups (crisis or no crisis) it will become clear in which direction the financial system variable will go, whether there is an increased probability of entering into a crisis when a country is market-based or bank-based. Given the fact that I want to look at the different types of crises, the following three equations exists which are specified below:

The currency crisis model

(1) P (Currency Crisis) = eb0 + b1X + b2X + b3X + b4X + b5X + b6X + b7X + b8X P (No currency crisis) + b9X

Where:

e = prediction error; B0 = intercept;

B1X = Financial system; B2X = GDP growth;

B3X = Domestic credit to the private sector; B4X = Interest rates;

B5X = Inflation;

B6X = International reserves; B7X = Real exchange rate;

The ability of the independent variables to improve the prediction of the model is not only related to its correlation with the dependent variable, but is also to the correlations of the other independent variables with each other. For that reason it is important to look at the correlations among independent variables. To maximize the prediction of the model, low correlation among independent variables is the goal, since when there is a high correlation, the unique variance explained by the independent variable decreases. In table 3.1 you can see the correlations among independent variables.

Table 3.1 Correlation matrix currency crisis variables

FS GDP DCre IR Inf IRes RXR M2/res TrBa

FS 1 GDP 0.036 1 DCre 0.438 -0.049 1 IR -0.012 -0.054 -0.035 1 Inf 0.026 -0.181 -0.080 0.264 1 IRes 0.205 -0.046 0.577 -0.014 -0.032 1 RXR -0.123 -0.039 -0.150 -0.007 -0.008 -0.072 1 M2/res 0.104 -0.022 0.119 -0.005 -0.016 0.092 -0.029 1 TrBa -0.183 -0.060 0.053 0.012 0.018 0.242 0.004 -0.075 1

FS (financial system); GDP (GDP-growth); DCre (domestic credit); IR (interest rate) Inf (inflation); Ires (international reserves); RXR (real exchange rate); M2/res (M2 to reserves); TrBa (trade balance)

The table shows no extremely high correlations among the variables. The two variables that show the most correlation with each other is the domestic credit variable and the international reserves. They correlate with a value of 0.577. In former research, authors also put these variables together in an equation and considering the importance proved in literature of both variables in predicting currency crises, both variables will be kept in the model. All other variables correlate less than (-)0.5 and in general correlations are very low, showing a good basis for these variables to be included in the analysis.

The banking crisis model

Where: e = Prediction error; B0 = Intercept; B1X = Financial System; B2X = GDP growth; B3X = Interest Rates; B4X = Inflation;

B5X = Deposit insurance scheme;

B6X = M2/Reserves (External vulnerability); B7X = Domestic credit to the private sector; B8X = Trade Balance;

B9X = Real Exchange Rate.

In table 3.2 the correlation matrix of the banking crisis independent variables is shown. Table 3.2 correlation matrix banking crisis variables

FS GDP IR Inf Dins M2/res DCre TrBa RXR

FS ₁ GDP _{0.036 1} IR _{-0.012 -0.054 1} Inf _{0.026 -0.181 0.264 1} Dins _{0.099 -0.157 0.038 -0.050 1} M2/res _{0.104 -0.022 -0.005 -0.016 0.090 1} DCre _{0.439 -0.049 -0.035 -0.080 0.217 0.119 1} TrBa _{-0.183 -0.060 0.012 0.018 0.037 -0.075 0.053 1} RXR _{-0.123 -0.039 -0.007 -0.008 -0.118 -0.029 -0.150 0.004 1}

FS (financial system); GDP (GDP-growth); IR (interest rate); Inf (inflation); Dins (deposit insurance scheme); M2/res (M2 to reserves); DCre (domestic credit); TrBa (trade balance); RXR (real exchange rate)

domestic credit variable is a proxy for the progress of financial liberalization and when this process occurs, a country is moving more towards a market-based system. The fact that this process goes parallel, may explain the correlation among these variables.

The debt crisis model

(3) P (Debt Crisis) = e b0 + b1X + b2X + b3X + b4X + b5X + b6X + b7X + b8X P (No debt crisis)

Where:

e = prediction error; B0 = Intercept;

B1X = Financial System;

B2X = External debt relative to GDP;

B3X = Short-term debt relative to international reserves; B4X = GDP growth;

B5X = Trade imbalances;

B6X = Domestic credit to the private sector; B7X = Real exchange rate;

B8X = Inflation.

In table 3.3 the correlation matrix on the debt crisis explanatory variables is shown. Table 3.3 Correlation matrix debt crisis variables

FS EDgdp STDres GDP TrBa DCre RXR Inf

FS ₁ EDgdp _{-0.145 1} STDres _{-0.119 0.548 1} GDP 0.040 -0.056 -0.125 1 TrBa _{-0.180 -0.011 0.005 -0.052 1} DCre _{0.439 -0.333 -0.269 -0.037 0.057 1} RXR _{-0.121 0.165 0.034 -0.037 0.004 -0.147 1} Inf _{0.026 0.108 0.109 -0.178 0.019 -0.079 -0.008 1}

Chapter 4 Results and Analysis

This chapter will discuss and analyze the results from the logistic regression analysis. The binary analysis is performed on 1014 observations on all crises types. When making the analysis a robust variance estimator was used (Huber/White), so that country specific variances are allowed. This specific method is also used by other researchers in the field of financial crises like Manasse et al. (2003) and Eichengreen et al. (1995). The chosen significance level will be 0.1 since it is widely used in the financial crises literature. The complete outputs of the results will be shown in appendix 7.

4.1 Currency crisis

Table 4.1 summarizes the results for the logistic regression analysis on currency crisis. When examining the direction of the coefficients, it can be seen that all the coefficients have the expected direction in the model. For an explanation why the independent variables influence the probability of a crisis, I refer back to paragraph 3.1.3 where the relationship between the explanatory variables and the occurrence of a crisis is explained.

The most important variable to examine here, financial system, has a positive sign, which implies that an increase in the financial system variable will result in an increase in the probability of the occurrence of a currency crisis. The increase in the value of the financial system variables stands for the movement towards a market-based system. This is because the higher the value for financial system, the more market-based a country is, based on the index numbers of Demirguc-Kunt and Levine.

Table 4.1 Logistic regression results Currency crisis

Variable Coefficient Standard error Z-statistic Prob.

Constant -0.997 0.204 -4.841 0.000

Financial System ** 0.312 0.125 2.492 0.013

GDP growth * -0.153 0.026 -5.898 0.000

Domestic credit ** -0.007 0.003 -2.126 0.033

Interest Rate 2.60E-07 1.68E-07 1.552 0.121

Inflation 0.0003 0.0003 1.069 0.285

International reserves -1.12E-11 1.44E-11 -0.0777 0.437

Exchange rate -4.42E-06 4.48E-05 -0.099 0.921

Trade Balance 6.97E-12 4.94E-12 1.409 0.159 Observations Dep=0 880 McFadden R2 0.0766

Observations Dep=1 134 Total Observations 1014

* Significant at 0.01 level ** Significant at 0.05 level

When further examining the model when looking at the z-statistic with the accompanying p-values it can be seen that only a few variables show significance in this model and that a large amount of explanatory variables, that were included in the model because they showed significance in other research are not significant in this research.

low overall model-fit also appears from the classification table in table 4.2 (the complete output of the table can be found back in appendix 7). The table shows to which extent the estimated equation model improves the prediction of the constant probability model (the model when only the intercept term is included). It also shows the correct and incorrect classification of the crisis variable in the model. From the table it can be seen that the estimated equation on currency crisis has a prediction gain of only 0.75 percent on the constant probability model. This is not a very large gain, which implies that the model-fit is not that high. However, it should be noted that the predicted model does predict 86.88 percent of the model correctly, which is a good score.

A reason that the overall model-fit of the model is not that high can be because of the great amount of factors that are influencing the occurrence of currency crises. In existing models on currency crisis, researchers include a large amount of different explanatory variables to estimate a model. Even after the inclusion of a wide range of indicators, the prediction of a crisis remains very difficult. This can be seen from the results of the research by Frankel and Rose (1996) and Eichengreen et al. (1995). These authors found a McFadden R2 of respectively 20 and 15 percent. Considering the fact that this equation is a simplistic version of variables found significant by a range of authors, this might be a reason why this model predicts a lower amount of the dependent variable than the research by Frankel and Rose (1996) and Eichengreen et al. (1995).

It should also be emphasized that it was not the goal of this research to predict a good-fit model for any type of financial crisis. The variables were mostly used to see how these variables work in relation to the financial system variable.

Table 4.2 Classification table for currency crisis regression Estimated equation

Total Constant probability Total

Total observations 1014 1014 Correctly predicted 881 880 % Correct 86.88 86.79 % Incorrect 13.12 13.21 Total gain * 0.10 Percent gain ** 0.75

The low model-fit does not keep other research on currency crisis from drawing conclusions from it, so this model will also be used to interpret the results and draw conclusions from, although doing it with caution.

What can be concluded from the results of this model is that the financial system variable does play a significant role in the prediction of currency crisis, since it is the second explanatory variable that enters the equation at a 0.05 significance level. This is the first implication that can be made from the model. The second implication is that this model also confirms that as when a country becomes more market-based, the probability of a currency crisis increases. This can be seen from the positive sign of the coefficient of the financial system variable.

4.2 Banking Crisis

The results from the logistic analysis on banking crisis are outlined in table 4.3. When examining the direction of the sign of the coefficients it can be seen that all coefficients show the expected direction. The explanation of the relationship between independent variables and crisis can be found back in paragraph 3.1.3. Again, the coefficient of financial system shows a positive sign, which implies, just as it was the case with currency crises, that the probability of a banking crisis increases when a country is more market-based. The interpretation for the direction of the deposit insurance variable deserves some extra attention. This explanatory variable is a dummy variable where 0 means that there is an implicit deposit insurance scheme or no scheme and 1 means that there is an explicit deposit insurance scheme, which serves as a proxy for moral hazard. The positive sign of the coefficient implies that when there is the presence of an explicit deposit insurance scheme, that there is an increased probability of the occurrence of a banking crisis. This is in line with literature. In a later stage of the analysis is topic will be discussed more thoroughly.

Table 4.3 Logistic regression results Banking Crisis

Variable Coefficient Standard error Z-statistic Prob.

Constant -1.139 0.217 -5.227 0.000

Financial System *** 0.230 0.124 1.859 0.063

GDP growth * -0.112 0.024 -4.645 0.000

Interest Rate 3.73E-06 2.54E-06 1.469 0.142

Inflation ** 0.001 0.0004 2.398 0.017

M2/reserves* -0.001 0.0003 -3.612 0.000

Domestic credit * -0.011 0.003 -3.630 0.000

Trade balance * 2.91E-11 6.75E-12 4.313 0.000

Exchange Rate*** 5.69E-05 3.10E-05 1.836 0.067

Observations Dep=0 841 McFadden R2 0.1160 Observations Dep=1 173

Total Observations 1014

* Significant at 0.01 level ** Significant at 0.05 level *** Significant at 0.1 level

When further examining the model and looking at the z-statistic and p-values, the importance of the independent variables can be found. The low GDP-growth is again the most important explanatory variable, followed by trade balance and the presence of an explicit deposit insurance scheme, and are significant at a 0.01 level. Since the presence of an explicit deposit insurance scheme is a proxy for moral hazard this finding is in line with the theory and motivation by different authors that asymmetric information and moral hazard is a leading factor in the existence of financial crisis. The other explanatory variables that are significant at a 0.01 level are domestic credit and M2/reserves. Inflation is significant at a 0.05 level. The financial system variable is the variable that enters the equation as the seventh most important explanatory variable and is significant at a 0.1 level. This variable has a probability of 0.063 with an accompanying z-statistic of 1.859. This implies that the financial system variable also plays a significant role in the occurrence of banking crises. The last explanatory variable that enters the model significantly at a 0.1 level is exchange rate. In this equation there is one variable that is not significant at a 0.1 level, which is the interest rate. However, it just falls out of the threshold value of 0.1 with a value of 0.142.

research since authors are not including these statistics in their analysis. This goes for Demirguc-Kunt and Detragiache (1998) and Hardy and Pazarbasioglu (1999). In both analyses, no levels of R2 are mentioned. Demirguc-Kunt and Detragiache do mention the overall classification accuracy. For their 1998 research this is 84 percent. The prediction accuracy in this model is 83.43 percent, as can be seen in table 4.4, so in that sense it is pretty much comparable. It should be assumed that the R2 in the above mentioned researches is higher than the one in this research since they take into account a wider variety of variables, and their research is specifically designed to create a warning system that predicts the occurrence of banking crises.

Table 4.4 Classification table for banking crisis regression Estimated equation

Total Constant probability Total

Total observations 1014 1014 Correctly predicted 846 841 % Correct 83.43 82.94 % Incorrect 16.57 17.06 Total gain * 0.49 Percent gain ** 2.89

* Change in ‘% correct’ from default (constant probability) specification ** Percent of incorrect (default) prediction corrected by specification

A conclusion that can be drawn from the analysis of the banking crises model is that most explanatory variables enter the model significantly. Most are significant at a 0.01 level. The financial system variable enters the equation as the seventh most important variable in explaining the occurrence of banking crises. This variable is significant at a 0.1 level. The conclusion that can be drawn is that financial system plays a significant role in the probability of the occurrence of banking crises. The second conclusion that can be drawn is that the probability of banking crises increases when a country is more market-based, since the coefficient of financial system shows a positive sign.

4.3 Debt Crisis