BANKING CRISES; DETERMINANTS AND CRISES‘ IMPACT ON FISCAL COST AND ECONOMIC OUTPUT
PURPOSE OF THE STUDY
This thesis‘ first empirical part studies determinants associated with the emergence of systemic banking crises. Low GDP growth, high inflation, low creditor rights, low GDP per capita and financial reforms are hypothesized to increase the probability of a banking crisis. I evaluate the determinants‘ value in explaining 80 banking crisis around the globe by employing a multivariate logit model, exactly as in Demirgüc-Kunt and Detragiache (1997). This thesis‘ second empirical part evaluates banking crises‘ effects on fiscal cost and economic output by regressing each of these two dependent variables separately on a number of new or updated explanatory variables in an ordinary least squares (OLS) regression, exactly as in Demirgüc-Kunt and Detragiache (1997). This study provides new insights into the determinants and effects of banking crises with new and larger datasets and time frames as well as better and further developed variables.
DATA
From the database by Laeven and Valencia (2008) I obtain the core data on banking crises, consisting of 80 banking crises in a sample of 120 countries during the period of 1980 to 2005. From IMF‘s IFS database and world economic outlook (WEO) I obtain the data for the remaining variables. The data finally consists of 2755 country-year observations for the empirical part I on the determinants of banking crises. For the empirical part II on banking crises‘ effects, I obtain the complete data on 42 banking crises and their explanatory variables, during the period of 1970 to 2007, from the database by Laeven and Valencia (2008).
RESULTS
The findings of the study on determinants of banking crises show that low GDP growth, highly developed institutional and regulatory environments and high GDP per capita increase the probability of a banking crisis statistically significantly. This thesis updated the study of Demirgüc-Kunt and Detragiache (1997) and turned their results around. The findings of the second empirical part on banking crises effects show that a parallel currency crisis and explicit deposit insurances increase the economic output loss. French and Socialist law countries experience lower output losses but more frequent crises. As in Honohan and Klingebiel (2003), I also observe that government interventions result in higher fiscal costs. Additionally, higher fiscal expenditure was not found to reduce output losses, exactly as Claessens, Klingebiel, and Laeven (2003) conclude. However, foregone tax revenues due to output losses might also impact fiscal cost, as suggested by Reinhart and Rogoff (2008). Thus, any conclusion about the relationship between fiscal cost and output loss is difficult due to a possible endogeneity.
KEYWORDS
Banking crisis, crises, determinants, economic output loss, fiscal cost,
Table of content
1.1 BACKGROUND AND MOTIVATION ................................................................................
1.2 EMPIRICAL PART I: DETERMINANTS OF BANKING CRISES .....................................
1.3EMPIRICAL PART II: BANKING CRISES‘ EFFECTS.......................................................
1.4 CASE STUDY .......................................................................................................................
1.5STRUCTURE OF THE STUDY ............................................................................................
2 LITERATURE REVIEW ................................................................................................
2.1FINANCIAL SYSTEMS AND INTERMEDIATION ...........................................................
2.2 UNSTABLE BANKING ........................................................................................................
2.3THEORY OF LIQUIDITY, FINANCIAL FRAGILITY AND CONTAGION ....................
2.3.1Cash-in-the-market pricing and financial fragility .......................................................
2.3.2Contagion .....................................................................................................................
2.4RETURNING PERIODS OF STABILITY AND INSTABILITY .......................................
2.5 BANKING CRISES .............................................................................................................
2.6DETERMINANTS OF BANKING CRISES........................................................................
2.6.1Financial liberalization and crises ...............................................................................
2.6.2 International shocks, exchange rate regime and crises ................................................
2.6.3 Bank ownership, structure and crises ...........................................................................
2.6.4 The role of institutions ..................................................................................................
2.6.5 The political system and crises .....................................................................................
2.7 EFFECTS AND IMPLICATIONS OF BANKING CRISES................................................
2.7.1The credit crunch hypothesis ........................................................................................
2.7.2Bank relationship and information loss hypothesis.......................................................
2.7.3Banking crises and economical costs ............................................................................
2.7.4Banking crises and fiscal costs .....................................................................................
3 EMPIRICAL PART .......................................................................................................
3.1 EMPIRICAL PART I: DETERMINANTS OF BANKING CRISES ...................................
3.1.1Banking crises determinants and hypotheses ................................................................
3.1.2Methodology: multivariate logit approach ...................................................................
3.1.3Data ..............................................................................................................................
3.1.4Empirical findings of determinants of banking crises ...................................................
3.2EMPIRICAL PART II: BANKING CRISES‘ EFFECTS.....................................................
3.2.1Determinants of banking crises’ costs and hypotheses .................................................
3.2.2Methodology: ordinary least squares regression ..........................................................
3.2.3Data ..............................................................................................................................
3.2.4Empirical findings of determinants of banking crises’ costs .........................................
3.2.5Time to next presidential elections................................................................................
4 CASE STUDY: THE U.S. CREDIT AND BANKING CRISIS OF 2007 ..................
4.1THE DEVELOPMENT OF THE U.S. BANKING AND CREDIT CRISIS OF 2007 .........
4.2THE ORIGINS OF THE U.S. BANKING AND CREDIT CRISIS OF 2007.......................
4.2.1Securitization ................................................................................................................
4.2.2 Institutional environment ..............................................................................................
4.2.3 Global account imbalances ..........................................................................................
4.2.4 Increasing financial fragility. .......................................................................................
4.2.5 Leveraged buy out ........................................................................................................
4.2.6The returning of hubris .................................................................................................
4.3 CASE STUDY CONCLUSION ...........................................................................................
CONCLUSIONS .............................................................................................................
REFERENCES .......................................................................................................................
APPENDICES ......................................................................................................................
List of tables
Table 1: Summary of the determinants of banking crises
Table 2: Summary of hypothesized and empirical findings of determinants‘ relation to banking crises
Table 3: Descriptive statistics of banking crises‘ determinants
Table 4: Correlation matrix of banking crises‘ determinants
Table 5: Determinants of banking crises results—panel excluding years after the first banking crisis
Table 6: Determinants of banking crises results—panel excluding years while the crisis is on-going
Table 7: Determinants of banking crises costs
Table 8: Descriptive statistics of banking crises‘ initial conditions
Table 9: Correlation matrix of banking crises‘ effects
Table 10: Banking crises effects‘ regression results for output loss and each variable separately
Table 11: Banking crises effects‘ regression results for fiscal cost and each variable separately
Table 12: Banking crises effects‘ regression results for output loss
Table 13: Banking crises effects‘ robustness test regression results for output loss
Table 14: Banking crises effects‘ regression results for fiscal cost
Table 15: Banking crises effects‘ robustness test regression results for fiscal cost
Table 16: Banking crises effects‘ regression results for robustness test II
Table 17: Banking crises effects‘ correlation matrix for robustness test II
Table 18: Time-to-election effects‘ OLS regression results for fiscal cost
CHAPTER ONE
1 INTRODUCTION
1.1 BACKGROUND AND MOTIVATION
The origins and effects of banking crises are currently, in the light of the U.S. credit and banking crises that started in 2007, often discussed. The public and media search for explanations but even more for scapegoats. However, during these discussions very fundamental aspects seem to be forgotten. The public‘s surprise about the possibility of a banking crisis shows that most have not understood the current financial system and its history. Therefore, I argue that understanding banking crises requires first to understand financial systems, banking and the history of financial crises.
Levine (1997) argues that banks satisfy random liquidity needs of lenders better (at lower risks) than if these lenders would invest their funds directly in the financial markets. However, in order to credibly commit to repay depositors, banks must choose a fragile bank capital structure; argue Diamond and Rajan (2001). Thus, the risk of a banking crisis may be a necessary disciplinary device in an imperfect market.
Liquidity is only provided if expected returns during times of low market liquidity are high in order to compensate providers for their large opportunity cost of holding idle capital for a long time. Low market liquidity however can trigger cash-in-the-market pricing which may lead to fire-sales, contagion and self-fulfilling prophecies of a banking crisis (Kane, 1989; Demirgüc-Kunt and Detragiache, 2002).
Research (e.g. Lindgren et al., 1996; Dooley and Frankel, 2003; and Collyns and Kincaid, 2003) identified various underlying macroeconomic origins of past banking crises such as inflation, cyclical output downturns, term of trade deterioration, exchange rate crashes, and currency as well as asset and real estate devaluations. Additionally, financial liberalization (Demirgüc-Kunt and Detragiache, 1998), flexible exchange rate regimes (Peria, 2003), public bank ownership (Caprio and Levine, 2001), a weak institutional environment (Demirgüç-Kunt and Detragiache, 1998) and an explicit deposit insurance (Demirgüç-Kunt and Detragiache, 2002) were all found to increase the probability of a banking crisis.
Once a banking crisis is ongoing, the credit crunch hypothesis predicts that decreased bank credits to firms decrease investments and expenditure, thus resulting in decreased economic output and demand (Eichengreen and Rose, 1998; Kaminsky and Reinhart, 1999; Demirgüç-Kunt and Detragiache., 2005). While fighting a banking crisis, governments face a trade-off between fiscal and economical costs. Higher fiscal spending on government interventions is expected to decrease the economical cost of crisis (Laeven and Valencia, 2008). However, earlier findings suggest that generous support to the banking system does not reduce the output cost of banking crises (Claessens, Klingebiel and Laeven, 2003).
Therefore, an important question to policy makers and economists is how to decrease the negative effects of a banking crisis most efficiently and effectively as well as how to prevent a banking crisis totally. Also highlighted should be the fact that ―econometric analyses of systemic banking crises are a new field‖ (Demirgüc-Kunt, Detragiache, 2005) and only limited research exists. Hence, more empirical research about banking crises, their origins and relation to economic and fiscal cost is needed.
I contribute to the research on banking crises by providing new insights into the determinants and effects of banking crises with new and larger datasets and time frames as well as better and further developed variables. Specifically, the dataset by Laeven and Valencia (2008) gives the opportunity to research banking crises‘ effects by taking into account government interventions. To my knowledge, earlier studies did not address broad government interventions and policies sufficiently, while researching banking crises‘ effects. Additionally, by taking a larger time frame, I will turn earlier results of Demirgüc-Kunt and Detragiache (1997) about determinants of banking crises around. Therefore, this thesis adds in-depth insight and new results to the research of banking crises‘ determinants and effects.
1.2 EMPIRICAL PART I: DETERMINANTS OF BANKING CRISES
The first empirical part studies the factors associated with the emergence of systemic banking crises. I hypothesize GDP growth, real interest rate, inflation, M2 to reserves, creditor rights, GDP per capita and financial reforms to have an impact on the probability of a banking crisis.
I evaluate the chosen determinants‘ value in explaining 80 banking crisis around the globe by employing a multivariate logit model, exactly as in Demirgüc-Kunt and Detragiache (1997). From the database by Laeven and Valencia (2008) I obtain the core data on banking crises, consisting of 80 banking crises within a sample of 120 countries during the period of 1980 to 2005. From IMF‘s IFS database and world economic outlook (WEO) I obtain the data for the remaining variables. The data finally consists of 2755 country-year observations.
Following variables are hypothesized to influence the possibility of a banking crisis. Low GDP growth is hypothesized to reflect adverse macroeconomic shocks that hurt banks through higher rates of non-performing loans (Demirgüc-Kunt and Detragiache, 1997). High short-term real interest rates affect banks negatively if they are not able to quickly pass the higher interests on to their customers (Demirgüc-Kunt and Detragiache, 1997). Real interest rates might proxy financial liberalization (Galbis, 1993) which is then again found to increase the likelihood of a banking crisis (Demirgüc-Kunt and Detragiache, 1997). The variable measuring inflation is expected to reflect macroeconomic mismanagement as high inflation is hypothesized to be a sign for a price bubble (Reinhart and Rogoff, 2008). The ratio of M2 to foreign exchange reserves of the central bank measures sudden capital in- or outflows and a country‘s vulnerability to balance-of-payment problems (Calvo, 1996). Thus, a high ratio of M2 to foreign exchange reserves is hypothesized to increase the probability of a banking crisis. The variable measuring creditor rights in each country is hypothesized to reflect the economy wide impact of a strong legal system and thus lower likelihood of a banking crisis (LLSV, 1999; Demirgüc-Kunt and Detragiache, 1997). The dummy measuring the introduction of a financial reform equals 1 for five years if the country was undergoing a large financial reform and thus has a larger risk of a banking crisis (Demirgüc-Kunt and Detragiache, 1998). Finally, I also measure the time that passed by since the last crisis as I expect that the probability of a crisis increases over time as new problems after the last crisis accumulate anew (Demirgüc-Kunt and Detragiache, 1997).
As a summary of the results, low GDP growth, highly developed institutional and regulatory environments and high GDP per capita were shown to increase the probability of a banking crisis statistically significantly. Therefore, this study updated the study of Demirgüc-Kunt and Detragiache (1997) and turned their results around. Their study‘s time frame is from 1980 to 1994 for which they find countries with weak institutional environments, low GDP per capita and low GDP growth to inhibit the largest risks of banking crises. The fact that the study by Demirgüc-Kunt and Detragiache (1997) has a different time frame is a limitation to comparing results. However, at the same time this is also an update of the results for most variables that are defined exactly as in Demirgüc-Kunt and Detragiache (1997). Thus, by
taking a larger time frame, this study turned earlier results of Demirgüc-Kunt and Detragiache (1997) around. In addition, according to my knowledge, no previous study has employed such a large time frame while studying determinants of banking crises. Therefore, this thesis adds in-depth insight and new results to the research of banking crises‘ determinants. Especially, we showed that most variables are statistically significantly correlated to at least one other variable. This raised questions about the endogeneity of these variables and whether the true determinants of banking crises are yet found. Future research is advised to focus on developing better determinants of banking crises.
1.3 EMPIRICAL PART II: BANKING CRISES’ EFFECTS
The second empirical part evaluates banking crises‘ effect on fiscal cost and economic output by regressing each of the two dependent variables separately on a number of explanatory variables in an ordinary least squares (OLS) regression. This event study methodology is performed exactly as in Demirgüc-Kunt and Detragiache (1997). I argue that financial and structural variables as well as government interventions explain fiscal costs and economic output loss of banking crises. From the database by Laeven and Valencia (2008) I obtain the complete data on 42 banking crises and their explanatory variables during the period of 1970 to 2007.
In summary, I report a parallel currency crisis and an explicit deposit insurance to increase the economic output loss. Countries with French and Socialist law origins experience lower output losses but more frequent crises. As in Honohan and Klingebiel (2003), I observe that government interventions result in higher fiscal costs. Also, higher fiscal expenditure was not found to reduce output losses, exactly as in Claessens, Klingebiel, and Laeven (2003). Reinhart and Rogoff (2008) suggest that a possible explanation could be found in output losses which result in foregone tax revenues and ultimately in higher fiscal costs. Thus, any conclusion about the relationship between fiscal cost and output losses is difficult due to a possible endogeneity. Finally, near-by elections are shown to decrease fiscal costs statistically significantly.
This study has limitations as financial crises are rare in number, thus this empirical part II is based on relatively few data points. In addition, every crisis might be different from previous ones, thus coefficients derived from in-sample estimation are of limited use out-of-sample (Demirgüc-Kunt and Detragiache, 2002). Also, information on intervention policies is rare and each intervention might depend on the skills of the policy makers. On the other hand, according to my knowledge, no previous research has used such a rich set of variables (including government intervention variables) while researching banking crises‘ effects.
1.4 CASE STUDY
Banking crises‘ determinants of my empirical study I would not have accurately detected the U.S. credit and banking crisis that started in 2007. Exactly as Reinhart and Rogoff (2008), I suggest therefore that every banking crisis might be different. Thus, I investigate the U.S. banking crisis that started in 2007, its development and origins, in a case study. I identify securitization, the institutional environment, account imbalances and hubris as major origins of this crisis. This case study adds to the literature a valuable and comprehensive review of the current literature on the U.S. credit and banking crisis‘ most important reasons and origins.
1.5 STRUCTURE OF THE STUDY
The remaining of the study is structured as follows. Section 2 reviews the literature on financial systems, banking, liquidity and banking crises. Thereafter, Section 3 inhibits two empirical studies of banking crises‘ determinants and costs. Section 4 looks at the U.S. banking crises of 2007 and Section 5 concludes this study with a summary of the findings.
