Macroeconomic, financial and institutional determinants of banking crisis, the money market pressure index approach

MMP: Money Market Pressure WGI: World Governance Indicator WB: World Bank IMF: International Monetary Fund IFS: International Financial Statistics ICRG: International Country Risks Guide

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

MACROECONOMIC, FINANCIAL AND INSTITUTIONAL

DETERMINANTS OF BANKING CRISIS:

THE MONEY MARKET PRESSURE INDEX APPROACH

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

Also, I would like to thank Dr Truong Dang Thuy for his guidance and advice ineconometric techniques, Dr Pham Khanh Nam for his encouragement and valuableadvice in the starting phase of my thesis research design.

My gratefulness is also extended to all of my lecturers and staffs of the Netherlands Program for their assistance during my first days in this programme.Besides, I would love to thank my parents and my families for their ceaselessencouragement and support during my study period Moreover, my special thanks to

Vietnam-my C.E.O – Mr Nguyen Huu Tram, who understands and gives me approval for Vietnam-mylong personal leave to finalize my thesis on time Without them, I would not haveopportunities and incentives to have my thesis finished

Finally, I would like to thank all my friends and other people who have had anyhelp and support for my thesis but are not above-mentioned

The thesis estimates a logit regression model by fixed effect with a combination ofsome macroeconomic and financial indicators from the work of Hagen and Ho(2007) and Worldwide Governance Indicators (WGI) from the updated database ofKaufmann (2013) as explanatory variables for binary dependent variable bankingcrises generated from the approach of money market pressure index (Hagen and Ho,2007) The monthly panel dataset, which is available in full range and easy ofapproach from International Financial Statistics CD-ROM (2011), of 18 countriesfrom Latin America and Asian over the scope of 2001 – 2010is applied Somespecific lag lengths of indicators are also applied according to the suggestion of

“flexibility in forecast horizon” of Drehmann et al (2011)

The crisis phenomenon of banking system seems to be well-described in light of thepresent of depreciation, former year crisis, high real interest rate in prior of 36months, growth of credit to GDP in prior 12 months Moreover, impact of inflationseems to support the school of thought that it is negative effect to crisis.Simultaneously, growth rate of bank deposits to GDP is likely useful to preventbanking systems from profitability risks exposure that leads to banking crisisprobability However, unfortunately, the indicators of growth of monetary base andgrowth of M2 to reserves give incorrect expected sign and negligible effect onbanking crisis Furthermore, the included institutional variables from WGI giveinsignificant statistic meaning Hence, another set of institutional indicators such asthat from International Country Risk Guide (ICRG) should be considered in futureanalysis to test for the relationship between Government health and banking crisisprobability

Despite, on one hand, there should be a more adequate research to be examined inthe future, this thesis attempts to contribute so-called new updates information onthe would-be banking crisis determinants Nevertheless, on the other hand, there islikely no proper explanation on the tranquil periods of banking system Hence, it is

suggested that thereshould be some assessment ofsuch time of banking system, which over a long time has beenneglected (Kauko, 2014).

Key words: banking crisis, tranquiltime, determinants, institutional indicators,

fixed effect logitregression.

TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION 1

1.1 Problem statement 1

1.2 Research objective 3

1.3 Research question 3

1.4 Structure of the thesis 3

CHAPTER 2: LITERATURE REVIEW 5

2.1 Defining banking crisis 5

2.2 Trends of banking crises researchtogether with crises mechanism 7

2.2.1 The first trend 8

2.2.2 The second trend 10

2.2.3 The third trend 14

2.3 Money Market Pressure (MMP) Index (Hagen and Ho, 2007) 19

2.4 Chapter summary 21

CHAPTER 3: METHODOLOGY, MODEL SPECIFICATION AND DATA 28

3.1 Model selection 28

3.2 Model specification 31

3.2.1 Macroeconomic indicators 33

3.2.2 Financial indicators 34

3.2.3 Institutional indicators 36

3.2.4 Use of lagged terms 37

3.3 Estimation strategies and relevant model diagnostics 40

3.3.1 Calculation of MMP for banking crisis assessment 40

3.3.2 Model estimation steps and diagnostics 41

3.4 Data scope and sources 43

3.5 Conceptual framework 46

3.6 Research Process 47

CHAPTER 4: RESUTLS AND FINDINGS 48

4.1 Descriptive statistics of explanatory indicators 48

4.2 Statistical tests for model 51

4.2.1 Model specification test 51

4.1.2 Goodness of fit test 51

4.1.3 Test for multicollinearity 51

4.3 Coefficients interpretation 53

4.3.1 Macroeconomic indicators 53

4.3.2 Financial indicators 55

4.3.3 Institutional indicators 57

CHAPTER 5: CONCLUSION, POLICY RECOMMENDATION AND LIMITATION 58

5.1 Conclusion 58

5.2 Policy recommendation 58

5.3 Limitation of the research 60

REFERENCES 61

APPENDICES 65

Table 2.1 Summary of literature reviewed 22

Figure 2.1 Mechanisms of banking crisis 27

Table 3.1 Data for MMP index calculation 44

Table 3.2 Data and sources of explanatory variables 45

Table 4.1 Banking crisis dates retrieved from MMP index 65

Table 4.2 Summary statistics of variables used in the regression 49

Table 4.3a The correlation on the sample observations 50

Table 4.3b The correlation on the sample observations 50

Table 4.4Linktest for specification error of logit model 66

Table 4.5 Goodness of fit test of model 67

Tabel 4.6 Full model multicollinearity test result 67

Table 4.7 Dropping significantly high correlated variables GE, RL: 68

Table 4.8 Dropping high correlated variables GE, RL and CC 68

Table 4.9 Using interactive term of GE and RL 69

Table 4.10 Full model 69

Table 4.11 Restricted model without GE, RL, CC 70

Table 4.12 Fixed effect model with lags 70

Table 4.13 Random effect model with lags 71

Table 4.14 Simple logit model with lags 72

Table 4.15Comparison of lagged terms of indicators in simple logit, FEM and REM 73

MMP: Money Market Pressure

WGI: World Governance Indicator

WB: World Bank

IMF: International Monetary Fund

IFS: International Financial Statistics

ICRG: International Country Risks Guide

FEM: Fixed Effect Model

REM: Random Effect Model

BC: Banking Crisis

CHAPTER 1: INTRODUCTION

1.1 Problem statement

Banking crisis in nowadays economies is not a new issue or even an old one that hasbeen given awareness to, discussed and researched from many angles andperspectives by applying many approaches from simple to complicate There havebeen three trends of banking system crisis researches from its first trend ofqualitative description by Friedman and Schwartz (1963) about US crisis over itspast decades to the second trend in which econometric analysis with panel data wereemployed according to relatively enough banking crises observations and to thethird trend since the 2007 “global financial turmoil” The trends of banking crisisresearch contribute most of important indicators related to macroeconomics andbanking sectors such as reserves, current account, real exchange rate (Kaminsky et

al, 1998) Despite the fact that the logistic regression approach focused more onquantitative economics model, it has seemed to be an important tool for anticipatingthe crisis signals and timing as well as significant indicators However, there wasalso some noise that could affect the effectiveness of this model Hence, it led to therise of further studies in terms of developing new method and other new criticalvariables

As suggested, there have been many criteria to help researchers with banking crisisidentification Amongst, money market pressure index from the work of Hagen and

Ho (2007), who expanded the literature of Eichengreen, Rose, and Wyplosz (1995,1996a, 1996b) for currency crisis, stands out to be convenient for understanding anddata collecting but still provide good judgment value for banking crisis symptom.Such index observed the periods that banking systems experience its liquidityproblem by considering simultaneously the phenomenon of both high central bankreserves demand and fluctuations of short-term real interest rate Originally, theindex provides the criterion to indicate whether there is a crisis or not under thescope analyzed

Banks relevant data, to some extent, seems to be difficult to obtain precisely due to

their sensitiveness Given those difficulties, the research will make use ofmacroeconomic indicators as suggested in a survey that emphasized “the analysis ofmacroeconomic variables is of some help for banking supervisors in order to fullyassess banks’ health” (Quagliariello, 2008) In accordance with both suggestionfrom Quagliariello (2008) and Hagen and Ho (2007), some availablemacroeconomic and financial variables such as inflation, growth of monetary base,depreciation, real interest rate, growth of private credit over GDP, growth ofdeposits over GDP and growth of M2 over reserves are examined In recent years,there has been the use of institutional signals (Kaufmann et al, 2008) to predict forthe probability of vulnerability and crisis occurrence besides quantitative economicindicators to enhance the limitation of the model by Kaminsky et al (1998).Moreover, being motivated by the work of Breuer et al (2006) on institutionalvariables and currency crisis, this research will take this idea together with thecombination with six updated world governance indicators (Kaufmann, 2013)namely voice and accountability, government effectiveness, political stability, rule

of law, regulatory quality, control of corruption to assess the role of “health” ofGovernment in the relationship with crisis time of the banking systems Last but notleast, the 12-month lagged term of banking crisis included into the regression model(Falcetti and Tudela, 2006) also give significant assessment

Nevertheless, it seems that most of relevant researches tend to try to explain thereasons for a banking crisis occurrence but not that why banking crisis does not takeplace in some situation over some period in some country The attempt tounderstand or even forecast the crisis is important on one hand But, on the otherhand, future researches should be carried out with the tranquil time of the bankingsystem, i.e the “non-crisis” situation, still has its important role which seems to bebelittled or even no need to be explained (Kauko, 2014)

Although there have been researches and studies on banking crisis, it seems thatthere are likely few works considering simultaneously the health of Government,macroeconomic and financial background in a same model Thus, the contribution

of this thesis is to employ a combination of MMP index approach with updated datafrom IMF – IFS over the year scope of 2001-2010 to analyze the somewhat overallbanking crisis phenomenon under the impacts of the macro-economy environment,the financial situation and institutional indicators The rationale of such approach isthat there may be more useful findings will be figured out for banking crisisanalyses as well as more awareness will be taken into account from the perspectives

of authorities’ management for banking sector, in particularly, and for the economy

in general

1.2 Research objective

This thesis, whose attempt is to contribute an updated research on benign periods ofbanking systems through the analysis of banking crisis, will focus on the objectiveswhich try to identify factors of macroeconomics, finance and institutions that areuseful for explaining the occurrence of banking systems crisis

1.3 Research question

Which are the macroeconomic, financial and institutional indicators that provideawareness for the crisis time of banking system?

1.4 Structure of the thesis

After the finish of Chapter 1 about thesis introduction, the rest of this thesis will becategorized as following chapters:

Chapter 2 introduces banking crisis definition, relevant literature reviews of trends

of banking crisis researches, money market pressure index which will be applied forbanking crisis dependent variable identification

Chapter 3 states the methodology, model choice and specification and data scopeused This chapter also gives readers clear arguments on explanatory variables used,suggested statistical diagnostics of significance of model and variables.Simultaneously, data scope and sources together with model conceptual frameworkand analytical framework are also declared

Chapter 4 interprets results and findings of thesis regression model

Chapter 5 concludes with policy recommendation, thesis limitation and further research suggestion.

CHAPTER 2: LITERATURE REVIEW

This section demonstrates the defining work of banking crisis and choice of theauthor for the appropriate definition from the perspective of understandability anddata availability Simultaneously, the research history of banking crises over timeare also introduced and discussed in terms of approaching methods applied,particular researchers, and dataset collections

Henceforth, this chapter includes four parts which will be introduced one by one inorder from the first part of banking crisis definition to the second part of theintroduction of three trends of banking crisis analyses The third part of this sectiongives detailed explanation and discussion on money market pressure index used byHagen and Ho (2007)and the last part will concludes all related literature of thischapter

2.1 Defining banking crisis

Banking crisis by the definition of IMF (1998) is the situation that “bank runs andwidespread failures induce banks to suspend the convertibility of their liabilities, orwhich compels the government to intervene in the banking system on a large scale”

In another work of Demirgtic-Kunt and Detragiache (1998), the concept of bankingcrisis was defined as event method whose conditions are that one or the entirefollowing phenomenon holds:

1) The existence of at least 10% of the ratio of non-performing assets over total assets in the banking system

2) Cost of the rescue packages reached at least 2% of GDP

3) Extensive nationalization of banks due to banking sector problems

4) Governmental regulation of deposit guarantee, large-scale bank runs, long holidays of banks, deposit freeze

However, this definition of banking crisis has some drawbacks Firstly, the cost ofrescue packages from the Government were unclear until after a crisis occurredleading to late identify of this crisis Long banks holidays, nationalization of banksseem to happen after the entire economy was hit by crisis Secondly, it is difficult to

determine the extent to which Government did intervene to help banks facing withcrises Thirdly, the intervention of authorities may be early or late, hence, theaccurate dates are often uncertain (Caprio and Klingebiel, 1996a) Finally, the eventmethod only classifies the crises when there are enough severities to acceleratemarket events Consequently, crises identification based on the events of policyresponses are biased in the nature of biased event selection This, with no doubt,limits the ability for banking crises likely determinants to prove their analyticvalues.

With the attempt to contribute an alternative identification for banking crisis, themoney market pressure index (MMP) was built up in the work of Hagen and Ho(2007) who were motivated by the ideas of Eichengreen (1995) on currency crisesanalyses Henceforth, the banking crisis is defined as “periods in which there isexcessive demand for liquidity in the money market” (Hagen and Ho, 2007) Therationale for this index to be born comes from the traditional assumption that theshort-term interest rate, i.e the opportunity cost for banking sector to hold reserves,has a negative relationship with its reserves demand for central bank Thehypothesis that “banking crisis is characterized by a sharp increase in the bankingsector's aggregate demand for central bank reserves” (Hagen and Ho, 2007, p.1039)can be analyzed through three reasons:

- Banks confront with increasing non-performing loans and/or significantdecline in bank loans quality leading to illiquidity, hence, a rising in demand

of reserves to retain liquidity

- When sudden withdrawals occur, there will be a pressure for banks to deal with interbank market and central bank to be refinanced

- Government bonds and other more guaranteed assets are favored by financialinstitutions rather than lending to those in troubled leading to “a drying up ofinter-bank lending”

With the attempt to react to the increasing demand for reserves, central bank, who isthe last lender, will enact two basic policies on either bank reserves targeting or

short-term interest rate targeting In the first scenario, short-term interest rate willincrease For the latter, an injection of reserves into the banking system through themechanism of OMO or discount window lending must be carried out As a result,the existence of either the symptom of drastically increasing of short-term interestrate or the amount of reserves of central bank, or even both, denoting money market

is under high pressure Thus, with a convincing reasoning, the index of moneymarket pressure may capture the vulnerabilities of banking sector and be defined as

“the weighted average of changes in the ratio of reserves to bank deposits andchanges in the short-term real interest rate The weights are the sample standarddeviations of the two components” (Hagen and Ho, 2007) The index can bedescribed by the equation herewith:

Where denotes reserves to bank deposits ratio which will, when money marketconfronts high tension, increase in the case of injecting reserves from central bank

to banking system or in the case there are withdrawals of depositors.r denotes term real interest rate, and are different terms of and , and are thestandard deviations of the two components respectively

short-The judgment for banking crisis (BC) will be shown below:

Ho (2007) as a conjunction for the Chapter 3

2.2 Trends of banking crises researchtogether with crises mechanism

Going through the history of banking system fragility, from the first popularly citedqualitative description of US crisis of Friedman and Schwartz (1963) to the so-called seemingly first banking crisis database of Caprio and Klingebiel (1996a,

1996b) and the widely cited works of Demirguc-Kunt and Detragiache (1998) andKaminsky and Reinhart (1999), banking panics or banking crises, on the whole,were and have been caused by somewhat similar factors such as the health economyand/or Government, the fragility of banking system itself, some contagion effectfrom the outside world/ economies, etc…Given those similarities in mechanism(s),each period has its own approaching method to the assessment of specific bankingsystem distress based on the availability of data, techniques and even support fromstatistical software packages The following words will introduce in details theexisting trends together with their relevant approach and the mechanism, if any,with the intention to provide readers with an overview of banking crises researchand analysis Some arguments on approaching methods are also discussed in thissection following the categorized suggestion of Kauko (2014).

2.2.1 The first trend

Description of specific historic events is mainstream of the first trend of bankingcrisis analyses The below words introduce some authors of this trend

Friedman and Schwartz (1963) in the work of “Monetary history of the United

States, 1867–1960” mentioned about bank run over the observations of an increase

of short-term interest rate and a decline in the ratio deposit over currency As cited

by Waldo (1985), bank tends to guaranty its withdrawal by selling long-termsecurities prematurely leading to a rise in yield of short-term assets In addition,with losses by the tradeoff between withdrawal readiness and the selling ofsecurities before maturity, bank has no choice but default some of its depositsmaking the depositors rush to shift their deposits into cash to somewhat self-protectthemselves against risks of bank-run Moreover, the banking crisis in October 1930supported for this point of view that some banks experienced failures making thepublic, on one hand, attempt to convert their deposits into cash On the other hand,this effect spread out to the whole banking system all over the country generating acollapse of the US banking system in December 1930 Not long after that, the

period from March to June 1931, the second wave of crisis occurred more severe because the banking system had been unhealthy during the former crisis.

Herrala (2011) contributes a description on Finnish crisis within the scope of 1865

– 1998 from the perspective of profitability of bank by using case studies of banks

in Finland The study shows that observations made by Herrala give evidence thatseries of event triggering banking crisis in Finland seem to go in line with otherformer studies using either data of others countries or international The studyconducts a definition of banking crisis under the condition of incidental occurrence

of negative profitability of banking sector By using available statistical data at thetime being, the study has made an attempt to figure critical characteristics ofbanking crisis and the crisis cycles which may deteriorate financial status ofbanking sector For the purpose of comparison, the study, then, take advantages ofthose findings from studies of international banking crises Indicators affecting theadvance phase of banking crisis cycle are sought by analysis of the periods whosefeatures are similar to those indicating typical case of banking crisis cycle whenfinancial conditions of banking sector are still healthy In addition to the mainexplanatory factor of bank profits over total assets, the study includes some otherstatistical descriptive factors such as growth of real GDP, investment, inflation,volume export change, stock money, exchange rate, interest rate, total assets

change, portion of bank deposits over loans, etc…

Gorton (1988)introduced econometric evidence on determinants of banking panics

in US before WW1, i.e U.S National Banking Era (1863-1914), by the analysis ofbanking panics and the depositors’ behaviors Moral hazard, i.e the role of agency,issue was also mentioned The research emphasized that the banking panics might

be caused by the changing in perceptions for risks of depositors Some indicatorswere taken into account such as deposits ratio, liabilities

Such econometric based researches made a link between the first trend and thesecond trend which will be introduced below

2.2.2 The second trend

In the condition of relatively adequate information of observations of banking crisesand relevant useful data, econometric researches have been deployed together withpanel data In this trend, banking crises were likely explained by the use ofmacroeconomic and financial factors Usually, researchers use the samples of paneldata with many countries over very long period of time, but the analyses seem tofocus on developed countries In addition, the crisis here only captured twoextremes of the situation whether there is crisis or not, this is the so-calleddichotomy nature as discussed in some papers of this trend

Being a highly attracted issue, banking crises phenomenon of this second trend

obtained an important contribution from Caprio and Klingebiel (1996a, 1996b)

whose work has been considered to be the first banking crisis database with crisisdates, countries and some economic explanatory variables together withobservations on policy measures The focus of this research was on the insolvency

of banks in the relation with readiness of more data could be collected such as GDP,inflation, monetary growth, fiscal balances, trade balances, real deposit rate,financial deepening, real credit/GDP, etc…from 69 countries over the period of late1970-1996 In-depth interviews with experts in this field were carried out to obtainepisodes of such crisis However, the work of Caprio and Klingebiel(1996a, 1996b)advised that it should be improved by more bank performance indicators which aredifficult to achieve (even in nowadays banking systems) and development indicatorswhich may contribute to the precision of crises occurrence predicting for individualbanks, on one hand, and for the whole system, on the other hand In addition, thepolitical economy researches for the phenomenon of bank insolvencies weresuggested to be a useful tool for Governments

Besides, in the trend of econometrically oriented analyses, the twin crisis wasintroduced as the simultaneous occurrence of both currency crisis and banking crisisbased on the signal-to-noise approach to judge for the situation of crisis or not, i.e.reach the alarm signal or not, in accordance to “the threshold values on an indicator-

by-indicator basis” (Kaminsky and Reinhart , 1999) Consequently, the thresholds

must be selected in the sense that could minimize the signal-to-noise ratio 16indicators from financial sector, external sector, real sector and fiscal sector wereemployed in this analysis of banking crisis individually and twin crises as a whole.However, there existed some drawbacks of wrong signaling in this method.Nevertheless, earlier signal are, to common sense, somewhat valuable informationfor the authorities Sample used in the research consists of 20 countries for theperiod 1970-mid-1995 This paper aimed to fill this void in the literature andexamine currency and banking crises episodes for a number of industrial anddeveloping countries including Denmark, Finland, Norway, Spain, Sweden,Argentina, Bolivia, Brazil, Chile, Colombia, Indonesia ,Israel, Malaysia, Mexico,Peru, The Philippines, Thailand, Turkey, Uruguay, and Venezuela This sample givesalso the opportunity to study 76 currency crises and 26 banking crises following thedatabase in the work of Caprio and Klingebiel (1996) Out of sample testing wasexamined with the twin crises in Asia of 1997

Dermirguc-Kunt and Detragiache (1998) used a large sample of both developed

and developing countries over their scope from 1980 to 1994 with a multivariatelogistic model to figure out the relevant factors of systemic banking crisesoccurrence This research pointed out that the crises seemed to burst under a weakmacroeconomic environment, i.e high inflation and low growth In addition, realinterest rate in its high status also contributed to problems in the banking sector, thesame evident finding for the role of vulnerable balance of payment was mentioned.Some institutional issues such as deposit insurance existence and weak lawenforcement were found to put risks to the banking systems The study emphasizedthe significance of low growth of GDP in the sense that it could make the bankingsector at risk On one hand, banks are the financial intermediaries, by nature, thatshould involve in risk taking manner; hence the vulnerability of outside economicenvironment should not be a worrying signal But, on the other hand, banks would,

to some extent, ignore the credit risk of domestic economy fluctuation and lend

overseas This activity of banking sector in developed countries benefited somedeveloping countries but put much pressure on the authorities to improve theinstitutional regulation on banking systems if they do not want to see the bankingsector fragility caused by the volatility from the expansion of cross-border bankingactivities There has been a debate for the role of financial liberalization in bankingsystem stability The study also showed some weak evidence for the likelihood ofbanking crises under the condition of controlled real interest rate in financialliberalization periods However, this study faced with some drawbacks related toestimation model, the tradeoffs between the macroeconomic, institutionalexplanatory indicators and the financial factors, i.e financial markets indicators,which might capture the banking system more entirely Some suggestions for furtherstudies on banking structural indicators, such as “degree of capitalization of banks,the degree of concentration and the structure of competition of the market for credit,the liquidity of the interbank market and of the bond market, the ownershipstructure of the banks (public versus private), and the quality of regulatorysupervision”, were also stated.

Broad new and old samples of banking crisis over different countries have beencombined in some researches However, once again, these analyses on focused ondeveloped countries

Bordo and Meissner (2012) submitted an analysis with a 14 advanced countries

over the scope of 1880-2008 to study the linkage between credit booms, inequalityand housing policy to banking crises Credit booms, whose explanatory factors stillhave not yet firmly indicated, are likely to contribute obvious evidence to bankingcrises By applying a logit model with and without countries fixed effect, theresearch found positive evidence between credit booms and banking instability, lagterm of credit booms indicator was also taken into the model for testing Althoughthe lag term of one year gave low probability of banking crisis, surprisingly, thefinding showed that there is a significant positive relationship that banking crisisoccurrence proceeded by a rise in real credit growth with its lag terms in prior to

two to five years Moreover, the research suggested that some factors such asincreasing of real income and fall in interest rate may be important to analyze creditbooms However, according to the existing dataset, the research could not find muchvaluable signal from the standpoint that a rise in income inequality and housingredistributive policy contribute to the probability of financial crises.

Schularick and Taylor (2012) analyzed the role of “hitherto unknown” credit

expansion in nowadays economy around the world The motivation of this is that astable relationship between money and credit found after the Great Depression andWorld War II still keeps hold to crisis today In addition, there is likelihood that,after the 1930s, complex macroeconomic environment and financial policies such asincreasing of fiat money, role of banks as the lender of last resort have beentriggering the credit to expand Moreover, the financial system with its particularstructural changes over a long period in the past has given credit an important role

in the macro-economy as a whole To this extent, the stated unlucky progress hasbeen making credit become more essential ever However, the raise of such credithas been debated to play no construction role for the monetary policy Nevertheless,from the perspective of lessons learnt from histories from both researchers andpolicymakers, the risk of this so-called credit accumulation was ignored Thereexists the likelihood that credit booms may contribute more risks to financial crises

in the future One could give some criticisms that this is not a perfect factor topredict the financial crises under some explanation that expansion of creditcontributes to the real economic growth; that some failures in terms ofoperations/regulations within the financial systems have decrease the role of creditexpansion Although there are many debates about the predicting power of creditbooms, the historic lessons of credit expansion and financial fragility still has itsvalue for more deep research in the future The role of credit in macro-economyshould be examined

Jorda et al (2011) analyzed the financial fragility in the relationship with external

imbalanced situations of the economy such as deteriorations of the current account,

growth of loans, volatility of interest rate, inflation, growth of GDP, etc…byapplying the logistic country fixed effect model over the wide scope of 140 yearsacross 14 developed countries A combination of descriptive statistics of financialfragility explanatory factors and that of econometrically oriented logit model werealso discussed The mechanism of macro-economic indicators was found out thatgrowth of loans played an important role in accelerating crises from both nationaland global perspectives Deteriorations of current account seemed evidentlycontribute to the run-up to crises for not only global but individual countries as well.Natural interest rate being under strong suppression gave signal to the phase of run-

up to crises especially in the cases of four global crises over 140 years of analysis.(i.e 1890, 1907, 1930-1931 and 2007-2008) Real interest rate and inflation alsogave similar predicting signal to this trend The conclusion of this research indicatedthat the built-up phase of crises should be paid more attention by policymakers byobserving and/or analyzing the activities of external macroeconomic imbalances.Moreover, the research emphasized that the credit growth and current accountintertwined significantly nowadays Hence, these factors might be good predictorsfor financial instability from both the viewpoint of clear historic event and recentobservations

2.2.3 The third trend

Cross-country analyses have been emerged since the 2007 global financial turmoiland the subprime Lehman crisis Mainstreams of this trend are those analyses on theimpact from the perspectives of financial sector, real economy to the employing ofvariety explanatory indicators on banking system fragilities Each of the researchaspect will be reviewed hereinafter separately

Financial sector perspective

Since the conjunction contribution of Gorton (1988) between the first and secondtrend of banking crises analyses, there have been raising the motivation foreconomists to study more on factors from both macro-economic and bankingsectors with the hope that their forecasting power for banking system fragilities will

be improved Kauko (2012) gave an analysis on the relationship of deficit current

account to banking vulnerabilities with main concentration in banking sector.Although the research did not deal with crisis occurrence probability, it focused onthe so-called direct factor causing crisis, namely the deterioration of credit quality.Non-performing loans of individual bank, on one hand, and of the whole system, onthe other hand, are attractive signal as they contribute to the losses of banksprofitability, i.e one among the indicators judging health of banks Over high creditgrowth may occur in prior to banking crises In addition, from the standpoint offinancial stability, a proxy for problematic credit growth is foreign debt In moredetails, the study analyzed how and to which extent the macro-economicenvironment reflected in the quality of credit in recent crises A dependent variable,indicated by the relative amount of non-performing loans of 2009 from IMF, wasapplied together with a cross-section of 34 advanced countries according to theclassification of IMF The outcome showed that credit growth in the combinationwith current account deficit act as an important predictor for financial crises

In the scenario of the well-known Global Crisis accelerated by 2007 financial crisis

in the US, the study of Aizenman and Pasricha (2012) seeks to understand the

dynamics of the spillover effect of this crisis to the rest countries of the world andthe crisis relevant factors The scope of analysis spread over 107 countries togetherwith their financial crises episodes from 2008-2009 In addition, a wide set ofindicators were taken into account by using the ordinary least square regression.There are six common variables that existed through all the regression models,namely per capita real GDP, international reserves-to-GDP ratio, an interaction termbetween international reserves-to-GDP ratio and a dummy variable indicatingwhether the country was a recipient of a swap line by the Federal Reserve, theEuropean Central Bank, or the People’s Bank of China, trade-to-GDP ratio, adummy variable for whether the country was a commodity exporter, and de jurerestrictions on capital flows measured by the Chinn-Ito index Moreover, manyother indicators were also applied such as those factors of external exposure,

institutions, financial development, banking sector health and competition, etc… theconcepts of internal and external financial stress were also distinguishedrespectively as pressure of capital outflows and pressure inducing declines in stockmarkets and expansions in central banks’ balance sheets The research showed thatthese two types of stresses shared some common factors Countries with greater defacto openness saw larger shocks, and countries with more competitive bankingsystems were less vulnerable if their banking systems were also better capitalized orbetter supervised In addition, countries with higher international reserves sawgreater external stress, and commodity exporters saw lower internal stress.

Real sector perspective

Berkmen et al (2012)shared the same concern of the contagion effect of financial

crisis in advanced countries to the rest of other countries around the world Theresearch stated that, although the severity of crisis may differ among countries, themacroeconomic background as well as institutional policies may play a role inreflecting the vulnerabilities of each financial system during the crisis shock Thedifferences between impacts of crisis in emerging and developing countries werealso discussed Variation in growth rate of the economy stands out to be aninteresting factor to capture the real economy activities Regarding tomethodologies, this research applied both descriptive statistic evidence and countrycross-sectional regression together with wide range of indicators to meet the aimthat the whole picture of which key indicators matter for shaping the varieties ingrowth The result showed that vulnerabilities of the financial systems seem to have

a significant role in the serious impact of growth Simultaneously, the resultindicated that a more leveraged system as well as more short-term debts seemed togenerate bigger losses for the country On the other hand, there is likelihood tosuffer shock for countries whose exchange regimes are pegged because the flexibleregime would be at good help to buffer the shocks Effective fiscal policies werealso found to be at good help for less severity with shocks However, the studymight be due to dataset problem, found no significance of other policy variables

Artha et al (2011)in another study has put a new brick to the analysis of financial

crisis but from a new perspective of the linkage between labor market, an importantentity of the real economy, and financial crisis to the impact of output losses in theeconomy A set of 56 countries over the scope of 2007-2009 first quarter with theoutput loss was examined through the declining of real GDP.Employing a cross-country model that includes control variables such as trade and capital marketintegration, financial development, monetary and fiscal policy, institutionaldifferences, and population growth, we find that lower hiring cost reduce the outputloss, notably so in high-income countries However, the duration of the crisis islonger in case of low dismissal cost, notably so in low-income countries

Rose and Spiegel (2011)are interested in the big amount of “fundamental” causes

of the global crisis of 2008-2009 and in the likelihood that there may be any linkagebetween the “Great Recession” and the actual crisis Using the cross-sectionalapproach based on their former dataset in 2010, the study examined the countriesand/or territories whose real GDP per capita higher than USD 10,000 as well asthose with at least USD 4,000 but their population has to be from one million Somestandard OLS regressions were also taken into account for this dataset Asmentioned, a wide set of indicators from many aspects were applied such as 7factors of cross-country crisis severity, i.e the GDP growth over times, growth ofconsumption, 8 indicators that widely used by other researchers such as exchangerate regime, current account, growth of trading partner, credit market regulation,short-term external debts, changes of house price, growth of bank credit andinternational reserves, etc…together with 6 different samples of country.Surprisingly, those large set of indicators and countries sample provided nosignificant linkage between the causes and the Great Recession A conclusion fromthis time of study is that causes of crisis may vary from country to country leading

to the fact that the cross-country models could not fit the data well even with sample test and they were not estimated with acceptable accuracy However, inanother following research, still a wide set of factors were employed using the

in-Multiple Indicator in-Multiple Cause (MIMIC) model introduced by Goldberg (1972)with a cross-sectional data from 107 countries to analyze causes of the 2008 globalcrisis The paper emphasized the importance of a wide range of indicators with therationale is obtaining as much as possible the explanatory ability of the data,

although those causes specification may be empirically unstructured (Rose and

Spiegel, 2012) Despite the big amount of fundamental indicators from financial

situation, macroeconomic background, institutions, geographic indicators andregulatory framework, this time of analysis still ended up with pessimistic outcomethat almost none of the indicators seemed to have strong significance in explainingthe cause of crisis in 2008 The paper, then, advised that there still exists somelinkage but the observed data may speak fewer things with the existing econometrictechniques and that the future crises seem difficult to be forecasted precisely

Frankel and Saravelos (2010) contributed a research motivated by the

cross-country incidence of the 2008-09 financial crisis using some approaches such asprobit model, signals method, combination of qualitative and quantitative method,etc… along with a dataset consists of 50 annual macroeconomic and financialvariables for 2007 or earlier from the World Bank World Development Indicatorsdatabase This source is augmented by monthly real effective and nominal exchangerate data from the IMF International Financial Statistics database Data availabilitydiffers by country, with the most data points available for the level and growth rate

of GDP (122 countries) and the least data available for various measures of term debt (67 countries) High frequency data for foreign exchange rates (156countries), stock market indices (77 countries), industrial production (58 countries)and GDP (63 countries) up to the second half of 2009 are sourced from Bloombergand Data stream for the financial and real data respectively The high frequency dataare used to define crisis incidence from the second half of 2008 onwards, asanalyzed in more detail below All the independent variables are dated from 2007 orearlier, minimizing endogeneity issues This paper conducted an extensive review ofthe early warning indicators literature, and found a number of variables to be

short-consistently useful in predicting financial crisis incidence across time, country andcrisis in earlier work These indicators were subsequently included in an empiricalanalysis of the 2008-09 crisis International reserves and real exchange rateovervaluation, the top two indicators identified in the review, stood out as usefulleading indicators of the current crisis Reserves were robust to a number of crisisincidence definitions as well as the inclusion of additional independent variables inmultivariate specifications using an exchange market pressure index as a measure ofcrisis incidence Past exchange rate overvaluation only proved useful for measures

of crisis incidence that defined a crisis in terms of the currency A number of othervariables appear as potentially useful leading indicators during the current crisis,though their robustness across different crisis incidence measures and specificationswas not as compelling Lower past credit growth, larger current accounts/savingrates, lower external and short-term debt were associated with lower crisisincidence There remains fertile ground for further research into the effectiveness ofearly warning systems in predicting the 2008-09 crises and beyond The findingsalso highlight the potential economic significance of reserve levels and exchangerate policy in affecting crisis vulnerability

2.3 Money Market Pressure (MMP) Index (Hagen and Ho, 2007)

The equation (2) illustrates the criteria for the identification of banking crisis usingMMP index Specifically, the index is calculated for each country separately and theperiods of banking crises identified when the index satisfies both of the twoconditions: (1) the index exceeds the 98.5 percentile of its sample distribution foreach country taken for computation; (2) the growth rate of the index is at least 5%.Following the explanation of Hagen and Ho (2007), the first criterion ensures thatonly distinctive episodes will be judged as crises; while the second criterionconsiders those countries confront no crisis during the scope of sample analysissince the first condition keeps hold for every sample distribution Empirical resultsindicate that, the first criterion once relaxed will lead to the probability that toomany crises are identified, while the tightened counterpart will lead to missing true

crisis In addition, the mentioned percentile has been changed to other values whichresulted in the decline of explanatory power of the regression model with lowerpercentile value of 95, while no significant change in the case of higher percentilevalue of 99.5 With same explanation, the tightened condition of the secondcriterion makes some true crises episodes missing Overall, the crisis defining here

is country specific, one may criticized the definition should be applied for the wholecountries under consideration by pooling all the data and applying the onlycalculation However, as a matter of fact, the magnitudes of fluctuation of MMPindex may vary among countries, the pooled data for computing may lead tomissing of true crises for countries whose fluctuations of the index are relativelylow After all, in terms of the first criterion, the percentile is preferred to themultiple standard deviations due to the “non-normal” nature of distributions of theMMP index

Nevertheless, there seemingly exist some drawbacks of this definition of bankingcrises: (1) banking crises are believed to occur in modern world due to asset-drivenrather than liability-driven mechanism, however, the increase in demand forreserves caused by the deteriorations of bank assets is out of the scope of study ofthis thesis; (2) the index is not suitable in the case of interest rate controlled by somecentral banks in some countries, yet one advantage is that the MMP index isindependent with the interest rate flexibility provided that managements on interestrate of central bank rely on market measures; (3) the MMP index can only indicatethe starting of the crises but not when the crises end as “identifying the end of abanking crisis is one of the more difficult unsolved problems in the empirical crisisliterature" (Kaminsky and Reinhart, 2000), but the identification of when the crisisstarts or ends is beyond the thesis

The sample taken into analysis spanned over the period of 1980-2001 with 47countries The reasoning of choice for countries sample is based on the availability

of data from IMF, but the study excluded Argentina and Brazil out of theconsideration due to their abnormal inflation and interest rate According to which

indicators were used, the total observations varied from 697-726 in which there are34-38 crisis episodes Thus, percentage of crises over the sample population is about5% The explanatory indicators were chosen from both former literature and theexistence of data The method of conditional countries fixed effect model wereapplied with banking crisis judged by MMP index as regressand, and regressorsinclude large sets of indicators such as: (1) Macroeconomic variable: Growth rate ofreal GDP, depreciation, over-valuation of real exchange, real interest rates, inflationrates, surplus/GDP, GDP, Dummy for severe recession (Growth <-5%), Dummy forhigh inflation (inflation >20%), Growth rate of monetary base, Growth rate of realdomestic credit; (2) Financial variable: Private credit/GDP, Cash/Bank, Changes instock market price Share prices; (3) Institutional variables such as: GDP/CAP (1000dollars/person), Dummy variable for existence of explicit deposit insurance,Dummy variable for financial liberalization The outcome shows that slowdown ofreal GDP, lower real interest rates, extremely high inflation, large fiscal deficits, andover-valued exchange rates tend to precede banking crises The effects of monetarybase growth on the probability of banking crises are found negligible.

Figure 2.1 demonstrates the mechanism leads to banking crisis from theperspectives of macro-economy, finance and institutions Generally speaking,banking system under weak macroeconomic environment, vulnerable financialbackground of banks and inefficient institutions is likely to be exposure high risk ofbanking crisis occurrence

Table 2.1 Summary of literature reviewed Author(s) Key indicators Methodology Data sample Findings

Friedman Bank run, Descriptive United States - Banks’ healthiness is likely considered and short-term analysis (1867–1960) as its withdrawal readiness in the expense Schwartz interest rate, of selling long-term securities prematurely (1963) bank deposit leading to a rise in yield of short-term

assets This action results in defaulting some of banks’ deposits making depositors rush to shift their deposits into cash to somewhat self-protect themselves against risks of bank-run.

- The later wave of banking crisis occurred more severe because the banking system had been unhealthy during the former crisis.

.Herrala Bank Descriptive Finland - Macroeconomic indicators such as real (2011) profitability analysis, case (1865-1998) GDP growth, change of volume in export,

studies investment, inflation, money stock,

exchange rate, interest rate, change in total assets and deposits as portion of loans were analyzed using descriptive statistics approach to figure out the characteristic of banking crisis.

- Banks’ profitability over total assets has been taken into account to analyze for the deterioration of banks’ financial condition which may lead to banking crisis cycles Gorton Banking panics Econometric United States - The research emphasized that the (1988) and the evidence (1863-1914) banking panics might be caused by the

depositors’ changing in perceptions for risks of

into account such as deposits ratio, liabilities.

Caprio and insolvency of In-depth 69 countries - Been considered to be the first banking Klingebiel banks, GDP, interviews over the crisis database with crisis dates, countries (1996a, inflation, period of late and some economic explanatory variables 1996b) monetary 1970-1996 together with observations on policy

balances, real health of banks tend to lead to banking

sector and fiscal sample testing was examined with the

- Volatilities of such indicators from financial aspect, external and real sector seem to contribute to the probability of banking crisis occurrence.

Dermirguc- Growth rate of multivariate a large - The research found that weak

Kunt and real GDP, logistic sample of macroeconomic background of the Detragiache change in term model 45-65 economy such as low growth rate of GDP (1998) of trade, countries in tends to trigger banking crises due to the

depreciation, IFS database risk taking nature of banks.

real interest from both - In addition, banking crises may rise rate, inflation, developed from problems of maturity transformation Government and of banking sectors as the consequences of surplus over developing high fluctuation of nominal interest rates GDP, M2 over countries in accompany with high inflation rate reserves, over their Banking sector stability should be benefit private credit scope from from inflation controlling policies such as over GDP, bank 1980 to 1994 restrictive monetary policies However, cash over such activities carried out in the context of assets, growth implementing of high inflation controlling rate of domestic may lead to the likelihood of banking credit, dummy crises occurrence through high real for deposit interest rate channel Weak banking insurance, GDP systems should be cautious about inflation per capita, law controlling policies and monetary policies and order - Moreover, institutional decision of

increase the likelihood of banking crisis Bordo and credit booms, logit model 14 advanced - Banking system instability, which may Meissner inequality and with and countries lead to banking crisis, was evidently found (2012) housing policy without over the to have a strong positive relationship with

to banking countries scope of lagged term of two to five years of credit crises fixed effect 1880-2008 booms.

- Inequality and housing policy are those factors of the economy were taken into account to test for impacts on the probability of financial crises Unfortunately, the results from existing dataset showed no evidence.

Schularick credit OLS linear 14 countries - Credit expansion contributes to real and Taylor expansion, probability, over the economic growth However, there is (2012) financial crisis logit years 1870– likelihood that credit booms may

2008 contribute more risks to financial crises in

the future due to failures in operations and/or regulations within the financial system.

- The role of such credit booms from the perspective of macroeconomics should be further studied as there have been the historic lessons of credit expansion and financial fragility.

Jorda et al the current logistic 140 years - The mechanism of macro-economic (2011) account, growth country fixed across 14 indicators to cause banking crisis was

of loans, effect model, developed found out that growth of loans played an volatility of descriptive countries important role in accelerating crises from interest rate, statistics both national and global perspectives.

growth of GDP evidently contribute to the run-up to crises

for not only global but individual countries as well Natural interest rate being under strong suppression gave signal to the phase of run-up to crises.

Real interest rate and inflation also gave similar predicting signal to this trend Kauko deficit current no banking cross-section The outcome showed that credit growth in (2012) account to crisis of 34 the combination with current account

banking probability advanced deficit act as an important predictor for vulnerabilities calculation countries financial crises.

with main model, OLS according to

concentration in regression the

banking sector, classification

GDP ratio, an crises stresses shared some common factors interaction term episodes - Countries with greater de facto openness between from 2008- saw larger shocks, and countries with international 2009 more competitive banking systems were reserves-to- less vulnerable if their banking systems GDP ratio and a were also better capitalized or better recipient of a supervised In addition, countries with

variable, factors saw lower internal stress.

Berkmen et wide range of descriptive 43 countries The contagion effect of financial crisis in

al (2012) variables of statistic from advanced countries to the rest of other

Trade linkages, evidence and Consensus countries around the world The research financial country Forecast, 141 stated that, although the severity of crisis linkages, cross- countries may differ among countries, the vulnerabilities/ sectional from WEO macroeconomic background as well as financial regression database over institutional policies may play a role in structures, the year of reflecting the vulnerabilities of each policy 2007, 2008, financial system during the crisis shock.

first quarter find that lower hiring cost reduce the

output loss, notably so in high-income countries However, the duration of the crisis is longer in case of low dismissal cost, notably so in low-income countries.

Rose and 7 factors of - cross- 6 different Those large set of indicators and countries Spiegel cross-country sectional samples of sample provided no significant linkage (2011) crisis severity, approach country between the causes and the Great

i.e the GDP - OLS 2008-2009 Recession Causes of crisis may vary from growth over regressions country to country leading to the fact that times, growth the cross-country models could not fit the

of consumption, data well even with in-sample test and

8 indicators that they were not estimated with acceptable

macroeconomic Cause countries to significance in explaining the cause of background, (MIMIC) analyze crisis in 2008

variables for method financial useful leading indicators of the current

2007 or earlier - combination crisis crisis Reserves were robust to a number from the World of qualitative of crisis incidence definitions as well as Bank World and the inclusion of additional independent Development quantitative variables in multivariate specifications Indicators method using an exchange market pressure index database as a measure of crisis incidence Past

exchange rate overvaluation only proved useful for measures of crisis incidence that defined a crisis in terms of the currency Lower past credit growth, larger current accounts/saving rates, lower external and short-term debt were associated with lower crisis incidence.

Hagen and - - 47 countries The outcome shows that slowdown of real

Ho (2007) Macroeconomic Identification over the GDP, lower real interest rates, extremely

variable: of banking period of high inflation, large fiscal deficits, and Growth rate of crisis using 1980-2001 over-valued exchange rates tend to real GDP, MMP index precede banking crises The effects of depreciation, Specifically, monetary base growth on the probability over-valuation the index is of banking crises are found negligible.

of real calculated for

exchange, real each country

interest rates, separately

inflation rates, and the

surplus/GDP, periods of

GDP, Dummy banking

for severe crises

recession identified

(Growth <-5%), when the

Dummy for index

high inflation satisfies both

(inflation of the two

>20%), Growth conditions:

rate of (1) the index

monetary base, exceeds the

Growth rate of 98.5

real domestic percentile of

credit its sample

- Financial distribution

variable: for each

Private country taken

credit/GDP, for

Cash/Bank, computation;

Changes in (2) the

stock market growth rate

price Share of the index

prices; is at least 5%.

- Institutional - Conditional

variables such countries

as: GDP/CAP fixed effect

good channel for increasing the risk of banking crisis In addition, the role of eachGovernment is significantly important to contribute to high or low exposure ofbanking system to crisis Mismanagement from in charged authorities may lead tobanking crisis directly or indirectly through the mechanism that makesmacroeconomic background go bad then banking crises occur Moreover, as bank isone element of the whole financial system, thus, the more unhealthy the financialsystem the higher risk of banking crises Figure 2.1 briefly shows a visualmechanism of factors that may turn banking sector to the risk of crisis.

Figure 2.1 Mechanisms of banking crisis

macroeconomic

background environment

financialsystem

CHAPTER 3: METHODOLOGY, MODEL SPECIFICATION

AND DATA

In the previous chapter, trends of banking crisis analyses have been introduced anddiscussed with the aims to provide readers with an overlook about banking crisisrelevant techniques and/or types of explanatory variables under consideration ofeach trend, each research model This chapter focuses on the following issues: (1)model selection, (2) specification of chosen model in terms of discussions ofeffective relationship of independent variables on dependent variable, i.e bankingcrisis and (3) the data sources and scope

3.1 Model selection

The judgment outcome of banking crisis is in binary format, i.e it takes the value of

“1” if there is a crisis and “0” if there is no crisis Hence, there are two methods mayoften be considered from empirical studies reviewed so far namely probit and logitregression model Moreover, it seems that most empirical studies reviewed abovetend to employ logit regression model for their analysis of banking crisis due to itsbinary nature, ease of understanding as well as interpretation (Dermirguc-Kunt andDetragiache, 1998, Hagen and Ho, 2007, Jorda et al., 2011, Bordo and Meissner,2012) Being motivated by this trend, the same technique of logistic regression will

be followed in this thesis The mathematical background of logit

regression is described in the below wordings

First, those may take a look at the logit model in its mathematic equation below:

Supposed that is the probability of banking crisis occurrence, then the state ofnon-crisis will be described by the term (1- )=1- = (2)

coefficient (in case there is more than one regressor in the model), subtract 1 from

it, and multiply the result by 100, the result will be the percent change in the oddsfor a unit increase in the jthregressor

The formula (4) under the conditions that time and entities are considered togethercan be rewritten as follow:

Gujarati (2004) indicated that the estimation (5) depends on the assumptions madefor its intercept, slopes of coefficients and the error term Thus, there are fiveassumptions: (1) the intercept and slopes keep hold regardless time and entities (i.e.countries) while letting the error term tell the difference; (2) slopes unchanged andintercept varies among countries; (3) slopes unchanged and intercept variesthroughout countries and time; (4) both intercept and slopes varies over countriesand (5) both intercept and slopes varies over countries and time As a result, theassumptions illustrate the increasing level of complexity and may coincide withmore reality For the purpose of analysis, only first two assumptions are discussedhereinafter

The case of assumption (1), normally known as pooled regression model, provides

an easy regression method together with high restrictions, therefore, the wholepicture of the relationship between the regressand and regressors may be distorted

As common sense, people are more interested in the specific nature of each country.This turns to the second case described by the estimation rewritten from (5) asbelow:

The estimation (6) with the intercept was rewritten as indicates the changingover countries for this item This is the so-called fixed effect model (FEM), inwhich its intercept keep changing across countries but still time invariant Thedifferences of intercepts indicate some specific features that one country has whilethe others may not have Hence, the choice of FEM seems to be preferable to pooledregression for the thesis

In addition, a formal statistical test, i.e the restricted F-test, is suggested for thechoice between restricted pooled model and FEM

Though the FEM is easy to be applied straightforwardly, this model is established inthe expense of the degrees of freedom if there are many cross-sectional units(Gujarati, 2004) One possible question rose for an alternative approach inconsidering the unknown information of the error term instead what have been donewith the intercept in FEM Consequently, random effect model (REM) was built up

as the expression below:

Where = + (7) denotes that the intercept now will be followed by a randomerror term with zero mean and variance The aim of REM is that it covers the ideathat individual countries have the same mean value of intercept but the differencescome from the error term

By substituting (7) into estimation (6), the REM estimation obtained below:

=

Another formal test introduced by Hausman (1978) for the choice between FEMand REM However, since FEM always gives unbiased estimation, FEM seems to

be most favored in this time of analysis

3.2 Model specification

As stated about the specification of regression model, Hagen and Ho (2007) andGujarati (2003) emphasized that the choice of any variable to be plugged into themodel has to be a combination of both theories and empirical researches However,

as a matter of fact, the risk of inappropriate structure leading to specification errormay still exist due to some reasons: (1) relevant variable(s) may not be consideredand/or irrelevant variable(s) may be taken into account, (2) inaccurate functionalform Regarding to the second reason, the thesis did argue and follow theassumption that logistic regression model is an appropriate model choice for thistime of banking crisis analysis In terms of the first reason, the thesis tends toinclude as many as possible the relevant variables in line with those already applied

by other previous researches and their data availability Since the aims of theanalysis are to recognize the more determinants the possible to the contribution toprobability of banking crisis occurrence, all variables will be included to test fortheir effectsthrough the use of logit model discussed in studies of Dermirguc-Kuntand Detragiache (1998), Hagen and Ho (2007), Jorda et al (2011) and Bordo andMeissner (2012)

Thus, general framework for predicting the occurrence of banking crisis wouldcomprise the all possible variables that we have Based on banking crisis theory anddata availability, the model is suggested as following:

denotes growth rate of monetary base to GDP of country i at time t

denotes the depreciation of domestic currency of country i at time t

denotes the short-term real interest rate at 36 months in advance

denotes the growth rate of domestic credit to GDP in 12 monthsearlier

denotes the growth rate of domestic deposits to GDP in prior 6months

denotes the growth rate of M2 over reserves of country i at time

t denotes voice and accountability

denotes political stability and absence of

violence denotes government effectiveness

denotes regulatory quality

denotes rule of law

denotes control of corruption

The relevant variables used in this thesis are classified into three variable groupsincluding macroeconomic indicators such as inflation, growth of monetary base,depreciation and past banking crisis The consideration of past banking crisis as afactor of macroeconomic background is due to suggestion that weakmacroeconomic environment may accelerate banking crisis (Berkmen et al., 2012).The other variable group includes financial indicators capturing such variables asgrowth of M2 over reserves, short-term real interest rate, growth of credit over