Early warning system for sovereign debt crisis in developing countries for period 1981 2010

The main findings are: i Solvency, measured by Public debt over GDP ratio, is positively correlated with sovereign debt crisis; ii Liquidity measures are highly associated with sovereign

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

EARLY WARNING SYSTEM FOR SOVEREIGN DEBT CRISIS

IN DEVELOPING COUNTRIES FOR PERIOD 1981-2010

Supervisor: Prof Dr Nguyen TrongHoai Student: Vu ThiLan Phuong

Class: MDE 20

Ho Chi Minh City, October 2016

ABSTRACT

This study constructed an Early warning system to explain and predict sovereign debt crisis

in 31 developing countries whose data is available through the period 1981-2010 at one year precedence by employing three-stage strategy with multinomial logit regression While three-stage strategy allowed selecting the best predictors among wide range of explanatory variables, multinomial logit regression solved “post-crisis” bias and thus, improved prediction quality The main findings are: (i) Solvency, measured by Public debt over GDP ratio, is positively correlated with sovereign debt crisis; (ii) Liquidity measures are highly associated with sovereign debt crisis When short-term debt to total external debt ratio grows up or reserves to total external debt ratio decrease, the likelihood of both entering into debt crisis and remaining in debt crisis rises; (iii) The macroeconomic fundamentals significantly affect sovereign debt crisis: while GDP per capita growth rate is negatively associated with both entering into and remaining in crisis , inflation only affect positively post-crisis period; (iv) The international liquidity, represented by the three-month the U.S Treasury bill rate, is highly associated with the first year of crisis and the following years of crisis rather than the initial year of crisis and (v) both external trade link and political institution measurements do not impact on sovereign debt crisis As a result, the study specified a multinomial logit Early warning model predicting sovereign debt crisis at one year precedence with six determinants namely Public debt over GDP ratio, Short-term debt

to total external debt ratio, Reserves to total external debt ratio, GDP per capita growth rate, Inflation rate and Three-month the U.S Treasury bill rate In addition, several policy implications are recommended for the countries to avert sovereign debt crisis

ACKNOWLEDGEMENT

First and foremost, I would like to take this opportunity to express my sincere gratitude to Prof Dr Nguyen Trong Hoai, my supervisor, for his strong support and useful advice from the very first days of my research work He was patient and sympathetic towards my delay His inspiration as well as prompt and intellectual guidance has encouraged me to finish the thesis

Finally, I would like to express my deepest grateful to my family, my colleagues and my friends for their continuous encouragement and support during the time I were busy with the study Thanks to their understanding and help, my thesis was finally completed

TABLE OF CONTENTS

CHAPTER I: INTRODUCTION 1

1.1 RESEARCH STATEMENT 1

1.2 RESEARCH OBJECTIVES 2

1.3 RESEARCH QUESTIONS 2

1.4 RESEARCH STRUCTURE 2

CHAPTER II: LITERATURE REVIEW 4

2.1 SOVEREIGN DEBT CRISIS AND EARLY WARNING SYSTEM 4

2.1.1 Sovereign debt crisis 4

2.1.2 Early warning system (EWS) 6

2.2 THEORETICAL LITERATURE ON DEBT DEFAULT 7

2.2.1 Model of debt overhang 7

2.2.2 Model of debt repudiation 9

2.3 EMPIRICAL LITERATURE ON SOVEREIGN DEBT CRISIS 11

2.3.1 Sovereign debt crisis and liquidity measures 11

2.3.2 Sovereign debt crisis and solvency measures 12

2.3.3 Sovereign debt crisis and macroeconomics fundamentals 13

2.3.4 Sovereign debt crisis and external trade link 13

2.3.5 Sovereign debt crisis and political institutions 14

2.3.6 Sovereign debt crisis and global liquidity 15

2.4 CHAPTER REMARKS 16

CHAPTER III: RESEARCH METHODOLOGY 18

3.1 MODEL SPECIFICATION 18

3.2 VARIABLES AND MEASUREMENTS 20

3.2.1 Sovereign debt crisis 20

3.2.2 Explanatory variables 22

3.2.2.1 Country characteristics 22

3.2.2.2 Exogenous factor 25

3.2.3 Analytical framework for the study 25

3.3 DATA SOURCES 27

3.4 ESTIMATION METHOD 27

CHAPTER IV: RESEARCH RESULTS 30

4.1 PRELIMINARY STATISTICS 30

4.2 REGRESSION RESULTS FOR EWS 33

4.2.1 Individual regression 33

4.2.2 Backward stepwise regression 35

4.2.3 Multinomial logit regression and final specification 35

4.2.4 Result discussion 39

4.2 CHAPTER REMARKS 44

CHAPTER V: CONCLUSION 46

5.1 CONCLUSION 46

5.2 POLICY IMPLICATIONS 48

5.3 RESEARCH LIMITATION 49

5.4 FURTHER RESEARCH RECOMMENDATION 50

REFERENCE 51

APPENDICES 54

LIST OF ABBREVIATIONS

CPIA Country Policy and Institutional Assessment

EWS Early Warning System

FDI Foreign Direct Investment

GDP Gross Domestic Product

IMF International Monetary Fund

M2 Broad Money Supply

NEER Nominal effective exchange rate

REER Real effective exchange rate

U.S United States

WDI World Bank Development Indicators

WEO World Economic Outlook Database

WGI Worldwide Governance Indicators

LIST OF TABLES

Table 3.1: Regime definition in Multinomial logit model 19

Table 3.2: Countries and sovereign debt crisis episodes during 1980-2010 21

Table 3.3: Analytical framework 26

Table 4.1: Summary on crisis occurrence during 1981-2010 30

Table 4.2: Average of explanatory variables 31

Table 4.3: 17 Variables passing first step 34

Table 4.4: Eight variables passing second step 35

Table 4.5: Correlation matrix of benchmark model 36

Table 4.6: Multinomial logit regression result 37

Table 4.7: Relative risk ratio of benchmark model 38

Table 4.8: Average marginal effect of explanatory variables 39

Table A1: Summary statistics of 27 explanatory variables 54

Table A2: Binominal logit variable-by-variable regression result 55

Table A3: Backward Stepwise regression result 56

Table A4: Benchmark model regression result – Full sample 57

Table A5: Benchmark model regression result – Restricted sample 58

CHAPTER I: INTRODUCTION

1.1 RESEARCH STATEMENT

The world has witnessed a numerous debt crises arising in recent years, for instance the 1997-1998 crisis in Asia region, the outright default of Argentina in 2001, the external crisis of Honduras prolongs from 1981 to 2010, and more recently, the crisis Eurozone such

as Greece, Spain, Ireland and Portugal since the year of 2009 The crises had disruptive influence on the performance of the in-trouble countries and even, the regional or global economy The frequent occurrence of debt crises and crisis’s serious consequences challenges the governments, especially developing countries where appear to be much more vulnerable compared to advanced economies It, in turn, has stimulated a controversy among various researchers on identification of leading indicators of debt crises, Kaminsky and Reinhart (1999), Manasse et al (2003) and Kraay and Nehru (2006) for example Those empirical studies attempt to construct an effective empirical model which helps to identify factors driving to debt servicing difficulties as well as predict the eve of debt crisis, the so-called Early warning systems However, due to unavailability of data on sovereign debt, only little work on establishment of Early warning system for debt crisis has been done compared to a wide range of researches on currency crisis and banking crisis

In addition, in almost previous empirical studies on establishment of Early warning system for debt crisis, binomial probit/ logit regression is highly preferred for example Manasse et

al (2003), Kraay and Nehru (2006) and Fuertes et al (2007) However, the result of binary logit/ probit regression result may be impacted by “post-crisis bias” This term, originally proposed by Bussiere and Fratzscher (2006), describes the bias caused by disregarding the difference in economics factors’ behaviors between “tranquil” state, the run-up to the crisis, and “recovery” state – the adjustment period after crisis Thus, discovery of more effective method tackling this kind of bias is open for researchers

Taken together above concerns, this study is carried out to answer the question which factors accounting for debt crisis in developing countries by applying more effective method – multinomial logit model – compared to empirical studies Specifically, this study constructs an Early Warning System for explaining sovereign debt crisis in 31 developing countries through the period 1981-2010 The results of the study will make a valuable contribution to the empirical literature on debt crisis determination, especially for

developing countries group By using multinomial logistic model which was proposed by the work of Bussiere and Fratzscher (2006) on predicting the onset of currency crisis in 20 open emerging markets during 1993-2001, the issue of post-crisis bias has been resolve As

a result, the study could work better at determining predictors for sovereign debt crisis in the developing countries More important, this study recommends several policies/ measures for the countries to prevent the occurrence/ prolongation of sovereign debt crisis based on the result analysis

principal objectives

Firstly, to determine an early warning system explaining for sovereign debt crisis in 31 developing countries whose data is available during the period from 1981 to 2010 at one year precedence

Secondly, to be made some recommendations for policy makers preventing the countries from debt service difficulties

The thesis is structured with five major chapters as follows

Chapter I introduces the research statement, research objectives and research questions

Chapter II reviews theoretical literature and empirical literature about heterogeneity of sovereign debt crisis definitions among studies and the nexus between sovereign debt crisis and potential explanatory variables which is useful for design of early warning model in this study

Chapter II provides the research methodology consist of model specification, response variables and explanatory variables and measurements together with research hypotheses, data sources and estimation method employed in the study

Chapter IV presents preliminary statistic of all variables and empirical regression results The result discussion in line with research’s questions is also provided in this chapter

Finally, Chapter V summarizes the main findings of the study as well as policy implications Besides, this chapter presents major concerns on the research limitation and proposal for further research

CHAPTER II: LITERATURE REVIEW

This chapter consists of your main sections The first section provides various definition of sovereign debt crisis among previous empirical studies as well as introduces the Early warning system The second part reviews theoretical literature on sovereign debt default while the third section review discusses empirical literature on sovereign debt crisis The last section presents the major approach of this study including sovereign debt crisis definition, early warning system method and the backbone of model explaining debt crisis

which was learnt from the relevant literature

2.1 SOVEREIGN DEBT CRISIS AND EARLY WARNING SYSTEM

2.1.1 Sovereign debt crisis

In general, sovereign debt crisis is term describing the event of debt servicing difficulties of

a country However, the very first notion is that there is no consensus on the exact

definition of debt crisis According to Sachs (1983), how the empirical studies define debt crisis depends on the specific research questions and also, the data availability

Many researches define debt crisis as sovereign defaults Rating agencies such as

Moody’s, Standard & Poor’s and Fitch’s often focus on default events For instance,

Standard and Poor’s defines default as “the failure of an obligor to meet a principal or interest payment on the due date (or within the specified grace period) contained in the original terms of the debt issue” (Chambers and Alexeeva, (2003)) including (i)outright defaults on external and/or domestic debt – in other words, scheduled debt services payment is not effected – for instance Argentina (2001) and Ecuador (1999) and (ii) rescheduling loan with new terms are less favorable to the creditors than the original loan

such as Ukraine (2000) and Uruguay (2003) Numerous empirical studies consider a country rated as “default” by those rating agencies is being in debt crisis status, Manasse et

al (2003) and Sy et al (2004) for example However, defining debt crisis as sovereign default solely is inadequate for the reason that sovereign default is not the only possible consequence of a serious debt servicing difficulties To illustrate, the crisis in Mexico in 1994-1995 was determined through a large scale financial aid package from IMF and turmoil of sovereign debt market instead of default

Secondly, debt crisis are considered as large arrears Detragiache and Spilimbergo

(2001) determine a country being in debt crisis if the country meets at least one of two

conditions (i) The arrears of principal or interest toward external commercial exceed five percent of total commercial debt outstanding (Indonesia in 1998 for example) (ii) There is

a rescheduling or debt restructuring agreement with commercial creditors which is recorded by World Bank’s Global Development Finance (for example Russia in 1998 and

Uruguay in 2001) The authors’ argument is that threshold of five percent helps to not miss the cases in which the share of defaulted debt is minor while the second criterion takes into account countries that are not technically in arrears as debt rescheduling or restructuring agreement have been arranged before actual default With the same approach but different threshold, De Paoli and Saporta (2009) define a country as in debt crisis when arrears exceed 15 percent of principal amount and five percent for the interest The authors also include the event of restructuring of a country’s external debt with its private creditors Defining debt crisis as a large arrears, however, do not capture the cases that some countries avoid an onset debt crisis by receiving a large official assistance from international financial institutions such as IMF (Manasse et al (2003))

Different from the above discussed approaches, there are studies using substantial IMF loans as an indicator of debt crisis Manasse et al (2003) argues that some countries

attempt to avert near default or rescheduling with its creditors through receiving large scaled financial support from official creditors and so, the authors count it into the debt crisis definition by considering if a country has non-concessional IMF loan exceeding 100 percent of its quota together with loan disbursement in the first year, the country is considered as being in crisis To illustrate, the case of Thailand in 1997-1998 was identified

by employing this definition In addition, the authors set the thresholds of 50 percent and

150 percent as a sensitivity analysis

Taking the turmoil of emerging bond markets into account, some other researchers define a

debt crisis as distress event More precisely, some studies consider the events that the

secondary market bond spreads are over a specific threshold in addition to sovereign defaults To illustrate, Sy et al (2004) determines that if a country faces sovereign bond distress which means “bond spreads are trading 1,000 basis points or more above the U.S Treasury securities” that country is being in debt crisis His argument is in reality, 1,000 basis points mark is considered as a “psychological barrier by market participants” Some

market participants consider securities reach the threshold when they have lost one-third of their value (Altman (1998)) Using the definition, the author identifies 140 debt distress episodes in 1994-2002

Finally, many papers combine the above proxies to construct comprehensive definition

of debt crisis such as Manasse et al (2003), Ciarlone and Trebeschi (2005) Ciarlone and

Trebeschi (2005) define a debt crisis occur when at least one of five following conditions

are met: (i) the country officially declares a moratorium on the payment of external and/or public debts, (ii) the country fails to pay principal and/ or interest of its external obligation

to creditors for an amount over five percent of the debt services ratio by end of the year, (iii) the country has accumulative arrears in interest and/ or principal to external creditors exceeding five percent of the debt services ratio by end of the year, (iv) the country arranged a rescheduling or restructuring agreement with its creditors and (v) the country receives non-concessional IMF loan exceeding 100 percent of its quota Under the sample

size of 28 emerging markets from 1980 to 2002, 44 debt crises with average length of seven years are identified

2.1.2 Early warning system (EWS)

In general and simplest words, EWS is the provision of timely and effective information about potential risk (“forecast” aspect) and then, allow the information users to take appropriate and effective action to mitigate or eliminate the occurrence and impacts of threats or undesirable events (“response” aspect) EWS comes from the basic idea that the

earlier and more accurate people are able to foresee the risk, the more likely people can manage and reduce negative effect of those unextend events Thus, EWS is widely used in many fields from predicting natural hazards, disease outbreaks and poverty to financial crash

In finance field, researchers often apply EWS to determine what factors drive to the onset of financial crisis For instance, Kaminsky and Reinhart (2000) build an EWS model to

predict the arrival of currency crisis at 24 months and 18 months preceding the crisis in emerging markets while banking crisis is forecasted with 12 month window For debt crisis

or debt default events, all the empirical studies predict the onset of debt crisis/ default at one year preceding, Detragiache and Spilimbergo (2001), Manasse et al (2003) and Sy et al (2004) are good examples

2.2 THEORETICAL LITERATURE ON DEBT DEFAULT

In the theory, sovereign default is the full or partial failure or refusal of a country to fulfill its debt obligations Debt obligations herein consist of principal and interest payment

In general, a country can experience periods of relative low and high income It could borrow in low income period with commitment to repay debt in period of high yield level Normally, the country honors obligation to repay its lenders because it expects that in future, it could possibly encounter undesirable periods in which its demand for borrow from other countries occurs If the country has a “bad” repayment history, it is obviously difficult for it to be granted new loan by the lenders besides other costs mentioned below Hence, the borrowing country has no incentive to default, except when it cannot repay This situation

signifies the so called “ability-to-pay” approach which was typically analyzed by Krugman

(1988) and hereinafter discussed

In reality, in some cases, the borrowing country choses to repudiate a debt obligation (although debt servicing is technically feasible) for few reasons: the borrower retains resources for other goals such as education, health, national security and defense and, accepts cost of default The costs associated with default are problematic in term of economics and diplomatic In first place, the cost is the debtor losses access to credit market after default (Eaton and Gersovit (1981), Sachs (1983)) Besides, given decision to default, the debtor may experience reduction in output which is caused by creditors’ retaliation measures through embargoing trade, seizing or freezing overseas assets of debtors and stopping aid for instance In spite of high cost, the country still decides to repudiate debt

based on optimization calculus It is so called “willingness-to-pay” approach which is first modeled by Eaton and Gersovit (1981) and hereinafter discussed

Using “ability-to-pay” approach, Krugman (1988) constructed a theoretical model that

underlines the importance of “debt overhang” in behavior of the lender at country level According to Krugman (1988), “a country has a debt overhang problem when the expected present value of potential future resource transfer is less than its debt” Consequently, the

country is unable to service existing debt fully without new borrowing

1

Debt overhang problem was first discussed by Myers (1977) The author investigates the determinants of borrowing at firm level

To illustrate the implication of debt overhang, considering a developing country has

inherited debt stock D Assuming that all debts mature in the first of two periods

Let denote x 1 , x 2 is potential resource transfer in period 1, 2 Thus, the country could repay

with maximum amount equal x 1, x 2 in each period

Considering the situation if D exceeds x 1 In that case, the debtor must engage in new loan with the amount (D - x 1) to avert immediate default However, the ability to attract new borrowing to service outstanding debt is contingent on the lender’s behavior

Supposing that the lender is risk neutral and has opportunity cost of funds i on financial

market

Model without uncertainty, x 1 , x 2 are known The lender is willing to grant new loan

equaling (D - x 1) (instead of experiencing immediate default) if this loan can be fully repaid

On the contrary, if x 1 + x 2 /(1+i) < D, the debtor is unable to service the debt It cannot borrow the additional (D – x 1 ) as the creditor’s loss is inescapable

Model with uncertainty, potential resource transfer in second period is unknown for the

reason that in fact, the world economic environment is uncertain or how well the country itself performs is unforeseen For simplification, the research assumes that:

x 1 is known

x 2 takes only two values: x B if the present value of potential resource transfer is

smaller than the inherited debt (bad outcome) and x G for the otherwise (good outcome)

Denote p = Pr(x 2 =x G ): probability of a good outcome where the present value of potential

resources transfer could cover the inherited debt

Thus Pr(x 2 =x B ) = (1 – p): probability of bad outcome

The creditor will grant new loan (D - x 1 ) at the interest rate r Thus, r is determined as

below

(D – x 1 )(1+r) =x G

r determined by (*) is the highest level of interest rate that the debtor can pay and thus the

creditor can be offered

Expected value of creditor’s earning is [px G + (1-p) x B ]/(1+i)

If the debtor could pay interest for additional borrowing, the creditor will lend their debtor

new loan (D - x1) if [px G + (1-p) x B ]/(1+i) > D – x 1 (2)

If (2) is satisfied, the country could borrow for debt repayment by paying an adequately high interest premium Otherwise, the country is unable to receive desired additional loan and so, it is insolvent In other words, default inevitably occurs if lender does not take any further action

Hence, the best solution for default avoidance is that the creditor immediately decreases the borrowing country’s obligation by forgiving any part of the loan (principal, interest), rescheduling, granting new concessional loan or a combining them By this, the debtor has incentive to use its potential resource in next period to service its debt

In brief, what relates to debt default could be learnt from Krugman’s work?

First, debt default is highly associated with borrower facing illiquidity, insolvency or both Second, quasi default might be averted by lender’s forgiving, rescheduling or additional lending It means that not only default, but the lender’s actions also capture debt servicing

difficulty of a borrowing country as well

2.2.2 Model of debt repudiation

Eaton and Gersovit (1981) introduced a theoretical analysis of international borrowing of a

sovereign country and potential debt repudiation by applying “willingness-to-pay”

approach

Specifically, the model considers the situation that a mall open economy raises capital by

borrowing b amount with the purpose of smoothing consumption

National income of the country is function of capital stock k and labor capital l:

y (b,t) = θF(k t , l t ) = θF[K(b t ), l t] (3)

Country’s spending c(the sum of consumption and investment) is calculated as follows

c t = y t + b t –p t where p t represents the debt service payment in the period t

If a country fails to repay, it has no opportunity to borrow from other countries Hence, the

borrower repays either p t = 0 or p t = d t where d t is the country’s debt services obligations in

period t (d t = R(b t-1 ) assuming that debt matures in one period)

The country decides whether to repay in order to maximize utility:

V = Max{V R , V D}

Where:

V R = U(y t + b t – d t ) + 𝛽E(V t+1 ) : utility value if the country repays

V D = U(y t + b t ) +𝛽E[U(y t+1 - P D ] : utility value if the country defaults

Default is optimal if and only if V D exceeds V R

Let denote the likelihood of default as π(b,t)

π(b,t) = Pr (V D > V R)

= Pr {U(y t + b t ) + 𝛽E[U(y t+1 )- P D ]> U(y t + b t – d t ) + 𝛽E(V t+1 )} (4)

By substituting θF[K(b t ), l t ] for y t in equation (3), (4) provides a strong evidence that the debt size affects the probability of default

In addition, an important theorem is proved and supplied by the authors It is “the probability of default (π t ) increases monotonically with debt service obligation (d t )”

The major conclusions were made through the model of repudiation are as follows

In the first place, sovereign default probability depends on the size of foreign debt

Secondly, the debt size – the amount a country borrows – is the minimum of the two figures – the desiring borrowing of the debtor (which represents quantity demanded) and the credit ceiling established by creditors (which implies quantity supplied)

Thirdly, both desiring borrowing and credit ceiling are positively influenced by income variance and, an increase in income growth rate leads to an increase in desiring borrowing amount while its effect on credit ceiling is ambiguous Other important property of the theoretical model is that when exogenous penalty rises, it will create an increase in the amount a country can borrow

In conclusion, the two theoretical models propose that probability of sovereign debt default

is impacted by (i) solvency and liquidity measures such as debt services and external debt stock, (ii) income variance and income growth rate Besides, for the reason that lender’s forgiving, rescheduling or additional lending could help borrowing country to avoid default (Eaton and Gersovit (1981)), if a country receives forgiveness, reschedule or new loan from its lender, the country should be considered as being in default

2.3 EMPIRICAL LITERATURE ON SOVEREIGN DEBT CRISIS

Compared to limited theoretical literature, there are a great number of empirical studies on determining which factors trigger sovereign debt crisis (which are to be discussed hereinafter) Furthermore, the empirical literature is closest to early warning system approach Thus, those studies play an important role in literature about early warning

system for sovereign debt crisis

As discussed in part 2.1 of this Chapter, the definition of debt crisis, however, varies among different researches With the aim to build a model to predict the onset of debt crisis at one year preceding, the section focuses on the possible linkage between debt crisis and explanatory variables pointed out by previous studies regardless of how debt crisis is defined in those studies

2.3.1 Sovereign debt crisis and liquidity measures

The term “liquidity” refers to a country’s ability-to-pay its short-term obligations If a country is illiquid, it has no sufficient cash to fulfill its on-due obligations even if the country is able to repay debt in future with future income In other words, illiquidity is temporary inability to pay debt Thus, the more a country illiquid, the more it faces debt servicing difficulties

The link between liquidity measures and debt crisis is strongly supported by a number of empirical studies such as Detragiache and Spilimbergo (2001), Ciarlone et al (2005), Sy et

al (2007) and McFaden et al (1985)

Detragiache and Spilimbergo (2001) investigated the correlation between debt crisis and external liquidity using the sample of 69 countries in 29 years from 1970 to 1998 The authors provided empirical evidence of more illiquidity is highly associated with debt crisis,

in which liquidity is indicated by short term debt, debt coming due and foreign exchange reserves McFaden et al (1985) examined probability of debt crisis in 93 countries during

period 1970-1982 and also found that the measure of liquidity – non-gold reserves – was a good indicator of debt crisis Other liquidity measures proved to be effective predictors of debt crisis, for example ratio of short term external debt to reserves, services of the external debt over reserves (Manasse et al (2003)), short term debt over total external debt (Ciarlone

et al (2005)) and short term debt in percent of GDP or of reserves (Sy et al (2004)) However, from the research result of Manasse et al (2003), the indicator M2 to reserves ratio, somehow, did not affect debt crisis in their sample of 47 countries having market access for the period 1970-2002

To sum up, the less liquid a country is, the higher probability of debt crisis the country faces This correlation is consistent with theory of debt overhang discussed above

2.3.2 Sovereign debt crisis and solvency measures

In addition to liquidity, solvency is an aspect of a country’s financial health Nevertheless, there is notable difference between two terms Whilst liquidity refers the ability to pay short term debt, solvency describes how well a country meets its long-term commitments Specifically, a country is solvent if the present value of future income is at least equal to its current debt stock In other words, a solvent borrower is able to pay debt with future income Likewise, an insolvent borrower is unable to repay its debt at maturity given future income Hence, measures of solvency are included into specification model and proved to

be a significant determinant of debt crisis For instance, Detragiache et al (2001) measured solvency by total debt in percent of GDP and for the sample of all countries which data is available in 1970-1998 (69 countries), the authors pointed out the higher this ratio, the more the crisis occurs The ratio again was proved to be positively correlated with debt crisis by

Sy et al (2007) Kraay and Nehru (2004) measured solvency by the ratio of current debt over export and found that the solvency measure triggered crisis in 98 low and middle income countries Other measure of solvency as total external debt as percent of GDP and

of reserves by Manasse et al (2003) and Fioramanti (2006) respectively with different dataset also has the positively impact on the occurrence of debt crisis event

To summarize, the empirical studies show that the probability of debt crisis is highly associate with insolvency and this conclusion is consistent with above mentioned theoretical literature

2.3.3 Sovereign debt crisis and macroeconomics fundamentals

Besides measures of liquidity and measures of solvency are key determinants of debt crisis, macroeconomic conditions also are proved to have significant influence on likelihood of entering into debt crisis

In more details, Kraay and Nehru (2006) found that real GDP growth, GDP per capita level and inflation have significant effect on debt distress Ciarlone and Trebeshi (2005) examined the situation in 28 emerging economies and found that macro variables proxied

by GDP growth, total private capital flows over GDP and annual inflation rate are significant determinants In addition, Detragiache and Spilimbergo (2001) discovered the negative correlation between crisis likelihood and foreign exchange reserves which reflects country’s resources for debt repayment The explanation for the above research result is that macroeconomic factors likely affect the ability to service debt obligation, especially for obligations with external creditors and thus, influence on the likelihood of being in debt crisis Higher level of GDP, foreign exchange reserves or economic growth indicates that the country has more resources for debt repayment and so, it less likely faces debt servicing difficulties Whilst inflation is positively connected with debt crisis as inflation deteriorates country’s income, ability to pay

Besides, exchange rate overvaluation has a positive linkage with debt crisis (Peter (2002) and Manasse et al (2005)) for some reasons First, overvaluation makes debt burden of foreign currency loan become more serious Second, overvalued exchange rate harms future export performance And thirdly, accordingly to Reinhart (2002), a debt crisis is preceded

by a currency crisis and change rate overvaluation is constantly significant determinants of currency crisis (Kaminsky et al (1999) and Reinhart (2002))

In summary, macroeconomic conditions measured by a wide range of indicators have a significant effect on probability of sovereign debt crisis

2.3.4 Sovereign debt crisis and external trade link

Empirical studies of Detragiache et al (2001) and Manasse et al (2003) supported the significant impact of current account balance on the likelihood of being in debt crisis Specifically, high current account balance help to reduce debt crisis probability The first reason is that, current account surplus reflects the development of a country in the relation with the external and the surplus means the country has more income from rest of the world

to repay debt In addition, current account could affect debt crisis through its impact on currency crisis In other words, current account deficit triggers currency crisis (Kaminsky et

al (1999) and Reinhart (2002)) and in turn, currency crisis is associated with debt crisis in the following year (Reinhart (2002))

The openness of an economy – measured as the sum of exports and imports over GDP ratio – is proved to be a significant determinant of debt crisis in several papers Detragiache et al (2001), Manasse et al (2003) and Sy (2007) for instance However, the sign of the effect is not homogeneous among studies More open helps a country to get “better position to service a large external debt through future export performance” (Detragiache et al (2001)),

By contrast, openness could influence on the cost of default – creditor’s trade sanction measure levied on defaulted debtor – thus, borrowing country’s willingness to pay

Terms of trade also affect debt distress in lower and middle income country, according to Kraay and Nehru (2004) The authors argued that adverse term of trade might cause lower level of export revenue, hurt debt repayment capacity, and hence, trigger debt crisis event

To brief, the trade position of a country measured by current account balance, trade openness and terms of trade for instance can determine whether a country enters into sovereign debt crisis or not While current account balance and terms of trade are positively correlated with debt crisis, the sign of trade openness’s impact is heterogeneous among researches

2.3.5 Sovereign debt crisis and political institutions

The very first works on examining the linkage between political institutions and debt default were conducted by McFadden et al (1985) and Aylward and Torne (1998) and Reinhart et al (2003) They simply used non-repayment history as a proxy of political institutions and concluded that poor debt repayment history indicates a high likelihood of debt servicing difficulties in the future The relationship could be explained by the role of banking and financial system In country with “bad” repayment history, a weak banking system lowers penalty to default Lower costs of disruption lead the country to default at lower threshold, further weakening its financial system, and prolong the time of debt servicing difficulty

Secondly, the inverse correlation between democracy and debt default is supported by Balkan (1992) and Kohlscheen (2005) The likelihood of default on external debt or

rescheduling is lower in country has higher level of democracy The reason is that compared to presidential system, parliamentary system has less constraint on executive Thus, democratic country might avoid “game of attrition and delay stabilization” (Drazen(1991)) and thus, lower tendency to default

Finally, Manasse et al (2003) find that the likelihood of debt crisis goes up in the year with presidential elections The association could be explained by the notion that in presidential elections reflect the political uncertainty and the latter, in turn, causes to instable macroeconomics – which is associated with debt crisis The relationship between political institution environment and debt crisis was affirmed again by Kraay and Nehru (2006) The authors used World Bank’s CPIA ratings as a measure of the quality of institutions and policies and found that CPIA index is adversely correlated with debt rescheduling In other words, higher quality of policies and institutions help to reduce likelihood of debt crisis Kraay and Nehru (2006) also emphasized that impact of this variable is much more significant in the lower income countries compared to the middle income countries

To summarize, the empirical studies highlighted the role of political institutions in determining debt crisis because it (i) reflects country’s policy credibility and captures the government’s willingness to pay and (ii) influence economic stability and so, ability to pay

2.3.6 Sovereign debt crisis and global liquidity

Empirical literature shows that not only country characteristics such as liquidity, solvency, macroeconomics condition and political institutions are significant determinants of debt crisis, but also the global factors namely global liquidity Manasse et al (2003) proxied the global condition by the U.S Treasury bill rate and found that this rate is individually high correlated with debt crisis event Specifically, the U.S Treasury bill rate increases in years preceding a crisis This relationship also emphasized by Sutton and Catão (2002) Besides, Haque et al (1996) found that international interest rate has adverse influence on country’s creditworthiness toward its creditors and thus, debt service difficulties The rationale is that high level of the advanced economies’ interest rate (the U.S Treasury bill rate, LIBOR for instance), implying an tight monetary condition in developed countries, might reduce the capital flow to developing countries and hence, make the debt servicing matter in those countries more serious

Reinhart and Rogoff (2003, 2010) define a country to be in debt crisis if the country in

external debt crisis or domestic debt crisis or both In which:

(i) External debt crisis is failure to meet a principal or interest payment on the due date (or within the specified grace period) External debt crisis includes outright default on payment

of external debt obligations (Argentina (2001) for instance) and repudiation (Mexico in 1867) This type of crisis also take into account reschedules debt with concessional terms compared to original contract (India in period 1958-1972)

(ii) Domestic debt crisis is defined in the same way as external debt crisis In addition, it includes the freezing of bank deposits and or forcible conversions of such deposits from foreign currency to local currency Argentina (1989-1990), Bolivia (1982-1984) and

Mexico (1994-1995) are the cases

By using this approach, the authors identify 231 external debt crisis episodes and 50 domestic debt crisis episodes of 70 developed and developing countries in the period1800-

2010

Manasse et al (2003, 2009) construct a comprehensive definition on sovereign debt crisis

by using quantitative approach Specifically, a country is defined as being in debt crisis if there is occurrence of at least one of below conditions

(i) The country is classified as being in default by Standard and Poor’s in which “sovereign

issuers in default when a government fails to meet principal or interest payment on an

external obligation on due date (including exchange offers, debt equity swaps, and buy back for cash)” (Standard & Poor’s (2002))

(ii) The country receives a large non-concessional IMF loan in excess of 100 percent of quota It is considered as a debt crisis indicator since the authors argued that a country can

prevent default by a substantial financial assistance from IMF

Manasse et al (2003, 2009) identify 54 debt crisis episodes in 47 market access countries for the period from 1970 onwards by applying the definition

Secondly, in term of Early warning system, this paper uses data of explanatory variables to predict the occurrence of sovereign debt crisis in next year In other words, this paper identifies determine factors driving to sovereign debt crisis at one year precedence

Thirdly, with regards to the backbone of empirical model, the theoretical and empirical literature proposes a set of major determinants of sovereign debt crisis as follows The below factors are to be taken into account in this study

 Debt solvency measures which reflect debt stock in relation with ability to fulfill debt obligations of country such as external debt, public debt to GDP or to exports

 Debt liquidity measures that assess borrowing country’s ability to honor its short term obligations, the ratio of short term debt, M2 or external debt services to foreign exchange reserves, exports or GDP for instance

 Macroeconomics factors such as GDP growth, inflation, reserves to GDP ratio

 External trade linkage indicators such as current account balance, imports ratio, exports ratio, trade openness and terms of trade

 Political and institutional factors which represented by presidential election event for example

 Global liquidity condition proxied by advanced economies’ interest rate such as LIBOR and the U.S Treasury bill rate

CHAPTER III: RESEARCH METHODOLOGY

Chapter III includes four major sections The first section proposes the empirical model specification The second part presents the model’s variables and measurements Moreover, analytical framework is provided in this section The next part describes the source of data

used in this study Final section explains in details the estimation method

3.1 MODEL SPECIFICATION

The multinomial logit model is applied in building the EWS model for explaining debt crisis in 31 developing countries over the period 1981-2010 Applying multivariate logistic regression to predict crisis is initially recommended by Bussiere and Fratzscher(2006) when investigating what drives to the onset of currency crisis of 20 open emerging markets during the period 1993-2001

According to the authors, compared to binomial logit model, multinomial logit model could improve forecast quality as it resolves the “post-crisis bias” Post-crisis bias is the bias occurs when no distinction is made between the “tranquil” state and “post-crisis” state In which, “tranquil” state refers to normal time prior to the occurrence of debt crisis when economic fundamentals are almost sustainable and the adverse fluctuation of the factors might trigger crisis “Post-crisis” refers to the crisis years other than the first year of the crisis and thus, the term captures the adjustment period after falling into crisis when a country might attempt to bring economic variables back to sustainable and exit from crisis Because right after the appearance of the crisis, economic variables do not immediately recover from crisis and reach a “tranquil” status but must go through an adjustment process before converging to new equilibrium status, it is important and necessary to distinguish clearly between “tranquil” and “post-crisis” period For the two reasons that (i) the two-stage logit model only compares the non-crisis observations with crisis observations In other words, logit model with two outcome combines of first year in crisis and post-crisis observations into crisis observations and so, the difference in behavior of economic variables in non-crisis state and post-crisis state cannot be observed and (ii) behavior of economics fundamentals is very different among tranquil, crisis and post-crisis states, significant bias will arise in binomial EWS model Multinomial EWS model tackles issue

of post-crisis bias by defining more than two outcomes for the whole sample

In more details, in binomial logit has only two outcomes “crisis” and “non-crisis” while in three outcome model, “crisis” is slit into first year in crisis (crisis state) and remaining years

of the crisis (post-crisis state) and “tranquil” state represented by “non-crisis” of binomial logit In remaining years of crisis or “post-crisis”/ recovery state in other words, the country

is still in crisis, however, behavior of the economy is significant different from the first year

in crisis as the government will make a great effort to exit from crisis

Specifically, in order to predict debt crises at one year precedence, multinomial logit EWS model is constructed with three regimes including:

i crisis state for the first year of crisis when crisis does not occur in preceding year

ii post-crisis/ recovery state for the remaining years of the crisis other than the first

year in crisis when crisis occurs in both observed year and previous year The case

non-crisis in year t but crisis occurs in both preceding year and following year, year t

observation is also counted as in “post-crisis”

iii tranquil state for otherwise, when crisis does not occur in observed year and crisis

event does not simultaneously appear in preceding year and following year

Table 3.1 presents how to identify debt status of a country in multinomial logit model

Table 3.1: Regime definition in Multinomial logit model

“non-crisis”

The above three-regime classification has two main strong points Firstly, it distinguishes between “tranquil” state and “post-crisis” state which help to solve post-crisis bias Furthermore, the splitting years-in-crisis into crisis and post-crisis/ recovery state allows to look closely into behavior of economic fundamentals in two period and helps to identify the factors driving to the crisis entry as well as determining the factors lead to remaining in debt crisis (or in other words, ability to escape from crisis)

There are three scenarios for an economy: tranquil period (j=0), crisis (j=1), or post-crisis/ recovery (j=2) The probability that an economy is in state j is given by below equations

3.2 VARIABLES AND MEASUREMENTS

3.2.1 Sovereign debt crisis

As discussed in the fourth part of Chapter II, the definition of sovereign debt crisis follows the definition of Reinhart and Rogoff (2003, 2010) and Manasse et al (2003, 2009) with the

aim to construct a comprehensive definition under the constraint of data In other words, a country is defined being in debt crisis when at least one of the two mentioned primary sources indicate so

Aggregating the dataset of Reinhart and Rogoff (2003, 2010) and Manasse et al (2003, 2009), 49 debt crisis episodes for 31 developing countries in 1980-2010 have been identified and listed in Table 3.2

Table 3.2: Countries and sovereign debt crisis episodes during 1980-2010

31 Zambia 1983-1994

3.2.2 Explanatory variables

Based on theoretical and empirical literature, this EWS model for explaining debt crisis in the developing countries is constructed with six explanatory variables which are divided into two groups: country characteristics and exogenous factor How explanatory variables are proxied depending on the availability of data Potential factors and expected sign of their influence are hereinafter described

3.2.2.1 Country characteristics

Solvency measure is the key determinant of debt crisis When a country faces solvency

issue, its future income is insufficient to honor its obligations towards creditors which is precisely debt default The relationship was emphasized in the theoretical literature of Krugman (1988) Numerous of empirical studies such as Detragiache et al (2001), Kraay and Nehru (2004) and Detragiache et al (2001) proved the role of solvency measures in reality: the less a country is solvent, the higher the probability of debt crisis is The paper uses several indicators for debt solvency: public debt to GDP ratio, external debt to exports ratio, external debt divided by GDP and total debt service to exports ratio These indicators are positively related with debt crisis as the high level of the ratios implies insolvency The following hypothesis will be tested to answer the question whether solvency is a factor driving to sovereign debt crisis

H1: The solvency measures are positively correlated with sovereign debt crisis

Liquidity measure, similar to debt solvency, is the principal factor driving to debt

servicing difficulty In this paper, debt liquidity is proxied by ratio of short term debt over reserves, M2 over reserves, debt services on external debt divided by GDP, debt services on external public debt to GDP ratio, short term debt as percent of total external debt and total reserves over to total external debt In which, ratio of total reserves to total external debt lower the probability of debt crisis because the higher the ratios indicate that the countries have sufficient cash or other liquid asset to repay its short term debt obligations Conversely, the others (except for M2/reserves ratio) are expected to have positive influence on the occurrence of debt crisis because the high burden of short term obligation toward creditors leads the countries to debt servicing difficulty The impact of the liquidity

measures is supported by the work of Kraay and Nehru (2004), Kraay and Nehru (2004) andFioramanti (2006) M2/reserves ratio reflects banking system’s ability in converting local currency to foreign as individuals’ demand for foreign currency increase in the existence of currency crisis Therefore, agreeing with empirical study of Lausev et al (2011), this ratio is expected to be positively correlated with debt service difficulty

The below hypothesis will be tested to provide the empirical evidence about the impact of liquidity measures on sovereign debt crisis

H2: The liquidity measures are positively associated with sovereign debt crisis

Macroeconomic fundamentals are controlling factors in the EWS model and measures by

several indicators

Firstly, macroeconomics environment is proxied by GDP per capita growth and inflation rate – the two typical indicators Growth rate helps to reduce debt crisis because it implies the country have more resources to pay debt By contrast, the higher inflation would decrease the real income that a country can utilize for repayment as well as discourage savings and encourage borrowing Thus, inflation rate has positively affect debt crisis probability The expectation is in compliant with the conclusion of Kraay and Nehru (2006) and Ciarlone and Trebeshi (2005)

Foreign exchange reserves growth is expected to lower debt crisis occurrence for the reason that reserves is the resource for debt repayment and if reserves increase over time, it guarantees for better capacity to pay The role of reserves in EWS model is supported by empirical studies of Detragiache and Spilimbergo (2001)

Furthermore, country’s expenditure and savings may affect debt crisis Savings precisely the sources for (i) debt repayment and (ii) for investment which is in turn generate more income in future and thus, gross domestic income to GDP can reduce the risk of debt crisis

In the contrary, expenditure proxied by gross government expenditure and gross national expenditure could harm the debt sustainability though increasing public debt and deteriorating income for investment The linkage between country’s expenditure and savings with its external debt service difficulty is pointed out by Hemming and Chalk (2000) and Lausev et al (2011)

In this paper, FDI net inflow to GDP ratio is also used as controlling factor for macroeconomics The ratio is expected to have negative correlation with debt crisis event as

for developing countries, the FDI inflow supports the economic development and so, capacity to pay of those countries This argument is empirically proved by Manasse et al (2003)

Other important indicator for macroeconomics condition is the exchange rate depreciation Weak local currency to the outside hurts the country’s exports performance Moreover, it makes debt burden of foreign currency loan become more serious Thus, depreciation in exchange rate worsens ability to pay and so, increase propensity of debt crisis The role of exchange rate is emphasized by several papers such as Reinhart (2002).In this study, exchange rate depreciation is represented by the growth of nominal effective exchange rate and real effective exchange rate (NEER and REER respectively) Thus, the higher the growth rates, the more serious debt servicing difficulty

To reach the research objectives of building an EWS model for explaining sovereign debt crisis, following hypothesis will be tested in terms of macroeconomic fundamentals’’ role

H3: The macroeconomic fundamentals significantly affect sovereign debt crisis

External trade link is highly related with debt crisis as empirical literature In case, current

account balance to GDP ratio, exports to GDP ratio and Term of Trade are used as a measure of commercial linkage with rest of the word, the relationship is negative (Detragiache et al (2001), Manasse et al (2003) and Kraay and Nehru (2004)) The rationale is current account surplus, high exports ratio and high Term of trade imply high income level and so, the country is less likely face difficulty in debt servicing In case trade link proxied by imports to GDP ratio, high level of this ratio in developing countries might decrease real income of those countries and so, the paper expects this ratio is positively associated with debt crisis event In addition, trade openness measured by the sum of exports and imports divided by GDP is expected to be a significant variable However, expected sign is indefinite for the reason that openness can help the country get more export revenues and make the country is more vulnerable at the same time

Regarding the influence of country’s commercial linkage with rest of the world on the sovereign debt crisis, it is necessary for the study to test the hypothesis as follows

H4: The external trade link is highly associated with sovereign debt crisis

Political institution factor was presented by the dummy for the year with election event as

recommended by the work of Manasse et al (2003) In which, the authors use dummy for

presidential election as measure of political institution factor and found that in the year with presidential elections, the likelihood of entering into and being in debt crisis is higher

Due to unavailable data, this paper uses a slight different proxy for political institution variable Specifically, dummy takes value of 1 for the year has legislature/ executive election, 0 for the otherwise In the year a country experiencing a legislature and/ or executive election, its economy might be unstable and so, harm the ability to pay Thus, it is expected that in those years, the probability of debt crisis is higher

To assess the significant role of political institution factor in EWS model for sovereign debt crisis, this study is to test following hypothesis

H5: The political institution is significant correlated with sovereign debt crisis 3.2.2.2 Exogenous factor

The empirical studies (Sutton and Catão (2002) for example) point out that international liquidity, represented by advanced economy’s interest rate, is a determinant of debt crisis in

developing countries This analysis here, following the empirical literature, uses month The U.S Treasury bill rate as a proxy and supposes that this rate positively impacts

three-on debt crisis The argument is high The U.S Treasury bill rate implies (i) tight international liquidity and (ii) high opportunity cost of lending and so, weaken the capital flow to developing countries, further trigger difficulty in debt servicing in those borrowing countries The positive nexus between the U.S Treasury bill rate and debt crisis is supported

by Manasse et al (2003)

In general, to identify whether the effect of international liquidity on sovereign debt crisis is significant, the study is to test the hypothesis as below

H6: The international liquidity is highly associated with sovereign debt crisis

3.2.3 Analytical framework for the study

<Insert Table 3.3>