Citation
Aspects of the external debt of developing countries

Material Information

Title:
Aspects of the external debt of developing countries
Creator:
Savvides, Andreas, 1957-
Publication Date:
Language:
English
Physical Description:
vi, 181 leaves : ; 28 cm.

Subjects

Subjects / Keywords:
Capital inflows ( jstor )
Commercial banks ( jstor )
Debt ( jstor )
Developing countries ( jstor )
Economic models ( jstor )
Interest rates ( jstor )
International loans ( jstor )
Library users ( jstor )
Loans ( jstor )
Mathematical variables ( jstor )
Debts, External -- Mathematical models -- Developing countries ( lcsh )
Dissertations, Academic -- Economics -- UF
Economics thesis Ph. D
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1983.
Bibliography:
Includes bibliographical references (leaves 174-180).
Additional Physical Form:
Also available online.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Andreas Savvides.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Andreas Savvides. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
11795501 ( OCLC )
0030528860 ( ALEPH )

Downloads

This item has the following downloads:


Full Text









ASPECTS OF THE EXTERNAL
DEBT OF DEVELOPING COUNTRIES








By

ANDREAS SAVVIDES


















A DISSERTATION PRESENTED TO THE
GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE DEGREE OF DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA

1983













ACKNOWLEDGMENTS


I would like to thank the chairman of my supervisory committee, Professor G. S. Maddala, for his advice and continual guidance and encouragement throughout the preparation of this dissertation. The other members of my committee, Professors W. Bomberger, R. Crum and D. Denslow, deserve particular recognition for providing freely their assistance and for their efforts in ensuring that the task was carried through to its completion. In the preparation of such a study, the contribution of colleagues and fellow students cannot be overestimated.

I would like to thank, in particular, my colleague and friend, Haralambos Sourbis, for finding the time to discuss problems and projects. The typing and editorial assistance of Betty Sarra has been outstanding throughout. Finally, I would like to thank my parents and dedicate this work to them. Their contribution to its completion is without measure.



















ii















TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ................................................... ii

ABSTRACT......................... ................................ v

CHAPTER

ONE DEVELOPING COUNTRY BORROWING IN THE INTERNATIONAL CAPITAL
MARKETS: AN OVERVIEW OF THE ISSUES..................... 1

Introduction .............................................. 1
Some Preliminary Comments and Definitions.................. 3
Distribution and Value of the External Debt of Developing
Countries ............................................ 9
The Distribution of Developing Country Debt............ 9
Real and Nominal Value of LDC Debt ..................... 13
Constraints on LDC Borrowing and the Role of International Reserves...................................... 18
Is a Credit Ceiling Imposed on LDC Borrowing? .......... 18 External Debt and International Reserves............... 21
Analyzing Country Risk: Debt Rescheduling and the Debt
Capacity of LDCs....................................... 25
Growth-cum-Indebtedness Models ......................... 26
Estimating the Probability of Rescheduling............. 29
Real Interest Rates and Commercial Bank Involvement in
NOLDC Lending.... ...................................... 44
The NOLDC Lending Process from the Commercial Banks'
Perspective......................................... 44
Movements in Real Interest Rates ....................... 54
External Borrowing, Consumption and Investment............ 59

TWO BANK LOAN RATE INDEXATION IN THE EUROCURRENCY MARKET...... 69

Introduction............................................. 69
The Model........ ...................................... 70
Theoretical Implications of the Model .................... 78
Empirical Results ..... ................................... 84
Summary.................................................. 90

THREE A DISEQUILIBRIUM MODEL OF EUROCURRENCY FINANCE TO
DEVELOPING COUNTRIES....................................... 92

Introduction..... ...................................... 92
The Rationale Behind Credit Ceilings in International
Banking................................................ 94


iii









Theoretical Arguments....... .......................... 94
Credit Ceilings in International Banking Practice....... 98
A Model of Credit Ceilings in International Lending to
Developing Countries.................................... 100
The Basic Framework.................................... 100
Elaborating on the Basic Framework...................... 105

FOUR ESTIMATION OF A DISEQUILIBRIUM MODEL OF EUROCURRENCY
FINANCE TO DEVELOPING COUNTRIES ........................ 109

Introduction............................................ 109
Disequilibrium Econometric Techniques: A Brief Methodological Review .. .................................... 110
Estimation of a Disequilibrium Model with Credit Ceiling... 112 Variables and Data Sources ................................ 116
The Maximum Likelihood Estimates ........................... 119
The Computer Program................................... 119
The Empirical Results.................................. 119
Conclusion..... ............................................ 123

FIVE A SIMULTANEOUS LOGIT MODEL FOR THE DETERMINATION OF
DEVELOPING-COUNTRY CREDITWORTHINESS AND CAPITAL INFLOWS.. 125

Introduction.............................................. 125
A Simple Equilibrium Model of the Supply and Demand for
External Capital..................................... 126
The Importance of Economic Indicators of Single-Equation
Models of Creditworthiness............................ 131
A Simultaneous Logit Model of Creditworthiness and Capital
Inflows................................................. 139
Conclusion................................................ 149

SIX SUMMARY AND CONCLUSIONS................................. 151

APPENDICES

A EFFECT OF CHANGES IN INTERNATIONAL RESERVES ON THE
COMPARATIVE STATICS OF THE LOAN RATE INDEXATION MODEL.... 153

B IMPLICATIONS OF THE CORRELATION BETWEEN THE DEBT-EXPORTS
RATIO AND THE AMOUNT OF DISBURSEMENTS.................... 157

C FIRST DERIVATIVES OF THE LOGARITHM OF THE LIKELIHOOD
FUNCTION FOR THE DISEQUILIBRIUM MODEL.................... 163

D DATA FOR THE SIMULTANEOUS LOGIT MODEL...................... 168

REFERENCES........................................................ 174

BIOGRAPHICAL SKETCH............................................... 181






iv














Abstract of Dissertation Presented to the
Graduate Council of the University of Florida
in Partial Fulfillment of the Requirements
of the Degree of Doctor of Philosophy


ASPECTS OF THE EXTERNAL
DEBT OF DEVELOPING COUNTRIES

By

ANDREAS SAVVIDES

December 1983

Chairman: Professor G. S. Maddala Major Department: Economics

This dissertation comprises a collection of essays focusing on

several aspects of the external debt of developing countries. Initially, the many facets of this complex subject are presented and analyzed.

The widespread practice of loan rate indexation in the Eurocurrency market is the first issue of concern. A theoretical banking model is introduced and the comparative static results reveal that contries with a lower level of international reserves or exports (usually the less creditworthy), are less likely to be rationed when the cost of funds to the bank rises. The empirical results support the theoretical hypothesis. The model directs attention away from studies investigating developing-country creditworthiness on the basis of country-wide economic indicators, and approaches the creditworthiness concept from the perspective of international commercial banks granting loans to developing countries.

The market for international debt is subsequently investigated and a disequilibrium model of supply and demand for international debt is

v








estimated. There is ample evidence to suggest that an explicit credit constraint exists in developing country borrowing. If the constraint is ineffective, the equilibrium values of the variables are observed; otherwise, the observations correspond to the supply function. The appropriate estimation procedure is maximum likelihood and the first derivatives of the likelihood function are given. A computer program was written for the estimation of the parameters of the model. The empirical results proved quite disappointing. Despite repeated attempts, the likelihood function failed to converge. However, the lack of results should not be taken as evidence of the failure of the econometric technique, but rather of the particular iterative procedure employed to obtain the estimates.

Finally, the issue of the simultaneous determination between creditworthiness and the amount of capital inflows into a country is taken up. The results reveal that whereas the amount of capital inflows exerts a negative impact on the likelihood of rescheduling when the simultaneity between the two variables is ignored, the impact is reversed when the simultaneity is taken account of. An explanation is ventured as to the reversal of the sign of the coefficient of capital inflows.

















vi













CHAPTER ONE
DEVELOPING COUNTRY BORROWING IN THE
INTERNATIONAL CAPITAL MARKETS:
AN OVERVIEW OF THE ISSUES


Introduction

The accumulation of external debt by developing countries has

emerged as one of the most pressing issues facing the world economy in the 1980s. As the effects of the tremendous increase in energy prices that wreaked havoc during the previous decade begin to loosen their grip on the world economy, another danger looms on the horizon and threatens the collapse of the international financial system. A number of developing countries have amassed debts to outside official and private creditors to such an extent that some of them cannot meet theirexternal obligations as originally contracted. At the same time, several international banks have increased their exposure, vis-a-vis individual countries, to the point where their profitability and, to some extent, their solvency rests on the ability of a handful of countries to continue making interest and amortization payments.

The World Bank (1981, p. iii) defines external debt as "debt owed to non-residents and repayable in foreign currency, goods or services that has an original or extended maturity of over one year." As Table 1 shows, the total debt of developing countries has risen from 109.4 billion dollars in 1973 to 529 billion in 1982, a fivefold increase. More importantly, however, the debt service owed (interest and amortization payments) has increased sixfold: from 16 billion dollars in 1973 to 95 billion in 1982.

1






2







Table 1
Public and Private Debt of Developing Countries 1973-1982
(Billions of U.S. Dollars) 1973 1975 1976 1977 1978 1979 1980 1981 1982a

Total Debt 109.4 161.9 195.5 240.1 298.8 352.4 404.8 462.1 529.0 Total Debt Service 16.0 23.0 26.1 33.1 47.9 62.3 70.4 83.0 95.0 Low-Income Africa 0.4 0.5 0.5 0.6 0.7 0.8 1.1 1.3 --Low-Income Asia 1.0 1.3 1.3 1.4 1.6 1.8 1.9 1.9 --Middle-Income
Oil Importers 9.7 14.0 15.4 18.8 27.9 34.6 40.8 49.8 --Oil Exporters 4.9 7.1 8.9 12.4 17.7 24.9 26.7 30.8 --aEstimates


Indicates data not available Source: Finance and Development, Vol. 20, No. 1, March 1983, p. 23.





3

Because of the magnitude of the external debt of developing countries as well as its implications for the stability and growth of the world economy, the subject has generated a considerable amount of interest on the part of economists. The intent of this introductory chapter is to present, in a brief manner, the major issues that have been investigated and the directions in which research in this area has proceeded. It is by no means intended as a comprehensive review of the literature. The sheer volume of writings on the subject1 makes a complete survey quite impossible.2 One can only hope to touch on the different aspects of a complex and multifaceted topic.


Some Preliminary Comments and Definitions


The decade just ended was one of unprecedented turmoil in world

economic relations and in particular the international capital markets. The price of oil quadrupled in 1973 and rose again sharply towards the end of the decade. A pattern, never encountered before, emerged in the balance of payments accounts of developed and developing countries. In particular, as has been pointed out by several writers (Sachs 1981; Solomon 1977, 1981), the pattern of moderate current account deficits of less developed countries (LDCs) matched by surpluses in the developed countries, changed dramatically. As Table 2 shows, the non-oil developing countries (NOLDSc) suffered large deficits in their current account after


In addition to the bibliography reviewed here, a number of books as well
as articles in journals and the financial press appear continuously offering different viewpoints and proposing various solutions. Some of the proposals will be discussed in the concluding chapter. 2For instance, the literature on intertemporal optimizing models, which focuses on deriving optimal borrowing criteria or the role of external funds in smoothing the consumption path over time, is completely ignored. McDonald (1982) provides a useful discussion of such models.






Table 2
Payments Balances on Current Account, 1973-81 (in Billions of U.S. Dollars)

1973 1974 1975 1976 1977 1978 1979 1980 1981b Industrial Countries 19.7 -11.6 17.6 0.2 4.6 30.8 7.8 -44.1 -29
Seven Larger Countries 14.1 3.7 23.2 9.1 8.0 34.2 4.7 -15.7 1
Other Countries 5.5 7.9 5.5 9.3 -12.6 3.4 -12.5 -28.4 -28

Developing Countries
Oil Exporting Countries 6.6 67.8 35.0 40.0 31.1 3.3 68.4 112.2 96 Non-oil Developing Countriesc -11.5 -36.8 -46.5 -32.9 -28.6 -37.5 -57.6 -82.1 -97
By Area
Africa 2.0 4.8 9.2 8.0 6.0 7.3 5.8 7.3 -14
Asiac 2.5 9.8 9.8 3.5 0.7 5.8 -14.1 -23.7 -25
Europe 0.3 4.3 4.7 4.1 7.6 5.2 8.5 -10.3 9
Middle East 2.6 4.5 7.0 5.4 5.2 6.3 8.3 7.7 8
Western Hemisphere 4.7 -13.4 -16.6 -11.9 .9:1 -12.9 -20.9 -33.1 -40

Totald 14.8 19.4 6.1 6.9 2.1 3.4 3.0 -14.0 -30

aOn goods, services, and private transfers.

bFund staff projections.

cExcludes data for the People's Republic of China prior to 1977.

dReflects errors, omissions, and asymmetries in reported balance of payments statistics plus balance of listed groups with other countries (mainly, the Union of Soviet Socialist Republics and other nonmember countries of Eastern Europe and, for years prior to 1977, the People's Republic of China).

Source: International Monetary Fund, Annual Report, Washington, D. C.: IMF, 1981, p. 18.





5

1973, accompanied by large surpluses for the oil-exporting countries. The developed countries switched from a surplus to a deficit in their current account, although a few countries West Germany, Japan and Switzerland continued to enjoy a surplus in their current account.

The increased current account deficits of the NOLDCs after 1973 were accompanied by large scale borrowing by the NOLDCs from private financial markets.3 The vast majority of the borrowing by NOLDCs has been in the form of bank loans. Issuing of bonds by developing countries has been relatively limited and the international bond market has been accessible to only a handful of them. As Table 3 reveals, international bank claims expanded by 130 billion dollars in 1979,of which industrial countries accounted for 71 billion and NOLDCs for 41 billion. In addition, new medium-term loan commitments reached 69 billion dollars in 1979, industrial countries accounting for 22 billion and NOLDCs for 35 billion. By comparison, net bond market lending amounted to only 29 billion in 1979. However, of this total industrial countries were responsible for 19 billion, whereas developing countries for only 3 billion. The international bond market has been, for the most part, an insignificant source of funds for developing countries.

The most striking feature of developing country borrowing, which has been pointed out by several writers (Friedman 1981; Long and Veneroso 1981; Sachs 1982), has been the increasing predominance of private creditors as suppliers of funds to developing countries.



It must be emphasized, at this stage, that no causality is implied between the two events. This point will be elaborated further in a subsequent section.





6







Table 3
Aspects of International Bank and Bond Financing 1976-1979
(Billions of U.S. Dollars)

1976 1977 1978 1979

Bank Lending

Increase in Net Claims 70 75 110 130
Industrial Countries 31 43 49 71 Oil Exporting Countries 9 10 17 7 Non-oil Developing Countries :21 15 30 41 Other 9 7 15 11 New Medium-Term Loan
Commitments 29 34 74 69 Industrial Countries 10 13 34 22 Oil Exporting Countries 4 6 10 7 Non-oil Developing Countries 13 13 27 35 Other 3 3 3 5 Net Bond Market Lending 30 31 30 29
Industrial Countries 20 21 18 19 Developing Countries 2 4 5 3 Other 8 6 7 7


Source: International Monetary Fund, International Capital Markets,
Occasional Paper No. 1, Washington, D. C.: IMF, 1981, p. 2.





7

The proportion of debt contracted from private sources (financial

institutions and other private creditors) rose from 50.1 percent in 1973 to 60.6 percent in 1982, whereas the proportion from official sources (governments and international institutions) fell from 49.9 to 39.4 percent during the same period.4 In addition as Table 4 reveals the proportion of the current account deficit financed by borrowing from private sources rose slightly for the low-income NOLDSc but increased substantially for the middle income NOLDCs. Since loans from private sources usually carry higher interest rates and shorter maturities, as compared to loans from-official sources, the debt-burden implications of the shift towards private lenders are all too evident for the WOLDCs and will be discussed in greater detail in the subsequent sections.

Whereas low-income NOLDCs have had to rely mostly on official aid to finance their current account deficit, the middle-income NOLDCs borrowed extensively in the international capital markets. The post-1973 surpluses of the Organization of Petroleum Exporting Countries (OPEC) provided the funds for recycling to NOLDCs. However, two developments in the international capital markets in the 1970s proved instrumental in fostering NOLDC lending, in the absence of which it is doubtful whether NOLDC borrowing would have grown so rapidly. The first was the practice of syndication, according to which several banks pooled their resources in lending to a particular country. Second, the shift towards flexible interest rates, enabled banks to charge a rate on the loan that is tied to a particular market interest rate and which is revised at predetermined


4
The figures quoted are from External Debt-The Continuing Problem, Finance and Development, Vol. 20, No. 1, March 1983, p. 23.






Table 4
Oil-importing Developing Countries' Current Account Deficit and Finance Sources, 1970-80
(Billions of 1978 Dollars)
Oil importers
Low-income Middle-income
Item 1970 1973 1975 1978 1980 1970 1973 1975 1978 1980 Current account deficita 3.6 4.9 7.0 5.1 9.1 14.9 6.7 42.8 20.4 48.9 Financed by:
Net Capital Flows
ODA 3,4 4.1 6.6 5.1 5.7 3.3 5.3 5.3 6.5 7.9 Private direct investment 0.3 0.2 0.4 0.2 0.2 ,3.4 5.1 3.8 4.6 4.5 Commercial loans 0.5 0.6 0.8 0.9 0.7 8.9 13.7 21.0 29.4 27.1

Changes in reserves and
short-term borrowingb -0.5 -1.1 -0.7 -1.1 2.4 -0.8 -11.7 12.7 -20.1 9.5

Memorandum item:

Current account deficit
as percentage of GNP 1.9 2.4 3.9 2.6 4.5 2.6 1.0 5.5 2.3 5.0

aExcludes net official transfers (grants), which are included in capital flows bA minus sign (-) indicates an increase in reserves.


Source: World Bank, World Development Report, Washington, D. C.: World Bank, 1981, p. 49.





9

intervals. Both practices contributed to the rapid growth of NOLDC borrowing and will be further analyzed in the following chapter.

The large scale tapping of financial markets by NOLDCs has raised a number of important questions. Before embarking on a discussion, it may be useful to distinguish a few terms: private and public debt, official and private creditors, short- and long-term debt. According to the World Bank (1981) definition debt from official creditors comprises loans from international organizations (multilateral loans) and loans from governments (bilateral loans). Debt from private creditors comprises loans from suppliers and financial markets and the latter includes loans from private banks and other private financial institutions and publicly issued and privately placed bonds. Public debt refers to the debt incurred by the central government or its agencies or a private debtor guaranteed by the government. Private debt refers to the external debt of corporations and private citizens. Debt of maturity longer than one year is designated as long-term and that of maturity less than one year as short-term. Most of the data available on LDC external debt refer to long-term public debt, but recently private debt data have become available from the World Bank and the Bank for International Settlements. The balance of this dissertation is concerned with the external public debt of developing countries contracted with private creditors.


Distribution and Value of the External
Debt of Developing Countries


The Distribution of Developing Country Debt

One theme that almost every writer has touched on concerns the total amount of LDC borrowing and its distribution. Conclusions based on aggregate figures for the LDCs as a group may be misleading insofar as





10

the distribution of debt among LDCs is ignored. As Table 5 reveals, the majority of the external debt owed to banks is concentrated on a small number of middle-income LDCs. As of the end of June 1982, the twenty-one major LDC borrowers shown accounted for 84 percent of the total debt to all banks reporting to the Bank for International Settlements (BIS). The concentration was even more pronounced as regards the debt owed to U.S. banks. Of the twenty-one countries, eight5 accounted for 57 percent of the total LDC debt to BIS-reporting banks,whereas Mexico and Brazil alone accounted for 34 percent. Similar figures were reported by Solomon (1981, p. 597), where the same-eight countries accounted for 71 percent of the total NOLDC6 debt in 1980; Mexico and Brazil alone accounted for 44 percent of the total.

In a recent paper, Long and Veneroso (1981, p. 508) point out that, while there does not exist a significant relationship between the overall debt level and the per capita income of LDCs, when the overall debt is broken down to commercial bank debt and official lending, the former is positively correlated with per capita income whereas the latter is negatively correlated. Thus the belief that commercial bank lending has been concentrated on a few middle income LDCs, while the "poorest of the poor" have had to rely increasingly on official aid, is confirmed. Eaton and Gersovitz (1981b, p. 5) state that the countries that were the major borrowers in the early 1970s, were heavily engaged in borrowing until the


5Argentina, Brazil, Chile, Mexico, Peru, Philippines, South Korea and Thailand.

6Excluding South Africa and countries in Europe.













Table 5
External Debt Owed to Banks by 21 Major LDC Borrowers (Billions of U.S, Dollars)

BIS-Reporting U.S. Banks
Banks All 9 Large
Latin America
Argentina 25.3 8.8 5.6 Brazil 55.3 20.5 12.3 Chile 11.8 6.1 3.3 Colombia 5.5 3.0 2.1
Eucador 4.7 2.2 1.3 Mexico 64.4 25.2 13.6 Peru 5.2 2.3 1.3 Venezuela 27.2 10.7 I 7.1
Subtotal 199.4 78.9 46.7

Asia
Indonesia 8.2 2.4 1.9
Korea 20.2 9.2 5.1 Malaysia 5.3 1.3 1.0 Philippines 11.4 5.3 3.7 Taiwan 6.4 4.4 2.7 Thailand 4.8 1.7 1.0
Subtotal 36.8 24.3 15.4

Middle East and Asia
Algeria 7.7 1.2 0.8 Egypt 5.4 1.5 1.0
Israel 6.1 2.6 1.4 Ivory Coast 3.2 0.5 0.4 Morocco 3.7 0.9 0.7
Nigeria 6.7 1.2 0.9 Turkey 4.0 1.4 0.9
Subtotal 36.8 9.2 6.2

Total of 21 LDCs 292.3 112.5 68.3 Total for LDCs 347.5 125.5 77.7 (Excluding Off-shore
Banking Centers)
21 LDCs as Percentage 84 90 88
of Total for LDCs

Source: World Financial Markets, New York: Morgan Guaranty Trust Company,
February 1983, p. 3.





12

end of the decade. The distribution of borrowing among LDCs, as measured by the Gini coefficient, has shown no trend towards equality.7

The concentration of bank lending on a small number of developing countries has raised two questions. First, it has led several writers (Sargen 1976; Solomon 1977, 1981) to argue that any future LDC repayment problems will be intimately linked to the economic performance of these countries. Second, the question is raised as to whether it is rational for banks to alter their lending policy towards LDCs and distribute funds more evenly.

As for the first question, forecasts of the economic performance of middle-income LDCs, which are heavy borrowers, have been favorable and Solomon (1977, p. 498) has argued "the performance and prospects of the major borrowers permit an optimistic judgment about their creditworthiness." While the problems and prospects of the major borrowers have received the most attention recently, the possibility of a relatively minor borrower, faced simultaneously with a number of adverse external shocks,8 defaulting and causing a chain reaction among other borrowers, must not be overlooked. The performance of all borrowers must be examined as regards the possibility of defaulting on their loans. Isolating a small number of countries, examining their economic performance and prospects and inferring that developing countries as a whole will not face serious repayment problems, is likely to result in erroneous conclusions. The commercial banks, in conjunction with their governments


7In fact, Eaton and Gersovitz (1981b) report that the coefficient comparing the distributions of debt to distributions of gross national product (GNP) rose from 0.448 in 1976 to 0.494 in 1978. 8Such as a drought, a drop in prices of its major exports or an increase in real interest rates.





13

and international organizations, will see to it that the circumstances that render profitable the default of a major borrower, do not arise; whereas they may not have as much influence or the required information concerning the smaller borrowers.

As far as the second question is concerned, rational behaviour

would imply that banks trade off diversification, by lending to a broad group of countries, including low-income LDCs, against riskiness, which is generally recognized to be higher for low-income LDCs.9 Eaton and Gersovitz (1980, pp. 14-19) have produced evidence to suggest that the export performances of large borrowers are mutually independent. Furthermore, Goodman (1981) has divided total bank risk into diversifiable and nondiversifiable and has argued, from her empirical results, that country specific (nonsystematic) risk has been relatively large in comparison to total (systematic) NOLDC risk. Therefore, debt problems among LDCs are likely to arise independently and commercial banks can achieve sufficient diversification among the middle-income LDCs, which are large borrowers, without having to lend to low-income LDCs. In conclusion, the lending behaviour of commercial banks, as regards the distribution of loans, appears economically rational. Real and Nominal Value of LDC Debt

The discussion of the external debt of developing countries has, so far, been conducted in terms of nominal values. However, if comparisons of debt levels over time are undertaken, the real value of debt is required. A rather awkward problem is encountered here. As Long



9Since these countries do not have the industrial base to support substantial loan payments and cannot compress imports to any great extent in the event of an adverse shock.





14

and Veneroso (1981, p. 505) and Long (1981,. p. 283) point out, debt is a financial asset of commercial banks representing future claims on goods and services. Whereas the real value of a basket of goods and services that remains unchanged over time can be calculated by deflating by an appropriate price index, there is no "appropriate" index in the case of debt, since it depends on what goods and services the debt is used to procure.

Furthermore, the real value of debt depends on whether it is calculated from the debtor's or the creditor's perspective. If calculated from the debtor's perspective, the nominal value should be deflated by an export or import index to take account of the real resources foregone to repay the debt. If calculated from the creditor's perspective, it must be deflated by an index of the prices of goods and services of the countries in which the shareholders and depositors of the banks making the loans are resident, since they are the ultimate beneficiaries. In this case an index of world prices would be appropriate.

Where these indices differ over time, they can give vastly different results regarding the real value of debt, as Table 6 shows. The nominal value of NOLDC debt rose by 199 percent between 1972 and 1977. During the same period the real value of NOLDC debt rose by 94 percent when deflated by the Organization for Economic Cooperation and Development (OECD) price index but only by 31 percent when deflated by the NOLDC export price index. At present, there is no widely accepted index for deflating nominal debt and for that reason Long and Veneroso (1981) suggest that it may be useful to look at the ratio of debt to GNP as a measure of the debt burden over time.





15






Table 6
Alternative Estimates for the Real Value of NOLDC External Debt
(Billions of U.S. Dollars, 1972 Prices)

1972 1973 1974 1975 1976 1977
Nominal Debt 62.6 76.9 102.6 130.5 159.0 187.0 Deflated by Price Index
U.S. GNP 62.6 72.7 88.5 102.6 118.8 132.1 OECD GNP 62.6 66.9 81.0 91.4 108.1 121.6 NOLDC Imports 62.6 60.8 54.0 65.3 79.5 91.6 NOLDC Exports 62.6 56.1 54.4 71.8 82.7 82.3 Nominal Debt/Nominal GNP 0.26 0.28 0.29 0.33 0.35 0.35 Nominal Debt/Nominal Exports 1.41 1.48 1.42 1.73 1.70 1.79


Source: Long and Verneroso (1981, p. 507)





16

The tendency for prices to rise over time has further implications for the current account and the debt position of developing countries. As has been pointed out by several writers (Long 1981, p. 290; Long and Veneroso 1981, p. 512; Sachs 1981, pp. 203-208; Solomon 1981, p. 595), during periods of rising inflation rates, the current account overstates the true position of debtors. The rise in nominal interest payments (assuming that nominal interest rates rise with inflation) is offset by the fall in the real value of the debt. However, since only the higher nominal interest payments are reported in current account statistics, the true debtor position is-overstated.

The reason for the overstatement, as\Long and Veneroso (1981, p. 512) point out, is that the part of the nominal interest payments corresponding to the inflation premium is a capital account item, as it is reimbursing the lender for the loss in principal value due to inflation, and should not be included as part of the current account. Sachs (1981, pp. 204-205) has estimated the current account of developed countries the NOLDCs and OPEC members adjusting for the inflation bias and has argued that these adjustments are of significant magnitude: by 1978, over half of the NOLDC debt could be attributed to the inflation bias and in 1978, while OPEC was nearly in balance on its nominal current account, when adjusted for the inflation bias it showed a deficit of 12.8 billion dollars.

Inflation plays an important part in determining the value of real interest payments made by LDCs. Here, the problem encountered earlier, of the appropriate deflator, resurfaces. Notwithstanding the question of the appropriate deflator, the choice between real and nominal interest payments plays an important part in determining a country's debt





17

capacity.10 Long (1980, pp. 494-5) has argued that when the nominal interest burden is compared to the real interest burden (under the prevailing maturity structure), the former, of course, reveals a more severe burden which, together with amortization charges, would push the ratio of debt service payments to exports, for many NOLDCs, to alarming levels. Such a distinction is crucial because "from a solvency and growth standpoint, it is the real interest burden that matters, but from the liquidity standpoint a country might find it difficult to finance such high outlays on external debt" (Long, 1980, p. 495).

The rise in nominal interest rates that normally follows a rise in the inflation rate, takes on added significance when considering the debt burden of developing countries. As Sachs (1981, p. 206) points out "a rise in inflation that is exactly matched by a rise in interest rates causes a rise in interest income for a creditor country that is exactly offset by greater capital losses." However, when nominal rates rise above the increase in the inflation rate a real burden (in terms of interest payments) is imposed on the debtor countries which, as has already been argued, is manifested as well in overstated current account deficits. Therefore, rising real interest rates impose a burden on developing countries in two ways: first by raising the value of nominal interest payments and secondly by overstating the extent of the current account deficit, necessitating higher borrowing LDCs to finance the

rising deficit.




10The concept of debt capacity for a country is explained in a further section of this chapter.





18
Constraints on LDC Borrowing and the
Role of International Reserves


Is a Credit Ceiling Imposed on LDC Borrowing?

In a series of papers, Eaton and Gersovitz (1980, 1981a, 1981b)

have argued that individual LDCs are not unconstrained in their borrowing, but face a ceiling on the amount of funds they can borrow from private financial sources. On the basis of a theoretical model (Eaton and Gersovitz, 1981a) they have shown that the ceiling is determined endogenously and is a function of several characteristics directly related to the costs and benefits of default. In particular, they have proposed that, on the one hand, the higher the variability of income of a country, the greater the penalty of default (by being excluded from future borrowing) and the higher the credit ceiling imposed. On the other hand, the probability of default is higher in periods of low income when the marginal utility of income is high. Therefore, a high variability of income, by increasing the probability of low incomes, increases the probability of default and therefore lowers the credit ceiling. In conclusion, the impact of the variability of income on the credit ceiling is ambiguous.

The long-term growth of income is negatively related to the credit ceiling because a higher growth rate means the country need not enter the capital markets in the future and is therefore more likely to default at present. Finally, the higher the dependence on trade (as measured by the ratio of imports to income) the less the likelihood of default, as the country would suffer a higher penalty from the disruption of its trade pattern, and therefore the higher the credit ceiling.

The proposition concerning credit ceilings is supported by Sachs (1982, p. 19), who shows that when default is precluded credit ceilings are fully consistent with perfect competition in the loan market.





19


Further on (Sachs, 1982, p. 21), he argues that in a two-period model, where default can occur, there is a maximum level of debt for each country and any amount greater will lead to default with certainty. No rational lender will be willing to loan any amount greater than the ceiling, irrespective of the risk premium. Therefore, "rationing will be a standard device in credit allocation to sovereign borrowers" (Sachs, 1982, p. 9).

On the basis of their theoretical work, Eaton and Gersovitz argue that the observed amount of LDC debt is the minimum of two quantities: the amount of desired debt or the credit ceiling. Thus, empirical estimation of demand and supply curves for LDC debt must be carried out within the framework of a disequilibrium model where the observed quantity of debt is the minimum of the amount desired or the credit constraint. In two papers (Eaton and Gersovitz, 1980, 1981a) they estimate a disequilibrium model, in which a country's demand and supply of debt are dependent on the economic characteristics described earlier. They show that, in the two years 1970 and 1974, 56 of 81 countries were constrained with probability greater than 0.5. They conclude that credit constraints are an important feature of the LDC loan market and any attempt to estimate a supply or demand function without incorporating the constraint explicitly is bound to lead to erroneous estimates.

In connection with the Eaton and Gersovitz work, it is worth

mentioning that the demand and supply curves estimated do not include the interest rate as an explanatory variable. As they correctly point out, the base ratell is the same for all countries during a particular


Which in the international loan market is the London interbank offered rate or LIBOR.





20

time period and any variation in the base rate between two years is captured by year-specific dummy variables. As far as the risk premium is concerned, they argue that "in competitive markets if lenders are risk-neutral or if default risk can be perfectly diversified the expected interest rate r is equal for all borrowers in each year" (Eaton and Gersovitz, 1980, p. 12). If their assumptions hold, then the risk premium is solely a function of borrower characteristics and can be correctly omitted from the supply and demand functions. Thus, the omission of the interest rate rests on how well the assumptions are met in practice.

A different approach to estimating the demand and supply of loans to developing countries is proposed in this dissertation. A disequilibrium model of the demand and supply of loans is specified whereby the interest rate variable is introduced as an explanatory variable and a quantity constraint is entered explicitly. The specification and estimation of the model is discussed further in chapters 3 and 4.

One final issue concerning the credit ceiling is related to the

penalties that lenders can impose on the borrower in the case of default. Eaton and Gersovitz (1981b, p. 33) put forward what may seem a strange proposition: both the lender and the borrower may be better off as a result of more severe penalties in the case of default. The lenders may be better off because they can loan a larger amount (by virtue of higher credit ceilings) than if the penalties were less severe. As far as the borrowers are concerned, on the one hand they are better off since higher penalties imply greater capital flows from the banks. On the other hand, greater penalties reduce the welfare of the borrower because they decrease the insurance aspect provided by default, in the event of





21

adverse economic conditions. Eaton and Gersovitz suggest that one possibility may be to let LDCs decide by themselves, in advance, the penalties to be incurred in the event of default. Of course, it must be remembered that all of the above arguments rest on the assumption that the penalties are actually enforced.

Essentially the same argument is made by Sachs (1982, p. 28). On

the basis of a two-period utility-maximization model, where a cooperative and a non-cooperative solution between banks and LDC borrowers are compared, he argues that under certainty, the utility from the cooperative solution is higher than-the utility from the non-cooperative solution and if default could be ruled out altogether,utility could be increased even further. It is therefore to the advantage of the borrowing country to seek higher penalties in the case of default. However, the argument does not always carry in the case of uncertainty. Although higher penalties increase welfare by raising the credit ceiling and therefore capital flows to LDCs, they lower the "insurance" aspect of default. Uncertainty opens up the possibility that the insurance aspect of debt is the dominant aspect and that penalties should not be increased. External Debt and International Reserves

One question that is of considerable importance from a public-policy standpoint is whether international borrowing is a substitute or a complement to a country's international reserves. As Eaton and Gersovitz (1980, p. 4) argue, there is no theoretical a priori reasoning either way. On the one hand, debt may be used to finance reserves while, on the other hand, both borrowing and reserves can be used as a means of smoothing the variability in the income stream of a country. In the first case they are complements; in the latter they are substitutes.





22

The question must, in the end, be an empirical one. Eaton and Gersovitz point out that the demand for reserves is a function of the same variables as the demand for debt. In their empirical investigations, they estimate separate reserve demand equations under two different regimes, depending on whether the country faces a ceiling on the amount it can borrow on the international capital markets or not. On the basis of their empirical estimates they find that debt and international reserves are substitutes. Therefore increased borrowing by LDCs from private sources should reduce the demand for international reserves.

On the same issue, Solomon (1977, p. 481) states that in 1976 NOLDCs borrowed more than their combined current-account deficit and added to their international reserves. The possibility arises that the NOLDCs borrow short-term in order to build up their reserves and repay the loans in the following years. The reserves can then be drawn down to meet unexpected fluctuations in the income stream when borrowed funds may be unavailable or too costly to obtain. This could very well have been the case with short-term borrowing from private sources which in 1974-75 amounted to five billion dollars or one fifth of the total borrowing of the NOLDCs. L'Heriteau (1979, p. 713) discusses one practice which was becoming increasingly common in the mid 1970s concerned with the depositing of the international reserves of NOLDCs in the Eurocurrency markets. These funds were, in turn, loaned back to the NOLDCs. Thus, the international banks acted as intermediaries channeling the funds of some NOLDCs to other NOLDCs and collecting interest, commissions and other fees in the process.

International reserves play an important part in determining the creditworthiness of developing countries and how that is perceived by





23


lenders in the international capital markets. The crucial role of international reserves as an indicator of creditworthiness has been discussed by several writers including Feder and Just (1977a, 1977b), Frank and Cline (1971), Goodman (1977) and Kapur (1977). In econometric studies the ratio of international reserves to import expenditures during a particular time period has been employed as an indicator of creditworthiness. The empirical results showed that, ceteris paribus, a larger volume of reserves improved the country's creditworthiness and reduced the likelihood that the country would seek to reschedule its external debt.

While a larger amount of reserves enhances the creditworthiness of a country and raises the likelihood of obtaining loans at more favorable terms, some developing countries have been motivated to borrow on the international capital markets by the desire to increase their international reserves and thus improve their creditworthiness. Friedman (1977, p. 18) has argued that "other developing countries, like Brazil, are borrowing both to finance external deficits and to rebuild reserves, profiting from the experience that a strong visible level of reserves improves creditworthiness as well as providing an additional cushion to meet contingencies." The notion that developing countries may have borrowed to increase their reserves and thus their creditworthiness so as to be able to borrow further, is also shared by Abbott (1981, p. 344) and by Griffith-Jones (1980, p. 215) who quotes the Chilean magazine, Gemines, as follows: "The Central Bank is ready to increase its foreign assets even more because the concentration of debts

to banks and financial institutions in a way obliges it to maintain available a relatively high level of reserves."





24

From a theoretical standpoint, beginning with the seminal paper by Bardham (1967),a substantial body of literature has developed on the optimal amount a country should borrow externally. The contributions in this area examine the question within the framework of a neoclassical growth model and the general conclusion arrived at is that borrowing should continue until the marginal product of capital becomes equal to the interest rate on external borrowing. Out of the numerous contributions, only the paper by Feder and Just (1979) has examined the relationship between the optimal amount of borrowing and the optimal level of reserves. Feder and Just postulate a two-sector open-economy model, where one sector produces an export good and the other a good which can be used either for consumption or investment. It is assumed that the interest rate paid on the external debt depends on two indicators of creditworthiness: first, the difference between export receipts and the average volume of debt service payments and second, the difference between international reserves and import expenditures.

The central result of the Feder and Just study is intuitively

plausible and holds that the level of international reserves should be adjusted until the marginal cost of borrowing to the economy is equal to the marginal benefit from holding reserves. The marginal cost of borrowing is higher than the interest rate on the external debt since the higher debt level reduces creditworthiness and raises the interest rate. The marginal benefit from holding international reserves is equal to the interest rate earned on the reserves plus the contribution to increased creditworthiness from holding a higher level of reserves. Thus, the importance of international reserves in the model enters through the impact of reserves on the creditworthiness of the country. If the impact





25

is insignificant then, according to the model, a zero level of reserves should be maintained.

One issue, not directly related to the above, but of considerable significance from a public-policy decision-making standpoint, is the relationship between debt from private and public sources. On the basis of their empirical estimates, Eaton and Gersovitz (1981b, pp. 19-20) have argued that they are complements, instead of substitutes as was expected. They attempt to explain their findings by recourse to the argument that funds obtained from official sources are not sufficient to complete the projects undertaken and that countries have to resort to private loans in order to complete these projects. One can cast some doubts on the validity of their explanation, especially in view of the fact that official sources of funds are usually of longer maturity and bear lower interest rates than private loans. Private funds are unlikely to complement official funds in project financing since it is quite possible that a project financed by official funds becomes unacceptable when funding is provided from private sources at higher interest rates. The question of the relationship between debt from private and public sources has, by no means, been settled.


Analyzing Country Risk: Debt Reschedulings
and the Debt Capacity of LDCs

One question that has received considerable attention, for obvious reasons, is that of debt repayment and the probability of default on the part of developing countries. There are two strands of modeling in this area. The first employs growth-cum-indebtedness models to calculate the value of such ratios as debt/income or debt/exports at the limit or the





26


speed at which debt accumulates, based on several assumptions about the parameters12 in the model. The second attempts to predict the probability an LDC will default on its external debt obligations, on the basis of several ratios. Because there has only been one outright default since World War II (North Korea), the models estimate the probability that a country will request a rescheduling of its debt obligations. Naturally, the two approaches can be combined, as in Feder (1980). Growth-cum-Indebtedness Models

In the theoretical development economics literature, considerable emphasis has been placed on the contribution of foreign funds to the process of economic development, particularly within the context of the "two-gap" models, as discussed in the pioneering works of McKinnon (1964) and Chenery and Strout (1966). According to these models, a target rate for the growth of income of a country is specified and it is assumed that a maximum amount of domestic savings is available and a minimum amount of imports is required during any particular period. The country's ability to export and its investment requirements define two gaps: the gap between maximum savings and investment (savings gap) and between exports and the minimum amount of imports (foreign exchange gap). It is therefore argued that, if the target rate of growth is to be achieved, foreign funds are required to fill the larger of the two gaps.

While "two-gap" models were originally developed to forecast the order of magnitude of foreign aid required to achieve different target rates of growth, it soon became clear that developing countries can obtain




Of course, the results are sensitive to the particular values chosen for the parameters.





27

foreign funds in the form of loans which have to be repaid and which entail a process of debt accumulation. Thus several models of the growthcum-indebtedness process have been developed. Beginning with the seminal work by Avramovic and his associates at the World Bank (Avramovic 1964) the earlier models assumed the savings gap to be binding (Ohlin 1966; King 1968; Solomon 1977, 1981; Feder 1980). The model of Feder (1981) investigated the pattern of debt accumulation when the foreign exchange gap is assumed binding.

Based on the earlier works of Avramovic (1964) and Ohlin (1966),

Solomon (1977) assumes that investment and savings increase exponentially over time and debt accumulates to fill the savings gap. The general conclusion drawn from the model holds that so long as the rate of growth of the economy is greater than the rate of interest charged on the loans, the debt/income ratio will converge to a maximum value. Solomon (1977, p. 495) calculates the limiting values for the debt/income, debt/exports and interest/exports ratios for several countries and argues that they are not excessive. The issue of what happens when the growth rate is lower than the interest rate on the debt is also addressed by Solomon (1981, p. 597). In the late 1970s, the rise in interest rates above the growth rates, for several countries, has meant that the process of debt accumulation is no longer self-limiting but is rather an explosive one.

One criticism of Solomon's procedure has been voiced by Greenspan in the discussion of his paper, who argues that a static framework has been imposed on what is a dynamic process. The values of the parameters in the model i.e., the incremental capital-output ratio, the growth of real and nominal net national product3 and savings ratio are all assumed



13Which are equal since prices are assumed constant in the model.





28

to remain constant over time, clearly an unrealistic state of affairs. Solomon (1977, p. 487) addresses this issue briefly, but provides no alternative to his analytical procedure.

The criticism is partially met by Feder (1980) and Long (1980,

1981), who conduct sensitivity analysis of their results. Feder (1980) simulates the number of years for debt to reach its maximum level, the maximal debt-service payments/exports ratio and the time it occurs, under different scenarios concerning the target rate of growth of GNP the marginal savings rate and the export growth rate. Long (1980, pp. 488497) calculates the debt/GNP ratio for different values of the real interest rate and the rate of growth of GNP as well as the number of years for the debt/GNP ratio to increase from 30 to 40 percent, again for different real interest rates and rates of growth of GNP. In a subsequent paper, Long (1981, pp. 296-299) calculates the number of years for the debt/GNP ratio to increase by ten percentage points, when a country suffers different adverse shocks4 of varying magnitudes.

Feder (1981) focuses attention on the foreign exchange gap as the binding constraint. He derives the intuitively obvious results that higher target rates of growth and higher interest rates imply higher levels of indebtedness whereas an increase in the rate of growth of exports leads to lower levels of borrowing. Therefore, export promotion provides an alternative means of attaining the target rate of growth without increasing indebtedness. However, in the model, the resources for



14Such as deterioration in the terms of trade, higher real interest rates, capital flight export decline and "disaster".





29

achieving the increase in exports are provided by reducing consumption and therefore the resulting decrease in borrowing eventually ceases when the savings gap is reached.

Further on, Feder provides projections for the highest debt/gross domestic product (GDP) ratio and the time it is reached, as well as the value of the debt service/exports ratio and the savings rate at that time. These projections are provided for different values of the target rate of growth, the rate of growth of exports and the proportion of consumptiongoods imports relative to total consumption. Another table provides similar information for the highest debt service/exports ratio. As Feder argues,such projections are useful as rough planning aids in the calculation of borrowing requirements and are, by no means, meant to provide an accurate description of the debt-accumulation process.

A general criticism of the growth-cum-indebtedness models is that they are solely concerned with the demand side of the market for loans and assume that funds are forthcoming at a constant average interest rate. Solomon (1977) recognizes this limitation and devotes several lines to it in his concluding remarks. As in any other market, both demand and supply are equally important and one cannot simply ignore one side of the market. In fact, if one is to believe Eaton and Gersovitz (1980, 1981a, 1981b), it could very well be that the supply side of the market plays a more important role than the demand side in determining the actual amounts borrowed by developing countries.

Estimating the Probability of Rescheduling

The second strand of modeling attempts to predict the likelihood of rescheduling by individual LDCs. One method, pioneered by Avramovic (1964), is the close examination of several indicators, usually in the





30

form of financial ratios, which are taken to be important predictors of repayment problems. The other is to incorporate these predictors in formal mathematical models and estimate probabilities of rescheduling.

The first systematic analysis of ratios was initiated by Avramovic (1964) at the World Bank. However, even earlier works focused on the ability of developing countries to service foreign capital.15 Finch (1951-52, p. 60) introduced the concept of an investment service ratio, which he defined as "the percentage of current foreign exchange receipts, exclusive of conpensatory financing, absorbed by investment income payments." He presented a-historical series for this ratio for a number of countries which by that time had developed sufficiently and others, which at the time, were still in the early stages of development. He demonstrated that for the developed countries the ratio was very substantial during the early development phase and declined to insignificant amounts during the later stages. No such trend could be discerned for the countries that had yet to reach the later stages of development. Finch concluded that the ratio was useful as a short-run indicator of a country's ability to meet its debt service obligations but not as a guide as to whether foreign capital will be to the long-run benefit of the country. The distinction between short-run disturbances that affect a country's ability to service foreign capital and long-run prospects, is also discussed, in an early work, by Alter (1963). He argues that a country can maintain its debtservicing capacity over the long run by way of a rising trend in per capita income.



15In this broader definition, foreign capital encompasses loans, bonds and other fixed obligations contracted with foreign parties, as well as foreign equity capital.





31

The work of Abramovic (1964) provided a systematic study of debtservicing capacity. A distinction was drawn between the liquidity (shortrun) and solvency (long-run) aspects of debt-servicing capacity. The ratio of debt service payments to export proceeds (or the debt service ratio as it has become known) was introduced as a short-run indicator, whereas per capita income and growth prospects have been identified as important long-run determinants of debt-servicing capacity.

Since the appearance of Abramovic's work, almost every study of the external debt of developing countries has included some discussion of financial ratios as indicators of debt-servicing capacity. Some of the more commonly used ratios are shown on Table 7. Despite the large disparities among individual countries which are masked by aggregate ratios, the figures reveal a gradual but steady deterioration in the debt-servicing capacity of NOLDCs throughout the 1970s. The debt service ratio increased by almost two-thirds between 1973 and 1982 whereas the debt/exports ratio increased by a quarter during the same period.

In addition to the ratios of Table 7, a number of others have been compiled by various writers. Dhonte (1975, p. 161-170) compared several ratios for 69 developing countries for the year 1969 to the ratios for 13 countries which renegotiated their debt obligations during 1959-1971. The ratios included in the analysis were debt outstanding/GNP, debt outstanding/exports of goods and services, the debt service ratio, a discounted debt service ratio, the ratios discounted debt service payments/debt outstanding, debt service payments/disbursements, disbursements/ imports and finally,the ratio of net transfer (disbursements minus debt service payments) to imports. Dhonte's comparison revealed that countries that renegotiated their debt borrowed heavily to secure a large net





32






Table 7
Long-Term Debt Ratios for NOLDCs, 1973-1982

1973 1975 1976 1977 1978 1979 1980 1981 1982 Ratio of External
Long-Term Debt
to Exports of
Goods and Services 88.7 97.7 102.6 104.9 111.2 101.9 92.9 103.2 109.1 Ratio of External
Long-Term Debt
to GDPa 16.6 18.3 20.5 22.1 23.7 22.7 21.8 24.3 25.2

Debt Service Ratiob 14.0 14.0 13.9. 14.0 17.3 18.1 16.3 21,0 22.3 aRatio for year-end debt to exports or GDP for year indicated. Debt service payments as percentages of goods and services.


Source: Finance and Development, Vol. 20, No. 1, March 1983, p. 23.





33

transfer, had a high and rising debt service ratio and sought large capital inflows at the same time that their debt accumulated on unfavorable terms.

Abbott (1981) examined the ratios shown on Table 7 for all NOLDCs and separately the low-income countries16 during the period 1970-1980. While cautioning against the indiscriminate use of such ratios, he concluded that for the NOLDCs as a whole the ratios have not shown any marked trend but have deteriorated (particularly the debt service ratio) for the low income NOLDCs. Further on, Abbott (1981, p. 347) computed the net transfer of resources7 to developing countries for the period 1972-1979. The results indicate a marked decline in net resource transfer: whereas in 1976 it amounted to 56 percent of gross disbursements, by 1979 the proportion had fallen to 36 percent.

Among the various formal models of rescheduling that have appeared in the literature, several indicators of debt-servicing difficulties have received prominence. They are discussed by Frank and Cline (1971) and Feder and Just (1977a). In addition to the debt service ratio already mentioned, others include the growth rate of exports, an export fluctuations index, the ratio of "non-compressible imports" to total imports, per capita income, the ratio of debt amortization to total outstanding debt, the ratio of imports to GNP and the ratio of imports to international reserves. Feder and Just (1977a, pp. 26-29) argue that it is virtually impossible to distinguish between "compressible" and "non-compressible" imports and in their study they introduce one further index: the ratio


16Identified as countries with per capita incomes of less than $300 U.S. dollars in 1978.

17Defined as gross disbursements less amortization and interest payments.





34


of capital inflows to debt service payments. Further discussion of these ratios is provided in Chapter 5.

Solomon (1977, p. 489) criticizes the debt service ratio as an

inappropriate measure of creditworthiness because maturities are usually bunched and not spread evenly over time. However, in his discussion of Solomon's paper, Greenspan argues that it is precisely the bunching of maturities that the debt service ratio attempts to measure. Eaton and Gersovitz (1981b, p. 27) make the point that a rising debt service ratio may not signal the deteriorating creditworthiness of a country but may in fact be a result of other factors which make the country a good risk and allow loans to be extended to it. Finally, it must be recognized that the debt service ratio considers a country's debt obligations during a particular year and neglects any future obligations which may be just as important in the country's decision of whether to reschedule at present or not.

Comparisons of debt service ratios over time may be misleading during periods of inflation, as pointed out by Solomon (1981, p. 598). When the nominal interest rate rises in accordance with the rise in the rate of inflation, the debt service ratio is biased upwards. A rise in both the interest rate and the inflation rate from 5 to 10 percent is a doubling of the interest rate but only a 5 percent increase in the price level. The debt service ratio, "exaggerates the increase in debt burden in periods when inflation accelerates and interest rates rise correspondingly" (Solomon, 1981, p. 598).

Along the same lines of ratio analysis, Euh (1979) devised a

creditworthiness index as the ratio of loan amounts supplied to a country by commercial banks to its external financing requirements (needs). The





35

latter were projected on the basis of a simple Harrod-Domar growth model. The index was regressed on a number of explanatory variables with data from 33 LDCs for the period 1971-75. Three variables were found to be significant and of the correct sign: the growth in GDP, the expected change in the debt service ratio and the debt/exports ratio. The sign of the coefficient of the other variables8 conformed to prior expectations, whereas the sign of the coefficient of the reserves/imports ratio proved consistently contrary to expectations. Euh's study shifted attention from other studies which focused on the short-run (liquidity) aspects of creditworthiness and pointed to the importance of the growth in GDP as a long-run indicator of creditworthiness.

There have been several approaches to formal models of the probability of rescheduling. One uses the method of discriminant analysis (Frank and Cline, 1971; Sargen, 1977; Black, 1981) another one,logit analysis (Feder, and Just 1977a; Mayo and Barrett, 1978; Feder, 1980; Feder, Just and Ross, 1981) and another the technique of principal components (Dhonte, 1975).

The Frank and Cline (1971) study investigated 13 debt reschedulings for 8 countries during the period 1960-68. It was found that a quadratic discriminant function in two variables19 gave the best results in terms of the fewest Type I and Type II errors.20 The Black (1981) study



181ncluded in the analysis were the growth in exports, per capita income, the debt service payments/net capital inflows ratio, a proxy for the willingness to pay and a political instability index. 19The debt service ratio and the ratio of debt amortization to total outstanding debt.

20Type I errors refer to the failure to predict countries which actually rescheduled, whereas Type II errors refer to falsely predicting a country to reschedule. Frank and Cline reported a Type I error rate of 23 percent and a Type II error rate of 11 percent in their study.





36

estimated four discriminant functions on the basis of 19 discriminating variables and compared the relative importance of each variable. In this study countries were assigned to five categories to reflect each country's creditworthiness in the international financial markets.

In an interesting paper, Sargen (1977) introduced two conceptual approaches to debt rescheduling. The first was the traditional debt service approach according to which repayment problems are exogenous to the economic management of the country and arise because of a shortfall in export proceeds or other adverse external factors. Such problems are short-term in nature and the debt service ratio was identified as an important indicator by the Abramovic (1964) study. Second, the monetary approach identifies debt repayment problems with factors endogenous to the management of the economic system.. Rapid expansion of the money supply and the resulting inflation, along with the maintenance of overvalued exchange rates, cause export demand to fall and import demand to rise and lead to debt accumulation.

The Sargen (1977) study employed six21 explanatory variables to

differentiate between rescheduling and nonrescheduling cases. Two variables figure prominently in the discriminant function: the adjusted debt service ratio (a debt-service approach variable) and the inflation rate (a monetary approach variable). Type I error rates varied from 15 to 54 percent according to the cut-off value, while Type II error rates ranged from less than 1 to 11 percent.22 An interesting result drawn from this



21The inflation rate, the growth rate of the money supply, the export growth rate, the adjusted debt service ratio the growth rate of real GNP and a measure of relative purchasing-power parity. 22Type II errors are to be compared to the naive rule of assigning all countries to the nonrescheduling group, which would produce an error rate of 5 percent.





37
study concerns the success of the adjusted debt service ratio in explaining reschedulings in South Asian countries associated with long-run debt problems. Exclusion of the debt service ratio from the discriminant function does not diminish its success in explaining reschedulings in South American countries associated with high inflation and short-run balance of payments crises. These results are not in accordance with the emphasis placed by Avramovic on the debt service ratio as an important indicator of short-term repayment problems.

Dhonte (1975, pp. 170-186) employed the technique of principalcomponents analysis, according to which a number of indicators (the ratios mentioned previously in connection with Dhonte's analysis, in addition to the growth rate of debt and the growth rate of exports) are combined linearly to derive a set of composite indicators or components. The importance of each component is measured by the percentage of the total sample information incorporated in the component. Four indicators are identified which account for 79 percent of the total information. Two indicators, which can be interpreted as, on the one hand, a measure of the involvement in debt and, on the other, the terms of the loans and the size of the debt service, account for 56 percent of the total sample information.

The conclusions of the analysis are arrived at by plotting the

correlation coefficients of the indicators with the components on a plane, whose axes are the two components, for both the standard sample of countries and the renegotiation cases. Two equilibrium relations are identifed: first, a balance must be maintained between the extent of the involvement and the borrowing conditions and second, the growth rate of debt must be maintained in line with the growth rate of exports, otherwise repayment problems will arise and are likely to lead to debt





38

renegotiations. The methodology employed in the study did not permit the testing of the second relation. Finally, the results of Dhonte's study do not fare well relative to the studies using the discriminant analysis technique: 33 percent of the renegotiation cases and 13 percent of the countries that did not renegotiate were incorrectly assigned.

The final technique that has been employed to identify rescheduling cases is that of logit analysis. Feder and Just (1977a) examined a sample of 21 reschedulings for 11 countries during the period 1965-72 and found six variables to be significant in the estimation of the "default" probability equation: the debt service ratio, the imports/reserves ratio, the amortization/debt ratio, export growth, per capita income and .he capital inflows/debt service payments ratio. A Type I error rate of 5 percent and a Type II error rate of 2.5 percent were obtained, the fewest of any of the other studies. On the basis of these six variables, Feder (1980) calculated the maximal probability of default, the time it occurs, as well as the probability of default in the first year, for high- and middle-income LDCs and for different values of the target growth rate of income, the marginal savings rate and the export growth rate.

The results of Feder and Just (1977a) were extended in Feder and Just and Ross (1981) who made use of a larger data base.23 Six explanatory variables were tested: the debt service ratio the ratios reserves/ imports, commercial foreign exchange inflows/debt service payments, net noncommercial foreign exchange inflows/debt service payments, exports/ GNP and real per capita GNP/U.S. per capita GNP. For the first three variables both the linear and the quadratic forms were tested and found



23Some 580 observations, of which 40 referred to cases of debt reschedulings.





39

to be significant. Type I error rates ranged from 8 to 33 percent and Type II error rates from 1 to 8 percent for the model which included the quadratic terms. The model was used to predict reschedulings with data from a period not covered by the sample used to obtain the original estimates. The Type I error rate remained unchanged whereas the Type II error rate increased somewhat when compared to the original-sample error rates.24

The technique of logit analysis has been employed by Mayo and Barrett (1978) as part of a larger country evaluation model at the Export-Import Bank. The other two parts of the model consist of a quantitative checklist system to assess "the medium- and long-range economic vitality of a country" and a qualitative checklist system to evaluate "the long-range strength and contribution of the country's human and natural resources" (Mayo and Barrett, 1978, p. 81). The analysis was performed with a relatively larger sample than that used by previous studies: 571 observations from 48 countries for the period 1960-1975,



24The logit analysis studies reviewed so far, estimate what can be termed as "objective" probabilities of default. The data used for the studies are drawn from actual developing country rescheduling experiences. In a further study, Feder and Just (1977b) estimate "subjective" probabilities, which are the international banks' own probability estimates. The authors develop an interest rate equation that depends on the cost of funds to the bank, the commitment period, the elasticity of demand, the probability of debt-servicing difficulties and the expected loss rate in the event of such difficulties. The probability of debt-servicing difficulties is described by an equation similar to that estimated by the objective studies. It is substituted in the interest rate equation giving a reduced form equation to be estimated. In addition to the variables found significant in the objective studies, an export fluctuation index and the ratio of imports to GNP are deemed important determinants of the subjective probability of default. The subjective probability concept is further explained in Chapter 2.





40

which included 28 instances of rescheduling in 11 countries. As in other studies of logit analysis, the dependent variable is binary,reflecting occurrences of default or not. The novelty of this study was in the way the dependent variable was defined to include reschedulings up to five years into the future. This enabled the authors to obtain predictions of debt-servicing problems for up to five years without having to forecast the explanatory variables.

The authors experimented with some 50 variables, six of which25 were chosen on the basis of correctness and consistency of the sign of the coefficient and the stability and significance of the t-statistics. It is worth pointing out that one of the variables found significant, the percentage change in the consumer price index, has been identified by Sargen (1977) with the monetary approach, while another, the ratio of gross fixed capital formation to GDP had not been included in any of the previous studies. The error rates Type I error rate of 24 percent and Type II of 13 percent do not compare favorably to those of previous studies, but no strict comparisons are possible because of the way the dependent variable was defined.

An interesting comparison between the Frank and Cline (1971) and the Feder and Just (1977a) model was undertaken by Manfredi (1981) with data from 60 countries for the period 1972-77. It was found that while there was no significant difference between the two models with respect to Type I errors, the Feder/Just model fared better in terms of Type II errors. The results were attributed to the fact that the Feder/Just model



25The ratios disbursed debt outstanding/exports, international reserves/ imports, imports/GDP, reserve position in the International Monetary Fund/imports, gross fixed capital formation/GDP and the percentage change in the consumer price index.





41

includes a capital inflows variable. It may be possible for countries to avoid rescheduling their debt if they can continue attracting foreign capital.26

The results of the formal models of rescheduling have been summarized and extended by Saini and Bates (1978) with data from 20 countries for the 1960-77 period. Two different dependent variables were defined: one was the traditional variable used in other studies consisting of official debt rescheduling and nonrescheduling cases; the other included both involuntary debt reschedulings and balance-of-payments support loans, but excluded voluntary debt reschedulings.27 The authors made use of both discriminant analysis and logit analysis and estimated models for both the whole of the period 1960-77 and for two separate subperiods, for each version of the dependent variable. They summarized their results in the following propositions: (i) when logit and discriminant analysis were compared no significant differences were observed in terms of the error rates and coefficient values, (ii) the modified dependent variable provided better results when compared to the traditional one, (iii) four variables proved more significant in terms of explanatory power: the consumer price index, the money supply growth (both variables suggested by the monetary approach), the cumulative current account balance to exports ratio and international reserve growth, (iv) the debt service



26Mexico could very well have been a case in point here. 27Saini and Bates (1978, p. 26) defined balance-of-payments support loans as foreign loans, in the absence of which, a rescheduling would have been necessary or arrears on external payments would have occurred; voluntary debt reschedulings represented cases where there were no apparent balance-of-payments difficulties.





42

ratio did not appear significant in explaining debt reschedulings,

(v) the explanatory variables proved more significant during the 1971-77 subperiod than during the 1960-70 subperiod, possibly because the former included most of the modification in the dependent variable (Saini and Bates, 1978, p. 15).

One final noteworthy issue concerns some problems of methodology. In connection with the discriminant analysis technique, Sargen (1977, pp. 28-29) points to some intriguing questions, a number of which are pertinent to the other techniques. In the first place, the instances of rescheduling are few in.relation to the total sample. The data for the rescheduling group suggest that they are not normally distributed; a theoretical assumption of the discriminant analysis technique. Second, the explanatory variables are serially correlated: a high debt service ratio in one year is followed by high ratios in subsequent years. The presence of serial correlation implies that a country which is misclassified (or correctly classified) in one year, is usually misclassified (or correctly classified) in other years. Third, the treatment of countries that have rescheduled more than once, poses problems. If one is interested in the factors that determine the process of rescheduling per se and not the times of rescheduling, then observations that relate to rescheduling countries in nonrescheduling years should be omitted. Fourth, in all the models, the implicit assumption is made that the parameters of the model do not change over time and no structural shifts affect the model. Because of the small number of rescheduling cases, the testing for structural shifts in the parameters is not feasible. One final shortcoming of all the models estimating the probability of default concerns the lack of any theoretical basis on which the selection of





43

explanatory variables is made. Far from testing theoretical hypotheses, the models search for statistical relationships that will provide the best results in terms of the fewest Type I and Type II error rates.

In conclusion, two issues neglected by the formal models delineated above appear worth mentioning. In the first place, as Eaton and Gersovitz (1981b, p. 29) point out, one must distinguish between exogenous variables determining the probability of rescheduling and endogenous variables determined simultaneously with the rescheduling of debt. In the Feder/ Just model, a decrease in capital inflows may increase the probability of rescheduling, but, at the same time, lower capital inflows may be the result of information that a country will soon seek to have its debt rescheduled. The simultaneous determination between the probability of rescheduling and the amount of capital inflows is taken up in Chapter 5, where a simultaneous equations formulation is estimated.

The second point concerns the role of savings and investment in the rescheduling process. Sachs (1981, 1982) has argued, both from a theoretical standpoint and with actual data, that the savings rate may be an important determinant of the probability of rescheduling. He provides evidence (Sachs 1981, p. 246) to show that five of six countries that rescheduled their debt in the 1970s had significant decreases in their savings rates in the period prior to the rescheduling. Econometric evidence is also available from Kharas (1981) to support the hypothesis that the probability of rescheduling decreases with increasing investment rates. In view of the evidence, it is rather surprising that only the study by Mayo and Barrett (1978) has included gross fixed capital formation as an explanatory variable. The importance of this variable will be further examined in Chapter 5.





44

Real Interest Rates and Commercial Bank
Involvement in NOLDC Lending

The NOLDC Lending Process from the Commercial Banks' Perspective

The NOLDC lending operations of international commercial banks have raised a number of questions. The first issue that must be tackled is why commercial banks have emerged as lenders to NOLDCs. The classical answer to the question of why banks arise is that they fulfill the function of financial intermediaries: they channel funds from surplus to deficit units. This is precisely their role in lending to NOLDCs. They channel funds from.countries with a surplus in their current account (the OPEC members and a small number of European countries that maintained their surplus in the 1970s) to countries in deficit in their current account.

The question arises as to why surplus countries do not lend directly to the deficit countries. Solomon (1977, p. 480) points out that since 1973 the OPEC countries have made some direct investments in NOLDCs and have also provided some grants, but these have been mainly motivated by political considerations. In fact, the majority of the flow of funds
28
from OPEC to NOLDCs has been on concessional terms. The Bank for International Settlements (1981, p. 97) reveals in its annual report that OPEC made long-term investments in developing countries of 4.9,

6.5, 9.6 and 6.6 billion dollars in the four years 1974, 1975, 1979 and 1980. These amounts are to be contrasted with current account deficits in the NOLDCs of 46.5, 32.9, 57.6 and 82.1 billion dollars (from Table 2).



2Shihata and Mabro (1979, p. 163) report that in 1977, of a total 7587.8 million dollars of net disbursements from OPEC to NOLDCs, 5740.9 million was on concessional terms and only 1846.9 on nonconcessional terms.





45

The emergence of commercial banks as lenders to NOLDCs and the

unwillingness of OPEC to lend directly to NOLDCs can be explained in terms of risk aversion on the part of OPEC. The OPEC members chose to invest their surplus funds in relatively liquid and riskless assets in western countries, mainly in the form of bank deposits and government securities. International commercial banks found their deposits swollen. Therefore, by lending to NOLDCs, they have assumed the role of risk takers for which they are better suited than OPEC. One other reason for the intermediation function of banks is information economies: banks can pool deposits from several sources and lend to a small number of countries, about which they are well informed and can monitor their performance closely.

Eaton and Gersovitz (1981b, p. 37) add that a further reason why OPEC members do not lend directly to NOLDCs is that, in the event of a default, the OECD countries would be able to impose relatively more severe sanctions than OPEC. Although this is quite probably true, it is not at all certain whether the OECD governments would be willing to impose sanctions on behalf of their banks. The historical record shows few instances of governments intervening to secure the assets of their commercial banks.29 Indeed, as the 1982 conflict between Britain and Argentina has shown, governments are more likely to use bank lending as a political weapon.

From a global standpoint, the commercial bank lending operations,

vis-a-vis the NOLDCs, raise the issue of whether the world's capital stock is optimally allocated between developed and developing countries. Aliber (1977) and Eaton and Gersovitz (1981b, p. 1) make a case for commercial



29In discussing this point, Sachs (1982, p. 36) provides some sporadic historical instances of government overseas intervention to force payment of debts.





46


bank lending to NOLDCs as the transfer of capital from developed countries, where capital is abundant and the rate of return relatively low, to developing countries where the rate of return is high. To quote Aliber: "The rates of return in the developing countries are substantially higher than those in developed countries, and rates of economic growth in the developing countries are higher than in the developed countries. Both suggest that world income is enhanced by the allocation of capital from the developed to developing countries."30 This view holds that, so long as differences in growth rates (proxies for the return on the marginal investment) between the'developed and developing countries are large in relation to the higher risks associated with the flow of funds to developing countries; then the world's welfare will be raised if larger credits are allocated to developing countries.

The role of commercial banks as financial intermediaries raises the question of whether they can be replaced by an international organization, such as the International Monetary Fund (IMF), if they are seen to be ineffective in their lending, charging too "high" an interest rate or lending too "little." Leaving aside the question of what constitutes too high an interest rate or too little lending, Solomon (1977, pp. 500-501; 1981, p. 605) answers the question in the affirmative but provides no explanation for his position. Later on, however, he argues (Solomon, 1981, p. 606) that the proper role of the IMF should not be to "bail out" banks but to supplement bank lending through its various facilities, enforce stabilization programs to revitalize the country's economy and provide the country with a "seal of approval."



30Cited in Long and Veneroso (1981, p. 514).





47

What the proper role of international institutions should be in the lending process to NOLDCs, was probably best summarized by Eaton and Gersovitz (1981b, p. 36). First, it is to collect and disseminate information concerning the economic conditions in each country, leaving the banks to undertake the risk-taking function for which they are optimally placed. Second, to organize lenders and provide a central coordinating body in the case of default by a NOLDC. Third, to act as a lender of last resort. They argue that the first two functions are potentially stabilizing whereas the last one is potentially destabilizing because it may encourage commercial banks to take on excessive risks in their lending to NOLDCs. As for the second function, the last few years have seen coordinated arrangements by commercial banks in rescheduling NOLDC debts under the auspices of the Paris Club, after a country has entered into negotiations with the IMF concerning an appropriate stabilization program to pursue.

One issue worthy of note is the geographic concentration of banks' lending to NOLDCs. The U.S. banks have tended to concentrate their lending in Latin American countries and European banks in African and Asian countries. Eaton and Gersovitz (1981b, p. 15) suggest three reasons for the concentration. First, drawing on their empirical results (Eaton and Gersovitz, 1980), they argue that only a small number of countries are needed for sufficient diversification. Second, economies of scale are derived from concentrating on the performance of only a small number of countries.31 Third, the governments of the lending banks can impose sanctions on different countries with varying degrees of effectiveness.



Here, the role of the IMF as a disseminator of information takes on added significance.





48


For instance, the U.S. government may be able to impose more effective sanctions in Latin America than in Asia. However, given the point raised earlier concerning the reluctance of governments to act on behalf of their commercial banks, this argument does not appear relevant. Finally, one reason that Eaton and Gersovitz (1981b) fail to mention but could be quite important, is the influence of tradition. Many African and Asian countries were formerly colonized by the European nations and today, as independent states, they continue their commercial ties with the countries they were once colonized by.

One very important issue concerning the relationship between international commercial banks and the borrowing countries, is the likelihood that a country would seek to default on its loan commitments or demand a rescheduling of its debt. Formal mathematical models attempting to estimate the probability of default, were reviewed in the previous section. From a somewhat different perspective, Sachs (1982) has examined the history of LDC defaults and has drawn the conclusion that the strategy of ommmercial banks vis-a-vis the borrowers can be modeled as a game: a noncooperative game before World War II, resulting in a series of defaults, and a cooperative game since the War, resulting in only one default (North Korea) and a number of reschedulings. Sachs (1982) and Sargen (1977, p. 21) provide an interesting insight concerning the end result of loan reschedulings: the commercial banks have emerged from rescheduling negotiations, having lost very little in terms of the value of their principal or interest. The same conclusion is reached by Feder and Ross (1982) who have examined the relationship between bankers' subjective estimates of rescheduling risks and the credit terms in the Euromarket. From their empirical results they have argued that the expected loss rates, both in





49

the grace period and the rest of the loan period, have been quite low, confirming the previous assertion regarding losses experienced as a result of loan reschedulings.

In the light of this analysis the assumption of Eaton and Gersovitz (1980, p. 5) that default by an LDC prevents its future reentry in the international financial markets, must be reevaluated for at least the postwar period. The banks have rescheduled loans rather than allow a country to default, so as not to have to face the issue of reentry after default. One reason may be the fear of one default sparking a whole series of defaults, much like a run on the deposits of domestic banks following the collapse of a single bank. Although the likelihoodiof default does not appear significant at present, the lessons from history must not be forgotten. The large scale defaults of developing countries in the 1930s, occurred precisely at a time when the financial community least expected them.

The question of widespread default among NOLDCs deserves further attention. Eaton and Gersovitz (1980, pp. 14-19) have argued that such an event is more likely if the exports of borrowing NOLDCs are highly correlated. Their empirical results32 demonstrate that the export performance of individual NOLDCs are not correlated and they conclude that default is likely to be confined to one individual country rather than involve a series of NOLDCs.

Widespread default may not only originate from the LDC side of the market but also from the supply side of the market, as suggested by Sachs (1982, p. 7). In the case of a panic following the default of one NOLDC



32Based on factor analysis and multidimensional scaling.





50


commercial banks may refuse to keep credit lines open to NOLDCs, pushing some of them to default. He suggests that supply elements may have been responsible for the widespread default of developing countries in the midst of the Great Depression in 1931 and 1932 and that history almost repeated itself in 1974. The question arises as to whether a lender of last resort would be able to prevent such widespread defaults. On this issue, Sachs is somewhat pessimistic and argues that it is doubtful whether an international organization, such as the IMF, could continue to keep credit lines open to NOLDCs in the case of a major panic.

One important development in the last decade, pointed out by Long (1980, p. 488), Long and Veneroso (1981, p. 515) and Kareken (1977) in his discussion of Solomon's (1977) paper, is the increasing share of NOLDC loans in the portfolios of many international banks. Long believes that as a result banks, on the one hand, will be reluctant to increase the NOLDC loan portion of their portfolio while, on the other, bank supervisory agencies and other bodies are likely to introduce rules and apply pressure on banks to lower their holdings to NOLDC loans.33 Kareken believes very strongly that commercial banks should cut back on their international lending to NOLDCs. He argues that the result would not be




33As witnessed recently in the U.S. by the joint regulatory package recommended by the Federal Reserve the Comptroller of the Currency and the Federal Deposit Insurance Corporation, which would require banks to set up a reserve against earnings on loans where the full interest has not been paid for six months or where there is no prospect of compliance with the IMF measures. The reserve would be 10 percent of all problem loans initially, rising to 15 percent later. Banks would also be required to spread loan fees across the maturity of the loan and increase their capital base.





51


a decrease in aggregate demand since the funds would be directed as loans elsewhere. Other writers, in their discussion of Solomon's paper, disagreed with Kareken, arguing that it would decrease aggregate demand, disrupting the development plans of many NOLDCs. Kareken also makes a strong case for the heavy regulation of the international lending of U.S. banks. Because the deposits of U.S. banks are insured, it induces them to take "excessive" risks in their lending. However, as Kareken adds, this is not a problem peculiar to the international lending of U.S. banks.

So far, what can be termed as the "orthodox" view of the role of commercial banks in the lending process to NOLDCs, has been outlined. It is quite clear that the indebtedness of developing countries it a subject of political economy proper, where economics and politics are inextricably linked. The very nature of the subject evokes suggestions, proposals and solutions, the origin of which can ultimately be traced to each writer's political convictions. While the remainder of this dissertation expounds on the orthodox view of the role of commercial banks, it may be worthwhile to devote a few lines to some alternative views.

Lichtensztejn and Quijano (1982) provide a lengthy discussion of the increasing privatization of the external debt of developing countries. The increasing involvement of private international banks has worsened the terms of indebtedness and has imposed a severe burden on the balance of payments of developing countries. They point to an important development in the market for international loans: a shift has occurred from evaluating countries on the basis os their ability to pay, to an overall evaluation of the policies of each country concerning its receptiveness towards foreign capital, the so-called "country risk" approach.





52

The shift in emphasis is important, in their estimation, as they view lending by private commercial banks as a mere extension of the penetration of transnational capital into developing countries, accompanying and financing direct productive investments. The lending by private banks has enabled them to increase their influence, either directly or under the auspices of the IMF, on the economic policies pursued by developing countries. These policies have promoted an export orientation to the development efforts of LDCs, so as to be able to obtain the foreign exchange to repay the loans granted to them. The involvement of private banks in the internal affairs of developing countries, has extended to the encouragement of the concentrationsof industrial capital and the denationalization of the country's industrial base. The authors provide examples and case studies of several Latin American countries to support their views.

Similar views can be found in L'Heriteau (1979), who argues that the financial capital loaned by commercial banks serves to strengthen the dominance of the "real" capital. The increasing involvement of private banks results in the ultimate subordination of national development plans to the exigencies of the banks and the IMF and the reorganization of domestic policies to promoting the country's exports. The increasing importance of private bank loans in developing country finance has resulted in a different composition of capital inflows into developing countries. While private direct investment comprised the majority of private capital inflows during the 1960s, loans were the major item in the 1970s. However, such a general account hides the disparities among individual developing countries. The countries most heavily indebted to the banks have seen their share of private direct





53

investments remain constant and even increase during the 1970s, because it is precisely in these countries that the international banking system can support the operations of multinational firms.

L'Heriteau makes one final noteworthy point concerning the origin of external indebtedness. Several writers have identified a "vicious" circle of indebtedness, according to which balance-of-payments deficits render necessary external borrowing to cover the deficit, which in turn increases the burden on the balance of payments on account of the debtservicing obligations. However, during the period 1970-77 several countries were in equilibrium or entertained a balance-of-payments surplus. For these countries, externaJ indebtedness did not result from balance-of-payments problems but rather from a conscious effort, on their part, to attract foreign direct investment. In Venezuela, the deficit in the balance of payments first appeared in 1977 but by the end of 1976, the country had found itself with an external debt exceeding 3 billion dollars. Such examples point to the opposite causation, between external debt and the balance-of-payments deficit, from that envisaged traditionally.

It is, perhaps, not surprising that there should exist disagreements and diametrically opposite points of view regarding the role of commercial banks in international lending. It is, after all, a topic that arouses the interest of various parties, be they politicians, administrators or economists. What may be somewhat surprising is the treatment of the subject by academic economists. While a huge amount of writings has appeared on the role of commercial banks in lending to NOLDCs, the approach has



34For example Gabon, Iraq, Jordan and Malta and other, more heavily indebted countries such as Taiwan, Iran,Syria and Venezuela.





54

been rather superficial, relying, on the whole, on general discussions of the involvement of commercial banks, illustrated by statistical tables pointing out general trends and different lending practices. There has been virtually no attempt at approaching the subject from a theoretical point of view, making use of existing banking theory models or developing new models to examine the lending behaviour of international banks and the implications of the banks' lending practices on the flow of funds to developing countries. The next chapter follows up on this theme. Movements in Real Interest Rates

The movement of real interest rates over the last few years has influenced, to a great extent, the pattern of lending to NOLDCs. :The decade of the 1970s saw a gradual but steady rise in real interest rates. While at the beginning of the decade real interest rates were very low and even negative, they rose sharply towards the end of the decade. As Figure 1 shows, real short-term interest rates rose in many developed countries' financial markets throughout the 1970s; the rate on threemonth Eurodollar and other Eurocurrency deposits rose steadily throughout the late 1970s as shown in Figure 2. Since increasingly many loans to developing countries are of the floating rate variety, where the rate is linked to those in the major capital markets, the rise in interest rates has imposed an increasing burden on borrowers. In addition to the obvious one of higher interest payments, the overstatement of the current account deficit was explained in a previous section.

Sachs (1981, pp. 238-241) has argued that nominal interest rates were kept low in the early 1970s for two reasons: first, the shift in strategy in developing countries from a policy of import substitution to one of export promotion and the subsequent liberalization of capital flows





55









%
s UnneOd S;js at Germnny






S Japan United Kingdom 4-K-4



\ l






Belgium Frnce -1






SILLLi L.L L _LLJrLLLLjL iJ
1977 1973 197 197 197 1972 1973 19734 17 95 1983 t
1971 1978 197 1 0 1 1T 1977 17 197 1 0 11


1972-76 --- 1977-81


Figure 1
Real Short Term Interest Rates


Source: Bank for International Settlements, Annual Report,
Basle: BIS, 1981, p. 59.








56


















% %
22 C as:22

20 Dierental: I- 20
Eurod_or rate minus US CD rate
18 Euro-COM ra:e minus Crman interbank rale 18
.. Euro-S.vss franc rat minus Swiss interbank rate

16 b US dollr 16 14 14 12 r :12 10 Deusche fMerk 10






-l
1Sw;ss franc 4


2- :. 2 0 ':.... ',".'"'"":__ 0___O 0


Diferpen oeals Over D Ra e






-2

1978 1979 1980 181










Figure 2

Interbank Rates on Three-Month Eurocurrency

Deposits and Differentials Over Domestic Rates




Source: Bank for International Settlements, Annual Report,

Basle: BIS, 1981 p. 102.





57

and second because of depressed investment opportunities in the developed countries. He attributes the reduction in investment opportunities to several factors. First, investments in many developed countries had reached diminishing returns by the early 1970s. Second, most developed economies have suffered from a sharp reduction in profits and an increasing share of labour in GNP and third an investment boom in raw materials took place in the NOLDCs.

The problem of rising real interest rates has been emphasized by

many writers. Solomon (1981) has examined the economic performance of the major borrowers in international financial markets on the assumption of a reduction in real interest rates and:has found them all creditworthy. However, as he points in his conclusion, if his assumption does not materialize, the economies of the borrowing NOLDCs will be severely disrupted with serious repercussions on the world economy.

Long (1980, p. 485) and Long and Veneroso (1981, p. 511) point out that while real interest rates have been rising, the proportion of loans made to NOLDCs on concessional terms has been declining and at the same time the proportion of floating rate loans has been rising, exacerbating the debt burden of NOLDCs. Long (1981, p. 296) shows that for a country like Peru, the debt of which is equal to one half of its GNP, a rise in real interest rates of two percent is equivalent to a tax of one percent of GNP, not an insignificant amount.

On the basis of a simulation exercise, Long (1980) has calculated
the limiting value of the ratio of debt to GNP and the number of years for the ratio to rise from 30 to 40 percent, both for different values of the real interest rate and the rate of growth of GNP. His results point to the crucial role of the real interest rate in the debt accumulation





58

process. In fact, a tradeoff exists between the growth rate of GNP and the real interest rate: approximately a 1.5 percentage point increase in the growth rate of GNP is required to offset a one percentage point increase in the real rate of interest, if the limiting value of the debt/ GNP ratio and the period for the ratio to increase by 10 percentage points is to remain constant.

The increase in nominal interest rates during the 1970s can be

attributed, to a large extent, to the increasing inflation rates in the major economies. When interest rates on loans are fixed, an increasing inflation rate benefits ,the borrower by decreasing the real value of the loan. However, as Nowzad (1982, pp. 166-67) points out, under variable interest rates, which incorporate an inflation premium, the higher interest rates resulting from the increase in the inflation rate may compensate for the drop in the real value of the loan. Actual interest payments, under variable interest rates, include a component that takes account of the effect of inflation on the real value of the loan and therefore the loan is amortized at a faster rate, in real terms, than was originally expected.

If the real interest rate remains unchanged, the long run creditworthiness of a country is unaffected by inflation and higher nominal interest rates. However, the faster effective amortization rate alters the short-run borrowing requirements of a country. In order to maintain the same real net resource transfer, higher gross borrowing becomes necessary, which worsens the debt burden of the country in the short run.

Several forms of indexation have been suggested as solutions to the problem of faster amortization. Financial amortization would adjust amortization payments by an interest rate index. The same principle of





59


changing the amortization schedule to offset the faster amortization rate lies behind the proposal of maturity indexation. Price indexation of amortization payments, combined with constant or floating real interest rates on loans, is a further possible solution, but the choice of the appropriate price index or the appropriate real interest rate present sizeable problems. Indeed, as Nowzad (1982, p. 168) has emphasized, all the indexation proposals have technical deficiencies, lack widespread support and do not contribute in any way in settling the uncertainties caused by inflation and interest rate movements; such uncertainties as regards the real or nominal debt service and the real or nominal borrowing requirements. The implementation of any of the indexation scheme' does not appear likely, at least in the near future. External Borrowing, Consumption and Investment

One crucial aspect of commercial bank loans to NOLDCs, is whether they are used to finance consumption or build up productive capacity. This issue is of extreme importance because it is instrumental in determining the amount loaned to a country, the interest premium charged and the overall credit standing of the economy. Loans used to sustain consumption levels that may have been sharply reduced due to adverse exogenous factors, will, in all likelihood, bear a higher interest rate and involve smaller amounts. On the contrary, loans which are used to invest in projects that earn a positive return at the interest rate charged, should provide a steady future income stream that can be used to repay the loans and will, therefore, bear a lower interest premium. Having drawn the distinction between consumption and investment loans, it may be useful to point out that in practice it is difficult to distinguish readily between the two types.





60

Sachs (1981) has argued that, since 1973, two events have made a profound impact on the current account: the drop in real interest rates during the beginning and middle years of the decade stimulated investment and the rise in oil prices inflated the oil bill out of proportion. On the basis of a theoretical model he demonstrates that the rise in oil prices will have a differential effect on the current account of the oil-dependent economies only when the rise in oil prices is perceived as temporary. If it is perceived as permanent, there should be no differential impact on the current account of oil-dependent economies as compared to oil-exporting countries.

The main proposition put forward-by Sachs (1981, p. 211) is that when perfect capital mobility is assumed, shifts in investment patterns between countries lead to corresponding shifts in the current account. While higher oil prices since 1973 may have forced countries to run higher current account deficits, these deficits also reflect borrowing by these countries in response to investment opportunities.35 Thus in the 1970s two effects on the current account of NOLDCs can be distinguished: first, the rise in oil prices and second the drop in real interest rates and the increase in investment opportunities. Therefore, the increased borrowing by NOLDCs may have been a result of higher oil prices and increased current account deficits or may have resulted from investment opportunities not available previously or some combination of the two. If the second hypothesis is valid then the higher borrowing will lead to a rising consumption path over time and there should be no



The previous section explained why investment opportunities shifted to NOLDCs in the 1970s





61


problem with loan repayments; if the first argument is valid, grave doubts can be raised concerning the repayment of NOLDC loans.

Sachs (1981, p. 233) argues that, between the two periods 1965-73

and 1974-79, the current account/GNP ratio of developing countries worsened more than that of developed countries as a result of increases in investment/GNP ratios in large LDCs36 which were in excess of the saving/GNP ratios; in addition, saving/GNP ratios rose for some of the LDCs and fell for others. By contrast, in most developed countries between the same periods both the investment/GNP and the saving/GNP ratios fell, the latter decline of greater magnitude. Sachs then goes on to argue that "much of the growth in LDC debt reflects increased investment and; should not pose a problem of repayment. The major borrowers have accumulated debt in the context of rising or stable, but not fall-ing, saving ratios" (Sachs, 1981, p. 243).

The argument of Sachs is supported by Solomon (1977, p. 498) who provides evidence in the form of a table of gross capital formation as a percentage of total absorption37 for ten major borrowers38 for the period 1970-76. In only two of the countries (Chile and Colombia) did the ratio decline after 1973. Solomon concludes that the fear that loans to developing countries are used to support consumption, is unfounded. In his discussion of Solomon's paper, Greenspan (1977) argues that aggregate


36The countries included in the analysis were Argentina, Brazil, Chile, Colombia, Mexico, Peru, Philippines, South Korea, Taiwan and Thailand. 37Defined as gross domestic product plus net imports of goods and services. 38The same countries as those investigated by Sachs (1981).





62

investment ratios may hide the fact that for some LDCs higher ratios may not represent investment in productive capacity, but rather military buildup.

In his follow-up paper, Solomon (1981, p. 601) presents further evidence, again in terms of ratios of gross fixed capital formation to total absorption for the period 1976-80, to support his earlier position. A different opinion on this issue is voiced by Eaton and Gersovitz (1981a, p. 301) who state, without providing any evidence, that the oil price rises in the 1970s and the recession in the developed countries prompted the LDCs to borrow for short-term adjustment purposes.

The question of whether loans are used for consumption or investment purposes must, in the end, be an empirical one. Sachs (1981, p. 245) has presented econometric evidence to support his position. The dependent variable is the spread over the London interbank offered rate (LIBOR) charged on loans to each country (i-LIBOR). The independent variable are the current account/GNP ratio (CA/GNP), the investment/GNP ratio (I/GNP), the per capita GNP (GNP/L) and the total debt/GNP ratio (D/GNP). Sachs' results are as follows:39

(i-LIBOR)79 = 1.35 1.83(I/GNP) 2.49(CA/GNP) (6.3) (-2.0) (-2.3)

-2
R = 0.16


39The numbers in parentheses represent t-ratios; the subscripts refer to years.





63

or


(i-LIBOR)79 1.38 1.89(I/GNP)78 3.17(CA/GNP)78 (5.7) (-2.2) (-3.2)

0.009(GNP/L)78 + 0.008(D/GNP)78
(-1.4) (0.2)

-2
R = 0.35


As can be seen, the coefficient of (I/GNP) in both equations is significant and negative indicating that a stronger investment performance implies a higher probability of repayment and therefore a lower interest premium is charged.

Two comments are pertinent to the equations estimated by Sachs, in particular the second one. First, the equation, as it stands, represents a supply of debt function and should form part of a simultaneous equations model supplemented by a demand for debt equation. Therefore the coefficient of (D/GNP) in the equation does not represent the coefficient of the supply function. This observation may help explain why the coefficient is not significant. Second, the level of investment will, in turn, depend on what the future interest premium charged on loans is expected to be.

A simultaneous equations model was estimated in an attempt to overcome these criticisms. The three-stage least squares estimates are as follows:40


40The first equation represents a supply of debt function. The second equation is an investment function where investment demand is a function of the expected interest premium in the following period and expectations are formed adaptively. The third equation is a demand for debt function. The variables endogenous to the model are (i-LIBOR), (I/GNP) and (D/GNP) whereas all the rest are exogenous or predetermined. All three equations are overidentified. The values reported in parentheses are asymptotic t-ratios and subscripts refer to years.





64


(i-LIBOR)79 = 1.626 2.636(I/GNP)78 + 0.677(CA/GNP)78 (3.728) (-1.794) (0.380) + 0.288(D/GNP)79 0.619DS78
(0.364) (-1.052)

(I/GNP)79 = 0.040 + 0.738(I/GNP)78 + 0.022(i-LIBOR)79 (0.864) (6.819) (0.804)

0.240(GB/GNP)79 + 0.248(GB/GNP)78 (01.383) (1.257)

(D/GNP)79 = 1.506 + 1.392(i-LIBOR)79 4.200(C/GNP)78 (1.284) (2.158) (-1.565)

2.023(CA/GNP)78 + 1.003DS78
(-1.724) (2.174)

where, in addition to the variables defined earlier, DS represents the debt service ratio, (GN/GNP) the ratio of the government budget surplus to GNP and (C/GNP) the ratio of private consumption to GNP. The threestage least squares results proved quite disappointing. Most coefficients are insignificant or of the wrong sign. In particular, the signs of the coefficient of (D/GNP) in the first equation and (i-LIBOR) in the third are the same (positive), indicating that the demand and supply equations have not been identified.41

One final question that needs to be investigated is whether the

risk premium (as measured by the spread over LIBOR) varies in a systematic



Two possible explanations come to mind concerning the disappointing nature of the results. First, the model is grossly misspecified and that variables assumed exogenous should be treated as endogenous. Second, if developing countries are constrained in their borrowing, as Eaton and Gersovitz (1980, 1981a, 1981b) have suggested, it makes absolutely no sense estimating models of the sort specified in this section without introducing a supply constraint explicitly.





65

fashion with the level of indebtedness of each country. Sachs (1982, p. 14) argues that in a world of perfect capital mobility there is very little tendency for interest spreads to rise as the indebtedness of a country increases. However, when the assumption of perfect capital mobility is dropped, he argues (Sachs, 1982, p. 19) that interest rate premia tend to rise as the level of debt for each country grows.

Empirical evidence to support Sachs' proposition is provided by Brittain (1977, pp. 377-80). The following equation was estimated:


(i-LIBOR) = 0.0541 + 1.1733DM + 0.0599(D/GNP) 0.0424(DM)(D/GNP) (Q.1650) (2.6616) (4.5339) (-2.6465)

-2
R = 0.5138 F(3,21) = 9.454


where DM is a dummy variable that takes on the value of 0 in 1974 and 1 in 1975 and 1976.4 Brittain confirms the hypothesis that the interest spread rises with the overall level of indebtedness. In addition he argues that a fundamental change occurred in the market in 1975-76 such that developing countries were viewed as greater risks.43

Brittain's equation was reestimated by the present author with data for the years 1974 and 1979 and the following results were obtained:


(i-LIBOR) = 0.0270 + 0.8392DM + 0.0590(D/GNP) 0.0615(DM)(D/GNP) (0.0692) (1.2816) (3.7764) (-1.8841) R2 = 0.5865 F(3,14) = 6.62



42Data from the World Bank is used for (i-LIBOR) and from Citibank for (D/GNP). F( ) is the value for the F-ratio statistic. 43The intercept increased from 0.0541 in the first period to 1.2274 in the latter period.





66

As can be seen, the results are similar to those obtained by Brittain. However, it must be pointed out that the positive coefficient obtained here for the variable (D/GNP) is a result of the strong positive correlation between (i-LIBOR) and (D/GNP) in 1974.44. When (i-LIBOR) is regressed on (D/GNP) for either 1975/76 or 1979 no significant correlation is obtained.45 Moreover, the same data from Euromoney used in Sachs' (1981) study does not provide any significant results when (i-LIBOR) is regressed on (D/GNP) for all developing countries. Eor the ten largest borrowers the coefficient is significant at the 10 percent level but not the 5 percent.46



44The simple ordinary least squares (OLS) regression of (i-LIBOR) on (D/GNP) for the year 1974 gave the following results:

(i-LIBOR) = -0.0279 + 0.0602(D/GNP) (-0.0442) (2.5156)
--2
S= 0.5133 F(1,6) = 6.33 Again t-ratios are shown in parentheses. 45The OLS regressions gave the following results for 1975/76:

(i-LIBOR) = 1.5908 + 0.0036(D/GNP) (11.1924) (0.9229)
--2
= 0.0505 F(1,16) = 0.85 and for 1979
(i-LIBOR) = 0.8662 0.0024(D/GNP) (5.2652)(-0.2713)
R2 = 0.0121 F(1,6) = 0.07
46The OLS regressions for all developing countries gave the following results:
(i-LIBOR) = 0.9270 + 0.0044(D/GNP) (7.6105) (1.1325)
R2 = 0.0353 F(1,35) = 1.28 and for the ten largest borrowers:
(i-LIBOR) = 0.4197 + 0.0187(D/GNP) (1.5697) (2.0397)
R2 = 0.3421 F(1,8) = 4.16.





67

The same criticisms that were voiced in connection with Sachs' model are appropriate here. The supply of debt function estimated by Brittain is part of a simultaneous equations system. Moreover, in the Brittain study the positive correlation obtained between the interest spread and the total level of indebtedness is a result of a strong relationship between the two variables during a particular year, but that relationship does not extend over time.

In a further attempt to examine the determinants of the risk premium, Sargen (1976, pp. 27-30) regressed the premium on a dummy variable used to distinguish between developed and developing countries, a year dummy (1974 or 1975) and the maturity of the commitment, for 67 loans to developed and 177 loans to developing countries from the 1974-75 period. He found that developing countries paid, on average, a premium of 140 basis points in 1974, whereas developed country borrowers paid about 25 basis points less on average. Maturity was negatively and significantly related to the risk premium but its magnitude was relatively small. Within the group of developing countries he found lower- and middleincome LDCs paid only about 10 basis points more than higher-income LDCs and Mexico paid about 25 basis points less than other higher-income LDCs. The coefficients of the debt service ratio and the inflation rate were both positive and significant but small in magnitude. Finally, the large trade deficits of developing countries in 1975 were reflected in higher premiums paid.

The models presented in this section serve to bring to the forefront an important issue namely the process of the determination of country risk





68

premiums. As of recent, interest premiums have not varied markedly among developing countries,47 despite obvious divergences in economic structures, potential and prospects among developing countries. Research in this area has not addressed this issue sufficiently and it is certain that further work is both desirable and necessary.



































47The data for developing countries from Euromoney show a range of weighted interest rate spreads over LIBOR of between 0.563 and 2.037 in 1979 and between 0.39 and 2.22 percentage points in 1981.












CHAPTER TWO
BANK LOAN RATE INDEXATION IN
THE EUROCURRENCY MARKET


Introduction

The previous chapter discussed the importance of the external debt of developing countries to the functioning of the international financial system and introduced it as one of the central themes shaping current economic relations between developed and developing countries. In addressing the different issues raised by the indebtedness of developing countries to commercial banks, it was pointed out that, despite the voluminous literature and the breadth of coverage, the vast majority of the contributions in this area, barring two exceptions,1 have centered on the ability of each country to service its debt from a "macroeconomic" perspective. The focus of attention has been the so-called "countryrisk" approach, according to which a country's prospects of repaying the funds loaned to it are gauged by a variety of economic indicators, usually in the form of aggregate ratios. These ratios are either closely examined to discover any trend that may exist, or are introduced into formal mathematical models estimating the probability of default. Apart from the two contributions cited, there has not been any attempt at approaching the question from the perspective of the international commercial banks granting loans to developing countries.


1
The two papers by Feder and Just (1977b, 1980).



69





70

The present chapter develops a model aimed at redressing this imbalance. It extends the models introduced by Jaffee and Modigliani (1969) and James (1982) to the international setting of loans made by commercial banks to sovereign borrowers, be they public or publicly guaranteed entities. The model introduces the risk of default as perceived by the lender,2 as an important determinant of the interest rate charged by commercial banks. In addition, it incorporates the practices of making loans under commitments and that of indexing loans to a particular rate, both of which are widespread in the Eurocurrency market.

The next section describes the model and its relevance to lending practices in the Eurocurrency market. The following section presents the theoretical results of the model with particular reference to the implications of loan rate indexation on the pattern of borrowing on commitments. The empirical results from testing the hypotheses derived from the model are presented in the following section and a final section summarizes the results.


The Model

Before a formal introduction to the model, it is worthwhile to explain, so far as possible, the rationale behind loan commitments. A loan commitment is an agreement between a bank and its customer, according to which the bank provides credit up to a maximum amount and on prespecified terms. There are usually two types of arrangements as regards the rate charged on the loan commitment: a fixed rate commitment



2What was termed, in chapter 1, the "subjective" probability of default.





71

stipulates a specific rate at the beginning of the contract period according to which borrowing takes place, whereas a fixed formula ties the borrowing rate to a particular rate in the market.3 There are also two different rules, depending on whether the bank or the customer determines the loan volume to be taken down, within the maximum amount specified by the agreement.

There seems to be some disagreement in the literature concerning the rationale behind credit commitments. Campbell (1978) argues that they provide insurance to the borrower from a change in credit standing.4 James (1982), on the other hand, views loan commitments as means of economizing on transaction costs, where there are "transaction specific assets"5 involved in the negotiation between the two parties, by the establishment of a long-term relationship. Regardless of the rationale behind loan commitments, some important implications concerning the degree of competition in the market and the pricing of loan commitments are evident. In particular, there is a high degree of competition in the market in the stages preceding the signing of the loan commitment, but once agreement is reached, the situation is that of bilateral monopoly: a single borrower faces a single bank.



3This is the practice of indexation, to which the remainder of this chapter is addressed.

41In the case of the fixed formula agreement, the premium charged (over the rate to which the loan is tied) remains constant. 5Such as the cost of establishing the credit standing of each borrower or the cost of initiating and processing loan applications from new customers.





72

The point of departure of the analysis developed in the present study, with that of the earlier works by Jaffee and Modigliani (1969) and James (1982), is that variables at the control of the borrower, namely the amount of exports and international reserves, are assumed to have an impact on the density function of the revenues generated by the borrower which are used to repay the loan, as that is perceived by the lender. In accordance with the earlier papers, the bank faces an uncertain return with respect to any one loan to a particular customer, due to the fact that adverse circumstances may imply that the borrower is unable to repay the full amount required under the contract agreement. Therefore, the amount of revenues generated by the borrower that can be used to repay the loan, e, is viewed as a random variable by the bank. However, it is assumed that the bank has a priori knowledge of the density function, f(.), of the random variable e.

In the international context of loans to sovereign borrowers, the density function f(.) is assumed to depend on two variables the borrower has control over: the amount of international reserves a country possesses pi and the amount of its exports xi. The importance of international reserves in determining the perceived creditworthiness of a country was discussed in Chapter 1. The studies of Feder and Just (1977a, 1977b), Feder, Just and Ross (1981), Frank and Cline (1971) and Kapur (1977) all point to the level of international reserves a country possesses as an important indicator of creditworthiness; a higher level of reserves implies a country is better able to meet its external debt obligations and therefore less likely to default. The same studies point to the role of exports, both as a static (the level of exports) and as a dynamic (growth of exports) indicator of creditworthiness. On the





73

importance of exports, Kapur (1977, p. 182) states: "The level and growth of exports in any individual economy have a bearing on its liquidity position, debt-service ratio growth in GNP and, as a combined result of these factors, ultimately the perceived creditworthiness of the country in terms of default risk." One further variable that the studies by Feder and Just (1977a) and Feder, Just and Ross (1981) have identifed as an important determinant of creditworthiness is the amount of capital inflows. However, as will be explained in Chapter 5, there exists a simultaneous determination between creditworthiness and the amount of capital inflows: while-a smaller amount of capital inflows may diminish a country's creditworthiness the smaller amount could be a result of knowledge that the country may seek a rescheduling of its debt. Therefore, the amount of capital inflows is omitted from the present analysis.

The density function for the ith borrower (as perceived by the bank) is denoted by fi(e; pi, xi) where


E[e] = f efi(e; Pi; xi)de = = g(pi, xi)


An increase in either pi or xi will move the entire density function to the right.6 The density function is assumed independent of the size of the loan made, although as Jaffee and Modigliani (1969, p. 852) point out, in the nonindependence case, the results would hold equally well, provided that the investment opportunity of the borrower is subject to decreasing returns.

The objective of the bank is to maximize expected profits by

choosing the amount of loan, Li, extended to the ith customer, given the



6Further details are given in Appendix A.





74

loan rate factor R By varying the loan rate factor, the supply function of bank credit is obtained. The expected profits to a bank from advancing a loan in the amount of Li, are given by Qi RiLi
i = RiL f fi(e; Pi', xi)de + f efi(e; Pi, xi)de pLi R.Li 0

Ci(L) (1) where

R. = 1 + r. = The interest rate factor
1
p = The cost of funds to the bank7

Qi = The upper bound on the revenues generated by the borrower

Ci(Li) = The administrative cost associated with a loan of Li,

Two assumptions are in order concerning the cost function: CI(Li) > 0 and C'(Li) > 0 i.e., the marginal cost is positive and increasing. It is worth noting that the term R Li represents the total amount to be repaid under the contract agreement.

In Equation (1), the first term represents the repaymentto the

bank if no default occurs, whereas the second term represents the repayment in the case of default. Maximization of (1) yields the following equation:

Ri[l Fi(RiLi' Pi', xi)] p C!(Li) = 0 (2) where




7Which is closely related to the London interbank offered rate (LIBOR).





75
RiLi
Fi(RiLi; pi' xi) = f. i(; Pi, x )de
0
is the probability of default. The following assumptions are advanced concerning the probability of default

DFi(.) aFi(.) aF.(.)
> 0 < 0 <0
DR iL Pi axi

Equation (2) can be solved implicitly to yield the loan supply curve as

Li = Li(Ri; p i' xi', ) (3)

As far as the demand side of the market is concerned, it is assumed that the demand for credit from the ith customer, Di, is inversely related to the interest factor Ri, the amount of international reserves pi and the level of exports x. as

D. = Di(Ri; pi' xi) (4)


3Di(*) 3Di() aD (*)
R i Pi xi

The demand function in Equation (4) is a residual demand after alternative sources of finance are taken into account.8 It is assumed that the country has limited access to alternative sources of finance, which is the reason for the negative slope (with respect to the interest factor) of the demand function.

A final, and rather crucial assumption, is put forward. It is held that an interest factor Ri exists that equilibrates the market


8This is the reason why the amount of international reserves Pi and the level of exports x. appear in the demand function.





76


Di(Ri', p, xi) = Li(Ri' pi' xi' P)


Substituting this equilibrium condition into (2), the following condition must hold true for the loan rate and amount set at the beginning of the loan commitment, to result in an efficient allocation of credit.


Ri[1 Fi(Ri[Di(Ri' Pi x)]; p i, x)] p


C (Di (R P, xi)) = 0 (5)


In (4) it was assumed that the demand curve for credit is downward sloping and therefore the bank has some degree:of monopoly power in setting the loan rate and amount. Thus, it may seem counter-intuitive for the bank to equilibrate the demand for credit with the supply. However, it must be remembered that before the loan commitment is negotiated the bank has no monopoly power; only after the agreement is signed, a situation of bilateral monopoly arises.

The equilibrium condition in (5) can be differentiated totally to yield some interesting and intuitively plausible results. In the two previously cited papers, Feder and Just (1977b, 1980) propose that the interest premium charged on loans to sovereign borrowers in the Eurocurrency market is directly related to the probability of default. Feder and Just proceed to argue that the probability of default is linked to the debt-servicing capacity of each country and that this link is most important in determining the interest premium charged on each loan. The debtservicing capacity, being an unobserved variable, is in turn determined by several economic variables including the ratio of imports to international reserves, the per-capita gross domestic product and the debt-





77

service ratio. Thus an indirect (and inverse) relation is established between the amount of international reserves and exports on the one hand, and the interest rate on loans on the other. This is an entirely plausible argument which will be shown, in this study, to result directly from profit-maximizing behaviour on the part of commercial banks.

In addition, the model outlined here extends the results obtained

by James (1982) for domestic borrowers, concerning the effects of indexing loans to a specific rate, to the international setting of sovereign loans. The model is particularly suited to the study of loans made to sovereign borrowers, because the interest rate charged on such loans is made up of two components: the London interbank offered rate (LIBOR) and anlinterest margin that reflects the perceived creditworthiness of the borrower. The loan is usually granted in the form of a revolving credit; it is renewed every six months for the duration of the commitment and whereas the interest margin remains constant, the base rate (LIBOR) changes in response to different conditions in the international capital markets.9

Finally, the model developed here provides a useful insight into the widespread practice of syndicating loans in the Eurocurrency market. As was stated earlier, the presence of transaction specific assets in loan negotiations has given rise to procedures aimed at reducing transactions costs. One such procedure is loan syndication, whereby one bank, known as the lead or agent bank, arranges for all the loan details and evaluates the creditworthiness of the borrower. It provides only part of the loan and for the remainder it organizes the participation of other



9A more complete discussion of lending practices in the Eurocurrency market is provided by Mohammed and Saccomanni (1973).





78

(usually smaller) banks. Loan syndication is one of the most important lending features in the Eurocurrency market and accounts for a substantial proportion of all loans generated in this market.10


Theoretical Implications of the Model

The results of specific interest to this study can be derived by total differentiation of Equation (5) to yield


[1 F.(Ri[Di(Ri; pi' x.)]; Pi' xi]dRi
1 1 1 1 1 1 1


R. {f.(Ri[Di(R ; pi' x.)]; pi' x.) (D.(R.; pi x.)
1 111 1 1 1 1 1 i1


Ri Di(Ri; pi' Xi)
+1 )dR.
DRi i


aD.i(Ri; Pi' xi)
+ fi(R [Di(R pi, x ) ; pi' xi)Ri Pi X d
Si i

aDi(Ri; pi' xi)
+ fi(Ri[Di(Ri' Pi' x.); pi, xi)Ri ax. dx}
1

aD.(R.; Pi x.) D.(Ri; pi' X)
C'(D.(R.i; p xi))[ dR. + Pi dpi


D.i(Ri; p ,i xi)
+ xi dx.] dp = 0 (6)
ax 1

The derivatives of interest are dRi/dp, dRi/dPi and dRi/dxi which can be computed by setting two of dp, dpi and dx. in (6) equal to zero to obtain
1 1


10A thorough discussion of the market for syndicated credits is provided by Goodman (1982) who points to the sixteenfold increase of the market from 4.7 billion dollars in 1970 to 76 billion in 1980.





79

dR.
i = 1/{[1 Fi(RiDi(*); pi, xi)]
dp 1


Rifi(RiDi(); pi', xi)Di(*)(l+DRi)

aDd(-)
C(Di())[ R ]} > 0 (7)


1
dR. Di(*) aoDi()
S = {R2fi(RiD (); i xi) + C ())[ 1 ] }/
dpi i pi P i

{[l Fi(Ri.D(.); Pi, Xi)] Rf i(R Di('); Pi, xi)Di')

BDi(*)
(+nDR C(D (*))[ aR ]} < 0 (8)
1 i

dR. aDRi() aD.(-)
dxi = { ifi(R .D (); pi, xi) x + 1(D. ))[ ax


[1 F.(RD (.); Pi, xi)] R.fi(R Di (); Pi, Xi)Di(*)

aD.(.)
(1+nDR.) C(D())[ R. ]} < 0 (9)
1 1
aD.(-) Ri
where nDR aR Di-If the demand curve for credit is assumed to be elastic11 so that

InDRi > 1, the signs of the derivatives follow from the earlier assumptions concerning the credit demand function and the cost function. There
is evidence from Feder and Just (1977b, p. 240), to suggest that the

demand elasticity for credit is indeed rather high.


11The same condition is required to hold true in the model of Feder and Just (1977b, p. 226), if profit maximization is to yield meaningful results.





80

The comparative static results of the model indicate that a rise in the cost of funds to the bank will prompt a bank to charge a higher interest rate on loans, a conclusion which is intuitively appealing and consistent with the results of James (1982).12 One further interesting conclusion derived from the model is that the interest rate charged on loans will be inversely related to the amount of international reserves and the level of exports of each country.

The inverse relation between the level of international reserves and exports on the one hand, and the interest rate charged on the loan on the other, has also been verified by Feder and Just (1977b, 1980). However, in their papers, by using the appropriate assumptions, they are able to prove that the probability of default (as perceived by the lender) does not depend on the interest rate charged or the loan amount. Subsequently, they derive an equation where the interest rate on the loan is directly proportional to the probability of default which, in turn, depends on the debt-servicing capacity of each country. It is further assumed that the debt-servicing capacity of each country is determined by several economic variables including the amount of international



12However, one must not forget that such a conclusion is predicated on the elasticity of demand being greater than one in absolute value. In fact, in the international banking context, one can give a plausible explanation for the case in which the sign of (7) is negative. If, for example, the cost of funds to the bank rises as a result of a decrease in the current account surplus of OPEC, the demand for credit from oil-importing LDCs is likely to decrease as their current account deficits are reduced. For the majority of LDCs, credit obtained in the international capital markets is directed, in one way or another, at financing the current account deficit. It could very well be that the decrease in credit demand is sufficient to decrease the interest rate on the loan even if the cost of funds has risen.





81

reserves and the level of exports. Thus, the link between the interest rate charged, on the one hand, and the level of international reserves and exports on the other, is established via the debt-servicing capacity of each country.

In this study, the inverse relationship between the variables concerned is derived explicitly within the framework of a profit-maximization model, where the probability of default is endogenous and depends on both the interest rate charged and the loan amount. The level of international reserves and exports enter the determination of both the demand and supply of credit, whereupon the claimed results are derived by assuming an equilibrating mechanism in existence in the market.

The model provides some interesting insights as regards the practice of loan indexation when, within the loan commitment, the bank determines the loan amount to be extended to the borrower. The implications of loan indexation on the pattern of rationing of different borrowers, in the event of an increase in the cost of funds to the bank, are examined.

Initially, it is assumed that there are two classes of borrowers:

prime and non-prime. In the international banking context prime borrowers are charged the LIBOR (or a very small premium over LIBOR). The perceived probability of default for prime borrowers is assumed lower than that of non-prime borrowers because of factors other than the level of international reserves or exports; for example, different perceptions concerning the political stability of each country. If loan contracts are indexed to a rate such as LIBOR, the changed in the loan rate associated with a change in cost of funds (p) is the same for all borrowers and is equal to the change in LIBOR, dRL/dp. From equation (7) it can be seen that with indexation, customers with a higher default probability than prime customers will have a lower interest rate adjustment, compared to a contract





82

without indexation. Therefore, if the cost of funds to the bank rises, these customers are likely to be rationed in the sense that the interest adjustment desired by the bank is greater than that possible under the indexed contract. A similar conclusion holds for borrowers with a lower elasticity of demand for credit where, again, the difference in elasticity is not the result of differences in international reserves or export proceeds. The conclusions reached are in agreement with those of James (1982) but not those of Blackwell and Santomero (1982), who showed that customers with the more elastic demand13 will more likely be rationed first.

The results obtained by James (1982) are seen not to extend when

one considers differences in international reserves. It is shown formally in Appendix A that, ceteris paribus, a country with lower international reserves (and hence a higher probability of default) is less likely to be rationed, when the cost of funds to the bank rises, provided the increase in the probability of default is small enough. This proposition is contrary to the results of James, who has argued that customers with a higher probability of default are likely to be rationed first. Countries with lower international reserves have more limited access to alternative sources of funds for their development needs and, ceteris paribus, one would expect their demand for credit to be less responsive to changes in the interest rate; alternatively their elasticity of demand for credit is expected to be lower. Indeed, Blackwell and Santomero have argued, within the context of a domestic-economy customer-relation model, that non-prime



13Blackwell and Santomero (1982) argue that such customers are likely to be the prime customers.





83

borrowers (those with lower international reserves here) will have a lower demand elasticity. They proceed to demonstrate that prime customers (with a higher demand elasticity) will most likely be rationed first.

In the present context, as shown in Appendix A, borrowers with a lower value of international reserves (non-prime) will be subject to a higher interest rate adjustment when loan contracts are indexed, compared to contracts without indexation. Therefore, when the cost of funds to the bank (p) rises, such customers are less likely to be rationed. The rationale behind this assertion is rather simple. A country with higher international reserves i.s perceived, according to the assumptions of the model, as having a lower probability of default. However, as shown in equation (8), it is charged a lower interest rate. Therefore, the loss in profit from rationing a country with higher international reserves is potentially smaller. If the decrease in probability of default from a higher level of international reserves is small enough, it will be least costly to the bank to ration such customers and will therefore be rationed first when the cost of runds to the bank rises.

All the above conclusions have been drawn with reference to the case where the bank decides on the loan amount to be advanced to each borrower. If the contract arrangements are such that, within the constraints of the loan commitment, the borrower can choose the loan amount, all the results follow suit except that there can be no credit rationing. Nonetheless, borrowers with lower international reserves will be worse off under indexation when the cost of funds to the bank rises as compared to a contract without indexation. Therefore such customers will be expected to reduce their borrowing on loan commitments during periods of rising cost of funds to the banks.





84

Regardless of the arrangements for determining the loan 'amount, the arguments put forward in this paper have some interesting empirical implications. Although one cannot advance any hypothesis concerning the total amount of loan commitments taken down by borrowers, it is possible to formulate a hypothesis as regards the composition of loan commitments taken down by all borrowers. In particular, as has been shown, the interest rate adjustment to borrowers with lower international reserves will be higher when loan contracts are indexed, than when contracts are not indexed. With LIBOR rises, the relative cost of borrowing on the loan commitments for borrowers with lower international reserves is higher (under indexation) than for borrowers with higher international reserves. Therefore, ceteris paribus, one would expect the proportion of loan commitments taken down by customers with lower international reserves to decline when LIBOR rises.14


Empirical Results

The theoretical results derived from the model in the previous section give rise to an interesting hypothesis which can be tested



Of course, it goes without saying that the argument is entirely
symmetric with respect to the other variable influencing the probability of default, the level of exports. While the whole of the analysis has focused on the absolute value of the variables exports and international reserves, it is evident, from the studies reviewed in Chapter 1, that the importance of these variables is gauged by reference to another variable in the form of ratios. In fact, in the empirical analysis which follows this section, developing countries are divided into two groups, more and less creditworthy, on the basis of two criteria: the ratio of total debt outstanding and disbursed to export proceeds and the ratio of imports to international reserves for each country.





85

empirically. In particular, it was argued that, if developing countries are divided into two groups (j = 1,2) on the basis of two ratios,15 then countries with lower values of international reserves or exports (and therefore higher values for either ratio) will be worse off when the cost of funds to the bank rises, in the sense that the relative cost of borrowing, on the loan commitments is higher when contracts are indexed compared to contracts without indexation. Therefore, when the cost of funds to the bank (LIBOR) increases, one would expect the proportion of loan commitments taken down by countries with higher values for either ratio to decline.

The hypothesis can be tested empirically by estimating the model


PROPtj = a + aD + a2LBRt + a3(LBRt)(D) + utj (10) where

PROPtj = proportion of disbursementsl6 through financial markets

to group j during time period t

D = dummy variable taking the value 1 for the group of countries with a low value for either ratio and 0 for the

group of countries with the high value.

LBRt = value of LIBOR during time period t.

Estimation of equation (10) is carried out twice: first, countries are divided into two groups according to the ratio of total debt outstanding



15The ratios are imports to international reserves and total debt outstanding and disbursed to export proceeds. 16Defined by the World Debt Tables of the World Bank (1981) as drawings on outstanding loan commitments.





86

and disbursed to exports and second the division is carried out according to the ratio of imports to international reserves. If the theoretical hypothesis of the model is to be verified empirically, then the coefficient of LBRt (a2) must be negative.

As far as the ratio of total debt outstanding and disbursed to

exports is concerned, annual data on disbursements were obtained, for a sample of 44 developing countries for the period 1971-1980, from the World Debt Tables17 of the World Bank.18 The countries were assigned to the high debt-export ratio group if the ratio of their total debt outstanding and disbursed to export.proceedsl9 exceeded 1.20 and to the low debtexport group if that ratio was less than 0.80. The data on LIBOR-were calculated as yearly averages of the end-of-month value of the six-month Eurodollar deposit rate, published by the Financial Times of London.

The model in (10) was estimated by ordinary least squares (OLS) for the period 1971-1980 and the regression results were as follows:20



17World Bank Publications EC-167/76, October 1976 and EC-167/81, December 1981.

18The 44 countries included in the empirical analysis accounted for 97 percent of the total disbursements through financial markets to all developing countries reported in the World Debt Tables during 1971 and for 94 percent during 1980.

19Data on debt outstanding and disbursed were obtained from the World Debt Tables and data for exports from the International Financial Statistics 1982 Yearbook, published by the International Monetary Fund. 20The t-statistic ratios are shown in parentheses.





87

PROP = 0.74 0.53D 1.43LBRt + 1.36(LBRt)(D) (10.06) (5.12) (1.85) (1.24)

--2
R = 0.93 F(3,16) = 66.43 D-!H 1.82

The empirical results support the main hypothesis of the study insofar as the coefficient of LBRt is negative and significant at the 0.05 level. The proposition that the proportion of loan commitments taken down by countries with low values of exports and high debt-export ratios declines when the cost of funds to the bank (LIBOR) increases, has been shown to hold true for a group of 44 developing countries during the 1971-1980 period. This result contradicts the earlier conclusion of James (1982). who has argued that countries with higher probabilities of default are more likely to be rationed first and that, under loan rate indexation, these countries are expected to increase their proportion of loan commitments taken down, when the cost of funds to the bank rises. Furthermore, although the coefficient of (LBRt)(D) is not significant at the 0.10 level, its sign is opposite to that of LBRt, indicating that the two groups of countries respond differently to changes in LIBOR.

As far as the group of countries with the low debt-export ratios (the more creditworthy) is concerned, it was argued in the previous section that when the cost of funds to the bank rises, the relative cost of borrowing remains unchanged for these countries, and is equal to dRL/dp. Therefore, the proportion of loan commitments taken down by this group is not expected to respond to changes in LIBOR. This assertion is corroborated by the empirical results which show that the proportion of




88

loan commitments taken down by this group responds only slightly21 to changes in LIBOR. In conclusion, it has been argued theoretically and demonstrated empirically that the proportion of loan commitments taken down by the group of countries with the higher debt-export ratios (the more creditworthy) does not respond to changes in LIBOR. However, for the less creditworthy group, the relative cost of borrowing on loan commitments is higher when loan contracts are indexed compared to contracts without indexation and, therefore, the proportion of loan commitments taken down by this group has been shown to decline when LIBOR rises.



The coefficient measuring the response of the proportion of commitments taken down to changes in LIBOR is -0.07 for the group of countries with low debt-export ratios compared to -1.43 for the group with the high ratios. 22The choice of cut-off points, according to which the countries are divided into two groups, can be criticized as arbitrary. However, the results do not appear to be sensitive to the choice of cut-off points. The countries were reassigned to the high and low debt-export ratio groups according to whether the ratio of their total debt outstanding and disbursed to export proceeds exceeded 1.50 or was less than 1.00. Equation (10) was reestimated to yield the following results:

PROPtj = 0.55 0.29D 0.58LBRt + 0.93(LBRt)(D) (6.74) (2.55) (0.68) (0.76)

2 = 0.73 F(3,16) = 14.20 D-W 2.03


Although the coefficients of LBR and (LBR )(D) are no longer significant at the 0.10 level, the signs of %he coeffi ients are in accordance with prior theoretical expectations.





89

One further point merits attention as regards the procedure used to estimate (10). Since the amount of total debt outstanding and disbursed in period t includes disbursements in period t as well as disbursements in previous periods, it is likely that the debt-export ratio for each country and the amount of disbursements are correlated; thus, the variable according to which observations are classified and the dependent variable are correlated. If such were the case, it would appear that application of OLS to (10) would yield inconsistent estimates. However, as shown in Appendix B, application of OLS to (10) yields consistent estimates. Nonetheless;- if the classification and the dependent variables are correlated, the constant term can no longer be interpreted as the same as that which would result from OLS estimation of (10) in the case where the correlation problem did not exist. Appendix B gives details of how the two constant terms are related.

Finally, countries were divided into two groups according to the ratio of imports to international reserves. Data on disbursements were obtained for 43 developing countries23 for the period 1971-1979.24 The countries were assigned to the high imports-reserves ratio according to




23The same data on disbursements as in the previous regression were used, except that Hong Kong was omitted due to the lack of data on international reserves.

24The year 1980 was omitted from the regression because several countries had unusually high imports-reserves ratios. The average imports-reserves ratio of 10.15 in 1980 was nearly twice as high as the average importsreserves ratio of 5.61 for the previous nine years. In the case of the Ivory Coast and Senegal, during 1980 the ratio showed a 23- and a 20-fold increase respectively over the average for the previous nine years.





90

whether the ratio of their imports to international reserves25 exceeded

5 and to the low imports-reserves ratio if the value of the ratio was less than 3,

The regression results were as follows:



PROPt = 0.22 + 0.33D 0.85LBRt + 0.79(LBRt)(D) (1.01) (1.07) (0.34) (0.23)


= 0,67 F(3,16) = 9.36 D-W = 2.61


As can be seen, neither of the coefficients is statistically significant. However, the signs and relative magnitudes of the coefficients support the main theoretical hypotheses advanced in this study.


Summary

This chapter presented a model to investigate the widespread

practice of loan rate indexation in the Eurocurrency market. The comparative statics of the model introduced the intuitively appealing proposition that the interest rate charged on loans to sovereign borrowers is inversely related to the amount of international reserves and the level of exports of each country. In contrast to previous work, it was

demonstrated both theoretically and (in the case of exports) empirically that countries with lower international reserves or exports (which are more likely to be the non-prime borrowers) are less likely to be rationed when the cost of funds to the bank rises. Furthermore, it was shown that the proportion of loan commitments taken down by these countries is expected to decline when the cost of funds to the bank rises.



25Data for both international reserves and imports were obtained from the International Financial Statistics 1982 Yearbook.





91

The implications of the results for the participants in the international loan markets are all too obvious. According to the conclusions of this study, a developing country which suffers a reduction in its international reserves or exports during a particular year, need not worry overly about its being rationed out of the credit markets. However, a crucial assumption must be met in order for this conclusion to hold true: the increase in the probability of default from the reduction in international reserves or exports must be sufficiently small. As for the international banks, knowledge of the response of developing-country disbursements to changes in the cost of funds to the banks, can better aid them in the optimal allocation of their portfolios.

While a great deal has been written about the ability of developing countries to service their external debt from a "macroeconomic" perspective, very little work has appeared that approaches the question from the perspective of the commercial banks granting loans to developing countries. The present chapter has contributed in the latter direction by extending a model of loan rate indexation, devised originally for the analysis of domestic banking practices, to the international setting of loans to sovereign borrowers in the Eurocurrency market.













CHAPTER THREE
A DISEQUILIBRIUM MODEL OF EUROCURRENCY
FINANCE TO DEVELOPING COUNTRIES


Introduction

The present and following chapters present and estimate an analytic model of a widely discussed, but little researched, facet of the external borrowing of developing countries, that of the setting of credit ceilings or country exposure limits on the amount each country may borrow from commercial banks. Apart from the contributions by Kapur (1977) and Eaton and Gersovitz (1980, 1981a, 1981b) there has not been any systematic analysis of this widespread practice among international banks granting loans to developing countries.

The setting of maximum limits on the amount of funds a bank will loan to a developing country has figured prominently in developingcountry lending. It has been described by various terms including country exposure limits, credit ceilings or quantity rationing. Some of the major international banks have well-established procedures for evaluating the creditworthiness of each country and establishing credit ceilings.

The work of Eaton and Gersovitz (1980, 1981a, 1981b) has been the

sole attempt at incorporating credit ceilings in an econometric estimation of the market for international borrowing by LDCs. Their study employed the method of maximum likelihood to estimate a model of the demand and supply of international debt by LDCs. Both demand and supply were formulated as functions of various economic characteristics of each


92





93

country. The novelty in their work was to assume that the actual amount of debt observed is the minimum of the amount demanded by each country or the quantity supplied (credit ceiling) by the bank. Thus, in their model, the credit ceiling was determined endogenously within the model. Using disequilibrium econometrics techniques, they obtained estimates of the demand and credit ceiling equations and demonstrated that 56 of 81 countries included in the sample were classified as constrained with probability greater than 0.5.

A different approach to incorporating credit ceilings is taken in this study. A demand for and supply of international debt by LDCs are postulated, which conform to the usual:assumptions of economic theory. However, the market is characterized by a ceiling on the maximum amount of credit a country can obtain, which is set exogenously by the banks. The actual quantity of debt observed represents either the equilibrium quantity of debt derived from the intersection of the supply and demand functions in the case where the credit ceiling is not effective, or the amount of the credit ceiling if that should prove to be less than the equilibrium quantity of debt.

The following section introduces the theoretical rationale behind actions taken by commercial banks to limit the amount of credit to a particular borrower. The practical implications of credit rationing on the setting of country exposure limits are then discussed. The model to be estimated in this study is outlined briefly in the following section and justification for its application to the market for international debt is provided. The chapter concludes by further elaborating on the model, which is to be estimated in the following chapter, and draws attention to some of the features that distinguish it from other work in this area.





94

The RationaleBehind Credit Ceilings in International Banking


Theoretical Arguments

The question of why a rational profit-maximizing lender would wish

to ration credit to a borrower has received considerable attention in the theoretical literature. The earliest attempts at explaining this phenomenon attributed credit rationing to sticky interest rates resulting from oligopolistic conditions in credit markets and to legal ceilings on interest rates. These explanations viewed credit rationing as a purely temporary phenomenon until the interest rate could adjut to a new equilibrium rate following a change in:borrower demand or actionsitaken by the monetary authorities to control the monetary aggregates.

The first systematic attempt at providing a theoretical explanation of credit rationing was offered in an early paper by Hodgman (1960) who based his argument for credit rationing on lender attitudes towards risk. He defined the risk of a loan as the ratio of the expected value of the payment to the lender, to the expected value of the payments below the amount originally contracted. The objective of the lender is to maintain the loan-risk ratio above a predetermined ratio. As the amount loaned increases, the loan-risk ratio can be maintained above the predetermined level by increasing the interest rate. However, above a certain loan size, raising the interest rate will not reduce the loan-risk ratio.




ICredit rationing is defined here as the practice of setting a maximum limit on the funds to be loaned to a specific borrower, beyond which no further funds can be obtained irrespective of the interest rate the borrower offers to pay.




Full Text

PAGE 1

ASPECTS OF THE EXTERNAL DEBT OF DEVELOPING COUNTRIES By ANDREAS SAVVIDES A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1983

PAGE 2

ACKNOWLEDGMENTS I would like to thank the chairman of my supervisory committee. Professor G. S. Maddala, for his advice and continual guidance and encouragement throughout the preparation of this dissertation. The other members of my committee. Professors W. Bomberger, R. Crum and D. Denslow, deserve particular recognition for providing freely their assistance and for their efforts in ensuring that the task was carried through to its completion. In the preparation of such a study, the contribution of colleagues and fellow students cannot be overestimated. I would like to thank, in particular, my colleague and friend, Haralambos Sourbis, for finding the time to discuss problems and projects. The typing and editorial assistance of Betty Sarra has been outstanding throughout. Finally, I would like to thank my parents and dedicate this work to them. Their contribution to its completion is without measure. i i

PAGE 3

TABLE OF CONTENTS Page ACKNOWLEDGMENTS ii ABSTRACT v CHAPTER ONE DEVELOPING COUNTRY BORROWING IN THE INTERNATIONAL CAPITAL MARKETS: AN OVERVIEW OF THE ISSUES 1 Introduction 1 Some Preliminary Comments and Definitions 3 Distribution and Value of the External Debt of Developing Countries 9 The Distribution of Developing Country Debt 9 Real and Nominal Value of LDC Debt 13 Constraints on LDC Borrowing and the Role of International Reserves 18 Is a Credit Ceiling Imposed on LDC Borrowing? 18 External Debt and International Reserves 21 Analyzing Country Risk: Debt Rescheduling and the Debt Capacity of LDCs 25 Growth-cum-Indebtedness Models 26 Estimating the Probability of Rescheduling 29 Real Interest Rates and Commercial Bank Involvement in NOLDC Lending 44 The NOLDC Lending Process from the Commercial Banks' Perspective 44 Movements in Real Interest Rates 54 External Borrowing, Consumption and Investment 59 TWO BANK LOAN RATE INDEXATION IN THE EUROCURRENCY MARKET 69 Introduction 59 The Model 70 Theoretical Implications of the Model 78 Empirical Results 84 Summary 90 THREE A DISEQUILIBRIUM MODEL OF EUROCURRENCY FINANCE TO DEVELOPING COUNTRIES 92 Introduction 92 The Rationale Behind Credit Ceilings in International Banking 94 i i i

PAGE 4

Theoretical Arguments 94 Credit Ceilings in International Banking Practice 98 A Model of Credit Ceilings in International Lending to Developing Countries 100 The Basic Framework 100 Elaborating on the Basic Framework 105 FOUR ESTIMATION OF A DISEQUILIBRIUM MODEL OF EUROCURRENCY FINANCE TO DEVELOPING COUNTRIES 109 Introduction 109 Disequilibrium Econometric Techniques: A Brief Methodological Review 110 Estimation of a Disequilibrium Model with Credit Ceiling... 112 Variables and Data Sources 116 The Maximum Likelihood Estimates 119 The Computer Program 119 The Empirical Results 119 Conclusion 123 FIVE A SIMULTANEOUS LOGIT MODEL FOR THE DETERMINATION OF DEVELOPING-COUNTRY CREDITWORTHINESS AND CAPITAL INFLOWS.. 125 Introduction 125 A Simple Equilibrium Model of the Supply and Demand for External Capital 126 The Importance of Economic Indicators of Single-Equation Models of Creditworthiness 131 A Simultaneous Logit Model of Creditworthiness and Capital Inflows 139 Conclusion 149 SIX SUMMARY AND CONCLUSIONS 151 APPENDICES A EFFECT OF CHANGES IN INTERNATIONAL RESERVES ON THE COMPARATIVE STATICS OF THE LOAN RATE INDEXATION MODEL.... 153 B IMPLICATIONS OF THE CORRELATION BETWEEN THE DEBT-EXPORTS RATIO AND THE AMOUNT OF DISBURSEMENTS 157 C FIRST DERIVATIVES OF THE LOGARITHM OF THE LIKELIHOOD FUNCTION FOR THE DISEQUILIBRIUM MODEL 163 D DATA FOR THE SIMULTANEOUS LOGIT MODEL 168 REFERENCES 174 BIOGRAPHICAL SKETCH 181 iv

PAGE 5

Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements of the Degree of Doctor of Philosophy ASPECTS OF THE EXTERNAL DEBT OF DEVELOPING COUNTRIES By ANDREAS SAVVIDES December 1983 Chairman: Professor G. S. Maddala Major Department: Economics This dissertation comprises a collection of essays focusing on several aspects of the external debt of developing countries. Initially, the many facets of this complex subject are presented and analyzed. The widespread practice of loan rate indexation in the Eurocurrency market is the first issue of concern. A theoretical banking model is introduced and the comparative static results reveal that contries with a lower level of international reserves or exports (usually the less creditworthy), are less likely to be rationed when the cost of funds to the bank rises. The empirical results support the theoretical hypothesis. The model directs attention away from studies investigating devel opi ng-country creditworthiness on the basis of country-wide economic indicators, and approaches the creditworthiness concept from the perspective of international commercial banks granting loans to developing countries. The market for international debt is subsequently investigated and a disequilibrium model of supply and demand for international debt is V

PAGE 6

estimated. There is ample evidence to suggest that an explicit credit constraint exists in developing country borrowing. If the constraint is ineffective, the equilibrium values of the variables are observed; otherwise, the observations correspond to the supply function. The appropriate estimation procedure is maximum likelihood and the first derivatives of the likelihood function are given. A computer program was written for the estimation of the parameters of the model. The empirical results proved quite disappointing. Despite repeated attempts, the likelihood function failed to converge. However, the lack of results should not be taken as evidence of the failure of the econometric technique but rather of the particular iterative procedure employed to obtain the estimates. Finally, the issue of the simultaneous determination between creditworthiness and the amount of capital inflows into a country is taken up. The results reveal that whereas the amount of capital inflows exerts a negative impact on the likelihood of rescheduling when the simultaneity between the two variables is ignored, the impact is reversed when the simultaneity is taken account of. An explanation is ventured as to the reversal of the sign of the coefficient of capital inflows. vi

PAGE 7

CHAPTER ONE DEVELOPING COUNTRY BORROWING IN THE INTERNATIONAL CAPITAL MARKETS: AN OVERVIEW OF THE ISSUES Introduction The accumulation of external debt by developing countries has emerged as one of the most pressing issues facing the world economy in the 1980s. As the effects of the tremendous increase in energy prices that wreaked havoc during the previous decade begin to loosen their grip on the world economy, another danger looms on the horizon and threatens the collapse of the international financial system. A number of developing countries have amassed debts to outside official and private creditors to such an extent that some of them cannot meet their external obligations as originally contracted. At the same time, several international banks have increased their exposure, vis-a-vis individual countries, to the point where their profitability and, to some extent, their solvency rests on the ability of a handful of countries to continue making interest and amortization payments. The World Bank (1981, p. iii) defines external debt as "debt owed to non-residents and repayable in foreign currency, goods or services that has an original or extended maturity of over one year." As Table 1 shows, the total debt of developing countries has risen from 109.4 billion dollars in 1973 to 529 billion in 1982, a fivefold increase. More importantly, however, the debt service owed (interest and amortization payments) has increased sixfold: from 16 billion dollars in 1973 to 95 billion in 1982. 1

PAGE 8

2 Table 1 Public and Private Debt of Developing Countries 1973-1982 (Billions of U.S. Dollars) 1973 1975 1976 1977 1978 1979 1980 1981 1982^ Total Debt 109.4 161.9 195.5 240.1 298.8 352.4 404.8 462.1 529.0 Total Debt Service 16.0 23.0 26.1 33.1 47.9 62.3 70.4 83.0 95.0 Low-Income Africa 0.4 0.5 0.5 0.6 0.7 0.8 1.1 1.3 Low-Income Asia 1 "O 1.3 1.3 1.4 1.6 1.8 1.9 1.9 Mi dd 1 eIncome Oil Importers 9.7 14.0 15.4 18.8 27.9 34.6 40.8 49.8 — Oil Exporters 4.9 7.1 8.9 12.4 17.7 24.9 26.7 30.8 — ^Estimates Indicates data not available Source: Finance and Development, Vol, 20, No. 1, March 1983, p. 23.

PAGE 9

3 Because of the magnitude of the external debt of developing countries as well as its implications for the stability and growth of the world economy, the subject has generated a considerable amount of interest on the part of economists. The intent of this introductory chapter is to present, in a brief manner, the major issues that have been investigated and the directions in which research in this area has proceeded. It is by no means intended as a comprehensive review of the literature. The sheer volume of writings on the subject^ makes a complete survey quite 2 impossible. One can only hope to touch on the different aspects of a complex and multi faceted topic. Some Preliminary Comments and Definitions The decade just ended was one of unprecedented turmoil in world economic relations and in particular the international capital markets. The price of oil quadrupled in 1973 and rose again sharply towards the end of the decade. A pattern, never encountered before, emerged in the balance of payments accounts of developed and developing countries. In particular, as has been pointed out by several writers (Sachs 1981; Solomon 1977, 1981), the pattern of moderate current account deficits of less developed countries (LDCs) matched by surpluses in the developed countries, changed dramatically. As Table 2 shows, the non-oil developing countries (NOLDSc) suffered large deficits in their current account after In addition to the bibliography reviewed here, a number of books as well as articles in journals and the financial press appear continuously offering different viewpoints and proposing various solutions. Some of the proposals will be discussed in the concluding chapter. 2 For instance, the literature on intertemporal optimizing models, which focuses on deriving optimal borrowing criteria or the role of external funds in smoothing the consumption path over time, is completely ignored. McDonald (1982) provides a useful discussion of such models.

PAGE 10

4 LO 0 — OD MO LO CTi 00 0 01 •I — CO C\J CM C3^ 1 — CM 00 3 1 1 1 1 1 1 1 1 fO (/I O3 0 QJ CO (/l QJ 1— o CM r~~ 00 00 r~ 0 0 Ct: Q. CO •1 — 0 00 CM CM 00 0 00 -M -l-J QJ Cl CM CD CM 00 (/) to Q_ 1 1 1 1 1 1 1 1 1 1 •I— +-> r— QJ ro ro -c CO CTl CO LO CO LO ro 0~i 0 +j r-M r-. c +J CD QJ QJ .— 1 — 1 CO CO CO LO 00 CO CM 00 CT) E I — CO >> > 0 o ro ro r~-. LO LO CM ro ro 0 +-> oo CO CO 00 1 — CL 00 CJ 1 — 1 1 1 1 1 1 %~ M— 4— 0 r— 0 0 •!— o l-D 0 CM U^ — QJ rt -P on CT) CO CM CO 0 LO CTl CM CJ 0 : E i00 CM sz •I1/) 0 ra 1 1 i 1 1 1 1 1 1 CD ro 4-} o r— 1 — :3 ro CT) o ro QJ E o c CM — CO 0 CTl 0 uo ^ Lf— O) •r— >> — ^1 S0 1 i. un C\J LCI 0 LO C\J CO KO J 0. CL C X! O) 3 4QJ •rE S1— O o CO LO LO <^ CD ro Sro ro 0 1 — CM ro =^ E Q. rci c 1 1 1 1 1 1 1 ^ cu QJ hO x: •r™ Q. o CJ t/1 0 •1 — CT) CO CO CO 00 00 LO 00 QJ S0) L(aj r3 ro CJ — cn — 00 cn 00 CTl s_ uj 3 c •1 — CO r~ iSE ro CQ 1 1 1 1 1 1 1 I QJ +-> C C E MQJ 3 Sro c c/) E 0 QJ CQ 1 — ro 1 — LO tXJ LO 0 LO 00 10 CO E 0 +J ro 23 >i oo o~i UO 1^ CM CM 0 CM >^ Cl i/i sro E 4_) QJ ro OJ LU £2 1 1 1 1 1 1 1 QJ QJ a -M Mcr 0 0 to ro ro QJ SI 00 n -M QJ jZ i/i •1 — -1 — c t/) 0 V) QJ 23 1 — 00 E QJ 00 CL 3 sl C_J ro f— QJ 00 3 0 s_ 0 0 0 (/) 40 CJ) S: E QJ £1 QJ rs c 0) 4J 0 CT) ^ • 0 •1 — 13 •r~ 0 1 — x: O) U QJ •r— S0 in SCJ O0. u QJ 0 0 X2 ro +-> +J QJ -M 0 to •1 — ^00 OJ E E ro £Z •1 — c: CP. •r— > 0 i. +-> QJ -1c :3 5c OJ -(-> B iSro 0 00 E JE t. o QJ +J 0 > 1/) ttJ QJ Q. +-> i•1E CJ dj o CJ) c: CJ -M QJ ro in LH ro Sr— 0 +-> SI c LO QJ cn -a E 0 Z5 Q) +-> QJ 0 >> CQ +-> c: X QJ 1 — TD Cl 3 c: OJ 0 0 ULU ct: ro E OJ 0 cn 1— ro 0 j2 ro tn

PAGE 11

5 1973, accompanied by large surpluses for the oil-exporting countries. The developed countries switched from a surplus to a deficit in their current account, although a few countries West Germany, Japan and Switzerland continued to enjoy a surplus in their current account. The increased current account deficits of the NOLDCs after 1973^ were accompanied by large scale borrowing by the NOLDCs from private 3 financial markets. The vast majority of the borrowing by NOLDCs has been in the form of bank loans. Issuing of bonds by developing countries has been relatively limited and the international bond market has been accessible to only a handful of them. As Table 3 reveals, international bank claims expanded by 130 billion dollars in 1979, of which indusytrial countries accounted for 71 billion and NOLDCs for 41 billion. In addition, new medium-term loan commitments reached 69 billion dollars in 1979, industrial countries accounting for 22 billion and NOLDCs for 35 billion. By comparison, net bond market lending amounted to only 29 billion in 1979. However, of this total industrial countries were responsible for 19 billion, whereas developing countries for only 3 billion. The international bond market has been, for the most part, an insignificant source of funds for developing countries. The most striking feature of developing country borrowing, which has been pointed out by several writers (Friedman 1981; Long and Veneroso 1981; Sachs 1982), has been the increasing predominance of private creditors as suppliers of funds to developing countries. It must be emphasized, at this stage, that no causality is implied between the two events. This point will be elaborated further in a subsequent section.

PAGE 12

6 Table 3 Aspects of International Bank and Bond Financing 1976-1979 (Billions of U.S. Dollars) Bank Lending Increase in Net Claims Industrial Countries Oil Exporting Countries Non-oil Developing Countries Other New Medium-Term Loan Commi tments Industrial Countries Oil Exporting Countries Non-oil Developing Countries Other Net Bond Market Lending Industrial Countries Developing Countries Other 1976 1977 1978 1979 70 75 110 130 31 43 49 71 9 10 17 7 21 15 30 ''41 9 7 15 n 29 34 74 69 10 13 34 22 4 6 10 7 13 13 27 35 3 3 3 5 30 31 30 29 20 21 18 19 2 4 5 3 8 6 7 7 Source: International Monetary Fund, International Capital Markets, Occasional Paper No. 1, Washington, D. C: IMF, 1981, p. 2.

PAGE 13

7 The proportion of debt contracted from private sources (financial institutions and other private creditors) rose from 50.1 percent in 1973 to 60.5 percent in 1982, whereas the proportion from official sources (governments and international institutions) fell from 49.9 to 39.4 4 percent during the same period. In addition as Table 4 reveals the proportion of the current account deficit financed by borrowing from private sources rose slightly for the low-income NOLDSc but increased substantially for the middle income NOLDCs. Since loans from private sources usually carry higher interest rates and shorter maturities, as compared to loans from -official sources, the debt-burden implications of the shift towards private lenders are all too evident for the '"NOLDCs and will be discussed in greater detail in the subsequent sections. Whereas low-income NOLDCs have had to rely mostly on official aid to finance their current account deficit, the middle-income NOLDCs borrowed extensively in the international capital markets. The post-1973 surpluses of the Organization of Petroleum Exporting Countries (OPEC) provided the funds for recycling to NOLDCs. However, two developments in the international capital markets in the 1970s proved instrumental in fostering NOLDC lending, in the absence of which it is doubtful whether NOLDC borrowing would have grown so rapidly. The first was the practice of syndication, according to which several banks pooled their resources in lending to a particular country. Second, the shift towards flexible interest rates, enabled banks to charge a rate on the loan that is tied to a particular market interest rate and which is revised at predetermined The figures quoted are from External Debt-The Continuing Problem, Finance and Development, Vol. 20, No. 1, March 1983, p. 23.

PAGE 14

8 -o c p o "r— lieu 4J O u u =: +-> o t/1 OD s_ 0) 1 n3 So sO C_) r— Q CO CO (/I r-. cu 0) o S+-> O o o to o O) c: O o •r— c n3 •1 — c 1 o. •1 — o Li. CQ > CU D CD •r-P 5O CL E rO CD -P 1O QO o cn cn Ln — Ln O CO Ol CO cn Ln 1 — CM CO ^ Ln o r-cn O CD CM S 1 — CM CM CM o 1 o c Ln CO ro 00 O Ln r— 1 cn CM Ln ro CM Ln O) CM -a •a ro ro r-^ o •f— CTl Ln Ln ro o ro cn CO <^ r-cri ro ro CO CD CM 1— o 1 — CO cn CO cn Ln E O Ln o o r-cn r--, 1 — o ro cn _J cn 1 — o 1^ cn ro ro -M *a cu "O +J c =J o o o fC c QJ ST3 O CO 2 O ro 4-> Cl ro C_) +-> r-. CM Ln o o ^— CM cn Ln o O UD CO LO O o CM IX) O o ro Ln co o O •4-> c CO QJ > c 1— lA sz +-> ro o o OJ -a ro •1 — CU CJ -p tro cu > E E CD si o O CJ CM o o I roxi CD to c (D -1> Scu cu o o •IScu to +-> CD I cn-p c: sro o o in E C ro Jo E cu Ln CM cn ro cn •pD_ cu CD a <+-p o 3 cu o ai u ro u pi ro c (U p> o c %cu QJ ro o 1 — M(— ro -p •1 — CL ro tj •1 — T3 OJ T3 =5 r— (J cu to ro OJ > j= u o cu CO cu s_ sz CO -p OJ d CO ro ro Scu CD 5O c CO •I— Scu <+ro CO CO ro cu p> pi ro u ro TD o •1 — p M1 O pi c: OJ CD •1 — to to cu to T3 £= O 'e X U-l cn CO cn ro cn o C_3 o c o p> cn c: CO ro p> so cu c QJ E Q. O QJ > CU Q So ro CQ o OJ o io

PAGE 15

9 intervals. Both practices contributed to the rapid growth of NOLDC borrowing and will be further analyzed in the following chapter. The large scale tapping of financial markets by NOLDCs has raised a number of important questions. Before embarking on a discussion, it may be useful to distinguish a few terms: private and public debt, official and private creditors, shortand long-term debt. According to the World Bank (1981) definition debt from official creditors comprises loans from international organizations (multilateral loans) and loans from governments (bilateral loans). Debt from private creditors comprises loans from suppliers and financial markets and the latter includes loans from private banks and other private financial institutions and publicly issued and privately placed bonds. Public debt refers to the debt incurred by the central government or its agencies or a private debtor guaranteed by the government. Private debt refers to the external debt of corporations and private citizens. Debt of maturity longer than one year is designated as long-term and that of maturity less than one year as short-term. Most of the data available on LDC external debt refer to long-term public debt, but recently private debt data have become available from the World Bank and the Bank for International Settlements. The balance of this dissertation is concerned with the external public debt of developing countries contracted with private creditors. Distribution and Value of the External Debt of Developing Countries The Distribution of Developing Country Debt One theme that almost every writer has touched on concerns the total amount of LDC borrowing and its distribution. Conclusions based on aggregate figures for the LDCs as a group may be misleading insofar as

PAGE 16

10 the distribution of debt among LDCs is ignored. As Table 5 reveals, the majority of the external debt owed to banks is concentrated on a small number of middle-income LDCs. As of the end of June 1982, the twenty-one major LDC borrowers shown accounted for 84 percent of the total debt to all banks reporting to the Bank for International Settlements (BIS). The concentration was even more pronounced as regards the debt owed to U.S. banks. Of the twenty-one countries, eight accounted for 57 percent of the total LDC debt to BIS-reporti ng banks, whereas Mexico and Brazil alone accounted for 34 percent. Similar figures were reported by Solomon (1981, p. 597), where the same-eight countries accounted for 71 percent of the total NOLDC^ debt in 1980; Mexico and Brazil alone accounted for 44 percent of the total In a recent paper. Long and Veneroso (1981, p. 508) point out that, while there does not exist a significant relationship between the overall debt level and the per capita income of LDCs, when the overall debt is broken down to commercial bank debt and official lending, the former is positively correlated with per capita income whereas the latter is negatively correlated. Thus the belief that commercial bank lending has been concentrated on a few middle income LDCs, while the "poorest of the poor" have had to rely increasingly on official aid, is confirmed. Eaton and Gersovitz (1981b, p. 5) state that the countries that were the major borrowers in the early 1970s, were heavily engaged in borrowing until the 5 Argentina, Brazil, Chile, Mexico, Peru, Philippines, South Korea and Thailand. Excluding South Africa and countries in Europe.

PAGE 17

n Table 5 External Debt Owed to Banks by 21 Major LDC Borrowers (Billions of U.S. Dollars) Latin America Argenti na Brazi 1 Chile Colombi a Eucador Mexi CO Peru Venezuela Subtotal Asia Indonesia Korea Mai aysi a Phi 1 i ppi nes Taiwan Thailand Subtotal Middle East and Asia Algeria Egypt Israel Ivory Coast Morocco Ni geri a Turkey Subtotal Total of 21 LDCs Total for LDCs (Excluding Off-shore Banking Centers) 21 LDCs as Percentage of Total for LDCs BISReportinq Banks 25.3 55.3 11.8 5.5 4.7 64.4 5.2 27.2 U.S. Banks 199.4 8. 20. 5. 11 6. 4. 36.8 7.7 5.4 6.1 3.2 3.7 6.7 4.0 36.8 292.3 347.5 84 All 8.8 20.5 6.1 3.0 2.2 25.2 2.3 10.7 78.9 2.4 9.2 1.3 5.3 4.4 1.7 24.3 1. 1 2. 0. 0. 1.2 1.4 9.2 112.5 125.5 9 Large 5.6 12.3 3. 2 1 13 1 7.1 46.7 1.9 5.1 1.0 3.7 2.7 1 .0 15.4 0.8 1.0 1.4 90 6.2 68.3 77.7 88 Source: World Financial Markets, New York: Morgan Guaranty Trust Company, February 1983, p. 3.

PAGE 18

12 end of the decade. The distribution of borrowing among LDCs, as measured by the Gini coefficient, has shown no trend towards equality.'^ The concentration of bank lending on a small number of developing countries has raised two questions. First, it has led several writers (Sargen 1976; Solomon 1977, 1981) to argue that any future LDC repayment problems will be intimately linked to the economic performance of these countries. Second, the question is raised as to whether it is rational for banks to alter their lending policy towards LDCs and distribute funds more evenly. As for the first question, forecasts of the economic performance of middle-income LDCs, which are heavy borrowers, have been favorable and Solomon (1977, p. 498) has argued "the performance and prospects of the major borrowers permit an optimistic judgment about their creditworthiness." While the problems and prospects of the major borrowers have received the most attention recently, the possibility of a relatively minor borrower, faced simultaneously with a number of adverse g external shocks, defaulting and causing a chain reaction among other borrowers, must not be overlooked. The performance of all borrowers must be examined as regards the possibility of defaulting on their loans. Isolating a small number of countries, examining their economic performance and prospects and inferring that developing countries as a whole will not face serious repayment problems, is likely to result in erroneous conclusions. The commercial banks, in conjunction with their governments In fact, Eaton and Gersovitz (1981b) report that the coefficient comparing the distributions of debt to distributions of gross national product (GNP) rose from 0.448 in 1976 to 0.494 in 1978. g Such as a drought, a drop in prices of its major exports or an increase in real interest rates.

PAGE 19

13 and international organizations, will see to it that the circumstances that render profitable the default of a major borrower, do not arise; whereas they may not have as much influence or the required information concerning the smaller borrowers. As far as the second question is concerned, rational behaviour would imply that banks trade off diversification, by lending to a broad group of countries, including low-income LDCs, against riskiness, which 9 is generally recognized to be higher for low-income LDCs. Eaton and Gersovitz (1980, pp. 14-19) have produced evidence to suggest that the export performances of large borrowers are mutually independent. Furthermore, Goodman (1981) has divided total bank risk into diversifiable and nondiversifiable and has argued, from her empirical results, that country specific (nonsystematic) risk has been relatively large in comparison to total (systematic) NOLDC risk. Therefore, debt problems among LDCs are likely to arise independently and commercial banks can achieve sufficient diversification among the middle-income LDCs, which are large borrowers, without having to lend to low-income LDCs. In conclusion, the lending behaviour of commercial banks, as regards the distribution of loans, appears economically rational. Real and Nominal Value of LDC Debt The discussion of the external debt of developing countries has, so far, been conducted in terms of nominal values. However, if comparisons of debt levels over time are undertaken, the real value of debt is required. A rather awkward problem is encountered here. As Long Since these countries do not have the industrial base to support substantial loan payments and cannot compress imports to any great extent in the event of an adverse shock.

PAGE 20

14 and Veneroso (1981, p. 505) and Long (1981, p. 283) point out, debt is a financial asset of commercial banks representing future claims on goods and services. Whereas the real value of a basket of goods and services that remains unchanged over time can be calculated by deflating by an appropriate price index, there is no "appropriate" index in the case of debt, since it depends on what goods and services the debt is used to procure. Furthermore, the real value of debt depends on whether it is calculated from the debtor's or the creditor's perspective. If calculated from the debtor's perspective, the nominal value should be deflated by an export or import index to take account of the real resources foregone to repay the debt. If calculated from the creditor's perspective, it must be deflated by an index of the prices of goods and services of the countries in which the shareholders and depositors of the banks making the loans are resident, since they are the ultimate beneficiaries. In this case an index of world prices would be appropriate. Where these indices differ over time, they can give vastly different results regarding the real value of debt, as Table 6 shows. The nominal value of NOLDC debt rose by 199 percent between 1972 and 1977. During the same period the real value of NOLDC debt rose by 94 percent when deflated by the Organization for Economic Cooperation and Development (OECD) price index but only by 31 percent when deflated by the NOLDC export price index. At present, there is no widely accepted index for deflating nominal debt and for that reason Long and Veneroso (1981) suggest that it may be useful to look at the ratio of debt to GNP as a measure of the debt burden over time.

PAGE 21

15 Table 6 Alternative Estimates for the Real Value of MOLDC External Debt (Billions of U.S. Dollars, 1972 Prices) 1972 1973 1974 1975 1976 1977 Nominal Debt 62 .6 76 .9 102 6 130 5 159 .0 187 0 Deflated by Price Index U.S. GNP 62 .6 72 7 88 5 102 6 118 8 132. 1 OECD GNP 62 .6 66 9 81 0 91 4 108 1 121 6 NOLDC Imports 62 6 60 8 54 0 65. 3 79 5 91. 6 NOLDC Exports 62 6 56 1 54. 4 71. 8 82 7 82. 3 Nominal Debt/Nominal GNP 0 26 0 28 0. 29 0. 33 0. 35 0. 35 Nominal Debt/Nominal Exports 1 41 1 48 1 42 1. 73 1 70 1 79 Source: Long and Verneroso (1981, p. 507)

PAGE 22

16 The tendency for prices to rise over time has further implications for the current account and the debt position of developing countries. As has been pointed out by several writers (Long 1981, p. 290; Long and Veneroso 1981, p. 512; Sachs 1981, pp. 203-208; Solomon 1981, p. 595), during periods of rising inflation rates, the current account overstates the true position of debtors. The rise in nominal interest payments (assuming that nominal interest rates rise with inflation) is offset by the fall in the real value of the debt. However, since only the higher nominal interest payments are reported in current account statistics, the true debtor position is -overstated. The reason for the overstatement, asiLong and Veneroso (1981, p. 512) point out, is that the part of the nominal interest payments corresponding to the inflation premium is a capital account item, as it is reimbursing the lender for the loss in principal value due to inflation, and should not be included as part of the current account. Sachs (1981, pp. 204-205) has estimated the current account of developed countries the NOLDCs and OPEC members adjusting for the inflation bias and has argued that these adjustments are of significant magnitude: by 1978, over half of the NOLDC debt could be attributed to the inflation bias and in 1978, while OPEC was nearly in balance on its nominal current account, when adjusted for the inflation bias it showed a deficit of 12.8 billion dol 1 ars Inflation plays an important part in determining the value of real interest payments made by LDCs. Here, the problem encountered earlier, of the appropriate deflator, resurfaces. Notwithstanding the question of the appropriate deflator, the choice between real and nominal interest payments plays an important part in determining a country's debt

PAGE 23

17 capacity.''^ Long (1980, pp. 494-5) has argued that when the nominal interest burden is compared to the real interest burden (under the prevailing maturity structure), the former, of course, reveals a more severe burden which, together with amortization charges, would push the ratio of debt service payments to exports, for many NOLDCs, to alarming levels. Such a distinction is crucial because "from a solvency and growth standpoint, it is the real interest burden that matters, but from the liquidity standpoint a country might find it difficult to finance such high outlays on external debt" (Long, 1980, p. 495). The rise in nominal interest rates that normally follows a rise in the inflation rate, takes on added significance when considering the debt burden of developing countries. As Sachs (1981, p. 206) points out "a rise in inflation that is exactly matched by a rise in interest rates causes a rise in interest income for a creditor country that is exactly offset by greater capital losses." However, when nominal rates rise above the increase in the inflation rate a real burden (in terms of interest payments) is imposed on the debtor countries which, as has already been argued, is manifested as well in overstated current account deficits. Therefore, rising real interest rates impose a burden on developing countries in two ways: first by raising the value of nominal interest payments and secondly by overstating the extent of the current account deficit, necessitating higher borrowing LDCs to finance the rising deficit. The concept of debt capacity for a country is explained in a further section of this chapter.

PAGE 24

18 Constraints on IPC Borrowing and the Role of International Reserves Is a Cr edit Ceiling Impo sed on IPC Borrowing? In a series of papers, Eaton and Gersovitz (1980, 1981a, 1981b) have argued that individual LPCs are not unconstrained in their borrowing, but face a ceiling on the amount of funds they can borrow from private financial sources. On the basis of a theoretical model (Eaton and Gersovitz, 1981a) they have shown that the ceiling is determined endogenously and is a function of several characteristics directly related to the costs and benefits of default. In particular, they have proposed that, on the one hand, the higher the variability of income of a country, the greater the penalty of default (by being excluded from future borrowing) and the higher the credit ceiling imposed. On the other hand, the probability of default is higher in periods of low income when the marginal utility of income is high. Therefore, a high variability of income, by increasing the probability of low incomes, increases the probability of default and therefore lowers the credit ceiling. In conclusion, the impact of the variability of income on the credit ceiling is ambiguous. The long-term growth of income is negatively related to the credit ceiling because a higher growth rate means the country need not enter the capital markets in the future and is therefore more likely to default at present. Finally, the higher the dependence on trade (as measured by the ratio of imports to income) the less the likelihood of default, as the country would suffer a higher penalty from the disruption of its trade pattern, and therefore the higher the credit ceiling. The proposition concerning credit ceilings is supported by Sachs (1982, p. 19), who shows that when default is precluded credit ceilings are fully consistent with perfect competition in the loan market.

PAGE 25

19 Further on (Sachs, 1982, p. 21), he argues that in a two-period model, where default can occur, there is a maximum level of debt for each country and any amount greater will lead to default with certainty. No rational lender will be willing to loan any amount greater than the ceiling, irrespective of the risk premium. Therefore, "rationing will be a standard device in credit allocation to sovereign borrowers" (Sachs, 1982, p. 9). On the basis of their theoretical work, Eaton and Gersovitz argue that the observed amount of LDC debt is the minimum of two quantities: the amount of desired debt or the credit ceiling. Thus, empirical estimation of demand and supply curves for LDC debt must be carried out within the framework of a disequilibrium model where the observed quantity of debt is the minimum of the amount desired or the credit constraint. In two papers (Eaton and Gersovitz, 1980, 1981a) they estimate a disequilibrium model, in which a country's demand and supply of debt are dependent on the economic characteristics described earlier. They show that, in the two years 1970 and 1974, 56 of 81 countries were constrained with probability greater than 0.5. They conclude that credit constraints are an important feature of the LDC loan market and any attempt to estimate a supply or demand function without incorporating the constraint explicitly is bound to lead to erroneous estimates. In connection with the Eaton and Gersovitz work, it is worth mentioning that the demand and supply curves estimated do not include the interest rate as an explanatory variable. As they correctly point out, the base rate"*^ is the same for all countries during a particular Which in the international loan market is the London interbank offered rate or LIBOR.

PAGE 26

20 time period and any variation in the base rate between two years is captured by year-specific dummy variables. As far as the risk premium is concerned, they argue that "in competitive markets if lenders are risk-neutral or if default risk can be perfectly diversified the expected interest rate r is equal for all borrowers in each year (Eaton and Gersovitz, 1980, p. 12). If their assumptions hold, then the risk premium is solely a function of borrower characteristics and can be correctly omitted from the supply and demand functions. Thus, the omission of the interest rate rests on how well the assumptions are met in practice. A different approach to estimating the demand and supply of loans to developing countries is proposed in this dissertation. A disequilibrium model of the demand and supply of loans is specified whereby the interest rate variable is introduced as an explanatory variable and a quantity constraint is entered explicitly. The specification and estimation of the model is discussed further in chapters 3 and 4. One final issue concerning the credit ceiling is related to the penalties that lenders can impose on the borrower in the case of default. Eaton and Gersovitz (1981b, p. 33) put forward what may seem a strange proposition: both the lender and the borrower may be better off as a result of more severe penalties in the case of default. The lenders may be better off because they can loan a larger amount (by virtue of higher credit ceilings) than if the penalties were less severe. As far as the borrowers are concerned, on the one hand they are better off since higher penalties imply greater capital flows from the banks. On the other hand, greater penalties reduce the welfare of the borrower because they decrease the insurance aspect provided by default, in the event of

PAGE 27

21 adverse economic conditions. Eaton and Gersovitz suggest that one possibility may be to let LDCs decide by themselves, in advance, the penalties to be incurred in the event of default. Of course, it must be remembered that all of the above arguments rest on the assumption that the penalties are actually enforced. Essentially the same argument is made by Sachs (1982, p. 28). On the basis of a two-period utility-maximization model, where a cooperative and a non-cooperative solution between banks and LDC borrowers are compared, he argues that under certainty, the utility from the cooperative solution is higher than -the utility from the non-cooperative solution and if default could be ruled out altogether, utility could be increased even further. It is therefore to the advantage of the borrowing country to seek higher penalties in the case of default. However, the argument does not always carry in the case of uncertainty. Although higher penalties increase welfare by raising the credit ceiling and therefore capital flows to LDCs, they lower the "insurance" aspect of default. Uncertainty opens up the possibility that the insurance aspect of debt is the dominant aspect and that penalties should not be increased. External Debt and International Reserves One question that is of considerable importance from a public-policy standpoint is whether international borrowing is a substitute or a complement to a country's international reserves. As Eaton and Gersovitz (1980, p. 4) argue, there is no theoretical a priori reasoning either way. On the one hand, debt may be used to finance reserves while, on the other hand, both borrowing and reserves can be used as a means of smoothing the variability in the income stream of a country. In the first case they are complements; in the latter they are substitutes.

PAGE 28

22 The question must, in the end, be an empirical one. Eaton and Gersovitz point out that the demand for reserves is a function of the same variables as the demand for debt. In their empirical investigations, they estimate separate reserve demand equations under two different regimes, depending on whether the country faces a ceiling on the amount it can borrow on the international capital markets or not. On the basis of their empirical estimates they find that debt and international reserves are substitutes. Therefore increased borrowing by LDCs from private sources should reduce the demand for international reserves. On the same issue, Solomon (1977, p. 481) states that in 1976 NOLDCs borrowed more than their combined current-account deficit and added to their international reserves. The possibility arises that the NOLDCs borrow short-term in order to build up their reserves and repay the loans in the following years. The reserves can then be drawn down to meet unexpected fluctuations in the income stream when borrowed funds may be unavailable or too costly to obtain. This could very well have been the case with short-term borrowing from private sources which in 1974-75 amounted to five billion dollars or one fifth of the total borrowing of the NOLDCs. L'Heriteau (1979, p. 713) discusses one practice which was becoming increasingly common in the mid 1970s concerned with the depositing of the international reserves of NOLDCs in the Eurocurrency markets. These funds were, in turn, loaned back to the NOLDCs. Thus, the international banks acted as intermediaries channeling the funds of some NOLDCs to other NOLDCs and collecting interest, commissions and other fees in the process. International reserves play an important part in determining the creditworthiness of developing countries and how that is perceived by

PAGE 29

23 lenders in the international capital markets. The crucial role of international reserves as an indicator of creditworthiness has been discussed by several writers including Feder and Just (1977a, 1977b), Frank and Cline (1971), Goodman (1977) and Kapur (1977). In econometric studies the ratio of international reserves to import expenditures during a particular time period has been employed as an indicator of creditworthiness. The empirical results showed that, ceteris paribus a larger volume of reserves improved the country's creditworthiness and reduced the likelihood that the country would seek to reschedule its external debt. While a larger amount of reserves enhances the creditworthiness of a country and raises the likelihood of obtaining loans at more favorable terms, some developing countries have been motivated to borrow on the international capital markets by the desire to increase their international reserves and thus improve their creditworthiness. Friedman (1977, p. 18) has argued that "other developing countries, like Brazil, are borrowing both to finance external deficits and to rebuild reserves, profiting from the experience that a strong visible level of reserves improves creditworthiness as well as providing an additional cushion to meet contingencies." The notion that developing countries may have borrowed to increase their reserves and thus their creditworthiness so as to be able to borrow further, is also shared by Abbott (1981, p. 344) and by Griffith-Jones (1980, p. 215) who quotes the Chilean magazine, Gemines as follows: "The Central Bank is ready to increase its foreign assets even more because the concentration of debts to banks and financial institutions in a way obliges it to maintain available a relatively high level of reserves."

PAGE 30

24 From a theoretical standpoint, beginning with the seminal paper by Bardham (1957), a substantial body of literature has developed on the optimal amount a country should borrow externally. The contributions in this area examine the question within the framework of a neoclassical growth model and the general conclusion arrived at is that borrowing should continue until the marginal product of capital becomes equal to the interest rate on external borrowing. Out of the numerous contributions, only the paper by Feder and Just (1979) has examined the relationship between the optimal amount of borrowing and the optimal level of reserves. Feder and Just postulate a two-sector open-economy model, where one sector produces an export good and the other a good which can be used either for consumption or investment. It is assumed that the interest rate paid on the external debt depends on two indicators of creditworthiness: first, the difference between export receipts and the average volume of debt service payments and second, the difference between international reserves and import expenditures. The central result of the Feder and Just study is intuitively plausible and holds that the level of international reserves should be adjusted until the marginal cost of borrowing to the economy is equal to the marginal benefit from holding reserves. The marginal cost of borrowing is higher than the interest rate on the external debt since the higher debt level reduces creditworthiness and raises the interest rate. The marginal benefit from holding international reserves is equal to the interest rate earned on the reserves plus the contribution to increased creditworthiness from holding a higher level of reserves. Thus, the importance of international reserves in the model enters through the impact of reserves on the creditworthiness of the country. If the impact

PAGE 31

25 is insignificant then, according to the model, a zero level of reserves should be maintained. One issue, not directly related to the above, but of considerable significance from a public-policy decision-making standpoint, is the relationship between debt from private and public sources. On the basis of their empirical estimates, Eaton and Gersovitz (1981b, pp. 19-20) have argued that they are complements, instead of substitutes as was expected. They attempt to explain their findings by recourse to the argument that funds obtained from official sources are not sufficient to complete the projects undertaken and that countries have to resort to private loans in order to complete these projects. One can cast some doubts on the validity of their explanation, especially in view of the fact that official sources of funds are usually of longer maturity and bear lower interest rates than private loans. Private funds are unlikely to complement official funds in project financing since it is quite possible that a project financed by official funds becomes unacceptable when funding is provided from private sources at higher interest rates. The question of the relationship between debt from private and public sources has, by no means, been settled. Analyzing Country Risk: Debt Reschedulings and the Debt Capacity of IPCs One question that has received considerable attention, for obvious reasons, is that of debt repayment and the probability of default on the part of developing countries. There are two strands of modeling in this area. The first employs growth-cum-indebtedness models to calculate the value of such ratios as debt/income or debt/exports at the limit or the

PAGE 32

25 speed at which debt accumulates, based on several assumptions about the parameters in the model. The second attempts to predict the probability an LDC will default on its external debt obligations, on the basis of several ratios. Because there has only been one outright default since World War II (North Korea), the models estimate the probability that a country will request a rescheduling of its debt obligations. Naturally, the two approaches can be combined, as in Feder (1980). Growth-cum-Indebtedness Models In the theoretical development economics literature, considerable emphasis has been placed on the contribution of foreign funds to the process of economic development, particularly within the context of the "two-gap" models, as discussed in the pioneering works of McKinnon (1964) and Chenery and Strout (1965). According to these models, a target rate for the growth of income of a country is specified and it is assumed that a maximum amount of domestic savings is available and a minimum amount of imports is required during any particular period. The country's ability to export and its investment requirements define two gaps: the gap between maximum savings and investment (savings gap) and between exports and the minimum amount of imports (foreign exchange gap). It is therefore argued that, if the target rate of growth is to be achieved, foreign funds are required to fill the larger of the two gaps. While "two-gap" models were originally developed to forecast the order of magnitude of foreign aid required to achieve different target rates of growth, it soon became clear that developing countries can obtain Of course, the results are sensitive to the particular values chosen for the parameters.

PAGE 33

27 foreign funds in the form of loans v;hich have to be repaid and which entail a process of debt accumulation. Thus several models of the growthcum-indebtedness process have been developed. Beginning with the seminal work by Avramovic and his associates at the World Bank (Avramovic 1964) the earlier models assumed the savings gap to be binding (Ohlin 1966; King 1968; Solomon 1977, 1981; Feder 1980). The model of Feder (1981) investigated the pattern of debt accumulation when the foreign exchange gap is assumed binding. Based on the earlier works of Avramovic (1964) and Ohlin (1966), Solomon (1977) assumes that investment and savings increase exponentially over time and debt accumulates to fill the savings gap. The general conclusion drawn from the model holds that so long as the rate of growth of the economy is greater than the rate of interest charged on the loans, the debt/income ratio will converge to a maximum value. Solomon (1977, p. 495) calculates the limiting values for the debt/income, debt/exports and interest/exports ratios for several countries and argues that they are not excessive. The issue of what happens when the growth rate is lower than the interest rate on the debt is also addressed by Solomon (1981, p. 597). In the late 1970s, the rise in interest rates above the growth rates, for several countries, has meant that the process of debt accumulation is no longer self-limiting but is rather an explosive one. One criticism of Solomon's procedure has been voiced by Greenspan in the discussion of his paper, who argues that a static framework has been imposed on what is a dynamic process. The values of the parameters in the model i.e., the incremental capi tal -output ratio, the growth of 1 3 real and nominal net national product and savings ratio are all assumed Which are equal since prices are assumed constant in the model.

PAGE 34

28 to remain constant over time, clearly an unrealistic state of affairs. Solomon (1977, p. 487) addresses this issue briefly, but provides no alternative to his analytical procedure. The criticism is partially met by Feder (1980) and Long (1980, 1981), who conduct sensitivity analysis of their results. Feder (1980) simulates the number of years for debt to reach its maximum level, the maximal debt-service payments/exports ratio and the time it occurs, under different scenarios concerning the target rate of growth of GNP the marginal savings rate and the export growth rate. Long (1980, pp. 488497) calculates the debt/GNP ratio for different values of the real interest rate and the rate of growth of GNP as well as the number of years for the debt/GNP ratio to increase from 30 to 40 percent, again for different real interest rates and rates of growth of GNP. In a subsequent paper. Long (1981, pp. 296-299) calculates the number of years for the debt/GNP ratio to increase by ten percentage points, when 1 4 a country suffers different adverse shocks of varying magnitudes. Feder (1981) focuses attention on the foreign exchange gap as the binding constraint. He derives the intuitively obvious results that higher target rates of growth and higher interest rates imply higher levels of indebtedness whereas an increase in the rate of growth of exports leads to lower levels of borrowing. Therefore, export promotion provides an alternative means of attaining the target rate of growth without increasing indebtedness. However, in the model, the resources for Such as deterioration in the terms of trade, higher real interest rates, capital flight export decline and "disaster".

PAGE 35

29 achieving the increase in exports are provided by reducing consLimption and therefore the resulting decrease in borrowing eventually ceases when the savings gap is reached. Further on, Feder provides projections for the highest debt/gross domestic product (GDP) ratio and the time it is reached, as well as the value of the debt service/exports ratio and the savings rate at that time. These projections are provided for different values of the target rate of growth, the rate of growth of exports and the proportion of consumptiongoods imports relative to total consumption. Another table provides similar information for -the highest debt service/exports ratio. As Feder argues, such projections are useful as rough planning aids in the calculation of borrowing requirements and are, by no means, meant to provide an accurate description of the debt-accumulation process. A general criticism of the growth-cum-indebtedness models is that they are solely concerned with the demand side of the market for loans and assume that funds are forthcoming at a constant average interest rate. Solomon (1977) recognizes this limitation and devotes several lines to it in his concluding remarks. As in any other market, both demand and supply are equally important and one cannot simply ignore one side of the market. In fact, if one is to believe Eaton and Gersovitz (1980, 1981a, 1981b), it could very well be that the supply side of the market plays a more important role than the demand side in determining the actual amounts borrowed by developing countries. Estimating the Probability of Rescheduling The second strand of modeling attempts to predict the likelihood of rescheduling by individual LDCs. One method, pioneered by Avramovic (1964), is the close examination of several indicators, usually in the

PAGE 36

30 form of financial ratios, which are taken to be important predictors of repayment problems. The other is to incorporate these predictors in formal mathematical models and estimate probabilities of rescheduling. The first systematic analysis of ratios was initiated by Avramovic (1964) at the World Bank. However, even earlier works focused on the 1 5 ability of developing countries to service foreign capital. Finch (1951-52, p. 60) introduced the concept of an investment service ratio, which he defined as "the percentage of current foreign exchange receipts, exclusive of conpensatory financing, absorbed by investment income payments," He presented a -historical series for this ratio for a number of countries which by that time had developed sufficiently and others, which at the time, were still in the early stages of development. He demonstrated that for the developed countries the ratio was very substantial during the early development phase and declined to insignificant amounts during the later stages. No such trend could be discerned for the countries that had yet to reach the later stages of development. Finch concluded that the ratio was useful as a short-run indicator of a country's ability to meet its debt service obligations but not as a guide as to whether foreign capital will be to the long-run benefit of the country. The distinction between shortrun disturbances that affect a country's ability to service foreign capital and long-run prospects, is also discussed, in an early work, by Alter (1963). He argues that a country can maintain its debtservicing capacity over the long run by way of a rising trend in per capita income. In this broader definition, foreign capital encompasses loans, bonds and other fixed obligations contracted with foreign parties, as well as foreign equity capital.

PAGE 37

31 The work of Abramovic (1964) provided a systematic study of debtservicing capacity. A distinction was drawn between the liquidity (shortrun) and solvency (long-run) aspects of debt-servicing capacity. The ratio of debt service payments to export proceeds (or the debt service ratio as it has become known) was introduced as a short-run indicator, whereas per capita income and growth prospects have been identified as important long-run determinants of debt-servicing capacity. Since the appearance of Abramovic' s work, almost every study of the external debt of developing countries has included some discussion of financial ratios as indicators of debt-servicing capacity. Some of the more commonly used ratios are shown on Table 7. Despite the large disparities among individual countries which are masked by aggregate ratios, the figures reveal a gradual but steady deterioration in the debt-servicing capacity of NOLDCs throughout the 1970s. The debt service ratio increased by almost two-thirds between 1973 and 1982 whereas the debt/exports ratio increased by a quarter during the same period. In addition to the ratios of Table 7, a number of others have been compiled by various writers. Dhonte (1975, p. 161-170) compared several ratios for 69 developing countries for the year 1969 to the ratios for 13 countries which renegotiated their debt obligations during 1959-1971. The ratios included in the analysis were debt outstanding/GNP, debt outstanding/exports of goods and services, the debt service ratio, a discounted debt service ratio, the ratios discounted debt service payments/debt outstanding, debt service payments/disbursements, disbursements/ imports and finally, the ratio of net transfer (disbursements minus debt service payments) to imports. Dhonte's comparison revealed that countries that renegotiated their debt borrowed heavily to secure a large net

PAGE 38

32 Table 7 Long-Term Debt Ratios for NOLDCs, 1973-1982 1973 1975 1976 1977 1978 1979 1980 1981 1982 88.7 97.7 102.6 104.9 111.2 101.9 92.9 103.2 109.1 16.6 18.3 20.5 22.1 23.7 22.7 21.8 24.3 25.2 14.0 14.0 13.9. 14.0 17.3 18.1 16.3 21.-0 22.3 Ratio for year-end debt to exports or GDP for year indicated. Debt service payments as percentages of goods and services. Ratio of External Long-Term Debt to Exports of Goods and Services Ratio of External Long-Term Debt to GDPa Debt Service Ratio^ Source: Finance and Development, Vol. 20, No. 1, March 1983, p. 23.

PAGE 39

33 transfer, had a high and rising debt service ratio and sought large capital inflows at the same time that their debt accumulated on unfavorable terms. Abbott (1981) examined the ratios shown on Table 7 for all NOLDCs 1 Fi and separately the low-income countries during the period 1970-1980. While cautioning against the indiscriminate use of such ratios, he concluded that for the NOLDCs as a whole the ratios have not shown any marked trend but have deteriorated (particularly the debt service ratio) for the low income NOLDCs. Further on, Abbott (1981, p. 347) computed the net transfer of resources''^ to developing countries for the period 1972-1979. The results indicate a marl^ed decline in net resource ; transfer: whereas in 1975 it amounted to 56 percent of gross disbursements, by 1979 the proportion had fallen to 35 percent. Among the various formal models of rescheduling that have appeared in the literature, several indicators of debt-servicing difficulties have received prominence. They are discussed by Frank and Cline (1971) and Feder and Just (1977a). In addition to the debt service ratio already mentioned, others include the growth rate of exports, an export fluctuations index, the ratio of "non-compressible imports" to total imports, per capita income, the ratio of debt amortization to total outstanding debt, the ratio of imports to GNP and the ratio of imports to international reserves. Feder and Just (1977a, pp. 26-29) argue that it is virtually impossible to distinguish between "compressible" and "non-compressible" imports and in their study they introduce one further index: the ratio Identified as countries with per capita incomes of less than $300 U.S. dollars in 1978. ^''pefined as gross disbursements less amortization and interest payments.

PAGE 40

34 of capital inflows to debt service payments. Further discussion of these ratios is provided in Chapter 5. Solomon (1977, p. 489) criticizes the debt service ratio as an inappropriate measure of creditworthiness because maturities are usually bunched and not spread evenly over time. However, in his discussion of Solomon's paper, Greenspan argues that it is precisely the bunching of maturities that the debt service ratio attempts to measure. Eaton and Gersovitz (1981b, p. 27) make the point that a rising debt service ratio may not signal the deteriorating creditworthiness of a country but may in fact be a result of other factors which make the country a good risk and allow loans to be extended to it. Finally, it must be recognized that the debt service ratio considers a country's debt obligations during a particular year and neglects any future obligations which may be just as important in the country's decision of whether to reschedule at present or not. Comparisons of debt service ratios over time may be misleading during periods of inflation, as pointed out by Solomon (1981, p. 598). When the nominal interest rate rises in accordance with the rise in the rate of inflation, the debt service ratio is biased upwards. A rise in both the interest rate and the inflation rate from 5 to 10 percent is a doubling of the interest rate but only a 5 percent increase in the price level. The debt service ratio, "exaggerates the increase in debt burden in periods when inflation accelerates and interest rates rise correspondingly" (Solomon, 1981, p. 598). Along the same lines of ratio analysis, Euh (1979) devised a creditworthiness index as the ratio of loan amounts supplied to a country by commercial banks to its external financing requirements (needs). The

PAGE 41

35 latter were projected on the basis of a simple Harrod-Domar growth model. The index was regressed on a number of explanatory variables with data from 33 LDCs for the period 1971-75. Three variables were found to be significant and of the correct sign: the growth in GDP, the expected change in the debt service ratio and the debt/exports ratio. The sign 1 8 of the coefficient of the other variables conformed to prior expectations, whereas the sign of the coeffi ci ent of the reserves/imports ratio proved consistently contrary to expectations. Euh's study shifted attention from other studies which focused on the short-run (liquidity) aspects of creditworthiness and pointed to the importance of the growth in GDP as a long-run indicator of creditworthiness. \ There have been several approaches to formal models of the probability of rescheduling. One uses the method of discriminant analysis (Frank and Cline, 1971; Sargen, 1977; Black, 1981) another one,logit analysis (Feder, and Just 1977a; Mayo and Barrett, 1978; Feder, 1980; Feder, Just and Ross, 1981) and another the technique of principal components (Dhonte, 1975). The Frank and Cline (1971) study investigated 13 debt reschedulings for 8 countries during the period 1960-68. It was found that a quadratic J • 19 discnminant function in two variables gave the best results in terms 20 of the fewest Type I and Type II errors. The Black (1981) study Included in the analysis were the growth in exports, per capita income, the debt service payments/net capital inflows ratio, a proxy for the willingness to pay and a political instability index. 19 The debt service ratio and the ratio of debt amortization to total outstanding debt. 20 Type I errors refer to the failure to predict countries which actually rescheduled, whereas Type II errors refer to falsely predicting a country to reschedule. Frank and Cline reported a Type I error rate of 23 percent and a Type II error rate of 11 percent in their study.

PAGE 42

36 estimated four discriminant functions on the basis of 19 discriminating variables and compared the relative importance of each variable. In this study countries were assigned to five categories to reflect each country's creditworthiness in the international financial markets. In an interesting paper, Sargen (1977) introduced two conceptual approaches to debt rescheduling. The first was the traditional debt service approach according to which repayment problems are exogenous to the economic management of the country and arise because of a shortfall in export proceeds or other adverse external factors. Such problems are short-term in nature and the debt service ratio was identified as an important indicator by the Abramovic (1964) study. Second, the monetary approach identifies debt repayment problems with factors endogenous to the management of the economic system. Rapid expansion of the money supply and the resulting inflation, along with the maintenance of overvalued exchange rates, cause export demand to fall and import demand to rise and lead to debt accumulation. 21 The Sargen (1977) study employed six explanatory variables to differentiate between rescheduling and nonreschedul i ng cases. Two variables figure prominently in the discriminant function: the adjusted debt service ratio (a debt-service approach variable) and the inflation rate (a monetary approach variable). Type I error rates varied from 15 to 54 percent according to the cut-off value, while Type II error rates ranged 22 from less than 1 to 11 percent. An interesting result drawn from this The inflation rate, the growth rate of the money supply, the export growth rate, the adjusted debt service ratio the growth rate of real GNP and a measure of relative purchasing-power parity. 22 Type II errors are to be compared to the naive rule of assigning all countries to the nonreschedul i ng group, which would produce an error rate of 5 percent.

PAGE 43

37 study concerns the success of the adjusted debt service ratio in explaining reschedulings in South Asian countries associated with long-run debt problems. Exclusion of the debt service ratio from the discriminant function does not diminish its success in explaining reschedulings in South American countries associated with high inflation and short-run balance of payments crises. These results are not in accordance with the emphasis placed by Avramovic on the debt service ratio as an important indicator of short-term repayment problems. Dhonte (1975, pp. 170-186) employed the technique of principalcomponents analysis, according to which a number of indicators (the ratios mentioned previously in connection with Dhonte's analysis, on addition to the growth rate of debt and the growth rate of exports) are combined linearly to derive a set of composite indicators or components. The importance of each component is measured by the percentage of the total sample information incorporated in the component. Four indicators are identified which account for 79 percent of the total information. Two indicators, which can be interpreted as, on the one hand, a measure of the involvement in debt and, on the other, the terms of the loans and the size of the debt service, account for 56 percent of the total sample information. The conclusions of the analysis are arrived at by plotting the correlation coefficients of the indicators with the components on a plane, whose axes are the two components, for both the standard sample of countries and the renegotiation cases. Two equilibrium relations are identifed: first, a balance must be maintained between the extent of the involvement and the borrowing conditions and second, the growth rate of debt must be maintained in line with the growth rate of exports, otherwise repayment problems will arise and are likely to lead to debt

PAGE 44

38 renegotiations. The methodology employed in the study did not permit the testing of the second relation. Finally, the results of Dhonte's study do not fare well relative to the studies using the discriminant analysis technique: 33 percent of the renegotiation cases and 13 percent of the countries that did not renegotiate were incorrectly assigned. The final technique that has been employed to identify rescheduling cases is that of logit analysis. Feder and Just (1977a) examined a sample of 21 reschedulings for 11 countries during the period 1965-72 and found six variables to be significant in the estimation of the "default" probability equation: the debt service ratio, the imports/reserves ratio, the amortization/debt ratio, export growth, per capita income and Jthe capital inflows/debt service payments ratio. A Type I error rate of 5 percent and a Type II error rate of 2.5 percent were obtained, the fewest of any of the other studies. On the basis of these six variables, Feder (1980) calculated the maximal probability of default, the time it occurs, as well as the probability of default in the first year, for highand middle-income LDCs and for different values of the target growth rate of income, the marginal savings rate and the export growth rate. The results of Feder and Just (1977a) were extended in Feder and Just and Ross (1981) who made use of a larger data base.^"^ Six explanatory variables were tested: the debt service ratio the ratios reserves/ imports, commercial foreign exchange inflows/debt service payments, net noncommercial foreign exchange inflows/debt service payments, exports/ GNP and real per capita GNP/U.S. per capita GNP. For the first three variables both the linear and the quadratic forms were tested and found Some 580 observations, of which 40 referred to cases of debt reschedul ings.

PAGE 45

39 to be significant. Type I error rates ranged from 8 to 33 percent and Type II error rates from 1 to 8 percent for the model which included the quadratic terms. The model was used to predict reschedulings with data from a period not covered by the sample used to obtain the original estimates. The Type I error rate remained unchanged whereas the Type II error rate increased somewhat when compared to the original -sample error 24 rates. The technique of logit analysis has been employed by Mayo and Barrett (1978) as part of a larger country evaluation model at the Export-Import Bank. The other two parts of the model consist of a quantitative checklist system to assess "the mediumand long-range economic vitality of a country" and a qualitative checklist system to evaluate "the long-range strength and contribution of the country's human and natural resources" (Mayo and Barrett, 1978, p. 81). The analysis was performed with a relatively larger sample than that used by previous studies: 571 observations from 48 countries for the period 1960-1975, The logit analysis studies reviewed so far, estimate what can be termed as "objective" probabilities of default. The data used for the studies are drawn from actual developing country rescheduling experiences. In a further study, Feder and Just (1977b) estimate "subjective" probabilities, which are the international banks' own probability estimates. The authors develop an interest rate equation that depends on the cost of funds to the bank, the commitment period, the elasticity of demand, the probability of debt-servicing difficulties and the expected loss rate in the event of such difficulties. The probability of debt-servicing difficulties is described by an equation similar to that estimated by the objective studies. It is substituted in the interest rate equation giving a reduced form equation to be estimated. In addition to the variables found significant in the objective studies, an export fluctuation index and the ratio of imports to GNP are deemed important determinants of the subjective probability of default. The subjective probability concept is further explained in Chapter 2,

PAGE 46

40 which included 28 instances of rescheduling in 11 countries. As in other studies of logit analysis, the dependent variable is binary, reflecting occurrences of default or not. The novelty of this study was in the way the dependent variable was defined to include reschedulings up to five years into the future. This enabled the authors to obtain predictions of debt-servicing problems for up to five years without having to forecast the explanatory variables. The authors experimented with some 50 variables, six of which were chosen on the basis of correctness and consistency of the sign of the coefficient and the stability and significance of the t-stati sties. It is worth pointing out that one of the variables found significant, the percentage change in the consumer price index, has been identified by Sargen (1977) with the monetary approach, while another, the ratio of gross fixed capital formation to GDP had not been included in any of the previous studies. The error rates Type I error rate of 24 percent and Type II of 13 percent do not compare favorably to those of previous studies, but no strict comparisons are possible because of the way the dependent variable was defined. An interesting comparison between the Frank and Cline (1971) and the Feder and Just (1977a) model was undertaken by Manfredi (1981) with data from 60 countries for the period 1972-77. It was found that while there was no significant difference between the two models with respect to Type I errors, the Feder/Just model fared better in terms of Type II errors. The results were attributed to the fact that the Feder/Just model The ratios disbursed debt outstanding/exports, international reserves/ imports, imports/GDP, reserve position in the International Monetary Fund/imports, gross fixed capital formation/GDP and the percentage change in the consumer price index.

PAGE 47

41 includes a capital inflows variable. It may be possible for countries to avoid rescheduling their debt if they can continue attracting foreign capital The results of the formal models of rescheduling have been summarized and extended by Saini and Bates (1978) with data from 20 countries for the 1960-77 period. Two different dependent variables were defined: one was the traditional variable used in other studies consisting of official debt rescheduling and nonreschedul ing cases; the other included both involuntary debt reschedulings and bal ance-of-payments support 27 loans, but excluded voluntary debt reschedulings. The authors made use of both discriminant analysis and logit analysis and estimated models for both the whole of the period 1960-77 and for two separate subperiods, for each version of the dependent variable. They summarized their results in the following propositions: (i) when logit and discriminant analysis were compared no significant differences were observed in terms of the error rates and coefficient values, (ii) the modified dependent variable provided better results when compared to the traditional one, (iii) four variables proved more significant in terms of explanatory power: the consumer price index, the money supply growth (both variables suggested by the monetary approach), the cumulative current account balance to exports ratio and international reserve growth, (iv) the debt service Mexico could very well have been a case in point here. 27 Saini and Bates (1978, p. 26) defined bal ance-of-payments support loans as foreign loans, in the absence of which, a rescheduling would have been necessary or arrears on external payments would have occurred; voluntary debt reschedulings represented cases where there were no apparent bal ance-of-payments difficulties.

PAGE 48

42 ratio did not appear significant in explaining debt reschedulings, (v) the explanatory variables proved more significant during the 1971-77 subperiod than during the 1960-70 subperiod, possibly because the former included most of the modification in the dependent variable (Saini and Bates, 1978, p. 15). One final noteworthy issue concerns some problems of methodology. In connection with the discriminant analysis technique, Sargen (1977, pp. 28-29) points to some intriguing questions, a number of which are pertinent to the other techniques. In the first place, the instances of rescheduling are few in. relation to the total sample. The data for the rescheduling group suggest that they are not normally distributed; a theoretical assumption of the discriminant analysis technique. Second, the explanatory variables are serially correlated: a high debt service ratio in one year is followed by high ratios in subsequent years. The presence of serial correlation implies that a country which is misclassified (or correctly classified) in one year, is usually misclassified (or correctly classified) in other years. Third, the treatment of countries that have rescheduled more than once, poses problems. If one is interested in the factors that determine the process of rescheduling per se and not the times of rescheduling, then observations that relate to rescheduling countries in nonreschedul i ng years should be omitted. Fourth, in all the models, the implicit assumption is made that the parameters of the model do not change over time and no structural shifts affect the model. Because of the small number of rescheduling cases, the testing for structural shifts in the parameters is not feasible. One final shortcoming of all the models estimating the probability of default concerns the lack of any theoretical basis on which the selection of

PAGE 49

43 explanatory variables is made. Far from testing theoretical hypotheses, the models search for statistical relationships that will provide the best results in terms of the fewest Type I and Type II error rates. In conclusion, two issues neglected by the formal models delineated above appear worth mentioning. In the first place, as Eaton and Gersovitz (1981b, p. 29) point out, one must distinguish between exogenous variables determining the probability of rescheduling and endogenous variables determined simultaneously with the rescheduling of debt. In the Feder/ Just model, a decrease in capital inflows may increase the probability of rescheduling, but, at the same time, lower capital inflows may be the result of information that a country will soon seek to have its debt rescheduled. The simultaneous determination between the probability of rescheduling and the amount of capital inflows is taken up in Chapter 5, where a simultaneous equations formulation is estimated. The second point concerns the role of savings and investment in the rescheduling process. Sachs (1981, 1982) has argued, both from a theoretical standpoint and with actual data, that the savings rate may be an important determinant of the probability of rescheduling. He provides evidence (Sachs 1981, p. 246) to show that five of six countries that rescheduled their debt in the 1970s had significant decreases in their savings rates in the period prior to the rescheduling. Econometric evidence is also available from Kharas (1981) to support the hypothesis that the probability of rescheduling decreases with increasing investment rates. In view of the evidence, it is rather surprising that only the study by Mayo and Barrett (1978) has included gross fixed capital formation as an explanatory variable. The importance of this variable will be further examined in Chapter 5.

PAGE 50

44 Real Interest Rates and Commercial Bank Involvement in NOLDC Lending" The NOLDC Lending Process from the Commercial Banks' Perspective The NOLDC lending operations of international commercial banks have raised a number of questions. The first issue that must be tackled is why commercial banks have emerged as lenders to NOLDCs. The classical answer to the question of why banks arise is that they fulfill the function of financial intermediaries: they channel funds from surplus to deficit units. This is precisely their role in lending to NOLDCs. They channel funds from. countries with a surplus in their current account (the OPEC members and a small number of European countries that maintained their surplus in the 1970s) to countries in deficit in their current account. The question arises as to why surplus countries do not lend directly to the deficit countries. Solomon (1977, p. 480) points out that since 1973 the OPEC countries have made some direct investments in NOLDCs and have also provided some grants, but these have been mainly motivated by political considerations. In fact, the majority of the flow of funds from OPEC to NOLDCs has been on concessional terms. The Bank for International Settlements (1981, p. 97) reveals in its annual report that OPEC made long-term investments in developing countries of 4.9, 6.5, 9.6 and 6.6 billion dollars in the four years 1974, 1975, 1979 and 1980. These amounts are to be contrasted with current account deficits in the NOLDCs of 46.5, 32.9, 57.6 and 82.1 billion dollars (from Table 2). Shihata and Mabro (1979, p. 163) report that in 1977, of a total 7587.8 million dollars of net disbursements from OPEC to NOLDCs, 5740.9 million was on concessional terms and only 1846.9 on nonconcessional terms.

PAGE 51

45 The emergence of commercial banks as lenders to NOLDCs and the unwillingness of OPEC to lend directly to NOLDCs can be explained in terms of risk aversion on the part of OPEC. The OPEC members chose to invest their surplus funds in relatively liquid and riskless assets in western countries, mainly in the form of bank deposits and government securities. International commercial banks found their deposits swollen. Therefore, by lending to NOLDCs, they have assumed the role of risk takers for which they are better suited than OPEC. One other reason for the intermediation function of banks is information economies: banks can pool deposits from several sources and lend to a small number of countries, about which they are well informed and can monitor their performance closely. Eaton and Gersovitz (1981b, p. 37) add that a further reason why OPEC members do not lend directly to NOLDCs is that, in the event of a default, the OECD countries would be able to impose relatively more severe sanctions than OPEC. Although this is quite probably true, it is not at all certain whether the OECD governments would be willing to impose sanctions on behalf of their banks. The historical record shows few instances of governments intervening to secure the assets of their 29 commercial banks. Indeed, as the 1982 conflict between Britain and Argentina has shown, governments are more likely to use bank lending as a political weapon. From a global standpoint, the commercial bank lending operations, vis-a-vis the NOLDCs, raise the issue of whether the world's capital stock is optimally allocated between developed and developing countries. Aliber (1977) and Eaton and Gersovitz (1981b, p. 1) make a case for commercial In discussing this point, Sachs (1982, p. 36) provides some sporadic historical instances of government overseas intervention to force payment of debts„

PAGE 52

46 bank lending to NOLDCs as the transfer of capital from developed countries, where capital is abundant and the rate of return relatively low, to developing countries where the rate of return is high. To quote Aliber: "The rates of return in the developing countries are substantially higher than those in developed countries, and rates of economic growth in the developing countries are higher than in the developed countries. Both suggest that world income is enhanced by the allocation of capital from the developed to developing countries This view holds that, so long as differences in growth rates (proxies for the return on the marginal investment) between the"developed and developing countries are large in relation to the higher risks associated with the flow of funds to developing countries; then the world's welfare will be raised if larger credits are allocated to developing countries. The role of commercial banks as financial intermediaries raises the question of whether they can be replaced by an international organization, such as the International Monetary Fund (IMF), if they are seen to be ineffective in their lending, charging too "high" an interest rate or lending too "little." Leaving aside the question of what constitutes too high an interest rate or too little lending, Solomon (1977, pp. 500-501; 1981, p. 605) answers the question in the affirmative but provides no explanation for his position. Later on, however, he argues (Solomon, 1981, p. 606) that the proper role of the IMF should not be to "bail out" banks but to supplement bank lending through its various facilities, enforce stabilization programs to revitalize the country's economy and provide the country with a "seal of approval." Cited in Long and Veneroso (1981, p. 514).

PAGE 53

47 What the proper role of international institutions should be in the lending process to NOLDCs, was probably best summarized by Eaton and Gersovitz (1981b, p. 35). First, it is to collect and disseminate information concerning the economic conditions in each country, leaving the banks to undertake the risk-taking function for which they are optimally placed. Second, to organize lenders and provide a central coordinating body in the case of default by a NOLDC. Third, to act as a lender of last resort. They argue that the first two functions are potentially stabilizing whereas the last one is potentially destabilizing because it may encourage commercial banks to take on excessive risks in. their lending to NOLDCs. As for the second function, the last few years have seen coordinated arrangements by commercial banks in rescheduling NOLDC debts under the auspices of the Paris Club, after a country has entered into negotiations with the IMF concerning an appropriate stabilization program to pursue. One issue worthy of note is the geographic concentration of banks' lending to NOLDCs. The U.S. banks have tended to concentrate their lending in Latin American countries and European banks in African and Asian countries. Eaton and Gersovitz (1981b, p. 15) suggest three reasons for the concentration. First, drawing on their empirical results (Eaton and Gersovitz, 1980), they argue that only a small number of countries are needed for sufficient diversification. Second, economies of scale are derived from concentrating on the performance of only a small number 31 of countries. Third, the governments of the lending banks can impose sanctions on different countries with varying degrees of effectiveness. Here, the role of the IMF as a disseminator of information takes on added significance.

PAGE 54

48 For instance, the U.S. government may be able to impose more effective sanctions in Latin America than in Asia. However, given the point raised earlier concerning the reluctance of governments to act on behalf of their commercial banks, this argument does not appear relevant. Finally, one reason that Eaton and Gersovitz (1981b) fail to mention but could be quite important, is the influence of tradition. Many African and Asian countries were formerly colonized by the European nations and today, as independent states, they continue their commercial ties with the countries they were once colonized by. One very important issue concerning the relationship between international commercial banks and the borrowing countries, is the likelihood that a country would seek to default on its loan commitments or demand a rescheduling of its debt. Formal mathematical models attempting to estimate the probability of default, were reviewed in the previous section. From a somewhat different perspective, Sachs (1982) has examined the history of LDC defaults and has drawn the conclusion that the strategy of ommmercial banks vis-a-vis the borrowers can be. modeled as a game: a noncooperative game before World War II, resulting in a series of defaults, and a cooperative game since the War, resulting in only one default (North Korea) and a number of reschedulings. Sachs (1982) and Sargen (1977, p. 21) provide an interesting insight concerning the end result of loan reschedulings: the commercial banks have emerged from rescheduling negotiations, having lost very little in terms of the value of their principal or interest. The same conclusion is reached by Feder and Ross (1982) who have examined the relationship between bankers' subjective estimates of rescheduling risks and the credit terms in the Euromarket. From their empirical results they have argued that the expected loss rates, both in

PAGE 55

49 the grace period and the rest of the loan period, have been quite low, confirming the previous assertion regarding losses experienced as a result of loan reschedulings. In the light of this analysis the assumption of Eaton and Gersovitz (1980, p. 5) that default by an LDC prevents its future reentry in the international financial markets, must be reevaluated for at least the postwar period. The banks have rescheduled loans rather than allow a country to default, so as not to have to face the issue of reentry after default. One reason may be the fear of one default sparking a whole series of defaults, much like a run on the deposits of domestic banks following the collapse of a single bank. Although the 1 ikel ihood.-of default does not appear significant at present, the lessons from history must not be forgotten. The large scale defaults of developing countries in the 1930s, occurred precisely at a time when the financial community least expected them. The question of widespread default among NOLDCs deserves further attention. Eaton and Gersovitz (1980, pp. 14-19) have argued that such an event is more likely if the exports of borrowing NOLDCs are highly 32 correlated. Their empirical results demonstrate that the export performance of individual NOLDCs are not correlated and they conclude that default is likely to be confined to one individual country rather than involve a series of NOLDCs. Widespread default may not only originate from the LDC side of the market but also from the supply side of the market, as suggested by Sachs (1982, p. 7). In the case of a panic following the default of one NOLDC Based on factor analysis and multidimensional scaling.

PAGE 56

50 commercial banks may refuse to keep credit lines open to NOLDCs, pushing some of them to default. He suggests that supply elements may have been responsible for the widespread default of developing countries in the midst of the Great Depression in 1931 and 1932 and that history almost repeated itself in 1974. The question arises as to whether a lender of last resort would be able to prevent such widespread defaults. On this issue, Sachs is somewhat pessimistic and argues that it is doubtful whether an international organization, such as the IMF, could continue to keep credit lines open to NOLDCs in the case of a major panic. One important development in the last decade, pointed out by Long (1980, p. 488), Long and Veneroso (1981, p. 515) and Kareken (1977) in his discussion of Solomon's (1977) paper, is the increasing share of NOLDC loans in the portfolios of many international banks. Long believes that as a result banks, on the one hand, will be reluctant to increase the NOLDC loan portion of their portfolio while, on the other, bank supervisory agencies and other bodies are likely to introduce rules and apply pressure on banks to lower their holdings to NOLDC loans."^"^ Kareken believes very strongly that commercial banks should cut back on their international lending to NOLDCs. He argues that the result would not be As witnessed recently in the U.S. by the joint regulatory package recommended by the Federal Reserve the Comptroller of the Currency and the Federal Deposit Insurance Corporation, which would require banks to set up a reserve against earnings on loans where the full interest has not been paid for six months or where there is no prospect of compliance with the IMF measures. The reserve would be 10 percent of all problem loans initially, rising to 15 percent later. Banks would also be required to spread loan fees across the maturity of the loan and increase their capital base.

PAGE 57

51 a decrease in aggregate demand since the funds would be directed as loans elsewhere. Other writers, in their discussion of Solomon's paper, disagreed with Kareken, arguing that it would decrease aggregate demand, disrupting the development plans of many NOLDCs. Kareken also makes a strong case for the heavy regulation of the international lending of U.S. banks. Because the deposits of U.S. banks are insured, it induces them to take "excessive" risks in their lending. However, as Kareken adds, this is not a problem peculiar to the international lending of U.S. banks. So far, what can be termed as the "orthodox" view of the role of commercial banks in the'lending process to NOLDCs, has been outlined. It is quite clear that the indebtedness of developing countries i^ a subject of political economy proper, where economics and politics are inextricably linked. The very nature of the subject evokes suggestions, proposals and solutions, the origin of which can ultimately be traced to each writer's political convictions. While the remainder of this dissertation expounds on the orthodox view of the role of commercial banks, it may be worthwhile to devote a few lines to some alternative views. Lichtensztejn and Quijano (1982) provide a lengthy discussion of the increasing privatization of the external debt of developing countries. The increasing involvement of private international banks has worsened the terms of indebtedness and has imposed a severe burden on the balance of payments of developing countries. They point to an important development in the market for international loans: a shift has occurred from evaluating countries on the basis os their ability to pay, to an overall evaluation of the policies of each country concerning its recepti veness towards foreign capital, the so-called "country risk" approach.

PAGE 58

52 The shift in emphasis is important, in their estimation, as they view lending by private commercial banks as a mere extension of the penetration of transnational capital into developing countries, accompanying and financing direct productive investments. The lending by private banks has enabled them to increase their influence, either directly or under the auspices of the IMF, on the economic policies pursued by developing countries. These policies have promoted an export orientation to the development efforts of LDCs, so as to be able to obtain the foreign exchange to repay the loans granted to them. The involvement of private banks in the internal affairs of developing countries, has extended to the encouragement of the concentration ^of industrial capital and the denationalization of the country's industrial base. The authors provide examples and case studies of several Latin American countries to support their views. Similar views can be found in L'Heriteau (1979), who argues that the financial capital loaned by commercial banks serves to strengthen the dominance of the "real" capital. The increasing involvement of private banks results in the ultimate subordination of national development plans to the exigencies of the banks and the IMF and the reorganization of domestic policies to promoting the country's exports. The increasing importance of private bank loans in developing country finance has resulted in a different composition of capital inflows into developing countries. While private direct investment comprised the majority of private capital inflows during the 1960s, loans were the major item in the 1970s. However, such a general account hides the disparities among individual developing countries. The countries most heavily indebted to the banks have seen their share of private direct

PAGE 59

53 investments remain constant and even increase during the 1970s, because it is precisely in these countries that the international banking system can support the operations of multinational firms. L'Heriteau makes one final noteworthy point concerning the origin of external indebtedness. Several writers have identified a "vicious" circle of indebtedness, according to which bal ance-of-payments deficits render necessary external borrowing to cover the deficit, which in turn increases the burden on the balance of payments on account of the debtservicing obligations. However, during the period 1970-77 several 34 countries were in equilibrium or entertained a bal ance-of-payments surplus. For these countries, external indebtedness did not result from bal ance-of-payments problems but rather from a conscious effort, on their part, to attract foreign direct investment. In Venezuela, the deficit in the balance of payments first appeared in 1977 but by the end of 1976, the country had found itself with an external debt exceeding 3 billion dollars. Such examples point to the opposite causation, between external debt and the bal ance-of-payments deficit, from that envisaged traditionally. It is, perhaps, not surprising that there should exist disagreements and diametrically opposite points of view regarding the role of commercial banks in international lending. It is, after all, a topic that arouses the interest of various parties, be they politicians, administrators or economists. What may be somewhat surprising is the treatment of the subject by academic economists. While a huge amount of writings has appeared on the role of commercial banks in lending to NOLDCs, the approach has For example Gabon, Iraq, Jordan and Malta and other more heavily indebted countries such as Taiwan, Iran, Syria and Venezuela.

PAGE 60

54 been rather superficial, relying, on the whole, on general discussions of the involvement of commercial banks, illustrated by statistical tables pointing out general trends and different lending practices. There has been virtually no attempt at approaching the subject from a theoretical point of view, making use of existing banking theory models or developing new models to examine the lending behaviour of international banks and the implications of the banks' lending practices on the flow of funds to developing countries. The next chapter follows up on this theme. Movements in Real Interest Rates The movement of real interest rates over the last few years has influenced, to a great extent, the pattern of lending to NOLDCs. ^The decade of the 1970s saw a gradual but steady rise in real interest rates. While at the beginning of the decade real interest rates were very low and even negative, they rose sharply towards the end of the decade. As Figure 1 shows, real short-term interest rates rose in many developed countries' financial markets throughout the 1970s; the rate on threemonth Eurodollar and other Eurocurrency deposits rose steadily throughout the late 1970s as shown in Figure 2. Since increasingly many loans to developing countries are of the floating rate variety, where the rate is linked to those in the major capital markets, the rise in interest rates has imposed an increasing burden on borrowers. In addition to the obvious one of higher interest payments, the overstatement of the current account deficit was explained in a previous section. Sachs (1981, pp. 238-241) has argued that nominal interest rates were kept low in the early 1970s for two reasons: first, the shift in strategy in developing countries from a policy of import substitution to one of export promotion and the subsequent liberalization of capital flows

PAGE 61

55 1972-76 -1 977-81 Figure 1 Real Short Term Interest Rates Source: Bank for International Settlements, Annual Reoort Basle: BIS, 1981 p. 59.

PAGE 62

56 Figure 2 Interbank Rates on three-Month Furocurrency Deposits and Differentials Over Domestic Rates source: J-^|e™ati„„a, ^ Settlements Annual Report.

PAGE 63

57 and second because of depressed investment opportunities in the developed countries. He attributes the reduction in investment opportunities to several factors. First, investments in many developed countries had reached diminishing returns by the early 1970s. Second, most developed economies have suffered from a sharp reduction in profits and an increasing share of labour in GNP and third an investment boom in raw materials took place in the NOLDCs. The problem of rising real interest rates has been emphasized by many writers. Solomon (1981) has examined the economic performance of the major borrowers in international financial markets on the assumption of a reduction in real interest rates and-has found them all creditworthy. However, as he points in his conclusion, if his assumption does not materialize, the economies of the borrowing NOLDCs will be severely disrupted with serious repercussions on the world economy. Long (1980, p. 485) and Long and Veneroso (1981, p. 511) point out that while real interest rates have been rising, the proportion of loans made to NOLDCs on concessional terms has been declining and at the same time the proportion of floating rate loans has been rising, exacerbating the debt burden of NOLDCs. Long (1981, p. 296) shows that for a country like Peru, the debt of which is equal to one half of its GNP, a rise in real interest rates of two percent is equivalent to a tax of one percent of GNP, not an insignificant amount. On the basis of a simulation exercise. Long (1980) has calculated the limiting value of the ratio of debt to GNP and the number of years for the ratio to rise from 30 to 40 percent, both for different values of the real interest rate and the rate of growth of GNP. His results point to the crucial role of the real interest rate in the debt accumulation

PAGE 64

58 process. In fact, a tradeoff exists between the growth rate of GNP and the real interest rate: approximately a 1.5 percentage point increase in the growth rate of GNP is required to offset a one percentage point increase in the real rate of interest, if the limiting value of the debt/ GNP ratio and the period for the ratio to increase by 10 percentage points is to remain constant. The increase in nominal interest rates during the 197Gs can be attributed, to a large extent, to the increasing inflation rates in the major economies. When interest rates on loans are fixed, an increasing inflation rate benefits.the borrower by decreasing the real value of the loan. However, as Nowzad (1982, pp. 165-67) points out, under variable interest rates, which incorporate an inflation premium, the higher interest rates resulting from the increase in the inflation rate may compensate for the drop in the real value of the loan. Actual interest pa^mients, under variable interest rates, include a component that takes account of the effect of inflation on the real value of the loan and therefore the loan is amortized at a faster rate, in real terms, than was originally expected. If the real interest rate remains unchanged, the long run creditworthiness of a country is unaffected by inflation and higher nominal interest rates. However, the faster effective amortization rate alters the short-run borrowing requirements of a country. In order to maintain the same real net resource transfer, higher gross borrowing becomes necessary, which worsens the debt burden of the country in the short run. Several forms of indexation have been suggested as solutions to the problem of faster amortization. Financial amortization would adjust amortization payments by an interest rate index. The same principle of

PAGE 65

59 changing the amortization schedule to offset the faster amortization rate lies behind the proposal of maturity indexation. Price indexation of amortization payments, combined with constant or floating real interest rates on loans, is a further possible solution, but the choice of the appropriate price index or the appropriate real interest rate present sizeable problems. Indeed, as Nowzad (1982, p. 168) has emphasized, all the indexation proposals have technical deficiencies, lack widespread support and do not contribute in any way in settling the uncertainties caused by inflation and interest rate movements; such uncertainties as regards the real or nominal debt service and the real or nominal borrowing requirements. The implementation of any of the indexation schemes does not appear likely, at least in the near future. External Borrowing, Consumption and Investment One crucial aspect of commercial bank loans to NOLDCs, is whether they are used to finance consumption or build up productive capacity. This issue is of extreme importance because it is instrumental in determining the amount loaned to a country, the interest premium charged and the overall credit standing of the economy. Loans used to sustain consumption levels that may have been sharply reduced due to adverse exogenous factors, will, in all likelihood, bear a higher interest rate and involve smaller amounts. On the contrary, loans which are used to invest in projects that earn a positive return at the interest rate charged, should provide a steady future income stream that can be used to repay the loans and will, therefore, bear a lower interest premium. Having drawn the distinction between consumption and investment loans, it may be useful to point out that in practice it is difficult to distinguish readily between the two types.

PAGE 66

60 Sachs (1981) has argued that, since 1973, two events have made a profound impact on the current account: the drop in real interest rates during the beginning and middle years of the decade stimulated investment and the rise in oil prices inflated the oil bill out of proportion. On the basis of a theoretical model he demonstrates that the rise in oil prices will have a differential effect on the current account of the oil-dependent economies only when the rise in oil prices is perceived as temporary. If it is perceived as permanent, there should be no differential impact on the current account of oil-dependent economies as compared to oil -exporting countries. The main proposition put forward;by Sachs (1981, p. 211) is :that when perfect capital mobility is assumed, shifts in investment patterns between countries lead to corresponding shifts in the current account. While higher oil prices since 1973 may have forced countries to run higher current account deficits, these deficits also reflect borrowing by these countries in response to investment opportunities.^^ Thus in the 1970s two effects on the current account of NOLDCs can be distinguished: first, the rise in oil prices and second the drop in real interest rates and the increase in investment opportunities. Therefore, the increased borrowing by NOLDCs may have been a result of higher oil prices and increased current account deficits or may have resulted from investment opportunities not available previously or some combination of the two. If the second hypothesis is valid then the higher borrowing will lead to a rising consumption path over time and there should be no The previous section explained why investment opportunities shifted to NOLDCs in the 1970s

PAGE 67

51 problem with loan repa^^ents; if the first argument is valid, grave doubts can be raised concerning the repayment of NOLDC loans. Sachs (1981, p. 233) argues that, between the two periods 1965-73 and 1974-79, the current account/GNP ratio of developing countries worsened more than that of developed countries as a result of increases in investment/GNP ratios in large LDCs which were in excess of the saving/GNP ratios; in addition, saving/GNP ratios rose for some of the LDCs and fell for others. By contrast, in most developed countries between the same periods both the i nvestment/GNP and the saving/GNP ratios fell, the latter decline of greater magnitude. Sachs then goes on to argue that "much of the growth in LDC debt reflects increased investment and' should not pose a problem of repayment. The major borrowers have accumulated debt in the context of rising or stable, but not falling, saving ratios" (Sachs, 1981, p. 243). The argument of Sachs is supported by Solomon (1977, p. 498) who provides evidence in the form of a table of gross capital formation as a percentage of total absorption^'' for ten major borrowers"^^ for the period 1970-76. In only two of the countries (Chile and Colombia) did the ratio decline after 1973. Solomon concludes that the fear that loans to developing countries are used to support consumption, is unfounded. In his discussion of Solomon's paper, Greenspan (1977) argues that aggregate The countries included in the analysis were Argentina, Brazil, Chile, Colombia, Mexico, Peru, Philippines, South Korea, Taiwan and Thailand. 37 Defined as gross domestic product plus net imports of goods and services. The same countries as those investigated by Sachs (1981).

PAGE 68

62 investment ratios may hide the fact that for some LDCs higher ratios may not represent investment in productive capacity, but rather military buildup. In his follow-up paper, Solomon (1981, p. 601) presents further evidence, again in terms of ratios of gross fixed capital formation to total absorption for the period 1976-80, to support his earlier position. A different opinion on this issue is voiced by Eaton and Gersovitz (1981a, p. 301) who state, without providing any evidence, that the oil price rises in the 1970s and the recession in the developed countries prompted the LDCs to borrow for short-term adjustment purposes. The question of whether loans are used for consumption or investment purposes must, in the end, be an empirical one. Sachs (1981, p. 245) has presented econometric evidence to support his position. The dependent variable is the spread over the London interbank offered rate (LIBOR) charged on loans to each country (i-LIBOR). The independent variable are the current account/GNP ratio (CA/GNP), the i nvestment/GNP ratio (I/GNP), the per capita GNP (GNP/L) and the total debt/GNP ratio (D/GNP). Sachs' results are as follows:"^^ (i-LIB0R)7q = 1.35 1.83(I/GNP)^p 2.49(CA/GNP),o (6.3) (-2.0) (-2.3) 72 R^ = 0.16 The numbers in parentheses represent t-ratios; the subscripts refer to years.

PAGE 69

63 or (i-LIBOR) = 1 .38 1.89(I/GNP).. 3.U{Cl\/QHP)^^ (5.7) (-2.2) 78 ^_3^2) '78 0.009(GNP/L)7j. + 0.008(D/GNP),Q (-1.4) 78 ^Q^2) '78 0.35 As can be seen, the coefficient of (I/GNP) in both equations is significant and negative indicating that a stronger investment performance implies a higher probability of repayment and therefore a lower interest premium is charged. Two comments are pertinent to the equations estimated by Sachs, in particular the second one. First, the equation, as it stands, represents a supply of debt function and should form part of a simultaneous equations model supplemented by a demand for debt equation. Therefore the coefficient of (D/GNP) in the equation does not represent the coefficient of the supply function. This observation may help explain why the coefficient is not significant. Second, the level of investment will, in turn, depend on what the future interest premium charged on loans is expected to be. A simultaneous equations model was estimated in an attempt to overcome these criticisms. The three-stage least squares estimates are as follows:^ Thefirst equation represents a supply of debt function. The second equation is an investment function where investment demand is a function of the expected interest premium in the following period and expectations are formed adaptively. The third equation is a demand for debt function. The variables endogenous to the model are (i-LIBOR), (I/GNP) and (D/GNP) whereas all the rest are exogenous or predetermined. All three equations are overidentified. The values reported in parentheses are asymptotic t-ratios and subscripts refer to years.

PAGE 70

64 (i-LIBOPO^Q = 1.626 2.636(I/GNP)7j, + 0.677(CA/GNP) (3.728) (-1.794) (0.380) 78 + 0.288(D/GNP)7Q 0.619DS (0.364) (-1.052) 78 (I/GNP) 79 0.040 + 0.738(I/GNP) (0.864) (6.819) 78 + 0.022(i-LIB0R) (0.804) 79 0.240(GB/GNP) (01 .383) 79 + 0.248(GB/GNP) (1.257) 78 (D/GNP) 79 1.506 + 1.392(i-LIB0R)7Q 4.200(C/GNP) (1.284) (2.158) (-1.565) 78 2.023(CA/GNP)7Q + 1.003DS (-1.724) (2.174) 78 where, in addition to the variables defined earlier, DS represents the debt service ratio, (GN/GNP) the ratio of the government budget surplus to GNP and (C/GNP) the ratio of private consumption to GNP. The threestage least squares results proved quite disappointing. Most coefficients are insignificant or of the wrong sign. In particular, the signs of the coefficient of (D/GNP) in the first equation and (i-LIBOR.) in the third are the same (positive), indicating that the demand and supply equations 41 have not been identified. One final question that needs to be investigated is whether the risk premium (as measured by the spread over LIBOR) varies in a systematic Two possible explanations come to mind concerning the disappointing nature of the results. First, the model is grossly misspecified and that variables assumed exogenous should be treated as endogenous. Second, if developing countries are constrained in their borrowing, as Eaton and Gersovitz (1980, 1981a, 1981b) have suggested, it makes absolutely no sense estimating models of the sort specified in this section without introducing a supply constraint explicitly.

PAGE 71

65 fashion with the level of indebtedness of each country. Sachs (1982, p. 14) argues that in a world of perfect capital mobility there is very little tendency for interest spreads to rise as the indebtedness of a country increases. However, when the assumption of perfect capital mobility is dropped, he argues (Sachs, 1982, p. 19) that interest rate premia tend to rise as the level of debt for each country grows. Empirical evidence to support Sachs' proposition is provided by Brittain (1977, pp. 377-80). The following equation was estimated: (i-LIBOR) = 0.0541 + 1.1733DM + 0.0599(D/GNP) 0. 0424(DM) (D/GNP) (a.l650) (2.6616) (4.5339) (-2.6465) = 0.5138 F(3,21) = 9.454 where DM is a dummy variable that takes on the value of 0 in 1974 and 1 42 in 1975 and 1976. Brittain confirms the hypothesis that the interest spread rises with the overall level of indebtedness. In addition he argues that a fundamental change occurred in the market in 1975-76 such that developing countries were viewed as greater risks. Brittain's equation was reestimated by the present author with data for the years 1974 and 1979 and the following results were obtained: (i-LIBOR) 0.0270 + 0.8392DM + 0.0590(D/GNP) 0.061 5(DM) (D/GNP) (0.0692) (1.2816) (3.7764) (-1.8841) R^ = 0.5865 F(3,14) = 6.62 Data from the World Bank is used for (i-LIBOR) and from Citibank for (D/GNP). F( ) is the value for the F-ratio statistic. The intercept increased from 0.0541 in the first period to 1.2274 in the latter period.

PAGE 72

66 As can be seen, the results are similar to those obtained by Brittain. However, it must be pointed out that the positive coefficient obtained here for the variable (D/6NP) is a result of the strong positive correlation between (i-LIBOR) and (D/GNP) in 1974.^^. When (i-LIBOR) is regressed on (D/GNP) for either 1975/76 or 1979 no significant correlation 45 is obtained. Moreover, the same data from Euromoney used in Sachs' (1981) study does not provide any significant results when (i-LIBOR) is regressed on (D/GNP) for all developing countries. .For the ten largest borrowers the coefficient is significant at the 10 percent level AC but not the 5 percent. 44 The simple ordinary least squares (OLS) regression of (i-LIBOR) on (D/GNP) for the year 1974 gave the following results: (i-LIBOR) = -0.0279 + 0.0602(D/GNP) (-0.0442) (2.5156) R^ = 0.5133 F(l ,6) = 6.33 Again t-ratios are shown in parentheses. 45 The OLS regressions gave the following results for 1975/76: (i-LIBOR) = 1.5908 + 0.0036(D/GNP) (11.1924) (0.9229) R^ = 0.0505 F(l ,16) = 0.85 and for 1979 (i-LIBOR) = 0.8662 0.0024(D/GNP) (5.2652)(-0.2713) R^ = 0.0121 F(l ,6) = 0.07 46 The OLS regressions for all developing countries gave the following resul ts : (i-LIBOR) = 0.9270 + 0.0044(D/GNP) (7.6105) (1.1325) R^ 0.0353 F(l,35) = 1.28 and for the ten largest borrowers: (i-LIBOR) = 0.4197 + 0.0187(D/GNP) (1.5697) (2.0397) R^ = 0.3421 F(l ,8) = 4.16.

PAGE 73

67 The same criticisms that were voiced in connection with Sachs' model are appropriate here. The supply of debt function estimated by Brittain is part of a simultaneous equations system. Moreover, in the Brittain study the positive correlation obtained between the interest spread and the total level of indebtedness is a result of a strong relationship between the two variables during a particular year, but that relationship does not extend over time. In a further attempt to examine the determinants of the risk premium, Sargen (1976, pp. 27-30) regressed the premium on a dummy variable used to distinguish between developed and developing countries, a year dummy (1974 or 1975) and the maturity of the commitment, for 57 loans to developed and 177 loans to developing countries from the 1974-75 period. He found that developing countries paid, on average, a premium of 140 basis points in 1974, whereas developed country borrowers paid about 25 basis points less on average. Maturity was negatively and significantly related to the risk premium but its magnitude was relatively small. Within the group of developing countries he found lowerand middleincome LDCs paid only about 10 basis points more than higher-income LDCs and Mexico paid about 25 basis points less than other higher-income LDCs. The coefficients of the debt service ratio and the inflation rate were both positive and significant but small in magnitude. Finally, the large trade deficits of developing countries in 1975 were reflected in higher premiums paid. The models presented in this section serve to bring to the forefront an important issue namely the process of the determination of country risk

PAGE 74

68 premiums. As of recent, interest premiums have not varied markedly among 47 developing countries, despite obvious divergences in economic structures, potential and prospects among developing countries. Research in this area has not addressed this issue sufficiently and it is certain that further work is both desirable and necessary. The data for developing countries from Euromoney show a range of weighted interest rate spreads over LIBOR of between 0.563 and 2.037 in 1979 and between 0.39 and 2.22 percentage points in 1981.

PAGE 75

CHAPTER TWO BANK LOAN RATE INDEXATION IN THE EUROCURRENCY MARKET Introduction The previous chapter discussed the importance of the external debt of developing countries to the functioning of the international financial system and introduced it as one of the central themes shaping current economic relations between developed and developing countries. In addressing the different issues raised by the indebtedness of developing countries to commercial banks, it was pointed out that, despite the voluminous literature and the breadth of coverage, the vast majority of the contributions in this area, barring two exceptions,"* have centered on the ability of each country to service its debt from a "macroeconomic" perspective. The focus of attention has been the so-called "countryrisk" approach, according to which a country's prospects of repaying the funds loaned to it are gauged by a variety of economic indicators, usually in the form of aggregate ratios. These ratios are either closely examined to discover any trend that may exist, or are introduced into formal mathematical models estimating the probability of default. Apart from the two contributions cited, there has not been any attempt at approaching the question from the perspective of the international commercial banks granting loans to developing countries. The two papers by Feder and Just (1977b, 1980). 69

PAGE 76

70 The present chapter develops a model aimed at redressing this imbalance. It extends the models introduced by Jaffee and Modigliani (1969) and James (1982) to the international setting of loans made by commercial banks to sovereign borrowers, be they public or publicly guaranteed entities. The model introduces the risk of default as perceived by the 2 lender, as an important determinant of the interest rate charged by commercial banks. In addition, it incorporates the practices of making loans under commitments and that of indexing loans to a particular rate, both of which are widespread in the Eurocurrency market. The next section cjescribes the model and its relevance to lending practices in the Eurocurrency market. The following section presents the theoretical results of the model with particular reference to the implications of loan rate indexation on the pattern of borrowing on commitments. The empirical results from testing the hypotheses derived from the model are presented in the following section and a final section summarizes the results. The Model Before a formal introduction to the model, it is worthwhile to explain, so far as possible, the rationale behind loan commitments, A loan commitment is an agreement between a bank and its customer, according to which the bank provides credit up to a maximum amount and on prespecified terms. There are usually two types of arrangements as regards the rate charged on the loan commitment: a fixed rate commitment What was termed, in chapter 1, the "subjective" probability of default.

PAGE 77

71 stipulates a specific rate at the beginning of the contract period according to which borrowing takes place, whereas a fixed formula ties the borrowing rate to a particular rate in the market. There are also two different rules, depending on whether the bank or the customer determines the loan volume to be taken down, within the maximum amount specified by the agreement. There seems to be some disagreement in the literature concerning the rationale behind credit commitments. Campbell (1978) argues that they provide insurance to the borrower from a change in credit standing.^ James (1982), on the other hand, views loan commitments as means of economizing on transaction costs, where there are "transaction specific assets" involved in the negotiation between the two parties, by the establishment of a long-term relationship. Regardless of the rationale behind loan commitments, some important implications concerning the degree of competition in the market and the pricing of loan commitments are evident. In particular, there is a high degree of competition in the market in the stages preceding the signing of the loan commitment, but once agreement is reached, the situation is that of bilateral monopoly: a single borrower faces a single bank. This is the practice of indexation, to which the remainder of this chapter is addressed. In the case of the fixed formula agreement, the premium charged (over the rate to which the loan is tied) remains constant. 5 Such as the cost of establishing the credit standing of each borrower or the cost of initiating and processing loan applications from new customers

PAGE 78

72 The point of departure of the analysis developed in the present study, with that of the earlier works by Jaffee and Modigliani (1969) and James (1982), is that variables at the control of the borrower, namely the amount of exports and international reserves, are assumed to have an impact on the density function of the revenues generated by the borrower which are used to repay the loan, as that is perceived by the lender. In accordance with the earlier papers, the bank faces an uncertain return with respect to any one loan to a particular customer, due to the fact that adverse circumstances may imply that the borrower is unable to repay the full amount required under the contract agreement. Therefore, the amount of revenues generated by the borrower that dan be used to repay the loan, 6, is viewed as a random variable by the bank. However, it is assumed that the bank has a priori knowledge of the density function, f(.), of the random variable e. In the international context of loans to sovereign borrowers, the density function f(-) is assumed to depend on two variables the borrower has control over: the amount of international reserves a country possesses p^. and the amount of its exports x^ The importance of international reserves in determining the perceived creditworthiness of a country was discussed in Chapter 1. The studies of Feder and Just (1977a, 1977b), Feder, Just and Ross (1981), Frank and Cline (1971) and Kapur (1977) all point to the level of international reserves a country possesses as an important indicator of creditworthiness; a higher level of reserves implies a country is better able to meet its external debt obligations and therefore less likely to default. The same studies point to the role of exports, both as a static (the level of exports) and as a dynamic (growth of exports) indicator of creditworthiness. On the

PAGE 79

73 importance of exports, Kapur (1977, p. 182) states: "The level and growth of exports in any individual economy have a bearing on its liquidity position, debt-service ratio growth in GNP and, as a combined result of these factors, ultimately the perceived creditworthiness of the country in terms of default risk." One further variable that the studies by Feder and Just (1977a) and Feder, Just and Ross (1981) have identifed as an important determinant of creditworthiness is the amount of capital inflows. However, as will be explained in Chapter 5, there exists a simultaneous determination between creditworthiness and the amount of capital inflows: while -a smaller amount of capital inflows may diminish a country's creditworthiness the smaller amount could be a result kif knowledge that the country may seek a rescheduling of its debt. Therefore, the amount of capital inflows is omitted from the present analysis. The density function for the ith borrower (as perceived by the bank) is denoted by f.j(e; p^. x.) where CO E[9] = / ef.(e; p.; x.)de =J= g(p., x-) — 00 An increase in either or x^j will move the entire density function to 6 the right. The density function is assumed independent of the size of the loan made, although as Jaffee and Modigliani (1969, p. 852) point out, in the nonindependence case, the results would hold equally well, provided that the investment opportunity of the borrower is subject to decreasing returns. The objective of the bank is to maximize expected profits by choosing the amount of loan, L., extended to the ith customer, given the Further details are given in Appendix A.

PAGE 80

74 loan rate factor R^.. By varying the loan rate factor, the supply function of bank credit is obtained. The expected profits to a bank from advancing a loan in the amount of L^, are given by ^i ^i = ^-^i ^ / ^^(e; P^-, x.)de + / ef.(e; p., x.)de pL. 1 1 W (1) where R^= 1 + r^. = The interest rate factor p = The cost of funds to the bank^ Q^= The upper bound on the revenues generated by the borrower C^-(L.) = The administrative cost associated with a loan of L,. 1 Two assumptions are in order concerning the cost function: Cl(L.) > 0 and C!j'(L.) > 0 i.e., the marginal cost is positive and increasing. It is worth noting that the term R.L. represents the total amount to be repaid under the contract agreement. In Equation (1), the first term represents the repayment to the bank if no default occurs, whereas the second term represents the repayment in the case of default. Maximization of (1) yields the following equation: R^Cl F.(R.L., p., X.)] p C!(L.) = 0 (2) where Which is closely related to the London interbank offered rate (LIBOR).

PAGE 81

75 R.L. F.(R.L.; p., X.) = / f.(9; p., x.)d9 is the probability of default. The following assumptions are advanced concerning the probability of default 9F.(-) 3F.(.) 9F.(.) WT^'' Equation (2) can be solved implicitly to yield the loan supply curve as L.(R.; p., X., p) • (3) As far as the demand side of the market is concerned, it is assumed that the demand for credit from the ith customer, D^. is inversely related to the interest factor R^. the amount of international reserves p. and the level of exports x^. as '• D^. = D.(R.; p., X.) (4) 3D.(-) 3D.(-) 3D.(.) The demand function in Equation (4) is a residual demand after alternative Q sources of finance are taken into account. It is assumed that the country has limited access to alternative sources of finance, which is the reason for the negative slope (with respect to the interest factor) of the demand function. A final, and rather crucial assumption, is put forward. It is held that an interest factor "R^. exists that equilibrates the market This is the reason why the amount of international reserves p. and the level of exports x^. appear in the demand function. ^

PAGE 82

76 D^-(R^, p., X.) = L.(R., p., X., p) Substituting this equilibrium condition into (2), the following condition must hold true for the loan rate and amount set at the beginning of the loan commitment, to result in an efficient allocation of credit. R^-Cl F.(R.[D.(R., p., X.)]; p., x.)] p C:(D.(R., p., X.)) = 0 (5) In (4) it was assumed that the demand curve for credit is downward sloping and therefore the bank has some degree.of monopoly power in setting the loan rate and amount. Thus, it may seem counter-intuitive for the bank to equilibrate the demand for credit with the supply. However, it must be remembered that before the loan commitment is negotiated the bank has no monopoly power; only after the agreement is signed, a situation of bilateral monopoly arises. The equilibrium condition in (5) can be differentiated totally to yield some interesting and intuitively plausible results. In the two previously cited papers, Feder and Just (1977b, 1980) propose that the interest premium charged on loans to sovereign borrowers in the Eurocurrency market is directly related to the probability of default. Feder and Just proceed to argue that the probability of default is linked to the debt-servicing capacity of each country and that this link is most important in determining the interest premium charged on each loan. The debtservicing capacity, being an unobserved variable, is in turn determined by several economic variables including the ratio of imports to international reserves, the per-capita gross domestic product and the debt-

PAGE 83

77 service ratio. Thus an indirect (and inverse) relation is established between the amount of international reserves and exports on the one hand, and the interest rate on loans on the other. This is an entirely plausible argument which will be shown, in this study, to result directly from profit-maximizing behaviour on the part of commercial banks. In addition, the model outlined here extends the results obtained by James (1982) for domestic borrowers, concerning the effects of indexing loans to a specific rate, to the international setting of sovereign loans. The model is particularly suited to the study of loans made to sovereign borrowers, because the interest rate charged on such loans is made up of two components: the London interbank offered rate (LIBOR) and an .-interest margin that reflects the perceived creditworthiness of the borrower. The loan is usually granted in the form of a revolving credit; it is renewed every six months for the duration of the commitment and whereas the interest margin remains constant, the base rate (LIBOR) changes in response to different conditions in the international capital markets.^ Finally, the model developed here provides a useful insight into the widespread practice of syndicating loans in the Eurocurrency market. As was stated earlier, the presence of transaction specific assets in loan negotiations has given rise to procedures aimed at reducing transactions costs. One such procedure is loan syndication, whereby one bank, known as the lead or agent bank, arranges for all the loan details and evaluates the creditworthiness of the borrower. It provides only part of the loan and for the remainder it organizes the participation of other A more complete discussion of lending practices in the Eurocurrency market is provided by Mohammed and Saccomanni (1973).

PAGE 84

78 (usually smaller) banks. Loan syndication is one of the most important lending features in the Eurocurrency market and accounts for a substantial proportion of all loans generated in this market. ""^ Theoretical Implications of the Model The results of specific interest to this study can be derived by total differentiation of Equation (5) to yield 0 F.(R.[D.(R.; p., X.)]; p., x.jdR. R/(f,.(R^.[&.(R.; P., X.)]; p., X.) (D.(R.; p., X.) R,-8D.(R.; p., X.) BR. )dR. 3D.(R. ; p., X.) f,(R,[D,(R,; p., X,)]; o, x^)R, \^,' dp, 9D. (R. ; p. X.) *f,(WR.; p., X,)]; p,. x.)R, ^ \/ dx.) i -q(o,,,.p,,x,,,,!MV!^ 3D.(R.; p., X.) + ^ dx.] dp = 0 (6) The derivatives of interest are dR./dp, dR./dp. and dR./dx. which can be computed by setting two of dp, dp. and dx. in (6) equal to zero to obtain A thorough discussion of the market for syndicated credits is provided by Goodman (982) who points to the sixteenfold increase of the market from 4.7 billion dollars in 1970 to 76 billion in 1980

PAGE 85

79 ^= !/{[! F.(R.D.(.); P^, X.)] Rifi(RiD^.(-); P^, x.)D.(-)(HnpR ) i 9D.(.) dR. 3Di(.) 9D,(-) {R?f,(R,D.(.); P,, X.) ^ q(D.(.))[-^] }/ dp. {[1 F.(R.D.(.); P^-, X.)] R.f.(R.D.(.); P,-, x.)D.(-) 3D.(-) } (HnoR.) q(D.(.))[-lR-]} < 0 (8) dR. 9D.(-) 8D.(.) (R|fi(RiD.(.); P,, X.) -1^ q(D.(.))[^]}/ dx. where n [1 F.(R.D.(.); P^., X.)] R.f.(R.D.(.); P^., x.)D.(.) 9D.(.) q(D.(.))[-g^]} < 0 (9) 3D.(.) R. 1 1 DR. 9R. DTTI If the demand curve for credit is assumed to be elastic''"' so that I^DR.I ^ '' signs of the derivatives follow from the earlier assumptions concerning the credit demand function and the cost function. There is evidence from Feder and Just (1977b, p. 240), to suggest that the demand elasticity for credit is indeed rather high. 1 1 The same condition is required to hold true in the model of Feder and Just (1977b, p. 226), if profit maximization is to yield meaningful results,

PAGE 86

80 The comparative static results of the model indicate that a rise in the cost of funds to the bank will prompt a bank to charge a higher interest rate on loans, a conclusion which is intuitively appealing and "I p consistent with the results of James (1982). One further interesting conclusion derived from the model is that the interest rate charged on loans will be inversely related to the amount of international reserves and the level of exports of each country. The inverse relation between the level of international reserves and exports on the one hand, and the interest rate charged on the loan on the other, has also been verified by Feder and Just (1977b, 1980). However, in their papers, by using the appropriate assumptions, they are able to prove that the probability of default (as perceived by the lender) does not depend on the interest rate charged or the loan amount. Subsequently, they derive an equation where the interest rate on the loan is directly proportional to the probability of default which, in turn, depends on the debt-servicing capacity of each country. It is further assumed that the debt-servicing capacity of each country is determined by several economic variables including the amount of international However, one must not forget that such a conclusion is predicated on the elasticity of demand being greater than one in absolute value. In fact, in the international banking context, one can give a plausible explanation for the case in which the sign of (7) is negative. If, for example, the cost of funds to the bank rises as a result of a decrease in the current account surplus of OPEC, the demand for credit from oil-iiTiporting LDCs is likely to decrease as their current account deficits are reduced. For the majority of LDCs, credit obtained in the international capital markets is directed, in one way or another, at financing the current account deficit. It could very well be that the decrease in credit demand is sufficient to decrease the interest rate on the loan even if the cost of funds has risen.

PAGE 87

81 reserves and the level of exports. Thus, the link between the interest rate charged, on the one hand, and the level of international reserves and exports on the other, is established via the debt-servicing capacity of each country. In this study, the inverse relationship between the variables concerned is derived explicitly within the framework of a profit-maximization model, where the probability of default is endogenous and depends on both the interest rate charged and the loan amount. The level of international reserves and exports enter the determination of both the demand and supply of credit, whereupon the claimed results are derived by assuming an equilibrating mechanism in existence in the market. The model provides some interesting insights as regards the practice of loan indexation when, within the loan commitment, the bank determines the loan amount to be extended to the borrower. The implications of loan indexation on the pattern of rationing of different borrowers, in the event of an increase in the cost of funds to the bank, are examined. Initially, it is assumed that there are two classes of borrowers: prime and non-prime. In the international banking context prime borrowers are charged the LIBOR (or a very small premium over LIBOR). The perceived probability of default for prime borrowers is assumed lower than that of non-prime borrowers because of factors other than the level of international reserves or exports; for example, different perceptions concerning the political stability of each country. If loan contracts are indexed to a rate such as LIBOR, the changed in the loan rate associated with a change in cost of funds (p) is the same for all borrowers and is equal to the change in LIBOR, dRj_/dp. From equation (7) it can be seen that with indexation, customers with a higher default probability than prime customers will have a lower interest rate adjustment, compared to a contract

PAGE 88

82 without indexation. Therefore, if the cost of funds to the bank rises, these customers are likely to be rationed in the sense that the interest adjustment desired by the bank is greater than that possible under the indexed contract. A similar conclusion holds for borrowers with a lower elasticity of demand for credit where, again, the difference in elasticity is not the result of differences in international reserves or export proceeds. The conclusions reached are in agreement with those of James (1982) but not those of Blackwell and Santomero (1982), who showed 1 3 that customers with the more elastic demand will more likely be rationed first. The results obtained by James (1982) are seen not to extend when one considers differences in international reserves. It is shown formally in Appendix A that, ceteris paribus a country with lower international reserves (and hence a higher probability of default) is less likely to be rationed, when the cost of funds to the bank rises, provided the increase in the probability of default is small enough. This proposition is contrary to the results of James, who has argued that customers with a higher probability of default are likely to be rationed first. Countries with lower international reserves have more limited access to alternative sources of funds for their development needs and, ceteris paribus one would expect their demand for credit to be less responsive to changes in the interest rate; alternatively their elasticity of demand for credit is expected to be lower. Indeed, Blackwell and Santomero have argued, within the context of a domestic-economy customer-relation model, that non-prime Blackwell and Santomero (1982) argue that such customers are likely to be the prime customers.

PAGE 89

83 borrowers (those with lower international reserves here) will have a lower demand elasticity. They proceed to demonstrate that prime customers (with a higher demand elasticity) will most likely be rationed first. In the present context, as shown in Appendix A, borrowers with a lower value of international reserves (non-prime) will be subject to a higher interest rate adjustment when loan contracts are indexed, compared to contracts without indexation. Therefore, when the cost of funds to the bank (p) rises, such customers are less likely to be rationed. The rationale behind this assertion is rather simple. A country with higher international reserves is perceived, according to the assumptions of the model, as having a lower probability of default. However, as shown in equation (8), it is charged a lower interest rate. Therefore, the loss in profit from rationing a country with higher international reserves is potentially smaller. If the decrease in probability of default from a higher level of international reserves is small enough, it will be least costly to the bank to ration such customers and will therefore be rationed first when the cost of runds to the bank rises. All the above conclusions have been drawn with reference to the case where the bank decides on the loan amount to be advanced to each borrower. If the contract arrangements are such that, within the constraints of the loan commitment, the borrower can choose the loan amount, all the results follow suit except that there can be no credit rationing. Nonetheless, borrowers with lower international reserves will be worse off under indexation when the cost of funds to the bank rises as compared to a contract without indexation. Therefore such customers will be expected to reduce their borrowing on loan commitments during periods of rising cost of funds to the banks.

PAGE 90

84 Regardless of the arrangements for determining the loan amount, the arguments put forward in this paper have some interesting empirical implications. Although one cannot advance any hypothesis concerning the total amount of loan commitments taken dov^n by borrowers, it is possible to formulate a hypothesis as regards the composition of loan commitments taken down by all borrowerSo In particular, as has been shown, the interest rate adjustment to borrowers with lower international reserves will be higher when loan contracts are indexed, than when contracts are not indexed. With LIBOR rises, the relative cost of borrowing on the loan commitments for borrowers with lower international reserves is higher (under indexation) than for borrowers with higher internattonal reserves. Therefore, ceteris paribus one would expect the proportion of loan commitments taken down by customers with lower international reserves to decline when LIBOR rises. ^'^ Empirical Results The theoretical results derived from the model in the previous section give rise to an interesting hypothesis which can be tested Of course, it goes without saying that the argument is entirely symmetric with respect to the other variable influencing the probability of default, the level of exports. While the whole of the analysis has focused on the absolute value of the variables exports and international reserves, it is evident, from the studies reviewed in Chapter 1, that the importance of these variables is gauged by reference to another variable in the form of ratios. In fact, in the empirical analysis which follows this section, developing countries are divided into two groups, more and less creditworthy, on the basis of two criteria: the ratio of total debt outstanding and disbursed to export proceeds and the ratio of imports to international reserves for each country.

PAGE 91

85 empirically. In particular, it was argued that, if developing countries are divided into two groups (j = 1,2) on the basis of two ratios,''^ then countries with lower values of international reserves or exports (and therefore higher values for either ratio) will be worse off when the cost of funds to the bank rises, in the sense that the relative cost of borrowing, on the loan commitments is higher when contracts are indexed compared to contracts without indexation. Therefore, when the cost of funds to the bank (LIBOR) increases, one would expect the proportion of loan commitments taken down by countries with higher values for either ratio to decl ine. The hypothesis can be tested emp-irically by estimating the model PROP^j = aQ + a^D + a2LBR^ + a3(LBR^)(D) + u^. (10) where PROP^j = proportion of disbursements'"^ through financial markets to group j during time period t D = dummy variable taking the value 1 for the group of countries with a low value for either ratio and 0 for the group of countries with the high value. LBR^ = value of LIBOR during time period t. Estimation of equation (10) is carried out twice: first, countries are divided into two groups according to the ratio of total debt outstanding T5 The ratios are imports to international reserves and total debt outstanding and disbursed to export proceeds. Defined by the World Debt Tables of the World Bank (1981) as drawings on outstanding loan commitments.

PAGE 92

86 and disbursed to exports and second the division is carried out according to the ratio of imports to international reserves. If the theoretical hypothesis of the model is to be verified empirically, then the coefficient of LBR^ (a2) must be negative. As far as the ratio of total debt outstanding and disbursed to exports is concerned, annual data on disbursements were obtained, for a sample of 44 developing countries for the period 1971-1980, from the World 17 18 Debt Tables of the World Bank. The countries were assigned to the high debt-export ratio group if the ratio of their total debt outstanding and disbursed to export .proceeds^ ^ exceeded 1.20 and to the low debtexport group if that ratio was less than 0.80. The data on LIBOR iwere calculated as yearly averages of the end-of-month value of the six-month Eurodollar deposit rate, published by the Financial Times of London. The model in (10) was estimated by ordinary least squares (OLS) for the period 1971-1980 and the regression results were as follows:^^ World Bank Publications EC-167/76, October 1976 and EC-157/81, December 1981 The 44 countries included in the empirical analysis accounted for 97 percent of the total disbursements through financial markets to all developing countries reported in the World Debt Tables during 1971 and for 94 percent during 1980. 19 Data on debt outstanding and disbursed were obtained from the World Debt Tables and data for exports from the International Financial Statistics 1982 Yearbook, published by the International Monetary Fund. The t-statistic ratios are shown in parentheses.

PAGE 93

87 PROP = 0.74 0.53D 1.43LBR, + 1 .36(LBR, ) (D) (10.06) (5.12) (1.85) ^ (1.24) ^ if = 0.93 F(3J6) = 66.43 D-H =1.82 The empirical results support the main hypothesis of the study insofar as the coefficient of LBR^ is negative and significant at the 0.05 level. The proposition that the proportion of loan commitments taken down by countries with low values of exports and high debt-export ratios declines when the cost of funds to the bank (LIBOR) increases, has been shown to hold true for a group of 44 developing countries during the 1971-1980 period. This result contradicts the earlier conclusion of James (1982). who has argued that countries with higher probabilities of default are more likely to be rationed first and that, under loan rate indexation, these countries are expected to increase their proportion of loan commitments taken down, when the cost of funds to the bank rises. Furthermore, although the coefficient of (LBR^)(D) is not significant at the 0.10 level, its sign is opposite to that of LBR^, indicating that the two groups of countries respond differently to changes in LIBOR.As far as the group of countries with the low debt-export ratios (the more creditworthy) is concerned, it was argued in the previous section that when the cost of funds to the bank rises, the relative cost of borrowing remains unchanged for these countries, and is equal to dRj^/dp, Therefore, the proportion of loan commitments taken down by this group is not expected to respond to changes in LIBOR. This assertion is corroborated by the empirical results which show that the proportion of

PAGE 94

88 loan commitments taken down by this group responds only slightly to changes in LIBOR. In conclusion, it has been argued theoretically and demonstrated empirically that the proportion of loan commitments taken down by the group of countries with the higher debt-export ratios (the more creditworthy) does not respond to changes in LIBOR. However, for the less creditworthy group, the relative cost of borrowing on loan commitments is higher when loan contracts are indexed compared to contracts without indexation and, therefore, the proportion of loan commitments taken down by this group has been shown to decline when LIBOR 22 rises. The coefficient measuring the response of the proportion of commitments taken down to changes in LIBOR is -0.07 for the group of countries with low debt-export ratios compared to -1.43 for the group with the high ratios. 22 The choice of cut-off points, according to which the countries are divided into two groups, can be criticized as arbitrary. However, the results do not appear to be sensitive to the choice of cut-off points. The countries were reassigned to the high and low debt-export ratio groups according to whether the ratio of their total debt outstanding and disbursed to export proceeds exceeded 1.50 or was less than 1.00. Equation (10) was reestimated to yield the following results: PROP = 0.55 0.29D 0.58LBR, + 0.93(LBR, ) (D) ^•^ (6.74) (2.55) (0.68) ^ (0.76) ^ R^ = 0.73 F(3,16) 14.20 D-W =: 2.03 • Although the coefficients of LBR and (LBR )(D) are no longer significant at the 0.10 level, the signs of the coefficients are in accordance with prior theoretical expectations.

PAGE 95

89 One further point merits attention as regards the procedure used to estimate (10). Since the amount of total debt outstanding and disbursed in period t includes disbursements in period t as well as disbursements in previous periods, it is likely that the debt-export ratio for each country and the amount of disbursements are correlated; thus, the variable according to which observations are classified and the dependent variable are correlated. If such were the case, it would appear that application of OLS to (10) would yield inconsistent estimates. However, as shown in Appendix B, application of OLS to (10) yields consistent estimates. Nonetheless;if the classification and the dependent variables are correlated, the constant term can no longer be interpreted as the same as that which would result from OLS estimation of (10) in the case where the correlation problem did not exist. Appendix B gives details of how the two constant terms are related. Finally, countries were divided into two groups according to the ratio of imports to international reserves. Data on disbursements were obtained for 43 developing countries^^ for the period 1971-1979.^^ The countries were assigned to the high imports-reserves ratio according to The same data on disbursements as in the previous regression were used, except that Hong Kong was omitted due to the lack of data on international reserves The year 1980 was omitted from the regression because several countries hadunusually high imports-reserves ratios. The average imports-reserves ratio of 10.15 in 1980 was nearly twice as high as the average importsreserves ratio of 5.61 for the previous nine years. In the case of the Ivory Coast and Senegal, during 1980 the ratio showed a 23and a 20-fold increase respectively over the average for the previous nine years.

PAGE 96

90 whether the ratio of their imports to international reserves exceeded 5 and to the low imports-reserves ratio if the value of the ratio was less than 3, The regression results were as follows: PROP, = 0.22 + 0.33D 0.85LBR, + 0.79(LBRJ(D) (1.01 ) (1.07) (0.34) ^ (0.23) ^ 0,67 F(3,16) = 9.36 D-W = 2.61 As can be seen, neither of the coefficients is statistically significant. However, the signs and relative magnitudes of the coefficients support the main theoretical hypotheses advanced in this study. Summary This chapter presented a model to investigate the widespread practice of loan rate indexation in the Eurocurrency market. The comparative statics of the model introduced the intuitively appealing proposition that the interest rate charged on loans to sovereign borrowers is inversely related to the amount of international reserves and the level of exports of each country. In contrast to previous work, it was demonstrated both theoretically and (in the case of exports) empirically that countries with lower international reserves or exports (which are more likely to be the non-prime borrowers) are less likely to be rationed when the cost of funds to the bank rises. Furthermore, it was shown that the proportion of loan commitments taken down by these countries is expected to decline when the cost of funds to the bank rises. Data for both international reserves and imports were obtained from the International Financial Statistics 1982 Yearbook.

PAGE 97

91 The implications of the results for the participants in the international loan markets are all too obvious. According to the conclusions of this study, a developing country which suffers a reduction in its international reserves or exports during a particular year, need not worry overly about its being rationed out of the credit markets. However, a crucial assumption must be met in order for this conclusion to hold true: the increase in the probability of default from the reduction in international reserves or exports must be sufficiently small. As for the international banks, knowledge of the response of developing-country disbursements to changes in the cost of funds to the banks, can better aid them in the optimal allocation of their portfolios. While a great deal has been written about the ability of developing countries to service their external debt from a "macroeconomi c" perspective, very little work has appeared that approaches the question from the perspective of the commercial banks granting loans to developing countries. The present chapter has contributed in the latter direction by extending a model of loan rate indexation, devised originally for the analysis of domestic banking practices, to the international setting of loans to sovereign borrowers in the Eurocurrency market.

PAGE 98

CHAPTER THREE A DISEQUILIBRIUM MODEL OF EUROCURRENCY FINANCE TOD£VEmPING COUNTRIES Introduction The present and following chapters present and estimate an analytic model of a widely discussed, but little researched, facet of the external borrowing of developing countries, that of the setting of credit ceilings or country exposure limits on the amount each country may borrow from commercial banks. Apart from the contributions by Kapur (1977) and Eaton and Gersovitz (1980, 1981a, 1981b) there has not been any systematic analysis of this widespread practice among international banks granting loans to developing countries. The setting of maximum limits on the amount of funds a bank will loan to a developing country has figured prominently in developingcountry lending. It has been described by various terms including country exposure limits, credit ceilings or quantity rationing. Some of the major international banks have well-established procedures for evaluating the creditworthiness of each country and establishing credit cei 1 i ngs. The work of Eaton and Gersovitz (1980, 1981a, 1981b) has been the sole attempt at incorporating credit ceilings in an econometric estimation of the market for international borrowing by LDCs. Their study employed the method of maximum likelihood to estimate a model of the demand and supply of international debt by LDCs. Both demand and supply were formulated as functions of various economic characteristics of each 92

PAGE 99

93 country. The novelty in their work was to assume that the actual amount of debt observed is the minimum of the amount demanded by each country or the quantity supplied (credit ceiling) by the bank. Thus, in their model, the credit ceiling was determined endogenously within the model. Using disequilibrium econometrics techniques, they obtained estimates of the demand and credit ceiling equations and demonstrated that 56 of 81 countries included in the sample were classified as constrained with probability greater than 0.5. A different approach to incorporating credit ceilings is taken in this study. A demand for and supply of international debt by LDCs are postulated, which conform to the usual .'assumptions of economic theory. However, the market is characterized by a ceiling on the maximum amount of credit a country can obtain, which is set exogenously by the banks. The actual quantity of debt observed represents either the equilibrium quantity of debt derived from the intersection of the supply and demand functions in the case where the credit ceiling is not effective, or the amount of the credit ceiling if that should prove to be less than the equilibrium quantity of debt. The following section introduces the theoretical rationale behind actions taken by commercial banks to limit the amount of credit to a particular borrower. The practical implications of credit rationing on the setting of country exposure limits are then discussed. The model to be estimated in this study is outlined briefly in the following section and justification for its application to the market for international debt is provided. The chapter concludes by further elaborating on the model, which is to be estimated in the following chapter, and draws attention to some of the features that distinguish it from other work in this area.

PAGE 100

94 The Rationale Behind Credit Ceilings in International Banking Theoretical Arguments The question of why a rational profit-maximizing lender would wish to ration"* credit to a borrower has received considerable attention in the theoretical literature. The earliest attempts at explaining this phenomenon attributed credit rationing to sticky interest rates resulting from oligopolistic conditions in credit markets and to legal ceilings on interest rates. These explanations viewed credit rationing as a purely temporary phenomenon until the interest rate could ad jut to a new equilibrium rate following a change inborrower demand or actions -taken by the monetary authorities to control the monetary aggregates. The first systematic attempt at providing a theoretical explanation of credit rationing was offered in an early paper by Hodgman (1960) who based his argument for credit rationing on lender attitudes towards risk. He defined the risk of a loan as the ratio of the expected value of the payment to the lender, to the expected value of the payments below the amount originally contracted. The objective of the lender is to maintain the loan-risk ratio above a predetermined ratio. As the amount loaned increases, the loan-risk ratio can be maintained above the predetermined level by increasing the interest rate. However, above a certain loan size, raising the interest rate will not reduce the loan-risk ratio. Credit rationing is defined here as the practice of setting a maximum limit on the funds to be loaned to a specific borrower, beyond which no further funds can be obtained irrespective of the interest rate the borrower offers to pay.

PAGE 101

95 Therefore, given the loan-risk ratio requirement, the lender will not extend credit beyond a certain limit regardless of the interest rate the borrower is willing to pay. Hodgman's analysis was concerned with what has been termed weak credit rationing: the interest rate on the loan varies directly with the amount of credit up to a maximum limit beyond which no further credit is extended. In a further study, Freimer and Gordon (1965) investigated both weak and strict credit rationing. According to the latter the borrower can obtain funds at a fixed interest rate up to a maximum amount beyond which it is impossible to obtain any further funds regardless of the interest rate paid. Their study considered two models of borrowing for investment purposes: fixed-size and open-end investment.^ They demonstrate that in the case of the fixed-size investment, a lender who maximizes the expected profit on the loan will practice weak credit rationing. In the case of the open-end investment, they argue that, for various parameter values of the borrower's utility function and investment opportunity, it is rational for lenders to practice strict credit rationing. A further important study of credit rationing by Jaffee and Modigliani (1959) attributes credit rationing to the lender's inability to discriminate perfectly among borrowers. Given each borrower's characteristics, a lender free to discriminate between borrowers by charging each customer a different rate will not ration credit. However, "... the pressure of legal restrictions and considerations of good will and social mores would make it inadvisable if not impossible ______ Fixed-size investment refers to a specified amount that must be invested whereas, under open-end investment any amount may be invested.

PAGE 102

§6 for the banker to charge widely different rates to different customers" (Jaffee and Modigliani, 1969, p. 860). Therefore, the actual structure of loan interest rates charged would lie within narrower limits than profit maximization would dictate and as a result credit rationinf would occur. The inability to discriminate perfectly among torrowers is crucial to the Jaffee and Modigliani argument for credit rationing. The theoretical arguments for credit rationing developed above hmt been applied to the international banking context. Kapur (1977) lists the important features of the international market for debt, which lead to credit rationing, as the lender's assessment of the probability of default by the borrower and the inability of the lender to perfectly discriminate between borrowers. The assessment of default risk by the lender could lead to credit rationing even under conditions of perfect competition in the loan market and rational economic behavior, in terms of profit maximization and risk aversion, on the part of lenders. The argument simply stated holds that where perfect discrimination is not possible differences in creditworthiness cannot be fully reflected in the interest and noninterest terms of the loan and the allocation of credit will be based on a system of quantity rationing. In general banks diversify their portfolios into various assets with different risks attached. The bank will increase its holdings of marginal assets to the point where the yield on the asset equals the banks' perception of marginal default risk. However, there exists a certain maximum limit beyond which borrowers cannot obtain any further funds regardless of the interest rate they are willing to pay because aft ifttretse in the interest m longer decreases default risk. Such practice is consistent with profit maximization and ensures that the profit/risk ratio is equalized over the bank's entire asset portfolio.

PAGE 103

97 Therefore, in order to equalize the default risk for the marginal borrower banks use credit ceilings. In conclusion "while there may be countries willing pay much higher costs for funds than the norm at that moment, they will be unable to get more funds if the banks' perceptions of their creditworthiness indicate that a certain exposure level (or rationing limit) has already been reached" (Kapur, 1977, p. 180). Credit rationing to developing country borrowers has been investigated to two studies by Sachs (1982) and Sachs and Cohen (1983). Sachs (1982, p. 19) shows that when default is precluded, credit ceilings are fully consistent with perfect competition in the loan market. However, in a two-period model where default can occur, a maximum level of'debt exists for each country beyond which default will occur with certainty. No rational lender would be willing to lend any amount greater than the ceiling regardless of the interest rate. Therefore, "rationing will be a standard device in credit allocation to sovereign borrowers" (Sachs, 1982, p. 8). Similarly, Sachs and Cohen (1982) demonstrate that the presence of default risk in international lending will lead to credit rationing and an upward-sloping supply schedule of loans until a loan ceiling is reached. The most thorough study of credit ceilings in international lending to developing country borrowers has been undertaken by Eaton and Gersovitz (1980, 1981a, 1981b). Their work will be discussed presently. A word concerning the implementation of credit ceilings in international banking practice is called for at this point.

PAGE 104

98 Credit Ceilings in International Banking Practice 3 The imposition of credit ceilings has been a part of international banking practice for a considerable time period. Brackenridge (1978, p. 60), a senior vice president at Morgan Guaranty Trust, suggests that one of the main reasons behind the banks undertaking studies to evaluate the creditworthiness of individual country borrowers, is to establish maximum exposure limits. He explains the difficulties associated with defining country exposure and how each bank defines exposure differently. Despite definitional problems, he reports that country exposure limits were first established at Morgan Guaranty Trust in the late 1960s. In a further article, Brackenridge (1977, pp. 8-11) describes in further detail how Morgan Guaranty sets country limits and discusses which factors are taken into consideration in arriving at maximum exposure limits for each country. These are divided into four categories policy factors, basic economic factors, external finances and politics. He concludes that "after a thorough discussion of all these considerations and with an eagle eye on the country-by-country distribution of exposure, the "maximum country exposure limit" and "limit for exposure over one year" are set by the division senior credit officer" (Brackenridge, 1978, p. 63). On the same theme. Hardy (1979, p. 191) reports that most banks set self-imposed limits on the amount of funds loaned to each developing country. She ascribes such limites to four factors: (i) requirements concerning capital adequacy, (ii) portfolio constraints, (iii) investor attitudes and (iv) regulatory environment. Long (1980, p. 488) argues Or maximum exposure limits as they are known in international banking.

PAGE 105

99 that some developing countries have encountered ceilings on the amount of credit they can obtain from international banks because the amount of developing country paper in banks' portfolios has increased very rapidly in the last few years and banks may be reluctant to increase their holdings any further. In addition, regulatory agencies have been alarmed by the rapid increase in developing country lending and will attempt to impose restrictions to any further increases in developing-country exposure. The importance of regulatory constraints in the setting of credit ceilings has also been discussed by Walter (1981, p. 89) who points to the October 1979 restrictions imposed by the Japanese Ministry of Finance on all foreign-currency lending by Japanese banks, exc^Dt export credits and loans for energy imports. Upon their return to international lending in June 1980, the Japanese banks imposed ceilings on Euroloans and on country exposure, as practiced by American banks. Referring to a survey carried out by the ExportImport Bank (Eximbank), Blask (1978) reports that of a broad cross section of 37 U.S. commercial banks surveyed, 26 banks or 70 percent of the sample use the results from country evaluation studies to set overall country exposure limits. Several of these banks also use the evaluation results to set limits for specific loan maturities and categories. The Eximbank survey results are also discussed by Angel ini et al (1979, pp. 124-5) who describe the country evaluation procedures of three major U.S. banks: Morgan Guaranty Trust, First National Bank of Boston and First National Bank of Chicago. Finally, with respect to country exposure limits, Walter (1981, p. 85) argues that such limits are set on the implicit assumption of zero correlation between different country risks. In other words, exposure limits are set on a country-by-country approach, without

PAGE 106

100 paying attention to the possible correlation of country risk between different borrowers. A Model of Credit Ceilings in International Lending to Developing Countries The Basic Framework The only attempts at an empirical investigation of the market for international debt have been the study by Kapur (1977) and a series of papers by Eaton and Gersovitz (1980, 1981a, 1981b). Kapur (1977) assumes that the loan supply to borrower i is a function of the risk perception of borrower i and the price of the loan. It is further assumed that the risk perception of borrower i is a function of the existing lender exposure to borrower i and a vector of indicators of relative creditworthiness.'^ By arguing that the price of loans does not determine the quantitative allocation of funds between borrowers because of the previouslydiscussed practice of credit ceilings, the price of the loan is omitted from the supply function which is estimated. The major drawback of the study, which Kapur (1977, p. 184) himself recognizes, is that by estimating solely a supply function and omitting the demand side of the market, he has assumed that all observations correspond to the supply function. Such an assumption requires all the developing countries in his sample to exhibit excess demand for loans so that they are all constrained in their international borrowing. Such a state of affairs is clearly inappropriate in modeling the market for international debt. The indicators employed by Kapur (1977) in the estimation of the supply function were the ratio of international reserves to imports of goods and services, the average annual rate of growth of real GNP, exports of goods and services and a projected debt service ratio.

PAGE 107

101 The deficiency in Kapur's work has been remedied in the papers by Eaton and Gersovitz (1980, 1981a, 1981b). On the basis of a theoretical model (Eaton and Gersovitz, 1981a) they demonstrate that where borrowing takes place under conditions of potential default, there exists a certain maximum limit (ceiling) of credit that a borrower can obtain during any particular period. They argue that the credit ceiling is a function of several economic characteristics of each country, which also influence the demand for debt on the part of individual LDC borrowers. The model which has been estimated by Eaton and Gersovitz is as follows:^ D. = min(D^, D.) (13) where is the demand for private debt, D"^. is the supply (credit ceiling) and is the actual debt of borrower i. X^^and X^^are the vectors of arguments of the demand for and supply of debt respectively, g-j and 3^ 5 In their empirical work Eaton and Gersovitz employ the following variables: the percentage variability of exports, the share of imports in GNP, the average growth rate of per capita GNP, total real GNP per capita, total population and the real per capita debt to official creditors. Actually the model estimated and the novelty of their econometric work involves the joint estimation of two separate demand-for~reserves equations with equations (11)-(13). One demand-for-reserves equation is estimated^ in the case where the country is constrained in its borrowing (D. < Dt), the other in the case where it is unconstrained (D* < D.). ^ ^

PAGE 108

102 are vectors of coefficients to be estimated and u-j^, u^^are the error terms. Therefore, according to this model, the actual amount of debt observed is the minimum of the amount demanded by each country or the credit ceiling. The credit ceiling is endogenous to the formulation of the model and the interest rate is excluded as an explanatory variable of o the demand and supply functions. A different formulation of credit ceilings in the international market for debt is proposed in this study. In particular, it is suggested that the following model be estimated: RD. = X.,.6., + a^Q. + u^ (14) RS. = X^.^^ + a^Q. + U2^ (15) R. = RS. = RD. if Qi < Q. (16) R. = RS. if Q. > Q. (17) where RD^. is the interest rate borrower i is willing to pay, RS^. is the interest rate funds are supplied to borrower i and R^. is the actual interest rate to borrower i, is the actual amount of debt of borrower i, Q^. is the credit ceiling to borrower i and a-^, are coefficients to be estimated. The rest of the variables are defined as in (11)-(13).^ ^Eaton and Gersovitz assume that the error terms are independently normally distributed with zero mean, variances and zero covariance "l "2 and that a o = a U-] u g The reasons for the exclusion of the interest rate will be discussed in the following section. 9 The error terms are assumed normally distributed with zero mean and variances a| and a^.

PAGE 109

103 The model differs from the Eaton and Gersovitz formulation in that the credit ceiling does not depend on economic characteristics of each borrower and the interest rate is included as an explanatory variable, both of which will be justified shortly. An explanatory note appears desirable initially. Equation (14) is the demand for debt from each borrower. It is a residual demand after other sources of finance have been taken into account. The next equation represents the supply of private bank funds to each developing-country borrower. It is an aggregate supply function derived by summing the individual supply functions of the major banks active in the international debt market. It depends on several va'riables which are important in determining the creditworthiness of each individual borrower. The credit ceiling, as formulated here, is the sum of the individual exposure limits from the different banks. Strictly speaking, the credit ceiling is not the simple sum of individual bank exposure limits because of the existence of externalities in lending decisions; an increase in the exposure limit of a major bank may act as a signal to other banks to modify their own limits accordingly. Kapur (1977, p. 183) sidesteps this issue and argues that the analysis is not altered by ignoring the difference between market exposure and the summation of individual bank exposure limits for any particular borrower. Since the estimation procedure for model (14)-(17) does not require quantitative knowledge of the credit ceiling or whether each country is constrained or not, this issue does not appear relevant provided one accepts the notion that during any particular period there exists a maximum limit on the amount of funds a developing country can solicit from private banks.

PAGE 110

104 The next question that arises concerns the fact that the credit ceiling is taken as a constant for each country and is not related to the economic characteristics of each country, as in the Eaton and Gersovitz formulation. As has been pointed by Angel ini et al (1979) and Blask (1978), the results from country evaluation studies which are instrumental in deriving the country exposure limits, are not made use of by banks in determining the interest rate to be charged on loans to each borrower i.e., the supply function. Different factors appear relevant in determining the interest rate to be charged to various borrowers and the exposure limits. In reviewing the results from the Eximbank survey, Blask (1978, p. 70) argues that "none bf the banks use the country evaluation results in determining interest rate or fees." Angel ini et al. (1979, p. 130) report that at the First National Bank of Chicago the two tasks are handled by different departments. In setting interest rates, marketing considerations, such as the desire to enhance market position and the need to maintain the existing position in the market, are examined by commercial banks in addition to the creditworthiness of each borrower. The variables that enter the determination of maximum exposure limits were reviewed in a previous section. Therefore, formulating the credit ceiling as exogenous to the market for international debt appears justified. Finally, it may be worthwhile to point out that the formulation in (14)-(17) results in a supply schedule which is upward sloping until the credit ceiling is reached, whereupon it becomes vertical. As discussed in a previous section, this was one of the main results drawn from the theoretical study of developing-country borrowing by Sachs and Cohen (1982).

PAGE 111

105 Elaborating on the Basic Framework Some of the features of the model introduced in the previous section require further explanation. In the first instance, the models estimated by Kapur (1977) and Eaton and Gersovitz (1980) omitted the interest rate as an explanatory variable. In justifying the omission of the interest rate, Eaton and Gersovitz (1980) point out that the base rate (LIBOR) is the same for all borrowers during any particular time period. As for any variation between the two time periods in which the model was estimated,""^ they are captured by year-specific dummy variables. As far as the risk premium is concerned, Eaton and Gersovitz (1980, p. 12) argue that "in competitive markets if lenders are risk neutral or if default risk can be perfectly diversified the expected interest rate r is equal for all borrowers in each year If their assumptions hold then the risk premium is solely a function of borrower characteristics and can be correctly omitted from the supply function. Furthermore, if borrowers' and lenders' expectations coincide, the expected interest rate a lender expects to receive is equal to the rate the borrower expects to pay and thus the interest rate can be correctly omitted from the demand function. Thus the exclusion of the interest rate from the demand and supply functions rests on the validity of their assumptions. The borrower characteristics which Eaton and Gersovitz (1980) consider in their study and on which the interest rate depends, include only economic variables. The inclusion of interest rates as explanatory variables in the present study relies on the argument that economic characteristics are not sufficient in determining the risk premium. Other political The model was estimated for a cross section of 45 LDC borrowers in two years, 1970 and 1974.

PAGE 112

106 and social factors enter the determination of the risk premium which render necessary the inclusion of the interest rate to capture differences in relative perceptions of the political and social stability of each borrower. The argument put forward here is that economic factors do not sufficiently proxy for the risk premium. The importance of political and social factors in determining the relative creditworthiness of each borrower has long been recognized by writers in this field, but has been omitted from empirical investigations of creditworthiness because of the lack of variables to employ as proxies for the borrower's political stability. Kapur (1977, p. 181) states that the indicators which determine a country's creditworthiness can bfe divided into five categories of exclusively economic characteristics. He then goes on to argue that "a factor which could be crucial, but extremely difficult to assess quantitatively, is bankers' views of the underlying political stability of the borrowing country." In his review of the Eximbank survey, Blask (1978) points to the importance of political factors in the country-creditworthiness evaluation process undertaken by commercial banks. The only study which has included an index of political stability in the creditworthiness estimation procedure is that of Euh (1979). In conclusion, the contribution of political and social factors, in addition to economic characteristics, to the determination of the risk premium to be charged on lonas to an LDC borrower, renders the inclusion of interest rates in the model desirable. One further issue that needs to be clarified is the exogeneity of the credit ceiling. Eaton and Gersovitz (1980) argue that the credit ceiling is a function of economic characteristics alone and is thus endoge nous in their model. An argument similar to the one for the risk premium i

PAGE 113

107 advanced here concerning the exogeneity of the credit ceiling. In other words, social and political factors, in addition to economic characteristics play an important part in determining the credit ceiling. The crucial nature of political and social factors has been pointed out by Brackenridge (1977, 1978) in discussing the procedure followed by Morgan Guaranty as regards the setting of maximum exposure limits for each country. According to the model of (14)-(17), the interest rate and quantity of debt observed are either the equilibrium amounts when the credit ceiling is inoperative or correspond to observations on the supply function when the ceiling is effective i.e., it is less than the equilibrium quantity of debt. The latter statement requires an explanation. The justification can be found in the degree of competition in the Eurocurrency markets, where the majority of loans to developing countries are negotiated. It is generally recognized that the Eurocurrency markets are characterized by a high degree of competition.''"' In their theoretical investigation into Eurocurrency lending, Feder and Just (1977b) model commercial banks as monopolistic competitors, whereas in a further paper Feder and Ross (1982) model them as perfect competitors. Under both assumptions, the representative bank is expected to earn no profits above normal. Therefore, if perfect competition is assumed in the market for lending to LDCs, banks are expected to charge an interest rate equal to their normal cost of funds when granting a loan in a specified amount. Thus one would expect to observe the quantity-interest rate combination corresponding to the supply function when the credit ceiling is operational. ^^An explanation for the high degree of competition in the Euromarkets can be found in Dufey and Giddy (1978).

PAGE 114

108 One final point worth mentioning is that the coefficients of the supply and demand functions are assumed to be the same for all the countries in the sample. The same assumption has been made by Eaton and Gersovitz in (11)-(13). The assumption is one of convenience since sufficient time-series data are not available to estimate the demand and supply functions for individual countries over time. Rejection of this assumption would render any estimation of the international market for debt intractable. What may perhaps be useful are tests for the stability of the coefficients across various groups of countries. However, before testing for the stability of the coefficients can be undertaken, the model of (14)-(17) needs to be estimated. The following chapter presents the appropriate estimation procedure.

PAGE 115

CHAPTER FOUR ESTIMATION OF A DISEQUILIBRIUM MODEL OF EUROCURRENCY FINANCE TO DEVELOPING COUNTRIES Introduction The present chapter expounds on the method of estimation of the model introduced in (14)-(17) of the previous chapter. As was previously mentioned, the estimation of the model employs disequilibrium econometrics techniques and relies on an iterative procedure to obtain estimates for the parameters of the model. The empirical applicability of the technique is intimately linked to the success with which the likelihood function, for the sample under consideration, converges to a maximum value. The next section provides a very brief review of the disqui 1 ibrium econometrics literature. The following section explains the estimation procedure for the model under consideration in this study. The choice of variables for the demand and supply functions and the sources of data are discussed next. The final section describes, at the outset, the computer program which was devised specifically for the estimation of model (14)-(17). The chapter concludes by discussing the empirical results obtained from the application of the computer program to the estimation of demand and supply functions for international debt. The disappointing nature of the empirical results is pointed out, insofar as the likelihood function did not converge for any of the different samples or different time periods which were investigated. Finally, as a further check on the applicability of this technique, it was assumed that knowledge of whether a country is constrained or not in its international 109

PAGE 116

110 borrowing from private sources, is available from another previous study. The model was reestimated, incorporating this information, but the results proved equally disappointing. Again, the likelihood function failed to converge for any of the different samples investigated. The usefulness of any econometric technique should utlimately rest on its ability to provide empirical estimates. The particular iterative procedure used in this study was unable to provide estimates for the model under consideration, as convergence of the likelihood function could not be achieved. However, the lack of success in obtaining estimates should not be seen as a failureof the econometric technique but rather of the particular iterative procedure chosen. Although it is quite impossible to know whether a different iterative procedure would perform better, the results of the present study should not discourage the exploration of other iterative procedures which may very well demonstrate the practical applicability of the theoretical technique. Disequilibrium Econometric Techniques: A Brief Methodological Review Beginning with the seminal work by Fair and Jaffee (1972), a substantial body of literature has developed which attempts to estimate supply and demand schedules when the price-quantity combinations observed are not always the equilibrium ones. In their paper, and subsequent work in this area, Fair and Jaffee make a key assumption which has become known as the minimum condition: the actual quantity observed is the smaller of the quantity supplied or the quantity demanded. The work of Fair and Jaffee has been followed by papers by Amemiya (1974), Fair and Kelejian (1974) and Maddala and Nelson (1974). Fair and Kelejian (1974) extended the original estimation method of Fair

PAGE 117

111 and Jaffee for markets in disequilibrium and discussed some maximum likelihood methods for such models. Amemiya (1974) discussed a two-stage least squares procedures which provides consistent estimates and argued that the maximum likelihood estimator of Fair and Kelejian is inappropriate as it maximizes the conditional likelihood function. Finally Maddala and Nelson (1974) provided the appropriate maximum likelihood techniques for models of disequilibrium. Their techniques are particularly well suited to the estimation of such models since they incorporate information contained in the model as to the likelihood of an observation belonging either to the demand or -supply function. All the models discussed so far assume that the nature of disequilibrium is, in a way, temporary and include a price adjustment equation such that the market always moves in the direction of equilibrium. In general, the inclusion of price adjustment equations is justified as, on the one hand, prices rising or falling in response to excess demand or supply or, on the other hand, prices not adjusting fully to the equilibrium levels, because of adjustment costs; thus excess supply or excess demand may exist in the market during any particular period. In contrast to the models of imperfect adjustment of prices, in a number of instances the source of disequilibrium can be traced to government or other regulation of a particular market. In other words, prices may be controlled by an outside authority such as the case with price ceilings, price floors or, in some markets, limits within which the price can fluctuate. In these markets the source of disequilibrium is exogenous to the functioning of the price mechanism. The econometric implications of the imposition of limits on the variation of prices in a particular market have been investigated by Maddala (1983a) where, the maximum likelihood methods of estimation of

PAGE 118

112 markets with price ceilings, price floors or limits on the fluctuation of prices, are discussed. In the case of a market with a price ceiling, the price-quantity combinations observed are assumed to be the equilibrium ones when the price ceiling is ineffective, or represent excess demand and correspond to the supply function when the ceiling is effective.^ Moreover, it is assumed that information exists as to which observations 2 correspond to equilibrium or excess demand. The estimation procedure employed here is an application of the maximum likelihood method suggested by Maddala for a market in which a price ceiling is imposed exogenously. However, in the market for international debt investigated in this study, the ceiling imposed is on the amount of credit which can be obtained during any particular period. Moreover, knowledge of which observations correspond to equilibrium or otherwise does not exist; in other words sample separation is unknown. The next section discusses the estimation procedure in some detail. Estimation of a Disequilibrium Model with Credit Ceiling The model which is to be estimated is repeated here for expositional convenience. Similar assumptions are made concerning price floors and limits on the variation of price. In addition the minimum condition assumption is maintained throughout. In the case of a price ceiling, it is assumed that the excess demand is rationed out and the quantity supplied at the ceiling price is made available for distribution. 2 What has been termed sample separation known.

PAGE 119

113 RD. = X'^j.B-, + a-jQ^ + u^. (14) RS. = X'2^e2 ^ "2^i ^2i ^^^^ R. = RS. = RD. if Q. < Q. (16) R. RS. if Q. > Q. (17) where the various symbols were defined in the previous chapter. The estimation procedure involves maximizing the likelihood function for the model. Therefore, the likelihood function of (14)-(17) is sought. The probability that for observation i the constraint is effective and therefore observation i belongs to the supply function is denoted by Then it must be true that CO r( g(RD,-,f!,)dRD^ p(R.|R. = RS.) = where g(RD.,RS^. ) is the joint density function of RD^. and RS^. Similarly, p(R. IR. = RS. = RD.) i i i i 1 -A where f(R^,Q^) is the joint density function of R^. and Q.^ Therefore, the likelihood function for the model under consideration can thus be written as: n oo L' = n [f(R.,Q.) + / g(RD.,R.)dRD.] i-1 ^ ^ R. Ill where n is the number of observations. The logarithm of the likelihood function is: L = z log[f(R ,Q ) + / g(RD.,R.)dRD.] z logG. i=l ^ ^ R. Til T (18)

PAGE 120

114 The estimates of the parameters which maximize the likelihood function can be found by an iterative procedure which employs the derivatives of the likelihood function. Therefore, it is required to find the derivatives of (18). The method followed is to obtain closed-form expressions for f(R.,Q.) and / g(RD ,R. )dRD. II 111 The first expression to be considered is that of f(R^-,Q^), the joint density of R. and Q^. It can be obtained from the joint density of u-jand U2^where the Jacobian of the transformation is equal to la-i-ct^l. Therefore, fCR^.Q^) = la-,-a2|h(u^ .,U2^-) o where h(u-| ^. ,U2^) is the joint density function of u-j^and U2^. It is assumed that h(u-j^,U2^-) is bivariate normal or 1 -, u,^ 2pu,Up h= 1 exp{-[ ^ ][_L.-^^ + ^]} (19) Z-no-^a^/T^ 2(l-p2) a| ^I'^Z where p = 0^2/(^1^12^ ^'^ correlation coefficient and a-j2 is the covariance between u-j and U2. Therefore, 1 1 ; 2pUt.u„. ul. f(Ri,Qi) = W^-a^\{i ^— )exp{— [-11 + ^]}} Zi^a^a^vT^ 2(l-p2) '^1^2 (20) It goes without saying that ^li = X^'.3^ a^Q. "2i ~~ ^^i ^i^Z 2Qi

PAGE 121

115 The second expression in (18) is now examined. The joint density function of RD^ and RS^, g(RD^. ,RS^. ) is equal to the product of the marginal density of RS^, gi(RS^) and the conditional density of RD^ given RS^. g2(RD^|RS^). Therefore, g(RD.,RS.) = g-,(RS.)g2(RD.lRS.) The conditional density g2(RD^. | RS^. ) is normal with mean M = X-jg^ + a^Q ^ (R X^B^ "2^) (21) "2 and variance J? Therefore R.-M. A / g (RD [RSjdRD. = 1 $(^— ^) where $(•) denotes the cumulative normal distribution. It follows that CO CO / g(RD.,RS.)dRD. = / (RS )g2(RD | RS )dRD R. R. CO = 91 (RS.) / g2(RD.lRS.)dRD. ^• The marginal density g-](RS^) is normal with mean y^2\^2 "z^i ^'^^ variance a 2Finally, CO H = / g(RD.,R.)dRD. R. TIT 1 {1 .r„..„r 1 V. o ^2 R.-M. 1 "1 }{exp[-i-(R. XJ.B2 a2Q^-) ]}{! 'S(^r~^)> /ZtT 20^ "U (23)

PAGE 122

115 From (20) and (23), the derivatives of the logarithm of the likelihood function (18) can be easily obtained and are shown in Appendix C, The iterative procedure which is used to obtain the maximum likelihood estimates is that byBerndtet al (1974). This procedure requires only the first derivatives of the likelihood function. In particular, if Og is the vector of initial estimates of the parameters of interest, the next round of estimates, e-j, is given by -1 Ql = + [Q(6q)] 5(9^) (24) where S(e) is a vector of the first derivatives which are shown in equations (C15)-(C21) of Appendix C and Q(e) is given by n 8logk(y. ,6) 8logk(y.,9) Q(e) .;^[ 36 86 ^ (25) n where logL(e) = z logk(y. ,e) is the logarithm of the likelihood function, i =1 The matrix [Q(e)]""' provides an estimate of the covariance matrix of the maximum likelihood estimates and e is the final (converged) value of e obtained from the Berndt et al iterative procedure. Further details, as regards the computation of the S(e) and Q(e) matrices, are provided in Appendix C. Variables and Data Sources The specification of the demand and supply functions encompasses, to a certain extent, a degree of arbitrariness in the choice of variables. It was argued in Chapter 2 that the demand for international debt is a residual demand after other sources of finance have been taken into consideration. Therefore, the amount of funds available from sources other than international borrowing, is expected to influence negatively

PAGE 123

117 the demand for borrowing; if more funds are forthcoming from alternative sources of finance, a country is expected to reduce its demand for external borrowing. Two of the major sources of foreign exchange for developing countries are exports and capital inflows; both of which are expected to influence the demand for debt negatively. As far as the supply function is concerned, the previous chapter argued that the major factors which influence the banks' decisions, as to the amount of funds to supply to a particular country, are the economic characteristics which determine a country's perceived creditworthiness. Two important economic characteristics that bankers examine when evaluating the creditworthiness of each country are the amount of exports and the level of international reserves, both of which are included in the supply function to be estimated. The importance of both variables in determining creditworthiness was discussed in detail in the introductory chapter. An increase in either of the variables enhances the creditworthiness of a country and thus bankers are willing to loan a higher quantity of funds to the country. Summarizing, the vector in (14) includes the variables exports and capital inflows while X2 in (15) includes the variables exports and international reserves. Thus, both equations (14) and (15) are exactly identified. Exports are defined as the total value of goods and services sold to the rest of the world. International reserves include a country's holdings of Special Drawing Rights (SDRs), its reserve position in the IMF, its holdings of foreign exchange and gold valued at year-end London prices. Data for both variables were obtained from the World Debt Tables

PAGE 124

118 (WDT) of the World Bank.'^ Capital inflows are defined here as long-term flows and include the following: direct investment less direct investment income repatriated, unrequited transfers and net loans from governments and international organizations (official creditors). Data for the first two items were obtained from various issues of the Balance of Payments Yearbook published by the International Monetary Fund, whereas data for the third item were obtained from the WDT. As far as the interest rate and the amount of debt are concerned, data were obtained from the WDT. The interest rate is defined as the average terms of interest on public debt, contracted during a particular year from private creditors. The amount of debt is defined as the total debt outstanding and disbursed. The sample contains all the LDCs included in the WDT for which data for all the variables could be found for each of the three years 1975, 1977 or 1978. All the variables (except the interest rate) were expressed in per capita terms to take account of the different size of the various countries participating in the international lending market. Finally the exogenous variables (those included in the and vectors) were lagged by one year on the assumption that the participants in the international capital markets make decisions at the beginning of the period, based on information available at the end of the previous period. ^World Debt Tables, Publication EC 157/81, Washington, D. C: The International Bank for Reconstruction and Development, December 1981. 5 Data for the population of each country were obtained from various issues of the United Nations Monthly Bulletin of Statistics.

PAGE 125

119 The Maximum Likelihood Estimates The Computer Program Since no computer program exists which will provide maximum likelihood (ML) estimates for the model considered in this study, a special program had to be written for the computation of the ML estimates of (14)-(17). The program was written using the PROC MATRIX procedure of the Statistical Analysis System (SAS) computer package. The iterative technique employed to estimate the model was that suggested by Berndt et al (1974) and described in (24) which makes use of the first derivatives of (18) given in Appendix C. The PROC MATRIX procedure is particularly well suited for handling the Berndt et al method since the values of the first derivatives at each observation can be written in matrix form and then manipulated to yield the Q(e) and S(e) matrices of (24). The Empirical Results The computer program was run for several samples of developing countries for each of the three years 1975, 1977 and 1978. As far as the starting values for the iterative procedure are concerned, the twostage least squares (2SLS) estimates from the estimation of the equilibrium model in (14)-(15) were employed. As has al ready been mentioned in the introduction to this chapter, the empirical results proved quite disappointing insofar as the likelihood function did not converge to a maximum for any of the samples investigated. In some instances, the procedure was run for as many as twenty iterations without convergence ever being achieved. Goldfeld and Quandt (1975) have argued that the likelihood function for disequilibrium models with sample separation unknown may in fact be

PAGE 126

120 unbounded so that successive iterations result in higher values for the likelihood function without ever converging. They suggest that setting the ratio of the error variances in (14) and (15) equal to a known constant may help overcome this problem. Several runs were executed with the ratio of the variances constrained to be equal to one or to the ratio of the 2SLS estimates of the variances. However, in neither of these cases did convergence obtain. A final attempt was made to accomplish convergence of the likelihood function. It has been claimed that where the likelihood function is fairly flat, the step taken at each iteration may be too large so that the values of the likelihood function at ea'ch iteration "jump" around -the maximum without converging to the maximum. It has therefore been suggested that instead of taking a step equal to,[Q(6)]~'' S(9) at each iteration, a step A[Q(e)]~'' S(6) is taken where x is a constant between 0 and 1. This gradual correction method was incorporated in the estimation procedure without any success; the likelihood function showed no signs of converging. Due to the failure of this method to yield empirical estimates of the demand and supply functions, an attempt was made to estimate a different model. In particular it was assumed that prior knowledge was available as to which countries were constrained or not in their international borrowing.^ Information as to which regime each country belongs to has been provided by the study of Eaton and Gersovitz (1980) in which, the probability of each country in their sample being constrained in its international borrowing during 1974, was calculated.'' Therefore countries The existence of knowledge as to which regime each country belongs to has been termed sample separation known. ^For Togo, the probability of being constrained was calculated for 1970.

PAGE 127

121 in the Eaton and Gersovitz sample constrained with probability greater than 0,5 were assigned to the constrained regime whereas the others to the unconstrained regime. The likelihood function for the model with sample separation known and its first derivatives are shown in Appendix C. The results from the model with sample separation known appear somewhat more encouraging than those of sample separation unknown. Convergence of the likelihood function was obtained for the two years Q 1976 and 1978 for a sample of 37 LDCs with a convergence accuracy of 10 The 2SLS estimates of the equilibrium model of supply and demand 9 for international debt were as follows: 2SLS Estimates for 1976 RD. = 8.046 0.993(X/N). + 3. 645(NKF/N) + 1.205(Q/N). ^ (0.100) (1.161) ^ (3.026) ^ (1.748) ^ RS. = 8.054 4.780(X/N). + 3.329(RES/N) + 6.774(Q/N). (0.168) (5.832) ^ (4.683) ^ (8.008) ^ 2SLS Estimates for 1978 RD. = 9.081 1 .695(X/N). + 8.982(NKF/N.) + 3.471 (Q/N). ^ (0.338) (1.109) ^ (6.820) ^ (1.631) ^ RS. = 8.567 16.088(X/N). + 1 1 .964(RES/N) + 21 .557(Q/N). (1.297) (24.014) ^ (19.714) ^ (31.236) ^ where (X/N). is per capita exports, (NKF/N)^. is per capita capital inflows, (RES/N)^. per capita reserves for country i and the other variables were previously defined. The 2SLS estimates were used as starting values for g Convergence accuracy refers to the maximum absolute value of the difference between two successive rounds of estimates. 9 The standard errors are shown in parentheses below the estimates.

PAGE 128

122 the iterative procedure. The ML results obtained were: ML Estimates for 1976 RD. = 7.682 0.067(X/N). + 1 1 706(NKF/N) 0.376(Q/N). ^ (0.062) (0.355) ^ (1.462) ^ (0.450) ^ RS. = 8.231 1.518(X/N). + 2.898(RES/N) + 2.155(Q/N). ^ (0.034) (0.220) ^ (0.152) ^ (0.346) ^ Value of the logarithm of the likelihood function = 521.1 Convergence obtained after 18 iterations. ML Estimats for 1978 RD. = 8.720 3.317(X/N). + 8.851 (NKF/N) + 8.570(Q/N). ^ (0.132) (0.216) ^ (1.380) ^ (0.829) ^ RS. = 10.352 5.225(X/N). + 1 .460(RES/N) + 7.679(Q/N). (0.014) (0.075) ^ (0.135) ^ (0.062) ^ Value of the logarithm of the likelihood function = 828.5 Convergence obtained after 9 iterations. As can be seen the signs of the coefficients of the ML estimates are the same as those of the 2SLS except for the ML estimate of (Q/N)^. in the demand function for 1976, the sign of which is the correct a priori negative sign in contrast to the positive sign of the 2SLS estimate. However, both the 2SLS and ML estimates for 1978 reveal a consistently positive sign for the coefficient of {Q/H). in both the demand and supply functions indicating that for that period the demand and supply functions have not been properly identified. ""^The numbers in parentheses below the estimates of the coefficients are the standard errors. The ML estimate divided by its standard error is asymptotically t-distributed. An estimate of the covariance matrix of the ML estimates is obtained from [Q(6)]~' where d is the converged values of the estimates and Q(e) was defined in (24).

PAGE 129

123 The sign of the coefficient of (X/N)^ in both the demand and supply functions is consistently negative for both the 2SLS and ML estimates which runs counter to prior expectations. It appears that the choice of exports as a determinant of the supply and demand functions has proven a poor one. The signs of the coefficients of the variables (RES/N)^. and (NKF/N)^ are in accordance with prior expectations for both the 2SLS and ML estimates but although the 2SLS estimates appear insignificant the ML estimates are highly significant. Finally, a note of caution appears in order at this point. The ML results must be interpreted, at best, as indicative of the demand and supply functions for international debt. The cutoff level for the convergence accuracy of 10"'' is very crude. Although there are no widely accepted standards as regards the level of accuracy of the convergence criterion, it is clear that 10""* does not represent an acceptable level. When a more stringent accuracy level was imposed, the likelihood function failed to converge. Therefore, the interpretation of the ML estimates as representing a global maximum of the likelihood function, instead of simply a local one, must be questioned. Conclusion This chapter presented the estimation method and empirical results from a study of a market in disequilibrium in which an explicit quantity constraint has been introduced. While the estimation technique is theoretically appealing, the empirical applicability of the technique has been questioned. The likelihood function failed to converge despite various attempts with different samples and time periods. The failure of the estimation method should not be construed as evidence for rejecting the usefulness of the theoretical technique, but

PAGE 130

124 rather as the failure of the particular iterative technique employed in this study. It could very well be that other iterative techniques, such as the Newton-Raphson or the quadratic hill-climbing methods both of which make use of the second derivatives of the likelihood function, may provide greater convergence accuracy."''' Further empirical work in this area is highly desirable before the practical applicability of this econometric technique can be ascertained. Eaton and Gersovitz (1980) employed the GRADX option of the GQOPT program at Princeton University, which is based on the quadratic-hill climbing method and obtained a convergence accuracy of 10-7.

PAGE 131

CHAPTER FIVE A SIMULTANEOUS LOGIT MODEL FOR THE DETERMINATION OF DEVELOPING-COUNTRY CREDITWORTHINESS AND CAPITAL INFLOWS Introduction The introductory chapter reviewed a number of studies whose focus centered around the ability of a developing country to service its external debt. Mathematical techniques such as discriminant analysis logit analysis and principal components have been employed to analyze the factors which determine a country's creditworthiness. These studies have identified a number of economic indicators, in the form of ratios, which determine a country's debt servicing capacity. Moreover the application of the technique of logit analysis has rendered feasible the calculation of the probability of a country encountering debt service difficulties during a particular time period. The present chapter focuses on one of the economic indicators these studies have identified as an important determinant of the debt capacity of each country, namely, the amount of capital inflows. The central theme of this chapter is that the amount of capital inflows cannot be taken as an exogenous indicator of the creditworthiness of each country but is in fact determined simultaneously with creditworthiness. A decrease in the amount of capital inflows may signal the deterioration of a country's creditworthiness but, at the same time, it may be the result of information that a country may soon seek to have its debt rescheduled. Therefore, the amount of capital inflows and the creditworthiness of a country are endogenously determined. A simultaneous 125

PAGE 132

126 equations formulation suggests itself as the appropriate estimation procedure. The remainder of this chapter introduces and estimates such a model The chapter is organized as follows. The next section describes a simple equilibrium model of the demand and supply for external capital which demonstrates that the equilibrium amount of capital inflows is a function of the probability of default. The causation, between the probability of default and creditworthiness, runs in the opposite direction to that traditionally assumed by the studies of creditworthiness reviewed in chapter one. Before.incorporating the joint determination of capital inflows and the probability of default into a simultaneous equations model, the importance of two variables, previously ignored by empirical studies, is examined. As discussed in the introductory chapter, although the investment rate and the rate of growth of the money supply have been identified as important determinants of a country's creditworthiness, they have received almost no mention in empirical studies of creditworthiness. Thus, the following section examines the empirical significance of these variables. The results from the simultaneous model of the determination of creditworthiness and capital inflows are discussed in the next section. The final section presents some concluding remarks on the general applicability of empirical models of creditworthiness. A Simple Equilibrium Model of the Supply and Demand for External Capital The role of capital inflows as an indicator of the capacity of developing countries to service their external debt, and therefore their creditworthiness, has received a considerable amount of attention in the literature. As Irvine et al (1970). have pointed out, capital inflows,

PAGE 133

127 in the form of loans, grants, direct investments and transfer payments, represent a substantial source of foreign exchange for many developing countries v/hich can be used to meet external debt obligations. Thus a larger amount of capital inflows should enhance the creditworthiness of a country and lower the probability of encountering debt service difficulties and requesting a rescheduling of its debt. The empirical significance of capital inflov.'s as an indicator of creditworthiness has been am.ply demonstrated in two studies by Feder and Just (1977a) and Feder, Just and Ross (1981).'' The sample for the first study com.prised 2-33 observations from 41 LDCs for the period 1965-1972 and demonstrated a negative -and significant impact exerted by the ratio capital inflows/debt service payments on a country's creditworthiness. The study by Feder, Justand Ross extended the results to 2 a mucn larger sample. Insofar as the capital inflows variable is concerned the study demonstrated that when the total amount of capital inflows was divided into commercial and noncommercial inflows, both variables (in relation to debt service payments) displayed a negative and significant impact on the perceived creditworthiness of each country. The empirical significance of capital inflows as a creditworthiness indicator appears v/ell founded. ^Both studies employed the logit analysis techniques according to which ~ l+exp(6'X) (26) where the probability of default P(-) is related to a vector of economic indicators X (which includes the amount of capital inflows) and 3' is a vector of fixed coefficients. 2 The study included some 580 observations from 56 countries covering the period 1965-1976.

PAGE 134

128 The main hypothesis advanced in this chapter is that the probability of default determines, in turn, the amount of capital inflows into a country, so that the amount of capital inflows cannot be taken as an exogenous indicator of the probability of default. To demonstrate this proposition assume a simple demand and supply model of external capital as follows: K= K-(r.) (27) K= K-(MPK., r., Y.) (28) s D where K. and K^. are the amounts of external capital supplied andidemanded by country i, respectively, r. is the interest rate, MPK. is the marginal product of capital and is per capita income in country i. The usual assumptions, concerning the derivatives with respect to the interest rate, are put forward in this model. Furthermore, it is assumed that the higher the marginal product of capital and per capita income the 3 greater the demand for external capital. The supply function (27) is examined at the outset. It is assumed that the yield from supplying one unit of capital to country i is (1+r^.), if no default occurs, or a fraction X^./K^. if default occurs. Therefore the expected return from advancing capital, in the amount is given by X. E(V) = (l+r.)(l-Tr.)K. + K. TT. = (l+r.)(l-^.)K. + X. TT. 3 The assumption concerning the derivative of the demand for capital with respect to per capita income will be further explored in a later section.

PAGE 135

129 where tt^. is the probability of default of country i. Furthermore, it is assumed that the return in the case of default, X^, depends on the rate of depreciation of the nominal exchange rate for country i (e^) such that a higher rateof depreciation results in a smaller return in the case of default. Assuming risk neutrality, the equilibrium amount of capital advanced must equal its expected return or Kf = (l+r.)(l-Tr.)K^ + ^.X. (e.) (29) where K| is the equilibrium amount of external capital. The solution of (29) for r^. yields the following: Tr.X.(e.) r. = _! LJ_J 1 (30) (l-^.)K^ It can be easily verified that the derivatives of r^ with respect to both ir^. and e^ are positive. Attention turns next to the demand function. Equilibrium in the market for external capital implies that K* = (MPK., r., Y.) and from (30) Equation (31) can be solved implicitly to yield the equilibrium amount of capital inflows as a function of the exogenous variables: K| = g(MPK., fe., Y.) (32)

PAGE 136

130 Given the assumptions of the model, it is trivial to observe that the derivatives of K| with respect to MPK^ and Y^. are positive, whereas those with respect to tt^ and e^ are negative. Equation (32) is a reduced-form equation for the determination of the amount of capital inflows. In conjunction with an equation of the form (26), they comprise a simultaneous equations system which is estimated in a subsequent section. The simple equilibrium model demonstrates theoretically the intuitively plausible proposition that information regarding the likelihood of default, on the part of a country, will naturally influence the amount of capital inflows into a country. The simultaneous equations model proposed here meets the criticism, raised by Eaton and Gersovitz (1981b, p. 29), concerning the distinction between exogenous country characteristics which determine a country's creditworthiness and endogenous variables determined simultaneously with the probability of default. As they correctly argue, the amount of capital inflows is clearly an endogenous variable. Finally, while stressing the importance of capital inflows in determining a country's creditworthiness, Irvine et al (1970) are quick to point to the importance of information regarding a country's intention to reschedule on the amount of capital inflows. With reference to the case of India, Irvine et al (1970, p. 46) state that "... debt reschedulings have already been arranged and more are in prospect. This does not create a favorable climate for the inflow of private capital." However, before examining the implications of the simultaneous determination between the amount of capital inflows and the likelihood of default, a brief discussion of two economic variables, which have previously been neglected by empirical studies of logit analysis, follows.

PAGE 137

131 The Importance of Economic Indicators 1n Single-Equation Models of Creditworthiness This section focuses on two economic indicators which have been identified by several studies as Important determinants of a country's creditworthiness and hence the likelihood of its seeking to reschedule its external debt obligations. The two variables are, on the one hand, the amount of savings or Investment undertaken by each country and, on the other, a measure of the rate of growth of the money supply or prices. While the importance of these variables has been emphasized by several writers in this field, they have, on the whole, been neglected by empirical studies. The aim of this section is to demonstrate the empirical significance of these economic indicators of creditworthiness. The role of savings and investment in the rescheduling process of developing country debt has been pointed out by Sachs (1981, 1982). He has argued both from a theoretical perspective and with statistical data that the savings rate is an Important determinant of the likelihood 4 of rescheduling. Higher savings and investment rates imply that the country is directing the borrowed funds into productive uses which should increase the level of income and consumption over time and thus reduce the likelihood of the country requesting a rescheduling of its external debt. His statistical evidence showed that five of six countries which rescheduled their debt in the 1970s, suffered significant decreases in their savings rates in the period prior to the rescheduling.^ Sachs' 4 Sachs (1981, p. 233) defined the savings rate as gross national product (GhP,) plus transfers from abroad minus private and public consumption expenditure as a percentage of GNP. 5 The five countries were Chile, Gabon, Peru, Sierre Leone and Zaire. The sole exception appears to have been Turkey.

PAGE 138

132 regressions results, demonstrating the importance of the investment rate in determining the perceived probability of rescheduling and therefore the risk premium on loans, were discussed in some detail in the introductory chapter. The relationship between the process of capital accumulation and debt accumulation has been examined from a theoretical standpoint by Kharas (1981). He derived the dynamic path of adjustment of the capital stock and the level of debt over time. Based on his dynamic analysis, a country was defined as creditworthy if it is on the path that leads to an ever-expanding capital stock. He then introduced a latent variable,^ creditworthiness, which he assumed to be a function of several economic characteristics.'' The introduction of a latent variable made possible the use of the method of probit analysis to estimate a probability of rescheduling equation. His empirical results demonstrated the significance of among other variables the level of investment; the ratio of fixed capital formation to GDP exerted a negative and highly significant impact on the rescheduling probability. Finally, the ratio of gross fixed capital formation to GDP was also found significant in the model of developing-country creditworthiness which Mayo and Barrett (1978) devised for the Export-Import Bank (Eximbank). This is the only empirically-based study which has validated the significance of the investment rate as an indicator of creditworthiness. The interpretation of latent variables in creditworthiness analysis will be examined presently. ''one of these, investment, was defined as gross fixed capital formation as a percentage of gross domestic product (GDP). The other indicators found to exert a significant impact on creditworthiness were the ratio of net capital flows to GDP, debt service payments to GDP and per capita GDP.

PAGE 139

133 The role of monetary variables in the rescheduling process has been pointed out by Sargen (1977). At the conceptual level, Sargen has argued that debt servicing difficulties arise as a result of internal economic mismanagement: excessive growth of the money supply and the resultant inflation along with the maintenance of overvalued exchange rates cause export demand to fall and import demand to rise and thus lead to debt servicing difficulties. Therefore, the monetary approach to debt rescheduling identifies the rate of growth of the money supply and prices as important indicators of creditworthiness. Sargen employed both the rate of growth of the money supply and consumer prices in his discriminant analysis study. Again, the only empirically-based study, which has incorporated indicators identified by the monetary approach to debt rescheduling, has been the study by Mayo and Barrett (1978) for Eximbank. They found that the percentage change in the consumer price index exerted a positive and significant impact on the probability of rescheduling. In order to test the empirical significance of the above two indicators, it was decided to estimate two probability-of-rescheduling equations: one employing the economic indicators suggested by Feder and Just (1977a) and another including, in addition to the variables of the Feder/Just study the two variables gross fixed capital formation and the rate of growth of the money supply. The variables identified by Feder and Just as indicators of the probability of rescheduling were the debt service ratio, the ratios imports/reserves, amortization/debt, capital inflows/debt service payments and the variables per capita income, export

PAGE 140

134 growth and growth of GDP. As Feder and Just explain the amortization of debt ratio was excluded from the final results because, in the first place, there are doubts about the direction of causality between the ratio and the probability of rescheduling and, second, because of measurement problems. This ratio was, accordingly, omitted from the present study. The definition and sources of data for the remaining variables are described in Appendix D. In accordance with the Feder/Just study, the estimation procedure employed is that of logit analysis. This method is particularly suited to modeling the rescheduling decision because, as Feder and Just (1977a, p. 26) point out "logit analysis assumes a discrete event takes place after the combined effect of certain economic variables reaches some threshold level" whereas the use of discriminant analysis implies that a rescheduling country suddenly becomes a "member of another species." Logit analysis assumes the existence of an underlying latent 9 variable y* which can be written as: y* = Xe + u (33) The ratio capital inflows/debt service payments is an inappropriate indicator of the probability of rescheduling, insofar as the direction of its impact depends on the sign of capital inflows: if capital inflows are positive, a positive coefficient is expected and vice versa if capital inflows are negative. However, since only 2 percent of the data refer to negative capital inflows the distinction does not appear relevant. The ratio capital inflows/debt service payments is maintained in this study for purposes of comparison with the Feder/Just study. 9 In the model considered here, y* can be thought of as a measure of the intention to reschedule or an inverse measure of creditworthiness: the higher y* the lower a country's creditworthiness.

PAGE 141

135 However, what is observed is not y* but y which takes the value 1 if default occurs and zero otherwise: y 1 if y* > 0 = 0 otherwise Therefore, the probability of rescheduling can be written as Pr(y=l) = Pr(u > -X3) (34) If u is assumed to follow the logistic distribution the probabilities in (34) can be calculated given estimates of the vector of coefficients 3. Estimates of the coefficients can-' be obtained by maximizing the appropriate likelihood function for the logit model. ""^ The method of logit analysis was applied to data for a sample of 43 LDCs covering the period 1965-1976. The logit estimation results made use of the PROC LOGIST procedure of the Statistical Analysis System computer package. Some 442 observations were included, 30 of which referred to instances of rescheduling. The countries, the number of years included for each, as well as the rescheduling years are listed in Appendix D. The logit estimates for the indicators included in the Feder/Just model are shown in Model (a) of Table 8. As can be seen, all the variables of Model (a) bear the correct a priori sign and are highly significant.''"' Thus, the empirical significance of the economic indicators identified by Feder and Just is confirmed in this study. ^'^For a discussion of the likelihood function for the logit model see Maddala (1983). ^Vhe variable growth of exports has been omitted because, as Feder and Just have argued, it is strongly correlated with the growth of GDP. In fact the correlation coefficient between the two variables is .43 as Table D2 of Appendix D reveals. When the growth of exports was included as an explanatory variable, an intuitively inplausible positive sign was obtained.

PAGE 142

136 Table 8 Logit Estimates for the Single-Equation Model^ Model Variabl e (a) (b) Intercept -4.6328 -3.7380 (1 .2841 } (1 .6378) Debt Service Ratio 25.8680 26.4800 (5.0453) (5.4642) Imports/Reserves •• 0.1879 0.1793 ^ (0.0714) (0.0826) Income/Capita -0.0071 0.0074 (0.0017) (0.0018) Capital Inflows/ Debt -0.8288 -0.6440 Service Payments (0.3795) (0.3846) GDP Growth -46.5436 33.8489 (16.6417) (18.4492) Gross Fixed Capital -12.2927 Formation/GDP (9.4462) Money Supply Growth 1.5859 (0.9241) Likelihood Ratio Statistic*^ 146.07 153.49 Figures in parentheses indicate standard errors; maximum likelihood estimators divided by the corresponding standard errors are asymptotically t-distributed. The likelihood ratio statistic is (asymptotically) chi-square distributed with k-1 degrees of freedom, where k is the number of parameters (including the intercept) estimated.

PAGE 143

137 In order to examine the significance of the investment and money growth variables, they were included in the logit model, in addition to the variables of Model (a). Since all other indicators are included in the form of either ratios or growth rate, investment is introduced into the model as the ratio of gross fixed capital formation to GDP, to 1 2 facilitate comparisons across countries. The results from the inclusion of these two indicators are shown in Model (b) of Table 8. As can be seen the signs of both the investment and money growth variables are in accordance with prior expectations and both coefficients are significant at the 0.10 level. In order to gauge an appreciation for the relative performance of the two models, one may calculate the implied probabilities of debt rescheduling using the coefficients of Models (a) and (b). It is assumed that for each of the observations, the predicted probability is known a priori whereas it is not known whether rescheduling will occur or not. Therefore, ex post given a critical probability P*, all countries with predicted probabilities higher than P* are classified as "rescheduling", whereas those with predicted probabilities below P* are classified as "nonreschedul i ng" For any given P*, two possible types of error can be distinguished. A Type I error occurs when, for a particular year, a country is classified as nonreschedul i ng when a rescheduling did take place. Conversely, a Type II error occurs when a country is classified as rescheduling, when a rescheduling did not actually take place. The number of Type I and Type II errors, for both models (a) and (b), are shown in Table 9. The definition and sources for both the investment and money growth variables are described in Appendix D.

PAGE 144

138 Table 9 Type I and Type II Errors for Models (a) and (b) Model (a) Model (b) p* Number of Type I Errors Number of Type II Errors Number of lype i trrors Number of Type II Erorrs 0.10 2 31 I 25 0.20 4 14 A 13 0.30 6 13 7 7 0.40 9 7 8 6 0.50 10 5 9 6 0.60 14 4 n 4 0.70 17 2 14 3 0.80 18 2 16 2 0.90 21 1 18 0

PAGE 145

139 As can be seen, in the case of Model (a), the minimum amount of both Type I and Type II errors is 15 and occurs with a critical probability level of 0.50. As for Model (b), the minimum total number of errors is 14 and occurs with either a critical probability level of 0.30 or 0.40. Moreover for eight of the nine critical probability levels. Model (b) performs better than Model (a) in terms of a smaller total number of 13 errors. Comparing the results of Model (b) with those of the Feder/ Just study, the Type I error rate ranges from 0.2 to 4.1 percent here, in contrast to error rates in the range of 0 to 2.9 percent in the Feder/Just model. As for Type II errors, the error rate ranges from 0 to 5.6 percent here and from 0.4 to 4.2 percent in the Feder/Just model. In conclusion, the error rates compare favorably to the Feder and Just (1977a) model. In addition, on the basis of the results of Table 9, Model (b) appears to fare better than Model (a), in terms of fewer Type I and Type II errors. The significance of the investment rate and money growth as indicators of creditworthiness, has thus been confirmed. A Simultaneous Logit Model of Creditworthiness and Capital Inflows The simple theoretical model of a previous section pointed to the simultaneous determination between the probability of default and the amount of capital inflows. However, as Nelson and Olson (1978) and 13 In the remaining case of the critical probability level of 0.50, the total number of errors of 15 is the same for both Models (a) and (b). Of course, in real world applications, the choice of the critical probability level will depend on the relative costs associated with Type I and Type II errors.

PAGE 146

140 Maddala (1980) have pointed out, in applying econometric techniques to estimate such models, it may make more economic sense to assume that, the economic indicators determine a latent variable^'^ as in (33) and that this latent variable is one of the determinants of the amount of capital inflows. The above considerations lead to the formulation of the simultaneous logit model as follows: y* = 3-1272 + y-i^i + Li^ (35) ^2 ^21^1 ^2^2 "2 where y| is the latent variable intention to reschedule which is unobserved; instead the variable y-| is observed such that =1 if yf > 0 y-| = otherwise 1 5 Moreover, y2 is the amount of capital inflows, x-j is a vector of exogenous 1 4 The perceived creditworthiness of each country in the model here. Actually the variable yf in (35)-(36) represents an inverse indicator of creditworthiness or the intention to reschedule. 1 5 In crediworthi ness analysis of the previous section, the economic indicators were expressed as ratios and therefore y2 is expressed as the ratio of capital inflows to debt service payments. When it comes to considering (36) where y2 is the dependent variable, if the exogenous variables in (36) are independent of, or affect debt service payments in the opposite manner to capital inflows, the results follow through when the ratio is used insteadof the actual amount of capital inflows. As will be discussed, the variables of vector X2 per capita income and growth of exports, are not expected to influence debt service payments, whereas exchange rate depreciation is expected to influence debt service payments in the opposite manner to capital inflows.

PAGE 147

141 economic indicators influencing creditworthiness, x,^ is a vector of exogenous variables determining the amount of capital inflows into a country, and are vectors of coefficients and g^^ g^-] a^e coefficients to be estimatedj^ The simultaneous model, as formulated in (35)-(36), states that creditworthiness (or the intention to reschedule) at the beginning of period t is determined by the expected amount of capital inflows during period t as well as a number of other exogenous economic indicators. In accordance with the Feder and Just (1977a) study, the expected amount of capital inflows in period t is proxied by the actual amount in period t-1 The expected amount o"f capital inflows during pefiod t is, in turn, influenced by the perceived creditworthiness at the beginning of period t, as well as a number of other exogenous variables. The appropriate estimation procedure for model (35)-(36), pointed out by Maddala (1980, p. 7-8), is the two-stage Nelson and Olson (1978) method. The procedure is as follows: (i) Estimate the reduced forms by ordinary least squares (OLS) or logit for completely observed and dichotomously observed variables respectively. (ii) Substitute the estimated values of the endogenous variables on the right-hand side of each of the structural equations (35) and (36). (iii) Estimate by OLS or logit the structural equations depending on whether the respective endogenous variable is completely or partially observed. The two-stage limited-dependent variable (2SLD\/) described will provide consistent estimates of the parameters of the structural equations. T6 All the variables refer to a particular time period t; the subscript t is omitted for simplicity.

PAGE 148

142 However, as Maddala (1980, p. 7) notes, the estimates of the coefficients from the 2SLDV do not represent actual estimates of the structural equations coefficients 3^2> ^21' "^1 ^Z' ^'^^^^ coefficients are estimable only up to a multiple. Since the focus of this study centers around the sign of the coefficients rather than their magnitude, this complication does not appear to pose a significant obstacle. The exogenous indicators determining creditworthiness were discussed in the previous section. As for the exogenous variables determining the amount of capital inflows, the theoretical model of a previous section identified three possible variables in equation (32): per capita income, the marginal product of capital and the rate 6f depreciation of the domestic currency. The importance of per capita income as a determinant of the amount of capital inflows has been pointed out in an early paper by Johnson (1955, p. 550), who has argued that per capita income may be used as an indicator of the level of economic development. Therefore, one would expect more advanced countries to attract higher private capital flows but lower official flows. Thus, the relationship between per capita income and capital inflows is ambiguous. Johnson speculated that, on the whole, one would expect to find a positive relation between per capita income and the amount of capital inflows. As for the rate of depreciation, Rhomberg (1966, p. 2) has argued that repeated devaluations or a continuously depreciating currency would reduce earnings as foreign investments in the country, create uncertainty and undermine confidence, thus discouraging capital inflows and encouraging capital outflov;s. Finally, Leff (1975) has argued that there may exist a positive relation between the marginal return on investment and the rate of growth of the

PAGE 149

143 economy. In the model here, the marginal product of capital is proxied by the growth rate of exports. The definition and sources of data for these variables are discussed in Appendix D. The results of the estimation of (35)-(36) using the 2SLDV method are as follows:''^ Model 1 (2SLDV) RESC = -17.4058 + 50.0041DSR + 1 .9582CAPF 46.1085 GGDP (6.7647) (20.7838) (1.1513) (16.0484) 0.0075 INC + 0.0435RESR + 2.1640MS (0.0017) (0.0769) (0.9322) Likelihood Ratio Statistic = 150.66 • CAPF = 3.8997 1.8237DEPR 2.9097RESC + 4.9533GEXP 0.0013INC (0.2826) (0.8930) (0.8810) (1.7615) (0.0005) R^ = 0.90 F(4,437) = 10.33 where RESC is the intention to reschedule, DSR the debt service ratio, CAPF the ratio of capital inflows to debt service payments, GGDP the growth rate of GDP, INC per capita income, RESR the ratio of imports to reserves, MS the rate of growth of the money supply, DEPR the rate of depreciation of the domestic currency and GEXP the rate of growth of exports. It must be noted, at this point, that the standard errors and other statistics produced by the application of the computer package to the second stage, are not the appropriate ones since the endogenous ^'^The investment variable was omitted from the creditworthiness equation because, when it was included, the second stage of the estimation resulted in the coefficients of several of the variables in the creditworthiness equation tending towards infinity. This could have been the outcome of the high correlation between investment and the variables growth of GDP per capita income and growth of exports, as shown in Table D2 of Appendix D.

PAGE 150

144 variables on the right-hand side have been replaced by their estimated values. Nonetheless, they provide reasonable approximations of the appropriate statistics. The coefficient of the imports/reserves ratio in Model 1 proved insignificant. It was therefore omitted and the model reestimated. The 2SLDV results were as follows: Model 2 (2SLDV) RESC = -18.3624 + 63.2993DSR + 2.1849CAPF 49.6471GGDP (6.5754) (20.2438) (1.1020) (15.5984) -0.0075INC + 2.2374MS (0.0017) (-.9239) < Likelihood Ratio Statistic = 149.89 CAPF = 3.8996 1.8240DEPR 2.9088RESC + 4.9536GEXP 0.0014INC (0.2826) (0.8933) (0.8831) (1.7616) (0.0005) R" = 0.09 F(4,437) = 10.31 As can be seen, the results of the two models are quite similar with the t-ratio values for Model (2) being somewhat higher. An interesting comparison to be undertaken, is between the results of (35)-(36) where the simultaneous determination between creditworthiness and capital inflows has been taken account of, and the single-equation (SE) methods of estimation of (35) and (36) where the simultaneity has not been considered. The SE results corresponding to Models 1 and 2 are shown as Models 1* and 2* below:

PAGE 151

145 Model 1*(SE) RESC = -4.9469 + 25.1258DSR 0.6901CAPF 44.1939GDP (1.3806) (5.1294) (0.3993) (17.4480) -0.0074INC + 0.1513RESR + 1.8392MS (0.0018) (0.0819) (0.8873) Likelihood Ratio Statistic 151.63 CAPF = 3.8228 2.1506DEPR 2.0634RESC + 5.1717GEXP 0.0013INC (0.2786) (0.8685) (0.6840) (1.7591) (0.0005) R^ = 0.08 F(4,437) 9.84 Model 2*(SE) RESC -4.2483 + 24.1-938DSR 0.5754CAPF 43.1298GGDP (1.2988) (4.8791) (0.4052) (17.6499) -0.0075INC + 2.1392MS (0.0017) (0.8476) CAPF = 3.8228 2.1506DEPR 2.0634RESC + 5.1717GEXP 0.0013INC (0.2786) (0.8685) (0.6840) (1.7591) (0.0005) R^ = 0.08 F(4,437) = 9.84 Comparisons of the SE and 2SLDV methods will be undertaken at two levels: first, with regard to the Type I and Type II errors implied by the different models and second, the economic implications of the two procedures will be discussed. The comparison of error rates is tackled first. The number of Type I and Type II errors for the 2SLDV Models 1 and 2 and the SE Models 1* and 2* are shown in Table 10. As can be seen, for only one of 18 critical probability levels does the two-stage simultaneous procedure result in a higher total number of errors than 1 8 the single-equation procedure. The number of errors generated by the 2SLDV is, in general, lower than that of the single-equation method. Moreover, the lowest number of errors (12) is generated by the This is for the comparison of Model 2 and 2* with a critical probability level of 0.10.

PAGE 152

146 4O 1 — t 1 — 1 CO dJ cu o > CL SQ E 1 1 — LU 00 zr CM — CNJ 4— O 1 — 1 CO CU S— "O cu dJ O o -Q CL SSI E 13 1 — LU 2: 4— CO O 1 — [ 1 — 1 CO CJ ^ CU CU o O -O CL Ss: ^ ^ >i s_ LU zs 1— LU It.-* oo I CNJ 4— CSJ O 1 — 1 CO "O CU CU cu o fD o j3 E >5 ^ UJ rs 1— LU 00 zz; cu 4O I— 1 >1 — I CO o S$— M— cu CU o CL Sl/l Q E >> ^ _J :3 (— LU o CO 2: CNJ SLjJ 4— O 1 — 1 r— 1 — 1 CO 1 — 1 CU SSTD cu oi o cu O _C1 CL SQST E 1 — LU >> 13 H2: "O ra 4O 1 — 1 1— ( 1 — 1 CO iS0) cu QJ O Q. JD QS>> E >^ ^ 1— LU f— LU 2r -K 4O 1 — 1 lA OJ SST3 cu CD O O JD CL SE =3 1— LU z: CL. CO CM LO cn r— cr> LO CM CO CM r— CM o CM CO CO CM 1— CO CD Cvj 00 Ln CD
PAGE 153

147 2SLDV procedure for Model 1 with critical probability levels either 0.40 or 0.50. It is interesting to note that this number of errors is even lower than those generated by the single-equation method of Model (b) in Table 9, where in addition to the economic indicators of Model 1, the ratio of gross fixed capital formation of GDP was included. On the whole, one may draw the general conclusion that the simultaneousequations models perform better than the single-equation models in terms of producing a fewer number of Type I and Type II errors. However, even more important than comparisons of error rates, the economic implications of the simultaneous-equations procedure loom large. The empirical results confirm prior expectations as regards the impact of the rate of depreciation and the growth rate of exports on the amount of capital inflows. The negative sign of the coefficient of per capita income contradicts Johnson's (1965) supposition. It could be that official capital inflows are more important than private or that the relationship between official and private capital inflows, on the one hand, and the level of development, on the other, does not conform to Johnson's arguments. The intriguing result of this study concerns the coefficients of the endogenous variables. The comparison of the singleand simultaneousequations models reveals that the sign of the coefficient of the intention to reschedule variable, in the capital inflows equation, is negative and significant in both cases, which corroborates prior expectations. However, when it comes to the sign of the coefficient of capital inflows in the creditworthiness equation, the single-equation models reveal a negative and significant coefficient, which is in agreement with the

PAGE 154

148 results of Table 8 and the previous single-equation studies of developingcountry creditworthiness by Feder and Just (1977a) and Feder, Just and Ross (1981). On the contrary the simultaneous-equations models reveal a 1 n positive and significant coefficient for the capital inflows variable. The reversal of sign calls for an explanation. There are two possible explanations which come to mind. In the first instance, the variable identifying the creditworthiness equation 20 is the rate of depreciation. If the rate of depreciation were negatively related to the probability of rescheduling it could account for the positive coefficient of captial inflows. One possible explanation for the negative correlation between depreciition and creditworthiness could run along the following lines. A developing country may attempt to peg its exchange rate to a major currency. As its currency becomes overvalued (depreciates), the country may run down its international reserves in order to reverse the exchange rate depreciation. However, the running down of reserves would increase the likelihood of rescheduling. Therefore, the attempt to reduce the level of exchange rate T9 The positive sign of the coefficient appears quite robust with regard to the specification of the simultaneous-equations model. Other specifications omitted both the investment and money supply variables or both the investment and per capita income variables from the creditworthiness equation; the positive sign of the coefficient of the capital inflows variable remained unchanged. 20 This is the variable omitted from the creditworthiness equation but included in the capital inflows equation. The growth of exports is also excluded but it is correlated with the growth of GDP so it does not contribute to the identification of the creditworthiness equation.

PAGE 155

149 21 depreciation results in a higher probability of rescheduling. However, at the same time, lower exchange rate depreciation results in higher capital inflows and may therefore account for the positive sign of the capital inflows variable in the creditworthiness equation. Another more likely explanation, pointed out by Dhonte's (1975) study of debt renegotiation cases, concerns the observation that countries which rescheduled their debt, sought at the same time large capital inflows, so as to be able to avoid the rescheduling. The single equation method of est-mation ignores actions taken by the country to compensate for lower capital inflows, resulting from less favorable perceptions of creditworthiness. In other words, a higher probability of rescheduling may result in lower capital inflows and thus force the country to adopt policy measures to increase the amount of capital inflows and enhance its creditworthiness. It may be that, the positive coefficient of the coefficient of the capital inflows variable in the creditworthiness equation captures the country's policy reaction to a more prominent expectation of rescheduling. Conclusion This chapter has focused on the role of capital inflows in determining a country's perceived creditworthiness. It has examined the technique of logit analysis as it is applied to the economic determinants of creditworthiness. The empirical significance of the two indicators investment and growth of the money supply, which were previously neglected, was emphasized. Moreover, a simultaneous equation model for the 21 In fact when the rate of depreciation was included in the creditworthiness equation (without the capital inflows variable with which it is correlated), a negative but nonsignificant coefficient was obtained.

PAGE 156

150 determination of creditworthiness and capital inflows was formulated and estimated. It was shown that a reversal of the sign of the coefficient of the capital inflows variable occurs when the simultaneous-equations model is estimated: the coefficient turns from negative, in the single-equation model, to positive in the simultaneousequations model. A brief explanation for the reversal was included. Before concluding, it may be worthwhile to point to some of the deficiencies stemming from the application of logit analysis to the investigation of creditworthiness. In the first instance, as pointed out by Eaton and Gersovitz '(igsib, pp. 29-30) and Kharas (1981, pp. 3-5) the selection of indicators is quite arbitrary: no underlying theoretical arguments exist for the choice of variables and the models search for statistical relationships rather than attempting to rest the significance of different indicators pointed out by theoretical analysis. Even when the variables are selected, it is not clear whether they should be introduced in the form of levels of growth rates. In relation to the measurement of variables, the question arises as to whether the variables should be lagged and, if so, how many time periods. Moreover, if a growth rate is used, a problem arises over how many years it should be calculated. The studies often use cross-sectional data for a group of countries over a number of years. The implicit assumption, when estimating such models, is that the coefficients of the explanatory variables are stable, not only across countries, but also over time. However, no tests have been proposed or carried out to ascertain the stability of these coefficients. Finally, it must be noted that the data for reschedulings used by these studies refer to reschedulings of official credits. One must therefore be careful in using the results from these studies to draw conclusions about the behavior of private banks in the rescheduling proces.

PAGE 157

CHAPTER SIX SUMMARY AND CONCLUSIONS This dissertation introduced and investigated several facets of the external debt of developing countries, which the literature has not previously discussed adequately, i One particular criticism of the literature on this subject has been the emphasis placed on the "macro-economic" approach to evaluating the creditworthiness of developing countries. In other words.the emphasis has been on tfie demand side of the market neglecting the issue of the supply of loanable funds to these countries. The dissertation has attempted to shift emphasis from the demand side of the market to investigating the market for international debt. A model investigating the supply of funds to developing countries on the part of international commercial banks was introduced. The probability of a developing country defaulting on its external debt, as that is perceived by commercial banks has been postulated as the main determinant of the expected profile to the bank. Thus attention is shifted to the factors relevant to commercial banks' decisions in lending to developing countries. The implicit assumption of studies evaluating the creditworthiness of developing countries has been that funds are forthcoming to these countries provided an "appropriate" interest rate is paid. The issue of the supply of funds to developing countries has been examined ina series of papers by Eaton and Gersovitz (1980, 1981a, 151

PAGE 158

152 1981b), In their work they have introduced an explicit constraint in external funds forthcoming to developing countries and have argued that there may be a ceiling on the amount of external borrowing developing countries can undertake during a particular time period. However, in their model they have introduced the credit ceiling as endogenous to their model and have omitted to include the interest rate as an explanatory factor. Both these assertions have been questioned in this dissertation. The interest rate is shown to be an important explanatory factor if it is assumed to capture political and institutional factors as well as economic factors. The same reasoningpoints to a formulation ofthe credit ceiling as exogenous to the model. Such a model was specified and estimated in this dissertation. The empirical results in this thesis suggest that the simple model of Eaton and Gersovitz (1980) is overly simplistic as well as unrealistic in that it assumes zero covariance between the error terms of the demand and supply equations and that the variances of the two equations are equal. Such simplifications may produce highly significant and interesting results following from overly simplified assumptions and constraints on the model. It has been indicated that once these simplifying assumptions are dropped, the results obtained by Eaton and Gersovitz (1980) become highly questionabl e.

PAGE 159

APPENDIX A EFFECT OF CHANGES IN INTERNATIONAL RESERVES ON THE COMPARATIVE STATICS OF THE LOAN RATE INDEXATION MODEL In this Appendix it is required to find the change in dR./dp from a change in the level of international reserves i.e., ~— ( j— By dp differentiating equation (7) of Chapter 2 the following equation is obtained: 3p/ dp^ 3F.(R.D.(.); P., X.) 3R. \p, i^fi(i(-)^ Pi' ^•)^-(-) 3f.(R D (•); p., X.) 3D.(.) -^-^^^•(•)--i-^-^-(Vi(-)^Pi'Xi) 3P,. 1 3^D.(.) 3D.(.) 3f (R D (.); P., X.) ^i^-(^-Di(-)^ Pi X.)^^ R^? ^ .,3P, 3p. 3R. 3D.(.) 2l^^^-(^-Di(-)' Pi' 3D,(-) 3D,(.) 3'D.(.) ^"(i(-))3R7^ / 3p. [1 F.(R.D.(.); P^-, X.)] 3R. 3D.(.) Rif,(RiD.(.); P,, x.)D.(.)(UnDR.) c;(Di(-))-^ (Al) The denominator of (Al ) is evidently positive so the sign of the numerator is sought. Several assumptions are in order. First as already argued an increase in pwill lower the probability of default 153

PAGE 160

154 9F(R D (•); P., X.) — Hr, — — ^ (^2) It is assumed that the density function of possible returns is bell-shaped and that an increase in will move the function to the right increasing its mean as shown in Figure A-1 below f(8) Figure A-1 Changes in Density Function When p^. Increases The bank is assumed not to advance a loan unless the probability of repayment is greater than one half. Under these assumptions it follows that 8fT(R^-D.(-); P^-, X.) — < 0 (A3) Furthermore assume that C"'(D(.)) = 0 and (A4)

PAGE 161

1SS 3'D.(.) wir> 0 (A5) The first of these implies that marginal cost is increasing at a constant rate, whereas the second argues that a country with lower international reserves, and therefore more limited access to alternative sources of funds, will be less responsive in its decrease in credit demanded following an increase in the interest rate charged, than a country with greater international reserves. Given the assumptions in (A2), (A3), (A4) and (A5), it can be shown that all the terms in the brackets in the numerator of (Al) are negative with the exception of the first term. .-Consider the following combination of terms: 9f.(R D (.); P,-, X.) 3D (.) 3f.(R.D.(.);. P., X.) L_LJ ] L.R.D.(.) R? —1L_LJ ] ^ = 3p^. 1 r 1 3R^ 3p^. 3fi(R-D.(.); p., X.) ' ^^R.D.(.)(l+nnn ) < 0 (A6) 3p^ '^i^i^ "DR. If, as was assumed in the text the interest elasticity of demand for credit is greater than one (or |n|^j^ 1 > l)j (A6) is negative. In a similar manner, 9R ^f,(R,D.(.); P,, x,)D,(.) MJ.R f (R DJ.); p^., X,) 3p^ 1' 1 r ^1' 1' ^' ^p. T 1' i i^ ^i' ''i 3R 3D.(-) '^^U^W-^' Pi' Xi^^i 1 ^ f^iR^D.{.); p., x.)D.(.)R^[ ][,^^_ + + 2nRp. npR_] < 0 (A7)

PAGE 162

156 where is the elasticity of the interest rate factor with respect to the level of international reserves 9D.(-) p. '^Dp. = DTT77 is the elasticity of the demand for credit with respect to the level of international reserves. It is easy to see from equation (8) in Chapter 2 that nf^p_ > provided that [np^ J > 2. Given that the demand for credit is sufficiently interest elastic and n^^p > n^p the sign^'of (A7) is negative. Finally, from (A4), (A5), (A6), (A7) and\he assumptions made in Chapter 2 concerning the cost function, it is obvious that all the terms in the brackets in the numerator of (Al), with the exception of the first term, are negative. If the change in the probability of default is sufficiently small, the sign of (Al ) is positive, which is the required resul t.

PAGE 163

APPENDIX B IMPLICATIONS OF THE CORRELATION BETWEEN THE DEBT-EXPORTS RATIO AND THE AMOUNT OF DISBURSEMENTS The purpose of this Appendix is to investigate the correlation, postulated in Chapter 2, between the debt-exports ratio and the amount of disbursements through financial markets to developing countries. In order to facilitate the exposition, it is worthwhile to point out that the equation estimated in Chapter 2 equation (10) can be formulated as a switching regression model of the general form: yt = \^lt + "lt ^'^^t^^l (Bl) where = debt-exports ratio in period t = proportion of disbursements through financial m.arkets during period t X^ = value of LIBOR in period t c-j = constants ^lt'^2t coefficients to be estimated '^lt''^2t ^'"1^01^ terms Since the amount of debt outstanding in period t (the numerator of z^) includes the proportion of loan disbursements in period t(y^) as well as that of previous periods (y^_^ for k = l,2,...,n) and since and ^ 157

PAGE 164

158 are expected to be correlated over time, then one may assume that and are correlated. Intuitively, a country has been granted and taken down a number of loan commitments in period t because it is considered creditworthy. The country has a high debt-exports ratio as a result of its being considered creditworthy and being granted simultaneously a large number of loans. Therefore in (Bl) and (B2), it is expected that and y, are positively correlated and by implication z^ and u,, and z^ and Uo^ ^ ^ r t It t 2t are also positively correlated. In order to estimate the model, the expected values of the residulas ^It ^2t ^"^^ required. It is assumed that u-^^, u^^^ and z^ have a trivariate normal distribution with mean vector zero and covariance matrix. 2 01 lz E = 2 02z 2 Z Therefore, for equation (Bl) = Vit + E( "itl^ In order to obtain E(u-|^ |z^>c 1 ) it must distribution of u^^ given z^ is normal with mean Iz t and variance a| Iz It follows that

PAGE 165

159 E{u,t!z^iC,) = E(a' a Iz 2 Z > a z z z z I z z where ^i-) denotes the density function.and $(•) the cumulative distribution function of the standard normal distribution. The last equality derives from a well-known formula for the mean of the truncated normal distribution^ Similarly for u,. a See Maddala (1983, p. 365) for a discussion of the truncated normal di stribution .,

PAGE 166

160 For expositional purposes define w-j and W2 as ^1 (a-) w 1 c w z 2 2 c z Therefore equations (Bl) and (B2) can be written as a yt = Vit + + ^-it ^ ^ = X^32, ^ W2 + ^ (84) where e-j^ and e^-t ^i^e the new residuals: ^Iz ^It = a7^1 "2t = ^2t + ^ ^2 It can be easily verified that the conditional means of e^^ and are equal to zero. The residuals e-j^ and e^^^ are not heteroskedasti c as is the usual case in switching regression models. The variances of e.|^ and e^^ are 2 given by The formulae for the variances make use of the moments of the truncated bivariate normal distribution. For the derivation of the variance see Maddala (1983, p. 225).

PAGE 167

16] z z a" z z which are constants for all values of t. Therefore, application of ordinary least squares to (Bl) and (B2) will produce consistent estimates. The simple model estimated in Chapter 2 can be written in the form of equations (Bl) and (B2) as follows: PROP^ = + e^LBR^ + u-,^ if >^ c-j ^ (B5) PROP^ = + 32 LBR^ + u^^ if < (B6) where z^ and u^^ and z^ and u^^ are correlated. However, instead of (B5) and (B6) the following equations are estimated: PROP^ = a* + e-jLBR^ + v^^ if z^ > c^ PROP^ = a* + 32LBR^ + v^^ ""^ ^ ^2 where from (83) and (B4) a 1 ^ t ^ '^l + (B7) I I 0_ I and '2z "2 "2 o"^ ^2 (^S)

PAGE 168

162 It was shown in connection with (B3) and (B4) that the conditional means of the residuals v-j^ and v^^ are zero and therefore application of ordinary least squares will yield consistent estimates. However, the constant terms obtained from the regression will be related to the constant terms of equations (B5) and (86) via equations (87) and (88).

PAGE 169

APPENDIX C FIRST DERIVATIVES OF THE LOGARITHM OF THE LIKELIHOOD FUNCTION FOR THE DISEQUILIBRIUM MODEL Sample Separation Unknown This Appendix provides the first derivatives of the likelihood function (18) in Chapter 3, For convenience (18) was decomposed into two parts and analytic forms derived for each in (20) and (23). In addition, the following matrices are defined: U = R X]B-, a^Q Z = R x^e2 2'^ uf 2puiu n\ w = + — 12 02 R-M V = — exp[(R X^Bg "2^)^^ The first derivatives of (20) are as follows: r: UX, pZX-, If= la^-a.lh {-J —1]} (CI) ^^1 ^ 2 (1-p^) o-,^ 1^2 f= h+la. ao|h {-J ^]} (C2) 163

PAGE 170

164 1^ = |a-,-aJh {— ^ ^]} (C3) aaf 2 2af 2(l-p^)a^^ 2(l-p^)2a| + (— )(^-^)]} (C5) 2(l-p^) o^a\ 9a| ^ 2a^ 2(l-p^)0^ 2(l-p2)^a^ + (— J )(--^)]} (C6) 2(l-p2) a^a^ ^= |a.-a2l{[ ^] + h[(— 3af2 2(l-p^)a|a^ 2(l-p^) afa^a^2 ( ^^ )]} (C7) 2(l-p^) afa^ where h was defined in (19). The first derivatives of (23) are given by: %=SiWV)][^] (C8) 3H 3a-j gi[*(v)][^] (C9) ~= g^C— ][i*(v)] + 9^LHym^^^'] (CIO) 2 0^ a^a\

PAGE 171

165 1^ = g^E^in $(V)] + g^[<^(V)][^] (Cll) — = gT[<^(v)]C-^] (C12) 1H„ = [1 $(V)][J + g [(•) denotes the cumulative normal distribution function and (}>(•) the normal density function. M and were defined in (21) and (22) respectively. The first derivatives of the logarithm of the likelihood function are as follows: 93-1 i=i ^ ^1 n 3G/3a-i ~-= ^ (C16) 332 i=l ^ I n 3G/3a^ „i n 36/30? — = z (C19) 3af i-1 ^

PAGE 172

166 .1 n 3G/3a^ ^ = E (C20) „, n 3G/3a.% — = Z P (C21) where G was defined in (18). The derivative of G with respect to any one of the variables is equal to the sum of the derivatives of f and H with respect to the same variable, given in (C1)-(C14). The first derivatives of the logarithm of the likelihood function, (C15)-(C21) are used in the Berndt et al (1974) iterative procedure to form the vector S(e) of equation (24) in Chapter 4. The first derivatives are also evaluated at each observation'' and the values are used to form the Q(e) matrix of equation (25) in Chapter 4. Sample Separation Known The likelihood function for the model (14)-(17), where knowledge is available as to which regime each observation belongs to, is given by: L* = n f(R.,Q.) n J g(RD. ,R. )dRD. i()2 R^and the logarithm of the likelhood function is logL* E logf + E log H In equations (C15)-(C21), the first derivatives are evaluated at each observation and then summed over all observations to form the S(6) matrix. In order to construct the Q(e) matrix, the first derivatives are evaluated at each observation and placed in a kxn matric where k is the number of parameters and n the number of observations. This matrix is post multiplied by its transpose to obtain the matrix Q(e).

PAGE 173

167 where \^-^ is the set of observations for which < Q"^ and ijj^ the set of observations for which Q^. = and f and H were defined in (20) and (23) respectively. The first derivatives of the logarithm of the likelihood function are therefore given by; 3logL* ^ ^^/^^l ^ ^H/^g] ^1 3logL* 9f/5a-j 9H/3a., da. f ^ H 8logL* ^^/^h ^^^^h 3Bo f ^ ^ H 3logL* 9f/8oi2 3H/3a2 3a7 ^ ~F ^ ^ ~H 3logL* f f ^ H 3logL* ^ ^^^^^^2 ^"^^^2 3a2 ^-^ il^2 3logL* ^ H!!^, ^"/^"12 3a-|2 ij^i 4^2 where the derivative of f with respect to any one of the parameters was shown in (C1)-(C7) and the derivative of H with respect to any one of the parameters was shown in (C8)-(C14).

PAGE 174

APPENDIX D DATA FOR THE SIMULTANEOUS LOG IT MODEL Sources and Definition of Variables Reschedul ina The latent variable creditworthiness or the intention to reschedule is unobservable. Instead what is observed is a binary-valued variable: if a country had its external debt rescheduled in a particular year the value of 1 is assinged to the observation and the value of 0 is assigned to non-rescheduling observations. A year was designated a rescheduling year if the country requested a debt rescheduling and proceeded to service its debt according to the renegotiated terms and/or built up serious external debt service arrears irrespective of whether a formal rescheduling took place later or not. If forecasted economic circumstances for year t were reported to have prompted the government to submit a rescheduling in year t-1 t was designated the year of rescheduling; conversely, a rescheduling submission in t+1 prompted by events in t, led to classification of year t as the rescheduling year. Reschedulings which were motivated by desire on the part of creditors to provide development aid were excluded from the sample. Data for this variable as well as for the debt service ratio and capital inflows/debt service payments variables were kindly made available to the author by Dr. Gershon Feder of the World Bank. Debt Service Rati o Debt service payments on public and publicly guaranteed debt in year t-1 divided by exports of goods and services. Where data were m

PAGE 175

169 available, the service on long-term private debt not guaranteed by the debtor's government was included. When a year was identified as a rescheduling case, the debt service originally due rather than the debt service actually paid vyas calculated. The source of the data is the same as that for the rescheduling variable. Capital Inflows/Debt Service Payments Capital inflovys in year t-1 include both commercial and noncommercial inflows. Non-commercial inflows include net mediumand longterm loans from governments and international organizations, capital grants, workers' remittances and net current transfers. Commercial inflows comprise net mediumand long-term loans from commercial sources and direct investment income. The source of the data is the same as that for the rescheduling variable. Imports/Reserves Imports for year t-1 divided by year-end reserves (in U.S. dollars). Reserves include gold holdings foreign exchange holdings special drawing rights (SDRs) and reserve position in the mv All the data were obtained from the May 1978 issue of the International Financial Statistics (IFS) of the IMF. Per Capita Income Data were available in U.S. dollars from various issues of the United States Statistical Yearbook (UNSY). GDP Growth Rate The growth rate was calculated as the average of the annual rates of growth between year t-4 and t-1. Annual growth rates were calculated from index numbers of per capita gross domestic product in constant prices. Data for the index number were obtained from various issues of the UNSY.

PAGE 176

170 Export Grovjth Rate The export growth rate was calculated as the average of the annual rates of growth between year t-8 and t-1 Data for the calculation of annual export growth rates were obtained from the 1980 IFS Yearbook. Investment Rate The variable was calculated as the ratio of gross fixed capital formation in year t-1 to GDP. Data for both variables were obtained from the May 1978 issue of the IFS. Money Supply Growth The variable was calculated as the average of the annual money supply growth rates between year t-4 and t-1. Data for the calculation of the annual growth rates were obtained from the 1930 IFS Yearbook. Exchange Rate Depreciation The variable was calculated as the average of the annual rates of depreciation of the nominal exchange rate betv;een year t-4 and t-1. Data for the nominal exchange rate of the domestic currency vis-a-vis the U.S. dollar were obtained from the 1980 IFS Yearbook.

PAGE 177

171 Table Dl Mean Standard Deviation and Range of the Variable Included in the Logit Analysis Variable Mean Standard Deviation Minimun Maximum 1) Debt Service Ratio 0.1202 0.1046 0.0030 0.7150 2) Capital Inflows/ Debt Service Payments 3.6274 3.5379 -1.4947 26.3394 3) Growth of GDP 0.0350 0.0452 -0.0470 0.6210 4) Growth of Exports 0.1051 0.0982 -0.0929 ^0.5425 5) Per Capita Income 371.7104 335.5991 41.00 3225.00 6) Imports/Reserves 4.5219 3.7150 0.4911 29.9992 7) Gross Fixed Capital Formation/GDP 0.1790 0.0646 0.0440 0.6270 8) Growth of Money Supply 0.2046 0.2791 -0.0050 2.8940 9) Nominal Exchange Rate Depreciation 0.0583 0.1985 -0.0340 2.9280

PAGE 178

172 CO o cr> o CM CNJ CO cn CNJ o CO m un cn 1 — o o 1 — o CO CO o CO o CO r— o CO o o C\J o o -a OJ -o at LO CO CO o UD LO (1) un cn 1X3 o CM o (O (/) o o •1 — CM 4O < CD (/) +-) 4-> 1 — JD C CD fa cu o 1— _l o ^ <++-> CU o o c: o +J OJ

PAGE 179

173 XI CD x: o v> tu I CTl > to CO o CO £Zj e: r— -O ^ 0) fc -a sro >u (/I QJ I o tjDLOr~^<-DLOLDLClLO KCi CTl CTl CTl 01 CTl CTl 01 0-1 cn CTl Oi cn CTl cn 0-1 LO LO LO LO Ln r-^ LO r\ t--, I I I I I LO LO CTl LO r-^ LO r-L£J LO LO LO 01 CTl CD sz +J r3 03 c 13 c > Ul 0 Ol ro ra ro LO 0 LJ ra r-M r3 >-, (73 >> u 0 SS00 cn J0) ra -r0 rn CD -tro 4-> Sr— X. 0 D5^ Ss0 _ro cu •Iro ro CD Cl. D. 0 ro LO LO CO cn CM CM C\J CM CM CM OJ ro CO to QJ CU ro c: i^ •r1 — o c; Cl. rO Dro •r:: ro _J !Z cu CD CI c c -I-a cu •.S3 ro -a ro ro ro >^ •1>, ro to cu =S •<^ ^ C S=3 ZJ Z5 S_ CO ro cn o rCO ro 0) =3 N QJ C QJ ro QJ CM CO 1/1 Sro QJ >>i S4-> CO 3 T3 QJ O lO GJ PC LO LO cn LO lO cn o CTl 10 LO LO I in CD o I LO LO cn o to rO QJ >O to 0) I o o LD'^LOLDLOOLlDLDLOmLOCnLOLOi — LOCO r^r^i^i — i^r^r^r^r^r~-r^Lor-^r^r-.r^LD I I I I I I I I I I I I I I I I I I I--. LD LO LO LO LO LO r-Ln LO LT-) CO LO Ln en CO LO LO LOLOLOLOLOLOLOLOLDLOLOLOLOLOLDr--LD "LOLOP^ cnaicncTicncncrichcricn cno-io-iCTicricncriLncncTiOi LO LO uo r~-~ ~ LO I I LO LO LD h~ r-~ r-^ I I Ln LO LO L O LO LO o"i cn cn 4-1 ro ro CJ c 3 O o ro c ro •.Q. ro Tro — "rcc Q) •r+J •.— _Q D-.: S-C:>'i-rOCUrOEro QJOJ-i-NEi — CO+J cn CDi — ro i-r•1 — to £ 1 — 5-oi-3_c:^ooQ CCCCCQCQ D3C_>(^OCOC:3 o o a +j ro O. =S >i O CTl LU UJ LU LU CD ro r— CL o c •o +J ro ro £1 ro C£i ro E OJ -M ro rs CD ro to •!— ro to QJ 3 cz a o o c CMrO':d-LnLOr^COCTlO CM 00 "^iLO CO cn to ro o C_) >^ ro ro O E S~ >
PAGE 180

REFERENCES Abbott, G. C. (1981), "International Indebtedness of Less Developed Countries: Structure Growth Indicators," Aussenwi rtschaft Vol. 36, No, 4, pp. 340-351. Aliber, R. Z. (1977), "Living With Developing Country Debt," Lloyds Bank Review No. 126, pp. 34-44. Alter, G. M. (1963), "The Servicing of Foreign Capital Inflows by Underdeveloped Countries," in H. S. Ellis, ed., Economic Development of Latin America ,. New York: St. Martin's Press, pp. 139-167. Amemiya, T. (1974), "A Note on a Fair.and Jaffee Model," Economefric a, Vol. 42, No. 4, pp. 759-762. Angel ini, A., E. Maximo and F. Lees (1979), International Lending, Risk and the Euromarkets London: Macmillan. Atramovic, D. (1954), Economic Growth and External Debt Baltimore, Maryland: Johns Hopkins Press. Bank for International Settlements, (1981), Annual Report Basle, Switzerland: BIS. Bardhan, P. K. (1967), "Optimum Foreign Borrowing," in K. Shell, ed., Essays on the Theory of Optimal Economic Growth Cambridge: MIT Press, pp. 117-128. Berndt, E. R., B. H. Hall and J. A. Hausman (1974), "Estimation and Inference in Non-Linear Structural Models," Annals of Economic and Social Measurement Vol. 3, No. 4, pp. 653-665. Black, S. W. (1981), "The Impact of Changes in the World Economy on the Stabilization Policies in the 1970s," in W. R. Cline and S. Weintraub, eds.. Economic Stabilization in Developing Countries Washington, D. C: Brookings Institution, pp. 43-77. Blackwell, N. and A. M. Santomero (1982), "Bank Credit Rationing and the Customer Relation," Journal of Monetary Economics, Vol. 9, No. 1, pp. 121-129. Blask, J. K. (1978), "A Survey of Country Evaluation Systems in Use," in S. H. Goodman, ed.. Financing and Risk in Developing Countries New York: Praeger, pp. 65-70. 1.74

PAGE 181

175 Brackenridge, A. B. (1978), "One Approach to Country Evaluation," in S. H. Goodman, ed.. Financing and Risk in Developing Countries New York: Praeger, pp. 60-64. Brittain, W. H. B. (1977), "Developing Countries' External Debt and the Private Banks," Banca Nazionale del Lavoro Quarterly Review Vol. 30, No. 123, pp. 365-381. Campbell, T. S. (1978), "A Model of the Market for Lines of Credit," Journal of Finance Vol. 33, No. 1, pp. 231-244. Chenery, H. B. and A. Strout (1966), "Foreign Assistance and Economic Development," American Economic Review Vol. 56, No. 4, pp. 679-733. Dhonte, P. (1975), "Describing External Debt Situation: A Roll-Over Approach," International Monetary Fund Staff Papers Vol. 22, No. 1, pp. 159-187. Dufey, G. and I. H. Giddy (1978), The International Money Market Engl ewood Cliffs, N.J.: Prentice-Hall, Inc. ; Eaton, J. and M. Gersovitz (1980), "LDC Participation in International Financial Markets: Debt and Reserves," Journal of Development Economics Vol. 7, No. 1, pp. 3-21. Eaton, J. and M. Gersovitz (1981a), "Debt With Potential Repudiation: Theoretical and Empirical Analysis," Review of Economic Studies, Vol. 48, No. 2, pp. 289-309. Eaton, J. and M. Gersovitz (1981b), "Poor-Country Borrowing in Private Financial Markets and the Repudiation Issue," Princeton Studies in International Finance #47 Princeton, N.J.: International Finance Section. Euh, Y, D. (1979), Commercial Banks and the Creditworthiness of Less Developed Countries UMI Research Press. Fair, R. C. and D. M. Jaffee (1972), "Methods of Estimation for Markets in Disequilibrium," Econometri ca Vol. 40, No. 3, pp. 497-514. Fair, R. C. and H. H. Kelejian (1974), "Methods of Estimation for Markets in Disequilibrium: A Further Study," Econometr i ca, Vol. 42, No. 1, pp. 177-190. Feder, G. (1980), "Economic Growth, Foreign Loans and Debt Servicing Capacity of Developing Countries," The Journal of Development Studies Vol. 16, No. 3, pp. 352-369. Feder, G. (1981), "Growth and External Borrowing in Trade Gap Economies of Less Developed Countries," Aussenwi rtschaft. Vol. 36, No. 4, pp. 381-395.

PAGE 182

176 Feder, G. and R. E. Just (1977a), "A Study of Debt Servicing Capacity Applying Logit Analysis," Journal of Development Eco nomics, Vol 4, No. 1 pp. 25-38. Feder, G. and R. E. Just (1977b), "An Analysis of Credit Terms in the Eurodollar Market," European Economic R eview, Vol. 9, No. 2, pp. 221-243. Feder, G. and R. E. Just (1979), "Optimal International Borrowing Captial Allocation and Credit-Worthiness Control," Kredit and Kapital Vol. 12, No. 2, pp. 207-220. Feder, G. and R. E. Just (1980), "A Model for Analyzing Lenders' Perceived Risk," Applied Economics Vol. 12, No. 2, pp. 125-144. Feder, G. and K. Ross (1982), "Risk Assessments and Risk Premiums in the Eurodollar Market," Journal of Fin ance, Vol. 37, No. 3, pp. 679-691. Feder, G. R. E. Just and K. Ross (1981), "Projecting Debt Servicing Capacity of Developing Countries," Journal of Financial and Quantitative Analysis Vol. 16, No. 5, pp. 651-669. Finance and Development (1983), "External Debt the Continuing Problem," Vol. 20, No. 1, March. Finch, D. (1951-2), "Investment Service of Underdeveloped Countries," International Monetary Fund Staff Papers Vol. 2, pp. 60-85. Frank, C. R. and W. R. Cline (1971), "Measurement of Debt Servicing Capacity: An Application of Discriminant Analysis," Journal of International Economics Vol. 1, No. 3, pp. 327-344. Feimer, M. and M. J. Gordon (1965), "Why Bankers Ration Credit," Quarterly Journal of Economics Vol. 79, No. 3, pp. 397-416. Friedman, I. S. (1977), "Most Eurocurrency Borrowers Remain Good Risks," Euromoney, March, pp. 16-21. Friedman, I. S. (1981), "The Role of Private Banks in Stabilization Programs," in W. R. Cline and S. Weintraub, eds.. Economic Stabilization in Developing Countries Washington, D. C. : Brookings Institution, pp. 235-268. Goldfeld, S. M. and R. E. Quandt (1975), "Estimation in a Disequilibrium Model and the Value of Information," Journa l of Econometrics, Vol. 3, No. 3, pp. 325-348. Goodman, L. S. (1981), "Bank Lending to Non-OPEC LDCs : Are Risks Diversifiable?," Federal Reserve Bank of New York Quarterly Review, Vol. 6, No. 2, pp. 10-20. ~

PAGE 183

177 Goodman, L. S. (1982), "Syndicated Eurocurrency Credits: Pricing and Practice," Federal Reserve Bank of New York Research Paper No, 8202, January. Goodman, S. H. (1977), "How the Big U.S. Banks Really Evaluate Sovereign Risks," Euromoney February, pp. 105-110. Greenspan, A. (1977), Discussion of R. S. Solomon, "A Perspective on the Debt of the Developing Countries," Brookings Papers on Economic Activity, No. 2, pp. 479-501. Griffith-Jones, S. (1980), "The Growth of Multinational Banking, the Euro-Currency Market and Their Effects on Developing Countries," The Journal of Development Studies Vol. 16, No. 2, pp. 204-224. Hardy, C. (1979), "Commercial Bank Lending to Developing Countries: Supply Constraints," World Development Vol. 7, No. 2, pp. 189-197. Hodgman, D. R. (1960), "Credit Risks and Credit Rationing," Quarterly Journal of Economics Vol. 74, No. 2, pp. 258-278. ~ International Monetary Fund (1981a), Annual Report Washington, D. C: International Monetary Fund. International Monetary Fund (1981b), International Capital Markets Occasional Paper No. 1, Washington, D. C. : International Monetary Fund. International Monetary Fund, International Financial Statistics Yearbook, Washington, D. C: International Monetary Fund. Irvine, R. J., Y. Moroni and H. F. Lee (1970), "How to Borrow Successfully, Columbia Journal of World Business Vol. 5, No. 1, pp. 42-48. Jaffee, D. M. and F. Modigliani (1969), "A Theory and Test of Credit Rationing," American Economic Review Vol. 59, No. 5, pp. 850-872. James, C. (1982), "An Analysis of Bank Loan Rate Indexation," Journal of Finance Vol. 37, No. 3, pp. 809-825. Johnson, H. G. (1965), "Some Aspects of the Theory of Economic Policy in a World of Capital Mobility," Rivista Internazionale di Scienze Economiche e Commerciali Vol. 12, No. 6, pp. 545-562. Kapur, I. (1977), "An Analysis of the Supply of Eurocurrency Finance to Developing Countries," Oxford Bulletin of Economics and St atistics, Vol. 39, No. 3, pp. 171^T88: ~~ Karekan, J. H. (1977), Discussion of R. S. Solomon, "A Perspective on the Debt of the Developing Countries," Brookings Papers on Eco nomic Activity No. 2, pp. 479-501.

PAGE 184

178 Kharas, H, (1981), "The Analysis of Long-Run Creditworthiness: Theory and Practice," Domestic Finance Study No. 73, The World Bank, July. King, B. B. (1968), Notes on the Mechanics of Growth and Debt Baltimore, Md: The Johns Hopkins University Press. Leff, N. H. (1975), "Rates of Return to Capital, Domestic Savings and Investment in the Developing Countries," Kyklos, Vol. 28, No. 4, pp. 827-851. L'Heriteau, M-F (1979), "Dette Exterieure et Modele de Developpement : La Place du liers Monde dans le Nouveau Dispositif Imperial iste," Revue Tiers-Monde Vol. 20, No. 80, pp. 703-724, (in French). Lichtensztejn, S. and J. M. Quijano (1982), "The External Indebtedness of the Developing Countries to International Private Banks," in J. C. S. Arnau, ed.. Debt and Development New York: Praeger, pp. 185-265. Long, M. F. (1980), "Balance of Payments Disturbances and the Debt of the Non-Oil Less Developed Countries: Retrospect and Prospect," Kyklos Vol. 33, No. 3, pp. 475-499. Long, M. F, (1981), "External Debt and Trade Imperative in Latin America," Quarterly Review of Economics and Business Vol. 21, No. 2, pp. 280-301. Long, M. F. and F. Veneroso (1981), "The Debt-Related Problems of the Non-Oil Less Developed Countries," Economic Development and Cultural Change Vol. 29, No. 3, pp. 501-516. Maddala, G. S. (1980), "Simultaneous Probit and Tobit Models with Latent Structures," Discussion Paper #15, Center for Econometrics and Decision Sciences, Gainesville, Florida, April. Maddala, G. S. (1983), Limited-Dependent and Qualitative Variables in Econometrics Cambridge: Cambridge University Press. Maddala, G. S. (1983a), "Methods of Estimation for Models of Markets with Bounded Price Variation," International Economic Review Vol. 24, No. 2, pp. 361-378. Maddala, G. S. and F. D. Nelson (1974), "Maximum Likelihood Methods for Models of Markets in Disequilibrium," Econo metri ca Vol. 42, No. 6, pp. 1013-1030. Manfredi, E. M. (1981), "Predicting Debt Reschedulings in Developing Countries," Agricultural Economics Research Vol. 33, No. 2, pp. 26-30. Mayo, A. L. and A. G. Barrett (1978), "An Early-Warni ng Model for Assessing Developing-Country Risk," in S. H. Goodman, ed. Financing and Risk in Developing Countr ies, New York: Praeger, pp. 81-87.

PAGE 185

179 McDonald, D. C. (1982), "Debt Capacity and Developing Country Borrowing: A Survey of the Literature," International Monetary Fund Staff Papers Vol. 29, No. 4, pp. 603-646. McKinnon, R. I. (1964), "Foreign Exchange Constraints in Economic Development and Efficient Aid Allocation," Economic Journal Vol. 74, No. 294, pp. 388-409. Mohammed, A. F. and F. Saccomanni (1973), "Short-Term Banking and EuroCurrency Credits to Developing Countries," International Monetary Fund Staff Papers Vol. 20, No. 3, pp. 612-639^ Nelson, F. D. and L. Olson (1978), "Specification and Estimation of a Simultaneous Equation Model with Limited Dependent Variables," International Economic Review Vol. 19, No. 3, pp. 595-709. Nowzad, B. (1982), "Some Issues and Questions Regarding Debt of Developing Countries," in T.-Killick, ed.. Adjustment and Financing in the Developing World: The Role of the International Monetary Fu nd, Washington, D. C: IMF, pp. 155.-169. 1~ Ohlin, G. (1966), Aid and Development Paris: Development Centre of the Organization for Economic Cooperation and Development. Rhomberg, R. R. (1966), "Private Capital Movements and Exchange Rates in Developing Countries," International Mon etary Fund Staff Papers, Vol. 13, No. 1, pp. 1-26. Sachs, J. (1981), "The Current Account and Macroeconomic Adjustment in the 1970s," Brookings Papers on Econo mic Activity, No. 2, pp. 201-269. Sachs, J. (1982), "LDC Debt in the 1980s: Risk and Reforms," manuscript, Harvard University, Cambridge, Mass., January. Sachs, J. and D. Cohen (1982), "LDC Borrowing with Default Risk," National Bureau of Economic Research, Working Paper No. 925, July. Saini, K. and P. Bates (1978), "Statistical Techniques for Determining Debt-Servicing Capacity for Developing Countries," Research Paper #7818, New York: Federal Reserve Bank of New York, September. Sargen, N. P. (1976), "Commercial Bank Lending to Developing Countries," Economic Review Federal Reserve Bank of San Francisco, Spring, pp. 20-31. Sargen, N. P. (1977), "Economic Indicators and Country Risk Appraisal," Economic Review Federal Reserve Bank of San Francisco, Fall, pp. 19-35.

PAGE 186

180 Shihata, I. F. I. and R. Mabro (1979), "The OPEC Aid Record," World Development Vol. 7, No. 2, pp. 161-173. Solomon, R. S.(1977), "A Perspective on the Debt of the Developing Countries," Brookings Papers on Economic Activit y, No. 2, pp. 479-501. Solomon, R. S. (1981), "The Debt of Developing Countries: Another Look," Brookings Papers on Economic Activity No. 2, pp. 593-605. Walter, I. (1981), "Country Risk, Portfolio Decisions and Regulation in International Bank Lending," Journal of Banking and Finan ce, Vol. 5, No. 1, pp. 77-92. World Bank, (1981), World Debt Tables Publication EC-167/81, Washington, D. C. : The World Bank, December. World Bank (1981a), World Development Report Washington, D. C. : World Bank. World Financial Markets (1983), Global' Debt: Assessment and Prescriptions, New York: Morgan Guaranty Trust Company, February.

PAGE 187

BIOGRAPHICAL SKETCH Andreas Savvides was born in Larnaca, Cyprus, on December 29, 1957. After attending public school in Cyprus, he completed the last two years of his secondary education at the United World College of the Atlantic in Great Britain. In 1975 he entered the University of Birmingham, England where he graduated in 1979 with First Class Honours in the Double Honors Degree of Bachelor of Science and Bachelor of Commerce in Engineering Production and Economics. In 1979, he entered the Graduate School of the University of Florida, where he is completing his Ph.D. in economics. He is currently a visiting faculty member at the University of Wisconsin-Milwaukee. 181

PAGE 188

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. G. S. Maddala, Chairman Graduate Research Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. d A. Denslow Associate Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Will iam Bomberger 0 Associate Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Roy LI Crum Professor of Finance This dissertation was submitted to the Graduate Faculty of the Department of Economics in the College of Business Administration and to the Graduate Council, and vas accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December 1983 Dean for Graduate Studies and Research