THE RELATIONSHIP BETWEEN AID AND DEBT IN DEVELOPING COUNTRIES.
Miles B. Cahill [*]Paul N. Isely [**]
Abstract
This paper constructs a model in which debt and aid are complementary. Specifically, the model shows how aid can be extracted from industrialized country governments by LDCs to finance debts. Policy implications for understanding debt crises are outlined. Using recent World Bank data, the fundamental equations of the model are estimated. While it is found that the model overestimates the actual amount of aid and debt, the relationship between aid, GDP and absorption; and debt, aid, GDP and absorption is of the predicted direction.
Introduction
Some recent papers have suggested a link between aid and external debt. This contrasts with traditional lending models in which debt is acquired to smooth consumption, substitute present for future consumption, build capital, or other reasons. [1] Aid has been associated with political goals, humanitarian goals, and debt relief, but few studies suggest that aid is specifically used to finance debt payments over extended periods. [2]
One model that does suggest that aid is used to finance debt is Bulow and Rogoff (1989c) (based on Bulow and Rogoff (1989a)). In this model, the debtor country is considered a developing country (LDC) because it is assumed that the discount rate of the LDC is above the world interest rate. This assumption also gives a rationale for the LDC to borrow. The Bulow-Rogoff model shows that debt can be supported by aid when the debtor country can extract side payments from an industrialized country's government. Side payments are given to ensure that any debt rescheduling agreements are settled in a timely manner, so trade is not interrupted. The industrialized country's government values trade because its constituents derive a consumer's surplus from consumed imported goods. When the side payments are anticipated, the debtor is able to borrow against future expected side payments. This gives the false appearance of a debt "overhang" problem, because while the developing country cannot sustain its loan payments wit h its own resources, it can rationally expect the aid extracted from the industrialized country's government. (Debt "overhang" is debt that a country is not expected to be able to repay.)
The Bulow and Rogoff model does have two drawbacks. First, any trading partner could extract side payments from the industrialized country's government by engaging in a debt agreement. Second, because the consumer's surplus derived from trade is not measured, the model cannot be directly tested.
We develop an alternate, testable framework in which aid is extracted for other equally compelling reasons. It is assumed that an industrialized country's government is willing to provide aid to keep the developing country stable, and stability is compromised if income is too low. We show that the LDC and creditors can extract the aid by engaging in a debt contract which apparently creates an overhang problem. Specifically, the debt contract specifies the LDC make payments so large that it would become unstable without the aid. In this way, the aid finances the overhang, and the overhang draws the aid payments. It is shown that this model has important policy implications for offering debt relief, as both "true" and "strategic" default announcements are possible.
Both the model in this paper and in Bulow and Rogoff (1989c) suggest there is a complementary link between aid and debt. This paper investigates this proposition by estimating the model with World Bank data. While the results suggest the model overestimates true debt and aid, the general nature of the complementary relationship is as the model predicts.
International Lending Models
There are two key issues that are particular to the market for loans to LDCs: the motivation for borrowing and the incentive to repay. [3] This section will briefly discuss these issues and review some of the associated literature. In doing so, some of the key assumptions of our model are placed in context.
Consider the question of why LDCs borrow. Several reasons have been suggested, including the hypotheses that LDCs borrow because they have a higher discount rate, wish to smooth consumption (in the face of shocks), or wish to increase their capital stock. In our study, we focus on the determinants of the size of loans rather than the motivations for borrowing. We use the simplest practical formulation to generate borrowing by assuming that the borrower's discount rate is higher than the world interest rate. This is consistent with many other studies, including Bulow and Rogoff (1989a, c).
A more complex issue is the incentives that induce the debtor to repay. Incentives are guided by the costs of default, the possibility of forgiveness or rescheduling agreements, the structure of rescheduling agreements, and the players in rescheduling agreements. Since the lender usually cannot hold collateral or seize a defaulting country's assets, moral hazard problems make this issue more complex than it is for a standard domestic lending agreement. [4]
The primary Costs of default fall into two categories: reputation costs and sanctions. In our model, the LDC borrows as much as possible at the onset, so reputation effects are largely irrelevant. [5] We employ trade sanctions as the cost of default. While a bank cannot seize the entirety of a defaulting country's assets, the appendix to Bulow and Rogoff (1989a) argues forcefully that banks can initiate trade sanctions to recover losses. They cite legal issues and empirical evidence to support this view.
Related to the issue of default is that of rescheduling. In the case where a debtor cannot--or will not--make promised payments, a rescheduling agreement is often Pareto improving. Rescheduling agreements entail lowering debt payments so that the debtor will resume making payments (thus improving the position of the lender over the default state) in exchange for the lifting of sanctions (improving the position of the borrower). The exact outcome of the negotiations depends on the bargaining power of each side, and whether a third party can be brought into the negotiations. Simple bargaining models assume that the lender's ability to inflict punishment or the borrower's ability to pay is enough to prevent default. [6]
More complex, and arguably more realistic, bargaining frameworks allow each party in the negotiations to have a role in the bargaining process. At a minimum, the debtor is allowed to stall negotiations in these models, effecting a loss of interest on the lender. Bargaining models derive the Nash equilibrium and subgame perfect Nash equilibrium ("perfect equilibrium" in Rubenstein's (1982) terminology) outcomes of bargaining to determine the size of lending agreements? Essentially, the assumption of a subgame perfect Nash equilibrium excludes cheap talk and idle threats. More formally, a Nash equilibrium is a set of strategies, where the strategy for each player is the "best response" to the optimal strategies of the other players. When a game has more than one round (or "stage") in which players must make choices (as is the case with bargaining situations), there are often more than one Nash equilibrium. A subgame perfect Nash equilibrium holds that the strategies in each subgame (e. g. later rounds of the g ame) are Nash equilibrium strategies. A central feature of subgame perfect bargaining models is that the borrowing country is unable to credibly commit to making the maximum payment possible, as its bargaining power allows it to negotiate a lower payment after loans are made.
Like other debt-bargaining models, we allow a third party--an industrialized country's government--to enter into the rescheduling agreement. As noted above, Bulow and Rogoff (1989c) suggests that an industrialized country's government will make trade-insuring side payments which result in larger loans for the LDC. In the model below, payments are made to ensure political stability, which is compromised in default. This leads to two important results: the benefits of the side payments accrue to the debtor, and the bank rationally lends more to the debtor than the debtor is able to repay. [8]
The Theoretical Model
The model consists of an LDC, a bank, and an industrialized country's government. The LDC produces output at a fixed rate (y), and trades the output at fixed terms of trade P, so that the trade value of GDP is Py. If consumption in the LDC falls below c (which need not be the subsistence level), the populace revolts and LDC becomes politically stable. This assumption is based on the observations made by Jeffrey Sachs (1989, p. 12), who attributed some of the high LDC debt levels in 1980s on the following: "In some of the cases ... the governments were too weak to resist the demands for spending from various highly mobilized social groups." [9] It is assumed that y [less than] C [less than] Py, so the LDC is unstable without either trade or outside aid.
When unstable, the government of the LDC suffers losses and repudiates all debts. In addition, the industrialized country's government suffers losses. These losses are caused by any number of specific reasons: costs of peacekeeping, loss of access to strategic areas, political pressure from voters, or any other reason. [10] Denote the value of stability to the industrialized country's government by [A.sup.*] Because it suffers losses, the industrialized country's government is willing to provide aid to prevent unrest. To minimize its aid bill, the most aid the industrialized country's government will provide is equal to the difference between the LDC's minimum consumption requirements (c) and autarkic income (y). This maximum aid payment is defined in equation (1).
A = c - y (1)
Aid is only provided when the value derived from stability ([A.sup.*]) exceeds A; this condition is assumed.
A bank is willing to loan the LDC only what the LDC can be expected to pay back, with interest. The interest rate on the loan is the world interest rate r. [11] The discount rate of the LDC is assumed to be [delta] [greater than] r, so the LDC seeks to borrow as much as possible at the onset. After the borrowed funds are received, the LDC may announce a default and call for a rescheduling agreement at any time, regardless of its ability to pay. When the LDC is in default, the bank seizes its trade as compensation. This forces the LDC to cease trade, which in turn forces the industrialized country's government to provide aid (A) to prevent instability. Thus, the LDC consumes c = y + A when in default.
The outcome of the rescheduling agreement is confined to the subgame perfect Nash equilibrium, as defined earlier. The structure of the meeting is as follows. The bank makes the first payment offer. The LDC then decides whether to accept or reject the offer. Rejection by the LDC results in a stalling period of default before the bank makes its next offer, causing losses to all parties. If the bank's offer requires the industrialized country's government to make payments, the industrialized country's government can also reject the offer (by not making the required payments), forcing a similar period of default. To avoid interest losses, the bank makes an offer that the LDC and industrialized country's government will accept in the first round, but only at the margin. The bank makes an offer that leaves the LDC with the same consumption as in default, and the industrialized country's government with the same aid bill as in a default. This offer, expressed as q percent of the LDC's Py income is [q.sup.*] = (Py + A - c) / Py. It requires the LDC to make payments at rate (P y + A- c), the industrialized country's government to make aid payments at rate A, and leaves the LDC consuming at rate c.
All parties are indifferent to accepting this agreement and forcing a default. The LDC consumes c whether it defaults or not, as a default results in a loss of trade and consumption of c = y + A. If the industrialized country's government does not provide the required A in aid, the LDC will default and cause the industrialized country's government losses of [A.sup.*]. [12] Because of this (marginal) indifference, all parties accept the agreement.
The LDC makes the maximum payment stream possible. This contrasts with a central feature of other game theory models, including Bulow and Rogoff (1989a) and (1989c). In these models, the borrower would like to commit to making as large a stream of payments as possible ex ante, but cannot as ex post it is able to negotiate a smaller payment. In the model above, the issue of instability, coupled with the assumption that the bank makes the first offer, gives the bank enough bargaining power to force the LDC to make its maximum possible payment.
As noted above, the bank will loan the present value of what the LDC can be expected to repay. The preceding paragraph showed that the bank can bargain for a minimum payment stream of (Py + A - c) if the LDC demands rescheduling. Thus, the debt level sustained by the LDC is equal to the present value of these payments, defined in equation (2).
D = (Py + A - c)/r (2)
Given that the difference between GDP and consumption is net exports, equation (2) can be rearranged to show that the debt level that can be sustained is equal to the present value of net exports plus aid.
Results of the theoretical model
The model has several implications for the understanding of debt and aid. First, it shows there is a complementary relationship between aid and debt. Second, the welfare of the LDC is improved by the apparent debt overhang. Third, the apparent debt overhang should not be remedied with forgiveness, unless a shock has changed the ability of the LDC to repay its debts. This last result suggests that there may be "strategic" and "true" default announcements. These results will now be detailed.
It is clear from equation (2) that aid and debt are complementary, and are not substitutes for one another. Higher aid leads to higher debt, and vice versa. However, the aid (and associated debt), is only given to politically important low income countries, as the purpose of the aid is to keep the LDC politically stable. Interestingly, the aid rate is dependent upon the minimum needs of the LDC, rather than its actual political value.
The debt agreement makes the LDC better off for two reasons. Primarily, since the loan forces the industrialized country's government to make aid payments that it would not have otherwise, the debt increases the LDC's overall consumption. In addition, as the LDC values present consumption more than the cost of the interest payments, the loan allows the LDC to shift future consumption to the present at a net higher value.
From the perspective of the industrialized country's government, the LDC has a debt overhang problem. The industrialized country's government is forced to finance the A/r portion of the loan to prevent default. While it would prefer the LDC to borrow a maximum of (Py - c)/r, the industrialized country's government is unable to force such an agreement as it is unable to regulate the lending behavior of all banks in the world. Interestingly, the LDC would be indifferent to making a payment stream of (Py - c) ex post. It is the bank's bargaining power that ensures that the larger contract prevails.
Despite the apparent overhang problem, and calls for debt relief, the industrialized country's government should not offer relief unless a shock affects the LDC's ability to make its payments. [13] If given relief, the LDC will take out new loans to maintain a debt level of (Py + A - c)/r; this debt level allows it to induce the aid payment stream of rate A. Put simply, every dollar of debt forgiveness will result in a new dollar of loans which will be financed by induced aid payments.
Thus, the model also emphasizes a distinction between strategic and true default announcements. [14] The LDC may have to make a strategic default announcement to signal its "overhang" situation to the industrialized country's government to obtain the necessary aid. Alternatively, it may make a default announcement to fool the industrialized country's government into providing relief. Such an announcement does not result in a default, as the default is averted as long as the necessary aid is received. [15]
A true default announcement will occur when any of the determinants of equation (2) (P, y, A, c, r) is shocked in the appropriate direction. [16] Such an announcement will induce a rescheduling agreement coupled with debt forgiveness equal to the full effect of the shock. That is, the bank is forced to forgive debt equal to the entirety of the present value of the income lost because of the shock. This contrasts with the results of standard game theory models, where only partial debt forgiveness is given. For instance, in Bulow and Rogoff (1989a), the LDC initially negotiates to pay only a fraction of its maximum possible payment, so a fall in income induces debt forgiveness equal to a fraction of the lost income. In a stochastic framework, the Bulow-Rogoff loans would be less risky than the loans in the above model.
Evidence suggests that the debt crisis of the 1980s was at least partly due to a rise in interest rates, a fall in the terms of trade for debtor nations, and the fact that much of the debt was not used productively, but instead used for consumption (Sachs 1989). As a result, the default announcements were in fact true defaults. The fact that the Brady Plan (and other rescheduling agreements) resulted in forgiveness lends further support to this hypotheses (Sachs 1990). While the great size of the loans can perhaps be attributed to strategic factors in the model, the defaults themselves are not, and the forgiveness agreements were unavoidable.
This simple model does not describe all issues related to lending and aid. For example, it does not address "aid" that is actually a payment for services, nor does it address the issue of economic growth. This latter shortcoming is important because banks may be willing to issue debt against future expected growth. At the same time, an increase in GDP also decreases the amount of aid received. This may create adverse incentives for growth, making an interesting project for future research. Yet, because of its simplicity, this model is able explore some basic, testable links between debt and aid. In estimating equations (1) and (2), the next section examines these relationships. Before presenting the results, data and methodological issues are addressed.
Empirical Results
Data
To estimate the equations, the 1997 World Development Indicators (WDI) from the World Bank is used. This data set has the advantage of reporting a wide variety of series estimates in a consistent manner. [17] The 1997 WDI reports
data for the period 1970-1995 in nominal U. S. dollars and market exchange rates, but not in purchasing power parity or chain weight index terms. All data are converted to real terms using the U. S. GDP deflator (reported in the WDI, 1987 = 100), and into per capita terms using WDI population estimates. [18] While it is accepted that these market priced data measures do distort the meaning of domestic statistics, it should be noted that there are disagreements over the validity of measures of purchasing power parity (see Srinivasan (1994) for a short critique).
To estimate equations (1) and (2), data are needed to represent debt (D), the trade value of national production (Py), the non-trade value of national production (y), aid (A), consumption (c) for each country-year. Complete sets of data exist for only a few countries for the 1970-1979 period, and as a result, this period was excluded. [19] Unfortunately, not all countries have observations for the full 1980-1995 period, either. All available country-years with complete data were included to allow the number of observations to be as large and diverse as possible. As the model is applicable to low income countries, we report results for three sets of countries: countries with real per capita GDP of less than $1500, $750, and $500. The list of country-years for each income group appears in the appendix.
Total external debt is used to measure debt (D). The debt figure is a present value calculation (by the WDI). [20] It should be noted that some debt in default is likely not reported at true market value, causing an upward bias on the figures. Furthermore, it is not clear that the present value calculation always uses the appropriate interest rate, adversely affecting the measures. Finally, these data do not include some funds borrowed by trans-national corporations for the purposes of direct investment, causing a downward bias to the figures.
To measure the trade value of national production (Py), GDP at market prices is used. There have been criticisms both of using market (rather than purchasing power parity) prices, and GDP itself, but the WDI GDP estimates provide the only estimates consistent with the other variables for each of the years in the study. However, in LDCs, a significant amount of production likely takes place outside of organized markets, placing a downward bias on GDP.
To measure the value of autarkic GDP (y), the GDP figures are adjusted for terms of trade using WDI Net Barter Terms of Trade data. Specifically, GDP is divided by an export price index and multiplied by an import price index. [21] This essentially converts the value of exports to the value of domestic consumption.
The most comprehensive aid (A) series was chosen, but covert and non-financial aid are not included, biasing the figures downward. For example, loan guarantees, preferential tariff rates, and subsidies to businesses that trade with (and work in) the LDC are equivalent to aid, but their value is not measured. On the other hand, some of the reported aid is likely a fee payment for services (and thus an export) rather than true aid, putting an upward bias on the figures. For example, an industrialized country's government might provide aid to an LDC's military to conduct security operations for the industrialized country's government.
Instead of national consumption (c), absorption (the sum of consumption, investment, and government purchases) data are used because it more accurately captures the notion of domestic expenditures, which is represented by consumption in the model (investment and government expenditures are not present in the model). Absorption is likely underestimated in LDCs for the same reason GDP is underestimated. Percentage-wise, the error is likely worse, since exports are more easily measured than absorption. The model predicts that absorption in countries that receive "stability" aid is held at the minimally acceptable level. Unfortunately, it is not possible to test this result, or separate countries that fit this criterion.
Methodology
The equations were tested with a least squares regression. However, there were several econometric problems that needed to be addressed. Because the data set consists of pooled data, heteroscedasticity is present. Heteroscedasticity is the presence of non-constant variance of the residual (error) terms over the observations in the regression results. While this does not affect the estimated coefficients, it makes testing the significance of the coefficients invalid. To address this issue, White's (1980) method of essentially dividing each of the observations by an estimate of the (non-constant) variance of the residuals is used. This method does not change the estimated coefficient values, and allows for hypothesis testing. In addition, because time series data is used, serial correlation is present. Serial correlation is the phenomenon where each error term is correlated with previous observations' error terms, and also invalidates hypothesis testing. To address the serial correlation, dummy (qualitative) v ariables were included in the regression for each year. [22] A dummy variable is simply an independent variable that takes the value of unity for certain specified observations (e. g. year = 1990), and a null value otherwise. Finally, because of the high correlation between GDP and absorption, near multicollinearity must also be considered a possibility. Near multicollinearity is the condition in which two or more variables are closely correlated, resulting in very high error and [R.sup.2] values. However, as will be seen below, neither standard errors nor the [R.sup.2] value are unexpectedly large, suggesting that multicollinearity is not important.
We must also acknowledge the issue that the data set contains observations for which the model is inappropriate. That is, the data set likely includes politically unimportant country-years, and country-years in which the autarkic value of production is sufficient to meet the minimum consumption demands of the public. To partially address this issue, we include dummy variables to represent five geographical regions; Europe, sub-Saharan Africa, North Africa and the Middle East, Central and South America, and Asia (excluding former Soviet Republics). While this will not perfectly filter out the effects of politically unimportant countries, it is impossible to find an objective measure of political importance. [23] To focus on low income countries, we perform separate regressions on countries with per capita income levels below $1,500, $750 and $500. We expect that the most accurate results be for the lowest income group. Summary statistics of all variables and a list of countries in each income group may be fou nd in the data appendix.
Results
Table 1 presents regression results for equation (1) [A = (c - y)]. When aid is regressed on a constant, absorption, and autarkic GDP, the model predicts that the constant be zero, the coefficient for absorption to be equal to 1, and the coefficient for autarkic GDP to be equal to negative 1. Furthermore, it is expected that the model best fits low income countries. Also included in the regression are dummy variables that correspond to time (to correct for serial correlation) and geographical areas. The most recent year, 1995, is the excluded year dummy variable. [24] For the per capital income group of less than $1500, Europe is the excluded geography dummy variable? For the per capital income groups of less than $750 and less than $500, no European countries are in the sample, so Asia (excluding former Soviet republics) is excluded.
For all three income groups, the coefficients on absorption and autarkic GDP are highly significant and have the expected sign. However, the coefficients are all less than 1 in absolute value, and the intercept is highly significant, and negative. This suggests that the model overpredicts the amount of aid that can be extracted. Alternatively, it may be that some countries in this sub-sample are not politically important or do not have the income constraints addressed in the model. The fact that the lowest income group has the best results (with the highest adjusted [R.sup.2] value and largest coefficients) suggests that the model is more descriptive of lower income countries, as expected. The dummy variables for years are significant and positive for the 1986-1993 period. This suggests that there was a change in aid policies in 1986. This corresponds to the period after the Brady plan was initiated, so may represent aid used for debt forgiveness in this period. The dummy variables for all geographic regions are significant, and relatively large; the mean aid per capita figure is approximately $50 for all years. This latter result suggests that there are regional factors that determine aid. These could reflect fees paid for services, or other factors.
Table 2 presents results for estimating equation (2), D = (P y + A- c)/r. Equation (2) predicts that when total external debt is regressed on a constant, GDP, aid, and absorption, the constant is expected to be zero, the coefficients on aid and GDP are expected to be positive and equal to the reciprocal of the interest rate, and the coefficient on absorption is expected to be negative and equal to the reciprocal of the interest rate. As above, also included in the regression are dummy variables that correspond to time (to correct for serial correlation) and geographical areas. Again, the most recent year, 1995, is the excluded year dummy variable, and for the per capita income group of less than $1500, Europe is the excluded geography dummy variable; for the per capital income groups of less than $750 and less than $500, so Asia (excluding former Soviet republics) is excluded.
The signs on the coefficients on aid, absorption, and GDP are highly significant and of the expected sign, and the intercept is not significant. The signs are also consistent with the Bulow and Rogoff (1989c) model, which predicted that debt is related to the consumers surplus derived from trade (net exports is the difference between GDP and absorption). However, the coefficients are again much smaller in magnitude then expected for the model in this paper. Since the coefficients are the reciprocal of the interest rate, the results imply an interest rate of 25% or higher, which is likely too large, even after considering risk premia. [25] Thus, the model apparently overestimates the true debt level for LDCs. The results are again best for the lowest income group, suggesting that the model fits low income countries better. Since the Bulow-Rogoff (1 989c) model links debt level to the size of trade, this result also suggests that the model in the paper is more appropriate than the Bulow-Rogoff model for analyz ing debt of developing countries. The year dummy variables are significant and negative in the period of the early-mid 1980s. Given that the year dummy variables on aid were positive and significant for the 1986-1993 period, this is to be expected. The model predicts that as aid rises (as in the mid-late 1 980s), so does debt. The geography dummy variables for Sub-Saharan Africa, North Africa and the Middle East, and Latin America are again significant and large (debt per capita ranges from $285 for the lowest income group to $468 for the highest income group). The significant geography dummy variable coefficients imply different regions receive debt at different proportions to extracted aid. This could be due to differing expectations of future growth in different regions, differing expectations of the variance of growth (or other key variables), differing amounts of debt guarantee agreements (which substitute for aid), region-specific measurement problems, or other issues.
These results likely reflect measurement error of the variables. As noted above, the model applies to politically important LDCs; however, the coefficients represent politically important and unimportant LDCs. In addition, the model specifies that consumption (absorption) be at the minimally acceptable level; for those countries for which absorption is above this level, the coefficient on absorption would be expected to be smaller in absolute value. Furthermore, if aid is underestimated by the WDI, this may explain the small coefficients on autarkic GDP and absorption in equation (1). However, if aid were measured correctly at a larger value, the coefficient on aid in equation (2) would likely fall, making the estimates worse. In addition, properly estimating GDP and absorption would likely result in smaller coefficients, as they are likely underestimated by official statistics.
Conclusion
This paper developed a model that adds a new dimension to the literature related to external debt and aid to developing countries. The novel features of the model are the assumptions that (1) political stability is compromised when national income falls too low, and (2) industrialized country governments are willing to give aid to ensure stability. We show that under these circumstances, politically important, low income LDCs can extract aid to service external debts, by engaging in a debt contract. The model differentiates between strategic and true default announcements, and suggests that even though LDCs have persistent debt overhang problems, debt forgiveness should only be given when the ability to make existing payments is compromised. There is also a general result that aid and debt are complements for one another, and are not substitute forms of financial assistance.
When the model is fitted to the data, it is found that the model overstates the true level of aid and debt in the developing world. Yet, the regression coefficients do have the predicted signs, the strongest results are for the lowest income group, as expected. Furthermore, while not showing causality, the results suggest that indeed, aid and debt are complements.
(*.) Assistant Professor of Economics, Department of Economics, College of the Holy Cross, Worcester, MA 01610. Telephone: 508/793-2682, fax: 508/793-3708. E-mail: mcahill@holycross.edu
(**.) Assistant Professor of Economics, Department of Economics, Grand Valley State University, Grand Rapids, MI 49504. Telephone: 616/336-7418, fax: 616/366-7445. E-mail: iselyp@gvsu.edu
The authors gratefully acknowledge John F. O'Connell, John R. Carter, Nicolas Sanchez, Jacob Paroush, Norman C. Miller, participants of a College of the Holy Cross research seminar, and an anonymous referee for helpful comments. Miles Cahill acknowledges financial support for a College of the Holy Cross Bachelor (Ford) Summer Faculty Fellowship.
NOTES
(1.) See Eaton (1993) for a review of debt models.
(2.) See Raffer and Singer (1996) for a review of aid issues.
(3.) See Eaton (1993) and Cline (1995) for reviews of this literature.
(4.) To avoid the problems of default, Fafchamps (1996) suggests that conditionality (e. g. conditions placed on the use of funds, or policies made by the borrowing nation) can be used to insure debt repayment. However, the majority of recent literature has focused on the costs of default as deterrence, and that is our focus here.
(5.) There is disagreement about the effectiveness of reputation in enforcing lending agreements. See Bulow and Rogoff (1989b) and Rosenthal (1991) for general models in which reputation alone will not impose discipline, and see Eaton (1993) and Thomas (1992) for counterexamples.
(6.) For example, Eaton and Gersovitz (1981) use a reputation cost, Cohen (1991) assumes that solvent debtors will always repay, and Fernandez and Rosenthal (1990) assumes that the banks have all the bargaining power.
(7.) See for instance Rosenthal (1991), Diwan (1989) and Bulow and Rogoff (1989a), (1 989c).
(8.) Another more recent paper showing rational lending in the presence of default is Egli (1997). This model shows a situation where the willingness of the LDC to make payments increases with debt, when debt is used purely for investment.
(9.) As further support, Aggarwal (1996) notes that debtor governments consider the possibility of domestic turmoil when choosing bargaining strategies. If payments are too high, the debtor government may lose power through election or revolution. Kaminsky and Pereira (1996) formulate a "political stability index" to examine the impact of political instability on sovereign default, but conclude it contains too many problems to be trusted. Siermann and de Haan (1993) show using a similar index that political instability is often the cause of unsustainable fiscal policy.
(10.) See Chase et al (1996) for a discussion of countries that are politically important to the U.S.
(11.) As there is no uncertainty, there is no risk premium on the loan.
(12.) This is the operative use of the subgame perfect assumption; "cheap talk" is not credible.
(13.) There is a literature that suggests that debt forgiveness may be Pareto improving; see for example, Sachs (1988), Froot (1989) and Egli (1997). In each of these models, debt is so large that it is in the best interests of the borrower to default. The defaulting country then changes its allocations of resources to minimize the impact of sanctions. In each case, debt forgiveness results in situations where the borrower makes payments, and does not default, so the borrower once again allocates its resources efficiently. In the model above, production in the developing country is fixed, so these factors are not applicable.
(14.) Note that both of these types of defaults are present in other models, including Bulow and Rogoff (1989a) and (1989c).
(15.) One possible example of such a strategic announcement is the aid and new loans Russia recently received after announcing defaults. The April 30, 1996 issue of the Wall Street Journal reported that after Russia pushed for long term rescheduling, the world's industrialized nations signed a debt relief package. The package included giving a six year grace period on principal payments due to industrialized country governments (essentially aid), rescheduling $40 billion in government-held debt over 25 years, and a $10.2 billion loan from the International Monetary Fund. Separately, Russia rescheduled $32.5 billion in privately held debt. In response, Russian officials were reported to have said they hope to issue as much as $2 billion in Eurobonds later in the year. The agreements were clearly designed to help the incumbent Boris Yeltsin win the next presidential election, and prevent the neo-Communists from taking power. (Liesman 1996 p. A15)
(16.) Note that new loans are granted if the shocks are in the opposite direction. Note also that unlike in the Bulow and Rogoff (1989a) and (1989c) models, the serviceable size of the loan in our model does not depend on [delta], and the size of the loan payment does not depend on r. In Bulow and Rogoff (1989a, 1989c), these factors determine the relative impatience of each side in the negotiations, and thus bargaining power. In our model, the guarantee of aid makes the LDC and industrialized country's government indifferent to a failure of the negotiations. Yet, all bargaining power rests with the bank, because the bank makes the first offer.
(17.) The World Bank is a data source commonly used by researchers when studying developing countries. When looking at debt, researchers often use the World Debt Tables, also published by the World Bank. The data for debt and aid that is relevant to our study appears in both the World Debt Tables and the WDI. For an empirical study using World Bank data, see Amoateng and Amoako-Adu (1996), which also gives a concise review of the empirical literature linking economic growth with export growth and debt servicing.
(18.) Henceforth, any reference to a variable will refer to its real, per capita value.
(19.) Many data series are not available for many countries in the 1970-1974 period, and terms of trade date are not available for 1975-1979. We did estimate equation (2) using 1975-1995 data, and the results were not significantly different from those of the 1980-1995 period.
(20.) Bulow and Rogoff (1991) use the corresponding data series from the World Debt Tables.
(21.) Sachs (1985) uses similar data to examine terms of trade shifts (in relation to debt problems).
(22.) A very similar method was used by DemirgucKunt and Detragiache (1994) Because the data is pooled, this is a superior method to using an autoregressive or Cochrane-Orcutt procedure. These procedures would essentially transform the data to a (quasi) first difference model, where for example, instead of regressing [Y.sub.t], on [X.sub.t], [Y.sub.t] -- [rho][Y.sub.t-1] is regressed on [X.sub.t]-[rho][X.sub.t-1], where [rho] is an estimated constant. This is inappropriate for pooled data, as [X.sub.t-1] may not be the previous date's value of [X.sub.t], but may be another country's value.
(23.) Demirgtic-Kunt and Detragiache (1994) use a similar method to account for country differences. It is difficult to define an objective measure of political importance, especially with respect to all LDCs and all industrialized countries. Chase et al (1996) reports a list of politically important nations to the U. S., but it cannot be considered an exhaustive list.
(24.) No dummy variable was used for 1995 or Europe because the inclusion of a dummy variable for every observation results in the dummy variable set replicating the constant term of the regression (the "dummy variable trap"), making a least squares regression impossible.
(25.) See Demirguc-Kunt and Detragiache (1994) for LIBOR and interest rates faced by highly indebted countries. The highest interest rate in the 1973-1989 period was never as high as 20%.
References
Aggarwal, V., 1996, Debt games: strategic interaction in international debt rescheduling, (Cambridge University Press, New York).
Amoateng, K. and B. Amoako-Adu, 1996, Economic growth, export and external debt causality: the case of African countries, Applied Economics 28, 21-27.
Bulow, J. and K. Rogoff, 1991, Sovereign debt repurchases: no cure for overhang, The Quarterly Journal of Economics 106, 12 19-1235.
Bulow, J. and K. Rogoff, 1989a, A constant recontracting model of sovereign debt, Journal of Political Economy 97, 155-78.
Bulow, J. and K. Rogoff, 1989b, LDC debt: is to forgive to forget?, American Economic Review 79, 43-50.
Bulow, J. and K. Rogoff, 1989c, Multilateral negotiations for rescheduling developing country debt: a bargaining-theoretic framework, in J. A. Frenkel, M. P. Dooley, P. Wickham (eds.), Analytical issues in debt, International Monetary Fund, Washington, 194-207.
Chase, R. S., E. B. Hill, and P. Kennedy, 1996, Pivotal states and U. S. strategy, Foreign Affairs 75, 33-51.
Cline, W. R., 1995, International debt reexamined, Institute for International Economics, Washington.
Cohen, D., 1991, Private Lending to Sovereign States, MIT Press, Cambridge, MA.
Demirguc-Kunt, A. and E. Detragiache, 1994, Interest rates, official lending, and the debt crisis: a reassessment, Journal of Development Economics 44, 263-285.
Diwan, I., 1989, Foreign debt, crowding out, and capital flight, Journal of International Money and Finance 8, 121-136.
Eaton, J., 1993, Sovereign debt: a primer, The World Bank Economic Review 7, 137-172.
Eaton, J. and M. Gersovitz, 1981, Debt with potential repudiation: theory and estimation, Review of Economic Studies 48, 289-309.
Egli, D., 1997, Optimal debt relief under threat of trade punishments, Review of International Economics 5, 272-83.
Fafchamps, M., 1996, Sovereign debt, structural adjustment, and conditionality, Journal of Development Economics 50, 3 13-35.
Fernandez, R. and R. Rosenthal, 1990, Strategic models of sovereign-debt renegotiations, The Review of Economic Studies 56, 331-49.
Froot, K., 1989, Buy-backs, exit bonds, and the optimality of debt and liquidity reliefs, International Economic Review 30, 49-70.
Kaminsky, G. L. and A. Pereira, 1996, The debt crisis: lessons of the 1980s for the 1990s, Journal of Development Economics 50, 1-24.
Liesman, 5., 1996, Russia's $40 billion debt is restructured, The Wall Street Journal 227, A15.
Raffer, K. and K. W. Singer, 1996, The foreign aid business: economic assistance and development co-operation, Edward Elgar, Brookfield, Vermont.
Rosenthal, R., 1991, On the incentives associated with sovereign debt, Journal of International Economics 30, 166-76.
Rubenstein, A., 1982, Perfect equilibrium in a bargaining model, Econometrica 50, 97-109.
Sachs, J., 1990, Making the Brady Plan Work, Foreign Affairs 68, 87-104.
Sachs, J. 1989, Introduction, in Jeffrey Sachs, (ed.), Developing country debt and the world economy, University of Chicago Press, Chicago, 1-33.
Sachs, J., 1988, The debt overhang of developing countries, in Ronald Findlay (ed.), Debt, Growth and Stabilization: Essays in Memory of Carlos Diaz Alejandro, Blackwell, Oxford, 80-102.
Sachs, J., 1985, External debt and macroeconomic performance in Latin America and east Asia, Brookings Papers on Economic Activity 2, 523-573.
Siermann, C. L. J. and 3. de Haan, 1993, On sustainability and political determinants of government debt in developing countries, Rivista Internazionale di Scienze Economiche e Commerciali 40, 81-92.
Srinivasan, T. N., 1994, Human development: a new paradigm or reinvention of the wheel?, American Economic Review (Papers and Proceedings) 84, 238-243.
Thomas, J. P., 1992, Sovereign Debt: Ignorance can be bliss, Journal of Development Economics 39, 389-96.
White, H., 1980, A heteroscedasticity-consistent covariance matrix and a direct test for heteroscedasticity, Econometrica 48, 817-838.
World Bank, World debt tables, various years.
World Bank, 1997, The world development indicators 1997 CD-ROM, Win*Stars version 4.0 (Print version available under title World Development Indicators).
Data appendix
Per capita income of $1,500 or less
Country-years
Angola (1985-1993), Burundi (1985-1993), Benin (1980-1993), Burkina Faso (1980-1993), Bangladesh (1980-1993), Bulgaria (1991-1993), Bolivia (1980-1993), Botswana (1980-1987), Central African Republic (1980-1993), Chile (1985), China (1982-1992), Cote d'Ivoire (1981-1993), Cameroon (1980-1993), Congo (1980-1993), Colombia (1983-1994), Comoros (1980-1993), Cape Verde (1980-1993), Costa Rica (1981-1983), Dominican-Republic (1982-1993), Ecuador (1984-1993), Egypt, Arab Rep. (1980-1993), Ethiopia (1981-1993), Fiji (1988), Ghana (1980-1993), Guinea (1986-1993), The Gambia (1980-1993), Guinea-Bissau (1982-1993), Equatorial Guinea (1987-1993), Guatemala (1982-1993), Guyana (1980-1985, 1992, 1993), Honduras (1980-1993), Haiti (1980-1993), Indonesia (1981-1993), India (1980-1994), Jamaica (1984-1988, 1991), Jordan (1989-1994), Kenya (1980-1995), Lebanon (1989-1992), Liberia (1981-1986), Sri Lanka (1980-1995), Morocco (1980-1994), Madagascar (1980-1993), Mali (1980-1993), Mozambique (1980-1993), Mauritania (1980-1993), Mauritius (1981-1986), Malawi (1980-1993), Niger (1980-1993), Nigeria (1981-1993), Nicaragua (1980-1993), Nepal (1980-1993), Pakistan (1980-1995), Peru (1983-1993), Philippines (1980-1993), Papua New Guinea (1980-1993), Poland (1990), Paraguay (1984-1993), Romania (1990-1993), Rwanda (1980-1993), Sudan (1980-1991), Senegal (1980-1993), Solomon Islands (1980-1988), Sierra Leone (1980-1993), El Salvador (1980-1993), Somalia (1980-1990), Sao Tome and Principe (1981-1993), Swaziland (1980-1993), Syrian Arab Republic (1986-1991), Chad (1980-1993), Togo (1980-1993), Thailand (1980-1991), Tunisia (1983-1991, 1993), Turkey (1983-1985), Tanzania (1981-1993), Uganda (1980-1993), Republic of Yemen (1990-1993), Zaire (1982-1989), Zambia (1980-1993), and Zimbabwe (1980-1993).
Per capita income of $750 or less
Country-years
Angola (1991-1993), Burundi (1985-1993), Benin (1980-1993), Burkina Faso (1980-1993), Bangladesh (1980-1993), Bolivia (1982-1993), Central African Republic (1980-1993), China (1982-1993), Cote d'Ivoire (1985, 1991-1993), Comoros (1980-1993), Cape Verde (1980-1993), Dominican Republic (1985, 1988), Egypt, Arab Rep.(1981-1993), Ethiopia (1981-1993), Ghana (1980-1993), Guinea (1986-1993), The Gambia (1980-1993), Guinea-Bissau (1982-1993), Equatorial Guinea (1987-1993), Guatemala (1990), Guyana (1982-1985, 1992, 1993), Honduras (1989-1993), Haiti (1980-1993), Indonesia (1982-1993), India (1980-1994), Kenya (1980-1995), Lebanon (1989-1990), Liberia (1981-1986), Sri Lanka (1980-1995), Morocco (1984-1985), Madagascar (1980-1993), Mali (1980-1993), Mozambique (1980-1993), Mauritania (1980-1993), Malawi (1980-1993), Niger (1980-1993), Nigeria (1986-1993), Nicaragua (1989-1993), Nepal (1980-1993), Pakistan (1980-1995), Philippines (1983-1993), Papua New Guinea (1985, 1990), Rwanda (1980-1993), Sudan (1980-1991), Senega l (1981-1993), Solomon Islands (1980-1988), Sierra Leone (1980-1993), Somalia (1980-1990), Sao Tome and Principe (1980-1993), Swaziland (1985-1986), Chad (1980-1993), Togo (1980-1993), Tanzania (1981-1993), Uganda (1980-1993), Republic of Yemen (1990-1993), Zaire (1982-1989), Zambia (1982-1993), and Zimbabwe (1984-1993).
Per capita income of $500 or less
Country-Years
Angola (1991-1993), Burundi (1985-1993), Benin (1981-1993), Burkina Faso (1980-1993), Bangladesh (1980-1993), Central African Republic (1980-1993), China (1983-1993), Comoros (1981-1993), Cape Verde (1981-1986), Ethiopia (1981-1993), Ghana (1981-1993), Guinea (1986-1993), The Gambia (1981-1993), Guinea-Bissau (1982-1993), Equatorial Guinea (1987-1993), Haiti (1980-1993), Indonesia (1986-1988), India (1980-1994), Kenya (1982-1995), Liberia (1986), Sri Lanka (1981-1993), Madagascar (1982-1990), Mali (1980-1993), Mozambique (1981-1993), Mauritania (1984-1993), Malawi (1980-1993), Niger (1981-1993), Nigeria (1987-1993), Nicaragua (1989-1992), Nepal (1980-1992), Pakistan (1980-1995), Rwanda (1980-1993), Sudan (1980, 1982-1984, 1988-1991), Senegal (1983-1985), Sierra Leone (1980-1993), Somalia (1980-1990), Sao Tome and Principe (1981-1985, 1988-93), Chad (1980-1993), Togo (1981-1992), Tanzania (1982-1993), Uganda (1980-1993), Republic of Yemen (1990-1993), Zaire (1983-1989), Zambia (1984-1993), and Zimbabwe (1991-1993).
 Printer friendly
 Cite/link
 Email
 Feedback
| |
| Author: | Cahill, Miles B.; Isely, Paul N. |
|---|---|
| Publication: | American Economist |
| Date: | Sep 22, 2000 |
| Words: | 7647 |
| Previous Article: | AN EMPIRICAL MODEL OF VOTING BEHAVIOR IN THE BULGARIAN PARLIAMENTARY ELECTIONS OF 1994. |
| Next Article: | THIRD DEGREE PRICE DISCRIMINATION: A PROFITABLE ARBITRAGE CASE, A NOTE. |