SOVEREIGN DEBT OWED TO PRIVATE CREDITORS grew substantially over the 1990s, but official debt (multilateral and bilateral lending) remains significant—and constitutes more than 40 percent of the total sovereign external developing country sovereign debt (Figure 6.1).[1] Moreover, multilateral debt, which accounts for 28 percent of total developing country sovereign external debt, remains important not only for its sheer size, but also because different theories regarding the role of official creditors, and multilaterals in particular, suggest that their influence extends well beyond the cash they provide. Hence, while it is a commonplace to note the declining market share of multilaterals in capital flows, they remain important players for sovereign debt management.

Multilateral lenders consist of the International Monetary Fund, the World Bank, the regional development banks (RDBs, which include the Inter-American Development Bank), and other smaller institutions. The multilaterals each have different mandates in accordance with their charters or articles of agreement, and this affects their lending policies. For example, the IMF’s mission of supporting adjustment to external payments imbalances should imply a specific pattern of lending flows, heavily influenced by the external financial position of its member countries. The World Bank seeks to enhance development in lower-income countries and to eradicate world poverty, and thus its flows should be determined by longer-term strategies and to a lesser extent by current financing needs. The RDBs each have a particular set of objectives closer in spirit to those of the World Bank. The World Bank Group and RDBs including the IDB also lend to the private sector, an area which is growing in importance, but the majority of the business of most multilateral development banks remains sovereign lending.[2]

Multilateral lenders offer concessional finance and aid as well as finance at nonconcessional rates—although in general these rates are still lower than market rates paid on commercial debt. Concessional finance is offered to countries in accordance with particular conditions related to per capita income and the level of national development. Official lending also includes bilateral lending through national development agencies such as the United States Agency for International Development (USAID), the United Kingdom’s Department for International Development (DFID), and Germany’s Kreditanstalt für Wiederaufbau (KfW) and Deutsche Investitions- und Endtwicklungsgesellschaft (DEG). The majority of this debt is concessional in nature. However, bilateral lending also includes government-guaranteed loans extended by export credit agencies such as Hermes (Germany), Companía Española de Seguros y Créditos a la Exportación (CESCE, Spain), Coface (France), and the United States Export-Import Bank. Lending by these agencies may be nonconcessional. For example, export credit guarantees granted by these institutions to developing country governments or to private borrowers with official backing are included in sovereign lending and may be on nonconcessional terms.

BROAD TRENDS IN SOVEREIGN EXTERNAL DEBT

Total long-term developing country sovereign external debt (excluding that financed by the IMF) rose sharply during the 1990s but since the late 1990s has been flat at about $1 trillion (Figure 6.2) (see Chapter 2 for a discussion of the evolution of external debt in Latin America). Sovereign external debt held by private creditors in the 1980s was largely held by commercial banks, but through the 1990s, in part as a result of the Brady restructurings and then through subsequent new issues, debt in the form of bonds grew substantially. Bonds only represented some 6 percent of sovereign debt owed to private creditors in 1989 but now represent as much as 46 percent of total sovereign external long-term nonconcessional debt.[3]

At the same time, within nonconcessional debt it is notable that multilateral debt has increased relative to bilateral debt in recent times. In 1991, bilateral nonconcessional debt accounted for about 20 percent of total debt, and it reached a peak of about 25 percent in 1994–1995 with the financial support offered to alleviate the Mexican crisis. It has now fallen back to about 14 percent. Multilateral debt represented only 11 percent of the total in 1984, was about 18 percent of the total in 1991, and has now risen to about 21 percent of the total. In the last decade or so, then, multilateral debt has lost market share relative to private debt but has gained it relative to bilateral debt, and the net effect has been if anything a slight rise in market share (the trends are roughly identical when Latin America is considered, as opposed to all developing countries; see Chapter 2).

Country recipients of multilateral finance are highly concentrated. The top 10 recipients accounted for 58 percent of the stock of multilateral lending as of 2004. However, these countries also accounted for about 69 percent of recipient GDP. But within this group there are very different cases. China, for example, accounts for 30 percent of recipient GDP but only 7.4 percent of the stock of multilateral lending.4 And India accounts for 13 percent of recipient GDP and only 3.5 percent of multilateral lending. By contrast, the remaining 8 countries in the top 10 account for 47 percent of multilateral lending but only 25 percent of recipient GDP.5 While these 8 countries then appear to account for a disproportionate amount of multilateral debt stocks according to this measure (and China and India too little), it is interesting to note that they also account for 51 percent of external private sector lending to recipient sovereigns.

Considering Latin America specifically, the concentration of multilateral finance mirrors the size of recipient economies more closely. The largest seven economies in the region accounted for 90 percent of recipient GDP and received 80 percent of the stock of multilateral lending as of 2004. Chile is something of an exception, accounting for 4 percent of recipient GDP but receiving only 1 percent of multilateral lending. The stock of multilateral financing also mirrors closely that of external private financing, with the seven largest economies accounting for some 87 percent of total private external lending to the region.

Figure 6.3 illustrates the importance of multilateral financing in Latin America (panel a) and the importance of concessional finance within multilateral finance (panel b) as of 2004. Panel (a) shows that the World Bank tends to provide a large share of lending to the large countries in the region, and the RDBs concentrate more on the smaller countries. For the largest seven countries, the World Bank provides 39 percent of multilateral finance, ranging from 9 percent in Venezuela to 56 percent in Mexico.6 For the smaller Latin American and Caribbean countries, the World Bank’s average market share of multilateral lending is 24 percent, with the range from 4 percent in Barbados to 51 percent in Haiti.

Multilateral finance is particularly important for the poorer countries of the region, including Bolivia, Haiti, and Guyana, and less important for countries such as Venezuela, Argentina, Panama, Brazil, and Mexico. And for those countries in which multilateral finance is particularly important, concessional finance tends to be an important component of multilateral lending. In fact, total concessional finance (to all developing countries) increased considerably over the 1980s and jumped further in 2003 to reach the current stock of about US$450 billion (Figure 6.4). Note that the majority of concessional finance comes from bilateral donors—about 62 percent in 2004—although multilaterals have been gaining market share. In 1990, bilaterals accounted for about 75 percent of the stock of concessional financing, and the figure was 80 percent in 1980. As Figure 6.5 shows, in Latin America, concessional finance peaked in dollar terms in 1995 at just under US$40 billion. Again, multilaterals have increased their market share considerably vis-à-vis bilaterals since the early 1990s, with that share reaching 44 percent in 2004 (debt relief may affect these figures significantly).[7]

There is a small literature on the multilaterals and official lenders more generally. A survey of Econlit (a comprehensive electronic catalogue of papers published in economic journals) reveals some 924 references to the IMF from 1969 to the present; about 52 percent of all references to the international financial institutions are about the IMF.8 However, the catalogue contains only 100 references in all to the regional development banks. In fact, there is almost no research on what makes multilateral lenders different or special or on how they may differ from private lenders or bilateral lenders.9 And hence from this point on, the chapter focuses on multilateral lenders, specifically the multilateral development banks, excluding IMF lending.

DIFFERENCES BETWEEN MULTILATERAL AND PRIVATE LENDERS

What makes multilateral lending different from private lending? First and foremost, multilaterals and private sector lenders have different objectives. The private sector is motivated by profits, whereas multilateral lenders have a general objective of promoting development and social welfare in the countries that borrow from them. This may lead multilaterals to lend more in support of development projects, to lend in riskier environments, and to lend more in hard times relative to private lenders. This characterization also supports the view that multilaterals may act countercyclically, while it is likely that private lenders would be procyclical (Ratha, 2001).

Second, it has been argued that multilateral lending has a greater potential to become a victim of political influence than does private lending. Most multilaterals are governed by boards of directors controlled by the richest countries.10 This composition may be reflected in lending priorities. Barro and Lee (2005) suggest that IMF loan frequency is affected by country voting rights in the IMF and the alignment of countries with the United States in terms of voting patterns in the UN assembly and trade patterns. This analysis is extended in Bobba (2004), which also finds support for politics in IMF lending. Faini and Grilli (2004) argue that the pattern of World Bank and IMF lending is affected by shareholders’ commercial relations.

Third, from the standpoint of the borrower, multilateral loans may be considered more costly in terms of red tape and conditionality (when the borrower sees the latter as a burden) but cheaper in terms of the interest rate charged. Borrowing from a multilateral generally involves detailed discussions about the intended use of the funds, conditions regarding promised economic reforms or other matters, and extended negotiations on many details of both the loan and possibly the macroeconomic environment.[11] On the other hand, a bond may have a higher interest rate than a loan from a multilateral, and while there are certainly administrative and legal costs associated with a bond issue that have to be paid, these may be less demanding in terms of time of senior officials. As a consequence, when countries have access to relatively inexpensive private funds, for example, in periods of abundant global liquidity, officials may prefer to borrow from the private sector, but when conditions are less favorable and private sector interest rate spreads rise, the additional red tape and conditionality is worth enduring to obtain the lower costs associated with borrowing from multilaterals. In the limit, in times of severe market dislocation, multilateral funds may be the only ones available.

Fourth, it has been argued that an important difference between a private lender and a multilateral is the structure of information (Rodrik, 1995). Information problems have been stressed as explanations of both excessive private lending and sudden stops and—potentially—enhanced sensitivity to fundamentals on the part of private lenders.[12] A close relative to these information issues is coordination problems between individual lenders. This may give rise to such crises as multiple equilibria phenomena, as in the classic models of bank runs. Multilaterals may be presumed to be exempt from such behavior.

A fifth difference is seniority. Multilateral institutions enjoy the status of a preferred creditor, which grants them legal priority over private creditors. (Note that this does not extend, in general, to bilateral official loans.) The interaction between official and private debts in times of debtor distress is, however, a complex one, as official loans are usually made available at such times, while short-term private financing may be withdrawn (Jeanne and Zettelmeyer, 2001; Demirgüç-Kunt and Fernández-Arias, 1992).

The preceding discussion suggests that multilateral and private flows are likely to affect each other. An attractive feature of multilateral lending is that it may be catalytic, namely, that it may provide incentives for private investors to lend to the country as well. Rodrik (1995) suggests that multilaterals may have better information on the economic fundamentals in a particular borrowing country and rationalizes their lending as “putting their money where their mouth is.” In the absence of such lending, statements from the multilaterals regarding the good health of a particular economy may not be considered credible. Multilateral lending is then seen as a signal to enhance the generally poor information available to private lenders.[13] An alternative view that has not been explored in the existing literature is that causality may run in the opposite direction. That is, private flows may affect the amount of multilateral flows that follow in later years. Two theories that support this alternative are as follows. The first is that a negative effect of private flows on multilateral flows could be due to the fact that countries that obtain ample private finance graduate from multilateral lending and hence should see the share of multilateral lending in total lending diminish over time. A second theory, which would instead predict a positive correlation between private and multilateral flows, might be that countries attract private flows precisely because they have enacted reforms that also attract multilateral interest. A third, more political view might be that countries that are borrowers on private markets, and hence are highly integrated into world financial markets, have greater negotiating power or have become “too big to fail” and hence also attract more multilateral financing, especially when times are hard.

MODELING MULTILATERAL LENDING

The foregoing discussion suggests a set of interesting questions: do multilaterals lend to the same countries as the private sector, and if not, what might explain the differences? Is multilateral lending explained by economic variables: do countries graduate from multilateral lending such that private lending is a substitute for multilateral, or are the two lending sources complements, and does multilateral lending decline as GDP per capita rises? Is multilateral lending pro- or countercyclical, and does GDP per capita matter for nonconcessional multilateral lending? Is multilateral lending influenced by political factors? What happens to multilateral lending when countries default? In times when countries have access to private finance, does demand for multilateral lending decline as world interest rates fall?

The group of four graphs in Figure 6.6 illustrates multilateral and private flows to developing countries worldwide over two time periods. Panel (a) illustrates that multilateral flows rose substantially after 1976 as per capita GDP growth fell, reaching a peak in 1983 when per capita growth fell to a low of just over 1 percent in developing countries. Panel (b) shows a similar pattern in the 1990s. Multilateral flows rose in 1992 and peaked in 1993, while per capita GDP growth fell to a low in 1992 and then recovered strongly beginning the next year. In the remainder of the 1990s, growth remained high and multilateral flows low, although there is substantial volatility in the series. At the end of the time series depicted in the panel, in 2003, per capita GDP growth picks up again, and multilateral flows fall.

Panels (c) and (d) of Figure 6.6 illustrate private sector sovereign lending flows over the period. As the panels show, private flows are much more procyclical than multilateral flows in the same period. In the period covered in panel (c), private flows fall sharply as growth falls, although they do not pick up as growth recovers, surely reflecting the fact that many countries, especially in Latin America, were in default over this period. In panel (d), covering the 1990s, private flows rose strongly as growth increased. Going beyond this graphical analysis is complicated by the dynamic nature of loans, as a loan extended in one year has consequences for future flows and for future stocks. Fernández-Arias and Powell (2006) use dynamic panel techniques designed to handle this dynamic nature. Their main findings in regard to the factors influencing patterns of multilateral financing can be summarized as follows (Figure 6.7):[14]

1. Lower per capita GDP growth is associated with larger multilateral lending, suggesting that multilaterals’ financing flows are countercyclical.
2. There is some mild support for the notion that countries that are more politically aligned with the United States receive a higher share of multilateral loans as a percentage of GDP, but the coefficient is only marginally statistically significant.
3. There is no support for the idea that closer economic ties to the United States bring about higher multilateral lending.
4. Countries that receive more private sector lending also receive more multilateral finance.[15] Hence, there is no evidence of “graduation” in these estimates, in the sense that the results indicate that multilateral lending and private lending have been complements, rather than substitutes.
5. The coefficient on the indicator for being in default is highly significant, suggesting that countries in default, and hence lacking access to private markets, receive more multilateral finance.
6. In nondefault periods, when U.S. interest rates rise, countries appear to draw more multilateral financing, indicating that when countries do have access to private markets, the opportunity cost of borrowing on such markets becomes an important determinant of the choice of financing source. As world interest rates fall, countries borrow less from multilaterals.

The preceding analysis provides salient information about patterns in multilateral lending, but it still does not answer the question, how is multilateral lending different from private lending? If multilaterals behave like private creditors, then the explanatory variables should affect both sources of lending in the same way. Any deviation from such a pattern would then be suggestive of a differential effect of the explanatory variables. To filter out commonalities, Fernández-Arias and Powell (2006) conduct an analysis of the difference in lending shares of private and official creditors. Figure 6.8 illustrates the main results. The negative coefficient on the indicator of GDP growth suggests that multilaterals are more countercyclical than private creditors in their lending. There is also a negative coefficient on total external debt, indicating that as debt rises, multilaterals tend to lend less than the private sector. However, the positive coefficient on the private default variable indicates that multilateral financing is made available when countries do not have access to private creditors.[16]

ARE MULTILATERALS CATALYTIC?

If multilaterals have an information advantage over smaller, uncoordinated private lenders, multilateral lending may provide a “seal of approval” type of signaling effect (Rodrik, 1995). Fernández-Arias and Powell (2006) test this idea by examining whether private lending flows are affected by multilateral flows and a set of other explanatory variables (Figure 6.9).[17] They find evidence that multilateral flows do indeed crowd-in private flows. The lagged multilateral flow variable in their study is statistically significant with a positive coefficient, although the significance level varies in alternative specifications. One possibility is that the positive coefficient arises from waves of privatization or liberalization reforms and hence higher inflows of both private and multilateral money. However, introducing indicator variables for such effects does not change the results, and the indicators themselves are not significant. Growth also has a positive coefficient, suggesting that private flows are procyclical, and being in default has a negative coefficient, highlighting the previously discussed differences with respect to multilaterals. Overall, Fernández-Arias and Powell’s conclusion is that the catalytic effect of multilateral flows on private lending flows is robust to alternative specifications. Naturally, the result is open to different interpretations; three possibilities are (1) the story advanced by Rodrik: that multilaterals have superior information and signal good housekeeping; (2) that multilaterals actually promote reforms (that are not picked up in other variables in the regression), enhancing the investment climate; and (3) that multilaterals facilitate private sector lending through other channels, for example, by improving infrastructure and the availability of human capital.

In sum, the empirical results show that there are indeed significant differences between multilateral (nonconcessional) flows and private flows. There is evidence that the private sector is procyclical, whereas multilaterals are countercyclical, with respect to recipient country growth. Multilaterals tend to increase their exposure during periods when a borrowing country is in default, when private sector flows are reduced. Recipient countries tend to reduce their borrowing from multilaterals when world interest rates are low and increase them when these rates rise. There is only weak and not particularly robust evidence that politics affects multilateral lending. There is evidence that multilaterals catalyze private sector flows.

DIFFERENCES BETWEEN MULTILATERAL AND BILATERAL LENDERS

Total official development assistance (ODA) to developing countries grew rapidly over the 1970s and 1980s and then stabilized in the mid-1990s, with a slight falloff at the end of that decade, followed by another increase in the early 2000s (Figure 6.10). Bilateral aid is an important component of ODA, as more than 70 percent of the total consists of bilateral aid from 22 OECD countries. Development assistance is less important for Latin America relative to other regions, although it is important for some individual Latin American countries.[18] Again the majority of development assistance (about 75 percent of the total) offered to Latin America comes from bilateral sources.

Rodrik (1995), in considering the question of why multilateral lending exists, asks what would drive a creditor country to lend through the intermediary of a multilateral rather than directly to a recipient country. He suggests that multilaterals may have a comparative advantage over individual countries in establishing and monitoring appropriate conditionality. This is important because recent research suggests that the macroeconomic policy environment in a country may be important for making aid to that country effective.[19] Multilateral loans also enjoy a preferred creditor treatment that bilateral loans do not receive. Moreover, a country may attempt to selectively default on one bilateral, hoping that relations with others will not be affected. The Paris Club, although only an informal group, attempts to avoid these problems by reaching consensus among creditor nations.

The issue of coordination among bilateral lenders is also an important consideration. It is likely that $1 million in aid administered through one agency will be more effective than the same sum administered through 20 agencies. The 20 agencies might suffer problems in coordinating project preparation and planning, in monitoring the use of money extended, or in monitoring projects themselves. If several agencies support a single project, it clearly makes sense to aggregate many functions rather than duplicate them. These arguments suggest that multilaterals may have an advantage in simply coordinating aid from a wide set of bilateral donors and also suggest that aid may be more effective in countries that have only one or two donors.

It could be argued that politics may be more important for bilateral lending decisions than for multilateral lending decisions or that politics may affect bilateral lending in a different way from multilateral lending. Alesina and Dollar (2000), for example, compare bilateral concessional flows to private FDI flows, arguing that the former are driven more by politics and the latter more by economics. A bilateral lender may face the choice of lending out of its own resources to a political ally versus the possibility of harnessing the greater resources of many bilaterals, through the medium of a multilateral lender, but perhaps seeing its particular interest diluted.

Figure 6.11 sorts OECD donors in terms of total aid extended and describes the determinants of their allocations of bilateral aid. As the figure shows, the United States is the largest donor among the OECD countries, accounting for about 23 percent of total aid from those countries over the period considered here. The other large donors are Japan (20 percent of total aid), France (13 percent of total aid), Germany (11 percent of total aid), and Great Britain (5 percent of total aid).

The figure divides the determinants of bilateral aid into four categories: economic ties, political ties, colonial connections, and GDP per capita. The bars in the figure illustrate the importance of each factor in assigning each donor’s aid to recipient countries (see Powell and Bobba, 2006, for details).[20]

The figure indicates that some donors extend significant percentages of aid to ex-colonies (France, Great Britain, and Portugal), countries with economic ties (Australia, Portugal, New Zealand, and Greece, and to some extent, the United States and Japan), and countries with political ties (Greece, New Zealand, and Switzerland). The data suggest that the United Kingdom, Holland, Italy, Canada, Sweden, Norway, Denmark, Belgium, Finland, and Ireland extend more financing to poorer countries (as measured by GDP per capita), with the United States, Japan, Spain, and Greece extending less assistance to poorer countries according to this measure. The observations on colonial connections and to a lesser degree on political ties support the conclusions of Alesina and Dollar (2000).

Powell and Bobba (2006) compare the behavior of multilateral and bilateral aid flows, and while they control for a large set of variables, they focus their analysis on two variables. The first is a measure of political ties, and the second is an index that captures whether aid is given by many donors or is extended by only a few. In order to build an index of political ties for multilaterals, Powell and Bobba (2006) add, across donors, the log of their measure of political ties between each recipient and donor, with weights that consider the voting power of that donor within a particular multilateral institution.[21] The authors use the Herfindahl index to measure the concentration of bilateral aid to each country.[22]

Figure 6.12 illustrates Powell and Bobba’s (2006) main findings and shows that political ties are very important for bilateral lending and also that where bilateral aid is concentrated in only a few donors, recipients tend to receive more aid.[23] However, the effect of politics is much smaller and often not statistically significant for the multilaterals, including the International Development Association (the concessional arm of the World Bank), and even less important for the regional development banks. Powell and Bobba (2006) also find that the Herfindahl index is not significant for multilaterals.

The interpretation is that bilateral aid may be more subject to political influence relative to multilateral aid and that a lack of donor coordination (where there are many bilateral donors) may result in donors’ restricting aid, presumably because their aid is less effective or because they are less confident that the aid extended goes where it is intended. Multilateral aid does not appear to suffer from the same problems, suggesting that multilaterals may help to resolve coordination problems, but perhaps at the cost of diluting a bilateral donor’s particular interests.

MULTILATERALS: CONCLUDING COMMENTS

What makes multilateral development banks different when it comes to sovereign lending? This question in fact may be decomposed into two parts: first, what makes multilaterals different from private lenders with regard to nonconcessional lending, and second, what makes multilateral lending different from bilateral lending when it comes to concessional lending and in particular, official development assistance? The chapter shows that there are significant differences between the two types of lenders in both respects.

First, based on a direct test of whether multilaterals are similar in behavior to private sector lenders, there is evidence that multilaterals are more countercyclical and that multilaterals are particularly helpful when countries are starved of liquidity by global markets. Second, there is evidence that multilateral flows are catalytic in that they appear to lead to higher future private sector flows.

Finally, there are broad differences between multilateral and bilateral lenders when it comes to concessional financing and specifically to aid flows. Multilateral aid flows are less affected by political or colonial ties or political alignment, whereas bilateral flows do tend to be affected by these factors. There is also evidence that bilateral flows are larger where they are more concentrated in a few donors. A potential interpretation of this last result is that many bilateral lenders face a problem in terms of coordinating aid.

Countries with many bilateral donors receive less aid in total. Intermediating aid through the multilaterals may resolve the problems of donor coordination, but at the cost of diluting the interests of a particular donor.