The second session of the seminar examined the merits of the case for making new allocations of SDRs under the provisions in the present Articles of Agreement of the IMF. Midiael Mussa opened the session by setting out his views on this issue, drawing on his experience at the Fund over the past five years. Three outside experts—Montek Singh Ahluwalia, Horst Siebert, and John Williamson—joined IMF Executive Director j. de Beaufort Wijnholds in offering a range of additional perspectives and critiques of Mussa’s paper.

The second session of the seminar examined the merits of the case for making new allocations of SDRs under the provisions in the present Articles of Agreement of the IMF. Midiael Mussa opened the session by setting out his views on this issue, drawing on his experience at the Fund over the past five years. Three outside experts—Montek Singh Ahluwalia, Horst Siebert, and John Williamson—joined IMF Executive Director j. de Beaufort Wijnholds in offering a range of additional perspectives and critiques of Mussa’s paper.

The Rationale for SDR Allocation Under the Present Articles of Agreement of the International Monetary Fund

Michael Mussa 1

Despite very substantial changes in both the international monetary system and the nature of the SDR, the criterion for allocating SDRs remains identical to the provision that was added by the First Amendment to the IMF’s Articles of Agreement in 1969:

In all its decisions with respect to the allocation and cancellation of special drawing rights the Fund shall seek to meet the long-term global need, as and when it arises, to supplement existing reserve assets in such manner as will promote the attainment of its purposes and will avoid economic stagnation and deflation as well as excess demand and inflation in the world.2

The Articles also stipulate that SDRs may only be allocated or canceled on the basis of a proposal by the Managing Director, concurred in by the Executive Board, which secures at least an 85 percent majority of the votes of the Board of Governors of the Fund.

In assisting the Managing Director and the Executive Board in formulating and endorsing proposals for SDR allocations, the staff of the Fund must be concerned with two key issues: (1) how to interpret the criterion of “long-term global need… to supplement existing reserve assets” in the context of the prevailing international monetary system, and (2) how to apply economic analysis and empirical evidence to assess whether and to what extent there is a need to allocate (or cancel) SDRs based on this criterion. For the most part, I believe that the impasse over SDR allocations during the past decade and a half has reflected lack of consensus about the concept of long-term global need. To some extent, however, there has also been disagreement over what to infer from the empirical evidence.

This paper presents my personal views on the concept of long-term global need, and on the extent to which the empirical evidence supports the case for SDR allocation. For the most part, these views broadly coincide with those of many other participants in internal discussions at the Fund. The paper also alludes, without attribution, to the key arguments against SDR allocation.

As background, the next section provides perspectives on how the architects of the SDR appear to have intended the allocation criterion to operate, and on how the international monetary system and the nature of the SDR have changed since that criterion was adopted. The following two sections then turn to relevant issues and empirical evidence relating, respectively, to the growth in the demand for reserves over the long run, and to the consequences of alternative means of meeting the demand for reserves. The next section takes up the macroeconomic effects of SDR allocations, and the final section provides concluding remarks.

Historical Background

Perspectives at Time of SDR Creation

As discussed in the paper by Robert Solomon in Chapter 3, the SDR mechanism was designed during the 1960s for an international monetary system that has since become extinct. Under the Bretton Woods system, countries used their international reserves to intervene in exchange markets to maintain the exchange rates of their national currencies within narrow ranges of their official parities; and gold was the fundamental underlying reserve asset of the international monetary system. The United States maintained the commitment to convert, on demand, U.S. dollars into gold for official holders. Associated with this commitment and with the preponderant position of the U.S. economy, official holdings of short-term dollar assets became increasingly important as reserves for other countries.

The rationale for creating the SDR in the context of the Bretton Woods system was linked to the recognition that the supply of gold was likely to remain limited. There was also the perception that, as suggested by Robert Triffin, the international monetary system could become subject to considerable instability if the ratio of official holdings of reserves in the form of U.S. dollars relative to the gold reserves of the United States continued to increase significantly.

In this context, the SDR was created as a form of “paper gold.” Initially, the value of one SDR was set equivalent to one U.S. dollar, but the value of one SDR was legally defined as equivalent to 0.888671 gram of fine gold; and the interest rate on the SDR was set significantly below market rates at 1.5 percent. It was clearly intended that the SDR should stand alongside gold as a primary reserve asset of the international monetary system. The U.S. dollar, which remained linked to gold, would continue to play a critical role as part of the reserve base of the system. But, with the possibility of allocations of SDRs, the system would be less dependent on increases in official holdings of U.S. dollars for meeting the expected growth in the demand for reserves. In such an international monetary system, it was even believable that by controlling the growth of supply of the basic reserve assets, it would be possible to influence the longer-term trend of the world price level and thus to provide some assurance that the world economy tended neither toward economic stagnation and deflation nor toward excess demand and inflation. In fact, for a number of countries, loss of reserves and the potential embarrassment of devaluation had been a constraint on the pursuit of excessively expansionary policies. Maintenance of the link to gold was expected to exert an important disciplining influence on excessively expansionary policy in the United States, and arguably it did so until the late 1960s.3

Necessarily, the architects of the SDR system could not provide ciear quantitative guidelines, written into the Articles of Agreement, concerning the rate at which it would be appropriate to create new SDRs to meet the purposes stated in the amendment to the Articles. Inevitably, quantitative assessments concerning the appropriate pace of SDR creation would depend on analyses of the probable growth of demand for reserves associated with noninflationary growth of the world economy, on expected increases in the supply of gold available for reserves, and on judgments concerning the extent to which it would be desirable to meet the demand for reserves by increases in holdings of national currencies.4 The criterion for SDR allocations written into the Fund’s Articles of Agreement by the First Amendment intended to motivate serious analysis and considered judgment on these issues.

In particular, it is noteworthy that the reference to a global need reflects the view that

[d]eliberate reserve creation, when decided upon, should be neither geared nor directed to the financing of balance of payments deficits of individual countries but should take place on the basis of a collective judgment of the reserve needs of the world as a whole.5

Thus, the concept of global need did not mean that to justify an allocation almost all members of the Fund would have to see an important direct reason for supplementing their own reserves through an SDR allocation. It was recognized that because countries with balance of payments deficits tend naturally to be balanced by countries with balance of payments surpluses, it would be a somewhat bizarre situation for almost all countries to feel a direct need for reserve supplementation simultaneously. Instead, the essential idea of a global need to supplement reserves was that the performance of the world economy and the functioning of the international monetary system and the balance of payments adjustment process would be improved if other sources of reserve growth were supplemented by SDR allocations. The reasonable expectation of such an important global benefit from an SDR allocation was the logical equivalent of the global need for such an allocation.

The emphasis on a “long-term” need meant that, despite explicit reference to avoidance of deflation and inflation as an objective for deliberate reserve creation, SDR allocations are not intended to respond to, or deliberately seek to ameliorate, cyclical fluctuations in the world economy. In fact, the Articles provide that decisions about SDR allocations are to be made for “basic periods” that normally last five years, and that allocations should normally occur at yearly intervals within these basic periods.

Several additional points can be noted about the perceptions and intentions that prevailed when the First Amendment was drafted. First, the criterion for SDR allocations did not embody the idea that an allocation should be based on the existence or likelihood of a “shortage” of reserves. Nor was the concept of “need” to be interpreted as a moral absolute—there was no pretense that the world economy must have allocations of SDRs to avoid serious and imminent risk of catastrophic consequences. Rather, the question posed by the criterion specified in the First Amendment was whether and to what extent allocations of SDRs would be useful as a means of meeting part of the demand for international reserves, particularly with reference to potential pressures for worldwide deflation or inflation and of the fundamental purposes of the International Monetary Fund (as specified in Article I of the Articles of Agreement).

Moreover, with regard to decisions about allocations under the criterion written into the Articles, the record reveals that there were perceived to be “a great many different ways, none of them very satisfactory, of setting about the measurement” of the need for reserve supplementation.6 Typically, attempts at quantification started by trying to characterize the overall need for reserves or reserve growth, treating as a second issue the appropriate amount to be provided through allocation. In concept, two approaches were distinguished.7 One was to estimate the amount of reserves that each country desired, and then to add up the results to get a world total. The other was to try to identify the amount of global reserve growth that could be regarded as “optimal” from the perspective of inducing countries, on balance, to adopt sound monetary and fiscal policies and to move toward liberalizing restrictions on trade and (perhaps) capital movements. But in practice, neither approach could provide for much precision in generating quantitative estimates of either the most likely or the most appropriate amount of global reserve growth during any basic period.

When drafting the allocation criterion, the architects of the SDR mechanism clearly expected the demand for reserves to exhibit an upward trend over time. Indeed, reserve holdings had increased fairly steadily for both the industrial and the developing countries from the early 1950s through the late 1960s, which was the period on which the Fund staff had concentrated its analysis of reserve developments.8 Yet the architects were also well aware of the variability of factors influencing the demand for reserves. Indeed, it was recognized that “payments imbalances vary so much from year to year that it seems pretty hopeless to try to arrive at any estimate of needed reserve growth for periods of less than four or five years at a time.”9

Finally, it should be emphasized that by adopting the First Amendment to the Articles of Agreement, the members of the Fund expressed a firm intention that the SDR would play an important role in the international monetary system. Otherwise, there would have been no point to investing the considerable effort needed to agree upon and ratify the amendment. The criterion for allocation contained in the amendment fell somewhat short of a biblical injunction “to go forth and allocate.” Nevertheless, the language of Article XVIII has an injunctive tone: “In all its decisions with respect to the allocation and cancellation of special drawing rights the Fund shall seek to meet the long-term global need, as and when it arises, to supplement existing reserve assets …” (emphasis added). Decisions about the appropriate pace of allocations or cancellations would necessarily depend on assessments of evolving economic circumstances, particularly with respect to the demand for international reserves and the consequences of using alternative means of meeting that demand. However, it was expected that regular discussions of the need for SDR allocations would usually lead to periodic decisions to allocate or, conceivably, to cancel. It was not intended that these discussions should deteriorate into a persistent debate about whether the SDR should play any meaningful role in the international monetary system. That issue of general principle was resolved, in the affirmative, by the adoption of the First Amendment.

Perspectives After Collapse of the Bretton Woods System

The events of August 15, 1971, and the final collapse of the par value system in March 1973 changed the fundamental character of the international monetary system. With the suspension of the gold convertibility of the U.S. dollar, the system no longer required stocks of outside assets to be held as an international base for national stocks of money. Consequently, with gold no longer playing a central role as the anchor of the system, the rationale for defining the SDR in terms of gold was weakened, and in June 1974 the SDR was redefined as a basket of currencies.10

With the changes in the international monetary system, a unified base linked to gold no longer existed; instead, there are multiple bases consisting of the national base monies of the countries that have elected floating exchange rates. The key outside monies of the system became the base monies of the largest industrial countries, and most important, the U.S. dollar. Official holdings of these key national currencies by other countries became the central source of international reserves.11 Although inflation rates in the major industrial countries tended to move up and down together under the influence of common shocks such as the two surges in world oil prices, national price levels over the longer run have shown substantial divergences. Since 1973, the general price level in the United Kingdom has risen by almost twice as much as in the United States, and the general price level in the United States has risen by about twice as much as in Germany. The long-run behavior of price levels measured in national currencies is fundamentally determined by national monetary policies, without constraint from the supply of some outside reserve asset.

In this new system, the SDR could no longer reasonably be regarded as part of the outside monetary base for the international monetary system. It retained value and usefulness as a reserve asset because holdings in the SDR Department of the Fund could (as before) be exchanged with other participants for national currencies.12 SDRs functioned essentially as “unconditional lines of credit” defined in terms of a basket of national currencies.13 Countries holding SDRs effectively had the right to “borrow” on these lines of credit by exchanging them for one or more of the major national currencies at a value determined by the value of the SDR basket. Countries borrowing in this way would necessarily hold fewer SDRs than they had been allocated, and they would pay interest at the SDR rate on the difference between their cumulative allocations and their holdings of SDRs. This interest rate was raised gradually from 1.5 percent to a level that fully reflected a weighted average of market interest rates on a basket of financial instruments denominated in the national currencies of the SDR basket. Other countries were effectively net lenders in the SDR system by holding more SDRs than their cumulative allocations, and they would receive corresponding interest on their loans at the SDR rate. The existence of this special mechanism for arranging borrowing and lending transactions might be a matter of significant convenience. However, the economic similarity of SDRs to other forms of credit suggests that changes in the volume of SDRs made available through allocations would not exert the powerful effects that might earlier have been expected from changes in the supply of the base reserve asset of the international monetary system.

It might be argued that this change in the role of the SDR was of more theoretical than practical significance—that changes in the supply of SDRs through allocations would not have had powerful effects even if the Bretton Woods system had continued to function. Arguably, under the Bretton Woods system, gold (with or without augmentation by SDRs) did not exercise an overwhelming disciplining force as the ultimate reserve base for supplies of national base monies, including the U.S. dollar. Indeed, the monetary authorities of many countries did regularly sterilize the effect of reserve outflows and inflows, including flows of gold, on their domestic money supplies. This was particularly true for the United States. Moreover, then as now, an allocation of SDRs does not have any automatic effect on national money supplies. Such effects occur only if national monetary authorities choose to react to allocations, or to gains and losses of SDRs through transactions in the SDR Department of the Fund, by altering the domestic monetary base.

On the other hand, limits on the supply of outside reserves of gold and U.S. dollars (and SDRs) were a constraining influence on the policies of many countries under the Bretton Woods system. And, under the Smithsonian agreement of December 1971, the U.S. dollar was officially devalued against gold and against the SDR. Certainly, the period following the collapse of Bretton Woods and ejection of gold from the international monetary system was one of much more rapid inflation in the world economy than had characterized the earlier part of the postwar era (or any other period). It is at least arguable that inflation would have been much lower if the authorities of the world’s largest economy had been constrained to operate in a system of fixed exchange rates and forced to face the periodic embarrassment of devaluations of their currency against the commonly accepted international reserve assets of gold and SDRs. The point here, however, is not to discuss the advantages of the continuation of a gold-linked system of fixed exchange rates, which I doubt would have been feasible.14 Rather, it is to emphasize that fundamental changes in the nature and functioning of the international monetary system have undermined the proposition that controlling the supply of SDRs would be an important mechanism for regulating the longer-term trend of inflation or deflation in the world economy. Decisions concerning allocations (or cancellations) of SDRs were not the cause of, and could not have averted, the worldwide upsurge of inflation in the 1970s, nor were they responsible for the much lower inflation rates of more recent years.

The collapse of the Bretton Woods system and the move to floating exchange rates among the world’s major currencies necessitated substantial changes in the IMF’s Articles of Agreement, which were embodied in the Second Amendment ratified in 1978. As previously noted, however, while the Articles pertaining to the SDR were modified in several respects, the record shows that there was no intention to depart from the interpretation of the allocation criterion that prevailed under the First Amendment (discussed above), despite changes in the character of the international monetary system. Moreover, new language was added (in Article VIII, Section 7, and in Article XXII) that committed each member to collaborate with the Fund in accordance with “the objective of making the special drawing right the principal reserve asset of the international monetary system.” This language was not intended to alter the criterion governing allocations;15 in particular, it did not imply that the SDR should be made to play a large role in total reserves by the brute force of large-scale allocations. But the language does indicate that, despite important changes in the international monetary system since the adoption of the First Amendment, members of the Fund reconfirmed the general principle that the SDR should play a meaningful role.

Under the Second Amendment, decisions about SDR allocations were still to be based on assessments of the “long-term global need … to supplement existing reserve assets.” Thus, as a matter of international law, reaffirmed by the Second Amendment to the IMF Articles of Agreement, the concept of long-term global need to supplement reserves retained its operational relevance with respect to decisions on SDR allocations. This operational significance was reaffirmed by the decision to proceed with an SDR allocation shortly after the Second Amendment was ratified.16 The analysis and evidence used in support of this decision, as well as earlier decisions about SDR allocations, necessarily provide critical guidance to Fund staff in considering the factors to be addressed in assessing further allocations under the criterion of long-term global need to supplement reserves.

The Long-Run Growth of Reserve Holdings

One of the factors that is undoubtedly relevant in assessing the long-term global need to supplement reserves is the projected longer-term growth of demand for reserves. As can be seen in Figure 1, at the end of the 1960s the Executive Directors and the staff of the Fund could look back over two decades of fairly steady growth in global reserve holdings, along with the expansion of the world economy and world trade. This growth of reserve holdings occurred both for industrial countries (excluding the United States) and for developing countries. As may be seen from Figure 2, for both groups of countries reserve holdings grew roughly in proportion with the growth of trade (measured by imports).

Figure 1.
Figure 1.
Figure 1.

Reserve Holdings of Countries That Were Members of the IMF in 1952

(End-of-year data; in billions of SDRs; logarithmic scale)

Source: IMF, International Financial Statistics.1 Gold holdings are valued at SDR 35 an ounce.
Figure 2.
Figure 2.
Figure 2.

Reserve Holdings, Measured in Weeks of Merchandise Imports, 1952-941

Source: IMF, International Financial Statistics.1 Ratio of end-of-year reserves to annual merchandise imports for all members of the Fund during each year. Gold holdings are valued at SDR 35 an ounce.2 For the United 5tates, gold reserves and total reserves (not shown before the early 1960s) represented, respectively, approximately 120 and 130 weeks of imports in 1952.

The United States was a dramatic exception to these developments. The absolute level of U.S. reserves fell persistently and substantially in the 1950s and 1960s, with the decline in U.S. gold reserves significantly outstripping accumulation of foreign currencies by the U.S. authorities. Meanwhile, foreign official holdings of U.S. dollars generally continued to rise. Indeed, the growth of reserves for other industrial countries and for developing countries reflected predominantly the acquisition of gold from U.S. reserves and the accumulation of official holdings of U.S. dollars. This was the Triffin dilemma in operation. With very limited increases in the supply of gold, if the reserves of the rest of the world were to continue to grow along with demand, the ratio of U.S. official liabilities (held as reserves by other countries) would have to continue to rise relative to the U.S. gold reserve. At some point, the credibility of the U.S. commitment to maintain official convertibility of the dollar into gold was bound to come into question. On the other hand, if the United States maintained very tight policies to restrain both gold outflows and accumulation of foreign official holdings of dollars, the rest of the world would not be able to obtain the reserves that it wanted to acquire in view of the expanding volume of economic activity and world trade. The practical relevance of this concern was demonstrated when the United States tightened monetary and fiscal policy in 1968-69; U.S. gold reserves rose and foreign official holdings of dollars declined, leading to a fall in reserves of other Fund members.

Experience with the growth of reserves in rough proportion to world trade during the 1950s and 1960s strengthened confidence that the demand for reserves would continue to expand in a growing world economy, and it raised serious concerns about how this rising demand would be met. Although not sufficient by itself, the clear perception of a rising long-term demand for reserves was undoubtedly necessary both to the impetus to create the SDR system and to the decision to proceed with the first allocations.

In the aftermath of the breakdown of the Bretton Woods system, it was initially unclear whether the demand for reserves would continue to grow as it had previously. After all, if countries no longer needed to use reserves to intervene in support of officially pegged exchange rates, the demand for reserves might reasonably be expected to shift downward, if not to disappear altogether. If this prediction had turned out to be correct, it would have been very difficult to make any reasonable case for further SDR allocations under the criterion specified in the Articles.

As can be seen from Figures 1 and 2, expectations of a sharp drop in the demand for reserves with the advent of floating exchange rates were not validated. The transformation of the international monetary system in the early 1970s was not accompanied by a downward shift in either the level or the rate of growth of global reserve holdings.17 Indeed, the rate of growth of nominal reserves increased sharply during the 1970s, when rates of inflation of general price levels also rose markedly, and continued to rise at the more rapid trend during the 1980s and 1990s, when inflation moderated substantially. As world trade continued to expand, the ratio of reserves to imports appears to have declined a small amount for industrial countries (excluding the United States), while it rose by a modest amount for developing countries.18 For the United States, this ratio fell to an exceptionally low level (about five weeks of imports) in comparison with most other countries. Since the mid-1970s, however, this ratio has stabilized and U.S. reserves, like those of other countries, have generally grown about in line with the expanding value of imports. The fact that reserve holdings continued to grow despite the shift to a system of floating exchange rates was clearly recognized and considered highly relevant in the decision to proceed with further allocations of SDRs in the period from 1979 to 1981.19

Accordingly, in its analysis of the case for allocation (or cancellation) of SDRs, the Fund staff has continued to pay careful attention to the reliability of projections concerning the long-term demand for reserves. As is apparent from Figures 1 and 2, we can today look back at four and a half decades of sustained growth in nominal reserve holdings and relatively stable ratios of nongold reserve holdings to the scale of merchandise imports.

Consistent with the impressions conveyed by Figure 2, an extensive econometric literature suggests that countries’ reserve holdings can be explained, to a large extent, by the scale of their international trade or economic activity.20 In multivariate log-linear time-series regressions of reserve holdings (for various country groups) on the level of imports, the average propensity to import, and a measure of the variability of payments balances, the (elasticity) coefficient on the level of imports is generally found to be highly significant and typically lies in the range between 0.6 and 1.1. In this connection, Figure 2 shows that after some sharp movements in the years surrounding the collapse of the Bretton Woods system in the early 1970s, ratios of reserves to merchandise imports were quite stable for both industrial countries and developing countries.

To be sure, in the many countries that manage their exchange rates, the behavior of reserves from year to year may largely reflect the cumulative effects of the authorities’ day-to-day transactions in foreign exchange markets. However, the stability of reserve-to-import ratios for broad groups of countries over substantial time periods belies the notion that such systematic behavior could be primarily accidental. With a very high degree of confidence, it may be concluded that the demand for international reserves may be expected to grow roughly in line with the growth of world trade. In all of the recent controversy about the justification for further SDR allocations, no one has challenged the important conclusion that the world economy faces a rising need for international reserves over the longer term.

As a practical matter, the Fund staff has usually relied on a simple relationship between reserves and imports to project the likely growth of demand for reserves and has emphasized that rather wide error bands should be placed around the projections.21 Even so, the projections are useful in suggesting the order of magnitude of probable increases in the demand for reserves. For example, taking the increase in world trade projected over the next five years in the IMF’s World Economic Outlook, and multiplying this projected increase by a reasonable estimate of the import elasticity of reserves (specifically, by 0.85), one arrives at the crude estimate that the world demand for reserves will expand by roughly SDR 350 billion from end-1995 through end-2000. Even allowing for wide error bands, one can confidently predict that the order of magnitude of the increase will be hundreds of billions. One may also conclude that recent proposals for SDR allocations, such as the Managing Director’s suggestion of an allocation of SDR 36 billion (spread over three to five years), would satisfy only a modest portion of the probable growth of demand for reserves.

Implications of Alternative Means of Meeting the Demand for Reserves

It has always been understood that a rising long-run demand for reserves is not a sufficient basis, in itself, to conclude that there is a “long-term global need” to supplement reserves. It also must be shown that there is likely to be some meaningful benefit, for the world economy or the international monetary system, from meeting part of the growing demand for reserves through an allocation of SDRs. Three questions naturally arise when considering alternative means of meeting a given (or growing) demand for international reserves: (1) Are there important qualitative differences among different means of meeting this demand? (2) Are there significant differences in cost among different means of meeting this demand? (3) What will be the distribution of the net benefits associated with the “seigniorage” derived from newly created SDRs?

With the SDR as it existed under the Bretton Woods system, all three of these questions were important. Qualitatively, expanding the supply of SDRs increased the supply of outside reserves in the international monetary system. It was like an increase in the supply of monetary gold; in important respects it was unlike an increase in the supply of reserves in the form of national currencies. In particular, an increase in reserves in the form of SDRs (rather than a corresponding increase in reserves in the form of U.S. dollars) would help to ameliorate the Triffin dilemma, while the same increase in reserves in the form of U.S. dollars would exacerbate the Triffin dilemma. Looking to costs, an increase in the supply of SDRs was undoubtedly cheaper than increasing the supply of monetary gold by digging it out of the ground and was clearly seen as more desirable than raising the monetary value of existing gold reserves through a general increase in the price of gold.23

Concerning “seigniorage,” with the SDR interest rate set well below market levels (at 1.5 percent), it was recognized that the receipt of new SDR allocations conferred a significant economic benefit. Various forms of the “link” proposal sought to direct these benefits toward developing countries, either by allocating SDRs initially to these countries or to the International Development Association (IDA, the concessional lending facility of the World Bank). These proposals were rejected at the time of the First Amendment in favor of a more neutral mechanism for allocating new SDRs in proportion to members’ quotas in the Fund. Nevertheless, discussions of link proposals continued into the 1980s when the SDR interest rate was raised to market levels. More recently, as discussed below, the controversy over the distribution of the benefits from SDR allocations has taken a slightly different form—countries facing significantly higher costs of holding reserves than the SDR interest rate are perceived to be the principal beneficiaries of SDR allocations.

With de-linking of the SDR from gold and the removal of gold from a central role in the international monetary system, qualitative differences between SDRs and reserves of national currencies arguably became less important. However, when it was decided to allocate SDRs shortly after the approval of the Second Amendment, it was emphasized that there was an important distinction between “owned reserves” created through an SDR allocation and “borrowed reserves.” When SDRs are allocated, a permanent addition is made to the stock of international reserves that cannot be extinguished except by an explicit decision to cancel outstanding SDRs. On the other hand, reserves created through international borrowing need to be periodically refinanced or they will disappear, from the country holding them and from the international monetary system, as reserves are used to pay down maturing debts. Such destruction of borrowed reserves is most likely to occur when a country faces sudden pressures to pay down its foreign debts, particularly in circumstances where its current account is in deficit. In a general crisis, several countries would be facing pressures of this kind simultaneously, and there could be a sudden drop in the world supply of reserves, perhaps leading to a downward spiral. Reserves in the form of SDRs are not subject to this risk. However, with SDRs constituting only a small portion of total reserves, it is difficult to believe that this qualitative advantage would be of much practical help in circumstances where it would be relevant. On the other hand, some have argued that modest allocations sufficient to keep the SDR alive as a small fraction to total world reserves are important to preserve an option for creation of “owned reserves” that might prove useful in some future crisis.

On the question of whether there are significant differences in the costs associated with alternative sources of reserves, the Fund staff has found substantial evidence that there are significant advantages for SDR allocations in meeting part of the growing demand for reserves.

Apart from SDR allocations, the two channels through which countries can build their international reserves are by “earning” reserves by running current account surpluses or by “borrowing” reserves either directly from international capital markets or indirectly by issuing claims to private domestic residents that show up as international capital outflows. For countries that enjoy some access to international capital markets, one measure of the cost of holding reserves acquired through either of these channels is the interest rate that the government must pay to secure international loans, less the interest rate received on holdings of reserves. However, for countries that face potentially rising costs if they wish to expand significantly their international borrowing, the marginal cost of external credit is above the rate on existing international loans. Also, in many countries that borrow in international markets, central banks find that they need to sterilize reserve gains associated with private capital inflows by selling domestic securities. In such situations, reserve holdings are effectively financed by selling domestic debt, which often has a significantly higher interest rate than the return on foreign exchange reserves. For countries that cannot borrow on international markets, it is necessary to impute a shadow interest rate and then to subtract the rate of return earned on reserves.

For the largest industrial countries that can easily issue debt denominated in their own currencies on international markets at interest rates close to those prevailing in domestic markets, the cost of holding reserves is quite low. For other countries that generally enjoy good access to world capital markets, the costs of holding reserves appear more significant. As shown in Table 1, the spreads between interest rates on international bank loans and yields on U.S. Treasury bill holdings have averaged between 1 and 2 percentage points (annualized) in recent years both for the industrial countries and for those developing countries that have enjoyed continuous access to new bank loans. For many of the developing countries in this category, however, the costs of sterilizing reserve inflows in recent years have sometimes added substantially to the cost of holding reserves. Costs of holding reserves have been much larger for developing countries without continuous market access, as reflected in the yields to maturity on secondary market debts. Specifically, for the 13 countries represented in column 3 of Table 1, it is estimated that effective costs have averaged more than 7 percentage points in recent years. It may reasonably be assumed that other developing countries, for which no information on debt yields is available, also face costs of holding reserves that average at least 7 percent a year.

Table 1.

Interest Rate Spreads1

(In percent per annum)

article image

Columns 1 and 2 reflect average spreads on international syndicated credits based on data from OECO Financial Market Trends. Column 3 is based on data from Salomon Brothers on the secondary market prices of the external debts of 13 countries (Argentina, Bolivia, Brazil, Chile, Côte d’Ivoire, Ecuador, Mexico, Morocco, Nigeria, Peru, Philippines, Uruguay, and Venezuela).

Based on data for part of the year.

In contrast, the economic cost of supplying reserves through allocations of SDRs is zero or very close to zero. Recipients of SDR allocations pay the SDR rate of interest (and a tiny assessment to cover the costs of running the SDR Department of the Fund) on their cumulative allocations and receive the same rate of interest on their total SDR holdings.23 Thus, for countries that hold onto their entire cumulative allocations, the net carrying cost of those reserves is almost zero, and the holding of these reserves has no effect on other countries. Moreover, it is arguable that this conclusion is not much affected if reserves acquired through SDR allocation are held as SDRs or exchanged for foreign currency reserves. For net creditors in the SDR Department, experience indicates that they are satisfied with the return received; they have generally been willing to expand their holdings of SDRs, and they were satisfied with this situation a decade or more ago when SDRs were more than twice as large relative to total world reserves. Most net users of SDRs maintain total reserves at levels that are generally much larger than their net use of SDRs. To the extent that the yields on reserve assets (including expected exchange rate changes) are independent of the currency denomination of the assets, a country that draws down its SDRs and replenishes its reserves with U.S. dollars, for example, can expect ex ante to earn the same yield on its dollar holdings as it would have expected to earn on the SDRs. However, reserves are held with the expectation that they might sometime be used, and there are potential congestion costs if many countries attempt to be net users of reserves simultaneously. Normally this problem is solved by allowing the price of credit to rise or fall in order to balance demand for the use of credit with the available supply. Since the SDR system is closed and the SDR interest rate is set outside the system, however, congestion costs might in principle be a significant worry. The fact that for many years the designation mechanism has not been needed to secure counterparties for countries that wished to use their SDRs indicates that such congestion costs have not materialized in practice.24

Given these estimates of the costs of holding reserves for different groups of countries, and taking account of the distribution of an SDR allocation across these groups, it is possible to calculate the cost savings from supplying part of the growing demand for reserves through an allocation. For example, an allocation of SDR 36 billion (as suggested by the Managing Director) would be associated with projected annual savings of about SDR 1 billion. Relative to annual world GDP, of course, such savings are modest; miraculous results cannot reasonably be expected from moderate measures. However, for an allocation that is essentially costless, annual savings of this magnitude imply a very handsome rate of return.

In addition, in support of SDR allocations, it has sometimes been emphasized that some countries face acute shortages of reserves and may, as a consequence, be forced into actions injurious to their own economies and with broader negative effects for the rest of the world. For these countries, the availability of additional reserves to use could bring substantial benefits. By itself, however, this argument is insufficient to support a finding of long-term global need. The architects of the First Amendment did not see the concentrated reserve needs of a limited group of countries as a problem of “global need,” and an allocation of SDRs that spreads broadly (in proportion to quota) across the membership of the Fund would not be an efficient means of responding to such concentrated reserve needs. Nevertheless, if a finding of global need can be justified on other grounds, it is relevant to note as an advantage of allocation that it would help to ameliorate the situation of countries facing acute reserve shortages.

In my experience, the Fund staff’s analysis of the cost advantages of providing for some of the growth of demand for reserves through SDR allocations has not been directly challenged. No one has seriously questioned the evidence that the effective costs of holding reserves for a wide range of countries are much higher than the cost associated with reserves provided through allocations. Instead, three objections have been raised about the relevance of this evidence. (1) It has been suggested that “costs” are not relevant in assessing “needs.” (2) It has been argued that high costs of holding reserves are somehow appropriate for countries that lack creditworthiness. (3) Without explicitly conceding that a problem of the high cost of reserves exists, some have judged that this problem is not sufficiently severe or widespread to constitute a problem of global need.

Concerning these objections, my views are the following. First, there is no rational economic basis for saying that the relative costs of alternative means of supplying reserves do not matter when assessing the merits of SDR allocations versus other means of meeting the growing demand for reserves. Therefore, in an appropriate interpretation of the criterion for allocation in the Articles, costs are clearly relevant in deciding whether reserves should be supplemented through SDR allocations rather than having demands met through other means.

Second, the evidence shows that the demand for reserves to hold tends to rise, along with the level of economic activity and trade, for the wide range of countries that face high costs of holding reserves. The market, however, cannot enforce a distinction between borrowing to finance the holding of reserves and borrowing to finance spending; once the funds are borrowed, there is no guarantee that they will not be spent. The market interest rate for these countries—if they can borrow at all—reflects this fact. However, from the broader perspective of the world economic system, it is a virtual certainty that even countries not generally regarded as creditworthy will increase their reserve holdings over time. From this perspective, it is rational to supply at least part of the expected increase in demand for reserves that will be held by these countries, and the economic opportunity cost of providing such reserves through SDR allocations is close to zero.

Third, it is a matter of judgment how severe and widespread the problem of excessive cost of reserves must be before it plausibly becomes an issue of global need.25 As previously noted, acute reserve stringencies affecting a significant number of countries but representing less than 10 percent of world economic activity are arguably too narrow a problem to represent, by themselves, a global need for SDR allocation within the meaning of the Articles. However, the vast majority of the Fund’s membership accounting for about half of Fund quotas faces net costs of holding reserves that are higher, in many cases significantly higher, than the true economic cost of creating reserves through SDR allocations. This, in my view, falls well into the range of a global need for SDR allocations to supplement other sources of reserve growth. Others might set a more stringent standard. However, the requirement for an 85 percent majority of total voting power to approve an SDR allocation was never meant to imply that a global need to supplement reserves must be directly and severely felt by countries with at least 85 percent of Fund quotas. In an organization devoted to international monetary cooperation, the attitude cannot reasonably be, “Not a problem for me, not a problem for the world.” Instead, there ought to be some appreciation for the sentiment, “There, but for the grace of God, go I.”

Macroeconomic Effects of SDR Allocation

In connection with the criterion of “long-term global need,” the Articles require that supplementation of reserves through an SDR allocation”… will promote the attainment of [the Fund’s] purposes and will avoid economic stagnation and deflation as well as excess demand and inflation in the world.” In this regard, the staff’s position is that the purposes of the Fund are clearly served by meeting part of the growing demand for reserves at a lower and more economically appropriate cost than would otherwise prevail; that the risks of economic stagnation and deflation are probably somewhat ameliorated by making additional reserves available to countries with acute reserve stringencies; and that the risk of adding to world inflation through an allocation of SDRs that would add modestly to total world reserves is insignificant.

On these issues, opponents of SDR allocations have raised three main concerns about the possibly harmful consequences of SDR allocations: the potential for adding to world inflation, the potential for allowing countries to pursue undesirable macroeconomic policies, and the potential for some countries to increase their absorption of goods and services at the expense of others.

With regard to each of these concerns, the scale of an SDR allocation is a relevant issue. The ratio of SDRs outstanding to total nongold reserves (SDR/NGR) stood at 7 percent in 1981 after the last series of allocations; it declined to 2½ percent in 1995. With this in mind, I would consider annual allocations of SDRs that stabilized the SDR/NGR ratio at 4 percent or less to be “small”; this would encompass annual allocations of up to SDR 5 billion over the next five years. Allocations that stabilized the SDR/NGR ratio in the range of 4-8 percent might be termed “modest”; this would imply annual allocations of about SDR 10 billion over the next five years. “Moderate” allocations might range as high as SDR 20 billion annually, with the prospect of gradually raising the SDR/NGR ratio to the range of 8-20 percent.26 I would reserve the terms “large” and “very large” for allocations expected to supply at least 20 percent, or at least 40 percent, of the annual growth of nongold reserves. Such allocations would, over time, raise the SDR/NGR ratio to at least 20 percent, or at least 40 percent, respectively.

It should be acknowledged that large or very large allocations of SDRs could pose inflationary risks because they might affect national monetary policies in the largest countries. Receipt of substantial SDR allocations would have no direct monetary effect, but rapid accumulation of reserves as a result of very large allocations might tempt even the most prudent authorities to monetize some of these gains to increase spending. In addition, other countries might want to convert newly allocated SDRs into the currencies of the major industrial countries. If the resulting increases in reserve inflows were so large that the major industrial countries were not prepared to sterilize them, monetary policies would become more expansionary in the countries whose national currencies form the foundation of the present international monetary system, and world inflation would rise.

However, the reserve flows to the major industrial countries resulting from modest or moderate SDR allocations cannot reasonably be expected to have a similar inflationary effect, even relative to the size of the allocation. For example, annual allocations of SDR 10 billion would not plausibly induce a reserve inflow for the United States that is anything more than about 10 percent of the normal annual increase in the U.S. monetary base.27 It is inconceivable that the U.S. Federal Reserve would choose to sacrifice the main objectives of U.S. monetary policy in order to avoid sterilizing such a modest reserve inflow. The same holds true for the central banks of the other major industrial countries. Thus, the effects of modest or moderate SDR allocations on world inflation via the monetary policies of the largest industrial countries may reasonably be presumed to be nil.

Aside from effects through influencing monetary policy, it is also arguable that there would be a boost to world aggregate demand if countries receiving SDR allocations used the increase in available credit to boost their spending, thus placing upward pressure on either world interest rates or the world price level. In a world economy with annual output approaching SDR 20 trillion and with total credit outstanding of similar magnitude, however, it would take very large allocations of SDRs to have more than trivial effects through this mechanism.

The scale of SDR allocation is also relevant to the potential danger that an abundance of reserves might provide scope for policymakers to postpone desirable policy adjustments and to pursue undesirably lax macro-economic policies. This danger, however, would be no more than trivially affected by small SDR allocations; it is unlikely to be increased significantly by modest or moderate allocations that would add comparatively small amounts to countries’ reserves. Also, on the other side, low levels of reserves, and the high costs of carrying them, can induce countries to maintain restrictions on international transactions or other undesirable policies and can increase the vulnerability of countries to adverse shocks. Recent analysis has shown that developing countries that typically face higher costs of holding reserves are generally subject to significantly larger relative disturbances to their balance of payments than industrial countries that typically face lower costs of holding reserves.28

The concern about SDR allocations leading to inappropriate demands on real resources relates partly to the operation of the SDR system itself; some net users of SDRs might fail to meet regular interest payments and, in the event of liquidation of the SDR Department, some net users might not repay their net use. Experience casts some light on these concerns. In the present SDR system, only a very few countries in extreme circumstances have allowed arrears to accumulate on their net use of SDRs. With about SDR 21.5 billion now outstanding, the cumulative buildup of arrears on interest payments due from net users amounts to less than ¼ of 1 percent of total allocations. This suggests that the risk of failures to meet obligations within the SDR system does not constitute a serious obstacle to further allocations of SDRs on a modest scale.

Given the reasonable expectation that obligations within the SDR system will be fully met by almost all participants, there is arguably no legitimate concern that countries would use SDRs to increase unduly their absorption of goods and services at the expense of others, even for large-scale allocations. Experience shows that countries that are net users of SDRs typically hold other reserves in much greater volume than their net use of SDRs and have shown increases in total reserve holdings over the years. If SDR allocations provide no more than a moderate portion of the increase in reserves that countries (on average) will decide to hold, in whatever form, it is arguable that such allocations do not finance unwarranted demands on real goods and services. Of course, if countries could not satisfy some of their rising demand for reserves from SDR allocations, they might run larger current account surpluses to acquire reserves, implying an increase in their net supply of real goods and services. Or, •these countries might use credit—often high-cost credit—to acquire reserves, and use of credit for this purpose might diminish its use to finance absorption of real goods and services. But, from a global perspective, what is the economic sense of forcing countries to run current account surpluses or utilize expensive credit for the purpose of meeting all of their rising demand for reserves? From a global perspective, the main problem is that countries facing high costs of reserves are forced to diminish unduly their net use of real resources in order to meet their rising demands for reserves. Even large allocations of SDRs would do no more than ameliorate this problem.

This analysis is not something new in economics that has been developed specifically to argue in support of SDR allocations. Adam Smith writes as follows:

It is convenient for the Americans, who could always employ with profit in the improvement of their lands a greater stock than they can easily get, to save as much as possible of the expence [sic] of so costly an instrument of commerce as gold and silver, and rather to employ that part of their surplus produce which would be necessary for purchasing those metals, in purchasing the instruments of trade, the materials of clothing, several parts of household furniture, and the iron-work necessary for building and extending their settlements and plantations.…(1937 edition, p. 893)

The principle exposited in this passage still applies today—in the analysis of alternative means of meeting the rising demand for international reserves for countries that perceive high costs of holding reserves, when the world possesses a means of supplying reserves at a much lower and economically appropriate cost. Of course, just as Adam Smith warned of the dangers arising from “the unskillfulness of the conductors of this paper money,” the Articles caution that decisions about SDR allocations must weigh carefully any economic risks from such allocations.


The international monetary system has changed in major ways since the inception of the SDR. Most notably, many countries have moved to floating exchange rates, gold has lost significance as an actively used reserve asset, and private capital markets have become substantially larger and more accessible. But the relationship between reserve holdings and the scale of international trade and economic activity has remained relatively stable, and the demand for reserves has kept growing.

In addressing the case for SDR allocation in the current international monetary system, and under the present Articles of Agreement, the main challenge is to interpret the concept of long-term global need. The impasse over SDR allocation during the past decade and a half has primarily reflected a lack of consensus on how to interpret this concept. This paper contends that relying on SDR allocation to meet a moderate proportion of the projected growth in the demand for reserves over the long run would have economically meaningful and widespread benefits for many countries and for the global economic system, and would not add to world inflation or create significant difficulties. Consistent with the meaning intended in the First Amendment of the Articles of Agreement and retained in the Second Amendment, there is a solid case for moderate allocations of SDRs in order to meet the long-term global need to supplement other sources of international reserves.

The case for SDR allocation does not rest simply on the prospect of continuing growth in the demand for reserves, although this is an important element. In my view, the central economic point is that the holding of international reserves by a country has little or no real cost for the world economy, and SDR allocations are a means of providing reserves at an economically appropriate level of cost. In contrast, in the current system most countries—except for a few very large industrial countries—face significant costs in holding reserves, and these costs rise as the demand to hold reserves grows along with economic activity and international trade. In this situation, it would make economic sense to provide for some of the growth in demand for reserves through allocations of SDRs. Moderate allocations of SDRs do not provide cheap credit to finance increased spending; they provide a means for countries to meet their rising demand to hold reserves at the negligible cost that the holding of reserves imposes on the world economy. Moreover, moderate SDR allocations would help to alleviate acute reserve stringencies and would provide some benefit by increasing the share of owned reserves in the international monetary system.

To the best of my knowledge and judgment, opponents of SDR allocations have never seriously challenged the analytical or factual basis of this argument. They have mainly ignored it and have raised instead some concerns about the potential inflationary effects and other horrors that they claim might follow from even modest allocations of SDRs. These concerns do not, however, stand up to scrutiny, at least for allocations in the small to moderate range.29

In the end, I am persuaded that the real difference of opinion over SDR allocations has little or nothing to do with the criterion of “long-term global need.” Given that the principle of the SDR as a source of international liquidity was established by the First Amendment, the criterion for allocation was supposed to motivate primarily a debate with a quantitative dimension: in the upcoming basic period, how large or how small an allocation of SDRs would be appropriate to meet the long-term global need to supplement reserves, or would cancellation of some outstanding SDRs be appropriate? Aside from occasional compromise proposals, there has been very little debate about the appropriate size of allocations. I suspect, therefore, that the real difference is not about the quantitative assessment of long-term global need, but rather about the general principle of whether the IMF should still be in the business of supplying reserves through the SDR mechanism.

This apparent difference over the principle of the SDR system may reflect views about the probable distribution of the global benefits from allocations. Countries facing high costs of holding reserves see significant direct benefits from SDR allocations, and they have generally been strong proponents of allocations. In general, however, these countries have been substantial and persistent net users of SDRs, despite holdings of reserves that are typically much larger than their net use of SDRs. Why this should be is not entirely clear. The rate of return on SDRs is not unattractive relative to other reserve assets—as witnessed by the voluntary accumulations of net creditors in the SDR system. Perhaps net users of SDRs have relations with banks and other creditors that motivate reserve holdings outside the SDR system. Perhaps some countries see an inconsistency between holding SDRs and simultaneously being debtors to the IMF in its conditional facilities where the rate of charge exceeds the SDR interest rate. More generally, perhaps, for some countries that see Fund credit as normally accompanied by a heavy burden of conditionality, the temptation to use SDRs as an unconditional source of credit from the IMF is simply too great—despite a return on SDRs that is competitive with reserves that are actually held. In any event, the credibility of the position of many proponents of SDR allocations is not strengthened by the fact that they are persistent net users of SDRs that have already been allocated.

On the other side, those industrial countries that have been the main opponents of allocations generally face very low costs of holding reserves and perceive little direct benefit from allocations of SDRs.30 Some of these countries have also been large net creditors in the SDR system. While the interest rate on SDRs is tied to that on comparable reserve assets and is generally quite attractive for net borrowers in the SDR system, it is not exceptionally attractive for net creditors. Moreover, net creditors in the SDR system may worry more than net users about arrears and about the risks of defaults if the SDR Department was liquidated—especially if outstanding SDRs and net use were to expand significantly through allocations.

In addition, the countries opposed to SDR allocations also tend to be net creditors in the General Resources Account of the Fund. These countries generally see an important role for the Fund in providing credit under firm conditionality from that account to countries undertaking strong economic stabilization and reform programs. They see a possible inconsistency between this conditional lending and the simultaneous provision of essentially unconditional credit to the same countries (on more attractive terms) through SDR allocations.

Although understanding why these considerations might cause some countries to question the general principle of the Fund as a supplier of reserves through the SDR mechanism, I fail to see that they are particularly relevant under a proper interpretation of the allocation criterion in the present Articles. In any event, the Fund staff necessarily takes the Articles of Agreement as settling the issue of general principle and analyzes the case for SDR allocations under the criterion of long-term global need. That analysis supports the case for moderate allocations of SDRs to meet at least a modest part of the growing demand for international reserves.

SDR Allocations and the Present Articles of Agreement Montek Singh Ahluwalia1

The purpose of this paper is to consider whether an SDR allocation can be justified in current circumstances on the basis of the Articles of Agreement of the Fund as they stand at present. The issue has assumed special importance because of the impasse on SDR allocations that surfaced at the Interim Committee meeting in Madrid in 1994 and that continues unresolved to this day. The Fund management, backed by the technical expertise of the Fund staff, has come to the conclusion that a modest SDR allocation of SDR 36 billion over a five-year period is needed. A majority of Fund members, including all developing countries, the countries in transition, and some industrial countries have supported this conclusion. However, some of the major industrial countries remain unconvinced.

The persistence of this impasse raises suspicions that it reflects not just technical differences on interpretation of the relevant Articles but more fundamental differences about the future role of the SDR in the international monetary system as it has evolved. The first section of this paper discusses the Articles of Agreement as they presently stand, identifying the key ambiguities that leave room for disagreement. The following section examines past practice with SDR allocations to see if it provides a guide for the future. The final section examines the case for an SDR allocation in present circumstances.

The Relevant Articles and Their Ambiguities

The Articles relevant for deciding on an SDR allocation are Article XVIII, Section 1(a), and Article XXII. These are reproduced below, with key phrases italicized for emphasis.

Article XVIII, Section 1(a)

In all its decisions with respect to the allocation and cancellation of special drawing rights the Fund shall seek to meet the long-term global need, as and when it arises, to supplement existing reserve assets in such manner as will promote the attainment of its purposes and will avoid economic stagnation and deflation as well as excess demand and inflation in the world.

Article XXII

In addition to the obligations assumed with respect to special drawing rights under other articles of this Agreement, each participant undertakes to collaborate with the Fund and with other participants in order to facilitate the effective functioning of the Special Drawing Rights Department and the proper use of special drawing rights in accordance with this Agreement and with the objective of making the special drawing right the principal reserve asset in the international monetary system.

Article XVIII, Section 1(a), indicates, in general terms, the conditions under which an SDR allocation should be made. Broadly, it is necessary to demonstrate a long-term global need to supplement reserves, and this must be done in a manner that avoids both stagnation and inflation. Some aspects of these criteria have been further clarified in the Fund Board discussions, but others remain imprecise:

  • (1) The reference to “long-term” needs implies that SDR creation is not meant to respond to cyclical shortages but rather to shortages over a longer period of, say, five years. An implication of this interpretation is that SDR creation need not be in response to a current or immediate need, but in anticipation of a need expected to arise in future.

  • (2) No specific method has been prescribed to quantify the extent to which reserve assets need to be supplemented by allocations of SDRs. Implicitly, quantification of the supplement required depends upon the ability to (a) quantify the demand for reserves, or the optimal level of reserves and (b) quantify the available supply of reserves from various sources. In this context it becomes relevant to consider whether the SDR is an instrument of last resort, to fill the residual gap, if any, after all other sources of supply of reserves are exhausted, or whether it is designed to fill the gap between the demand for reserves and the optimal supply from various sources. Reference to optimality in relation to the supply from various sources is important because it may be possible to achieve any given level of reserves by relying heavily on borrowed reserves, but this may not always be optimal. Even when a reserve buildup through borrowed reserves is feasible, achieving the same objective through an SDR allocation may improve the composition or quality of reserves. Recognizing this aspect, it has been clarified in past discussions that there may be a case for an SDR allocation even in situations where the required level of reserves could be obtained from other sources.

  • (3) The interpretation of the term “global need” also requires some clarification. SDR allocations are clearly not to be triggered in response to the needs of individual countries, or even a small group of countries. But does global need have to be defined in terms of a universal shortage of reserves or can it be defined to cover situations of reserve shortages affecting a sufficiently large number of members? This is obviously a crucial issue in assessing the case for an SDR allocation in current circumstances.

  • (4) Whatever the case for SDR allocations on grounds of reserve shortages, the decision to allocate SDRs must keep in mind both the need to avoid economic stagnation and deflation as well as the need to avoid inflation. In practice, the concern with avoiding inflation has been repeatedly emphasized by industrial countries.

Article XXII is a general exhortation to members to cooperate with the Fund on matters relating to SDRs with the objective of making the SDR “the principal reserve asset” in the system. Developing countries have consistently argued that this objective provides an umbrella justification for periodic allocations of SDRs. The principal reserve asset objective need not be defined crudely in terms of some target share for SDRs as a proportion of world reserves, but equally, it surely implies that cumulative SDR allocations must at least bear some reasonable relationship to aggregate reserves if it is to be taken seriously at all. In practice, this creates a presumption in favor of a steady process of SDR allocation, in line with growing levels of world reserves, though perhaps varying contracyclically with inflationary pressures.

Implications of Past Practice

Ambiguities in the letter of the law (in this case the Articles) are normally resolved by reference to case law, which establishes how the law has been interpreted in the past. Unfortunately, the case law available to us in the matter of SDR allocation is very limited as there have been only two occasions in the past when there was agreement on SDR allocations.

The first allocation of SDRs took place shortly after they were created in 1969, when a total of SDR 9.5 billion was allocated over the three-year period 1970-72. However, the circumstances prevailing then were wholly different from those prevailing today. The par value system was in existence and the dominant concern about managing the international monetary system was the fear of global reserve shortages arising out of the twin problems of a limited supply of gold and the inherent instability of the only alternative, which was to rely upon growth in dollar reserves as a source of liquidity. The SDR, with a constant gold value, glamorously described as “paper gold,” was expected to provide a viable source of reserve expansion, free of the instability associated with expanding dollar reserves based on perpetual U.S. deficits.

This original role for the SDR was overtaken by events almost immediately after the first allocation. The par value system collapsed, and the world shifted to floating exchange rates and multiple currency reserves. The emergence of international capital markets created the possibility of building reserves by borrowing in private capital markets, thus eliminating the link between dollar reserve accumulations and U.S. deficits. These structural changes in the international monetary system necessarily implied that the rationale for future SDR allocations would have to be very different. There was in fact an explosion in world reserves in the 1970s and a significant acceleration in world inflation. These developments were bound to discourage fresh allocations, and there was no agreement on SDR allocations until 1979.

The second allocation of SDRs took place shortly after the Second Amendment to the Fund Articles in 1978, and this allocation has much more relevance to the current situation. Floating exchange rates had become a reality and international capital markets provided many countries with access to reserves if needed. Although the original role envisaged for the SDR had evaporated, the process of reforming the international monetary system, which had been under way since 1973, clearly envisaged a continuing need for the SDR. One of the elements in the reform proposal, later incorporated in the Second Amendment to the Articles in the form of Article XXII and Article VIII, Section 7, was to make the SDR the principal reserve asset in the international monetary system. The precise meaning of this objective was never made clear, and the perceptions of industrial countries and developing countries were very different. Industrial countries were probably concerned primarily with the need to avoid the exchange rate instability that might result from any large scale movement out of dollar reserves and felt that a substitution of dollar reserves by SDRs may have a stabilizing effect. Developing countries had long argued that what they needed was not so much reserves as a steady flow of resources to finance development, and in this context various proposals were advanced for a link between SDR creation and development financing. The idea of a link in all its many variants was firmly rejected by the industrial countries, but they did concede that one of the objects of reform of the monetary system should be to increase the flow of resources to the developing countries. There was no commitment that this increased flow should be through SDR allocations, and it could be argued that the reference was to other flows, including the flow of conditional resources.

Despite these developments, agreement on an SDR allocation was not easily achieved. The major industrial countries opposed an allocation on the grounds that there was no demonstrable global need for reserves. The following extract from de Vries (1985, Vol. II, p. 879), describing the views of Executive Directors on SDR allocations in 1978, bears an uncanny resemblance to the current debate.

The main question relevant to SDR allocations in this situation was whether the global need for reserves should continue to be met mainly through increases in holdings of reserve currencies or whether part of the need for additional reserves could and should be met by allocations of SDRs.

The Executive Directors could not agree on the answer. Some argued that no problem arose with respect to world liquidity if increases in reserves continued to take the form of national currencies. No allocations of SDRs were necessary. Others took the position that any further increases in world liquidity through the Fund should be conditional, that is, in the form of larger Fund quotas rather than of SDR allocations. These Executive Directors argued that, were members to receive more SDRs, some might postpone needed balance of payments adjustment. Recourse to conditional liquidity, that is, to use of the Fund’s resources, would require them to take measures to reduce their balance of payments deficits.

In the end the Managing Director was able to achieve a consensus on SDR allocations as part of a “package” involving not just SDRs but also a 50 percent increase in Fund quotas, which would provide the basis for expanded conditional liquidity from the Fund, an increase in the SDR interest rate, and the use of SDRs to pay for part of the increase in subscriptions resulting from the quota increase. An important feature of the 1978 decision was that it recognized that a decision to allocate SDRs can be justified even though most countries are in a position to achieve the desired levels of reserves from international financial markets. The Managing Director’s proposal also included specific reference to the objective of making the SDR the principal reserve asset as a relevant consideration supporting an allocation. A total allocation of SDR 12 billion was made in three installments in the last three years of the third basic period, 1978-81.

On the face of it, the 1978 decision was taken in circumstances not too dissimilar from the current situation and could be a relevant precedent for the future. However, it appears to have been an isolated event. No further SDR allocations were made in the 1980s despite considerable evidence of reserve stringency in many developing countries. These were the debt crisis years, and most developing countries experienced severe balance of payments difficulties, with serious import compression in many cases. Demands for fresh allocations of SDRs were regularly repeated in successive communiqués of the Group of Twenty-Four but no agreement could be reached. Fund staff papers in this period appeared to endorse the need for an allocation. The ratio of cumulative allocations of SDRs to aggregate nongold world reserves declined from 6.5 percent in 1982 to an estimated 2.5 percent at the end of 1995. This is much below the level of 3.8 percent prevailing in 1978, when the second allocation was made. Like the “dog that didn’t bark,” the failure to reach agreement on SDR allocations in the 1980s has become part of the case law that indicates how member countries interpreted the Articles in this period.

There are two possible explanations for the prolonged SDR drought in the 1980s. First, this was the period of structural adjustment following the debt crisis in Latin America and sub-Saharan Africa, and the dominant perception of the international community was that the solution to the problems facing these countries lay in the implementation of strong structural adjustment policies, supported by provision of conditional rather than unconditional resources. Second, it was also a period when industrial country policies focused heavily on controlling inflation. Inflation was much lower in the 1980s than in the 1970s, but perhaps not low enough for fears of inflation to abate.

In retrospect, we know that although adjustment and economic reforms were undoubtedly needed, many structural adjustment programs, implemented under strict conditionality in the 1980s, were much less successful than expected. This is not to deny the importance of conditional resources provided from the various facilities of the Fund. Such resources undoubtedly have a major role to play in facilitating adjustment in developing countries, but this does not warrant a conclusion that all resources must necessarily be conditional. If limited expansion in unconditional resources is justified on merit, it should not be jeopardized by an insistence upon conditional resources.

The resistance to an SDR allocation because of the concern with control of inflation is more understandable. Most industrial countries had experienced a sharp increase in inflation in the 1970s, which continued into the 1980s. The rate of inflation declined steadily in the first half of the 1980s from 12.4 percent in 1980 to 4.4 percent in 1985. It fell to 2.6 percent in 1986, but then edged up again to 5 percent by 1990. The experience thereafter was, however, quite different, and the rate of inflation has been below 3 percent since 1992. In retrospect, it is difficult to believe that a modest allocation any time after the mid-1980s would have had any significant effect on inflation.

Part of the reason for lack of agreement on an SDR allocation after 1981 probably lies in the change that took place over this period in the attitude of industrial countries toward the Fund, In 1978, there was a much more universal recognition among industrial countries of the role of the Fund in ensuring international monetary stability. The collapse of the Bretton Woods system was still relatively recent, and confidence in the new system had yet to evolve. In 1978, the financing role of the Fund was not viewed, as it is today, as relevant only for developing countries. Both the United Kingdom and Italy had entered into stand-by arrangements with the Fund in 1977, and in 1978 the United States drew on its reserve tranche as part of the Carter Administration’s package for stabilizing the dollar. In the 1980s, the perceptions of industrial countries about the role of the Fund changed considerably. Although problems of coordination of macroeconomic policy among major countries remained serious, the Fund was not always seen as having an active role to play in this process. Increasingly, the Fund was seen as relevant primarily to handle problems of developing countries and, more recently, those of the countries in transition. This inevitably led to a reduction in the bargaining power of developing countries and to a greater focus on conditionality attached to Fund resources.

The Case for an SDR Allocation Today

We now turn to an assessment of the case for an SDR allocation in present circumstances. The arguments for an allocation are well known and have been recently summarized in Buira and Marino (1995). They are reviewed below with an attempt to identify those elements of the argument on which there is disagreement and consider how they stand up against established practice in interpreting the Articles of Agreement.

One element of the argument that is generally agreed is that there will be a substantial growth in demand for reserves, in proportion to world trade or only slightly more slowly. Buira and Marino (1995) estimate an increased demand for reserves of about SDR 400 billion over the next five years, whereas Mussa in his paper for this seminar puts the figure a little lower, at SDR 350 billion. These estimates are broadly accepted, allowing for the usual margins of error. The contentious issue is whether it is necessary to allocate SDRs to meet any portion of this increase. Opponents of SDR allocations argue that there will be no difficulty in obtaining the necessary increase in reserves from the financial markets. Experience suggests that this may well be so for total reserves. Between the end of 1981 and 1994—a period in which no fresh SDRs were allocated, the ratio of nongold reserves to merchandise imports for all countries increased from 19 percent to 27 percent.

Special Needs of Some Countries

The case for a fresh SDR allocation therefore rests not on the expected shortfalls in the potential supply of aggregate reserves but rather on the distribution of global reserves in the absence of SDR allocations. The apparent adequacy of aggregate reserves at present hides considerable variation across countries, with many countries suffering from inadequate reserves. About a third of the developing countries and more than half of the transition countries had reserves below 8 weeks of imports in 1992. If the more conventional criterion of 12 weeks of imports is used to identify reserve shortages, almost two thirds of developing countries and three fourths of transition countries have inadequate reserves.

These countries do not have adequate access to world capital markets and cannot rely on that source to build up reserves at a reasonable cost. They would either have to borrow at high cost to build reserves or more likely could only rebuild reserves to adequate levels by adjusting their current account positions. In practice, this often means following contractionary policies aimed at import compression, which have often led to a prolonged slowdown in growth. This imposes a high cost not only on these economies but to some extent also on the rest of the world, if we take into account the collective impact of such policies in many developing countries. These costs have to be compared with the cost of a fresh SDR allocation, which is demonstrably quite low.

Industrial countries have consistently opposed allocation of SDRs to meet the reserve needs of a group of countries on the grounds that the Articles refer to allocations to remedy global reserve shortages and not shortages affecting groups of countries. Opponents of an allocation recognize that some countries do not have access to financial markets, but they argue that is so because of deficiencies in those countries’ policies. The denial of credit by financial markets to such countries is therefore a desirable market signal emanating from well-functioning markets, which serves to encourage these countries to adopt necessary adjustment measures. Provision of low-cost liquidity on an automatic basis through SDR creation in these circumstances is seen as a distortion of the market mechanism, weakening market signals and encouraging postponement of adjustment. An extension of this argument is that if some developing countries are indeed facing resource shortages, these resources are best provided through one of the conditional facilities of the Fund and not through the allocation of SDRs.

Some of these arguments are clearly based on misperceptions. One relates to the issue of whether SDR allocations imply an unjustifiable subsidy. As pointed out in Coats and others (1990), an SDR allocation implies a net grant element in favor of net users of SDRs only if the interest rate paid to SDR holders is not competitive with other reserve assets, or if the rate of charge on allocations does not include an adequate risk premium. As it is generally conceded that SDR interest rates are competitive with the return available on other comparable reserve assets, there is no subsidy on this account. For the rate of charge, it is certainly true that most net users of SDRs would pay significantly higher rates when borrowing in private markets, reflecting the market perception of risk premium. However, this does not necessarily imply that the rate of charge on the SDR is subsidized. It can be argued that net users of SDRs are highly unlikely to default on net interest payments on their SDR use, because they value their status as members cooperating with the Fund. If members do indeed display a greater commitment to meet their obligations arising out of SDR use, there is no case for charging a risk premium. The lower cost of reserves obtained from an SDR allocation is therefore not at the expense of other participants in the system. It reflects a genuine efficiency gain to the world economy arising out of the status of the Fund as an international institution and the greater compliance that it can evoke from members.

The concern that SDR allocations will weaken market signals also does not stand up to detailed scrutiny. As Polak (1988) pointed out, there is no reason to begrudge countries an SDR allocation that is otherwise justified simply because it might loosen the tight grip of conditionality. There are other favorable outcomes that might have the same effect, and we do not react to them in the same way. In any case, the impact on the incentive to adjust depends upon the size of the allocation. An allocation of SDR 36 billion over a five-year period amounts to only 25 percent of Fund quotas, or an average increase of 5 percent of quota a year. Additional availability of unconditional resources of this size is unlikely to have any impact on the willingness of countries to undertake adjustment, as adjustment-based programs can make available much larger amounts of Fund resources, up to 110 percent of quota.

To some extent the recent proposal of the Group of Seven countries for a special allocation on grounds of equity in favor of the countries in transition itself implies that additional allocation of SDRs justifiable on merits need not be seen as a weakening of the adjustment programs. The same argument can be extended also to a general allocation.

The Quality of Reserves

Even if we accept that many developing countries may be able to build up adequate reserves by accessing international financial markets, a case can be made for SDR allocations on the grounds of the quality of reserves. Any given stock of reserves acquired through a greater reliance on borrowing in international financial markets implies a greater vulnerability to changes in market conditions. Borrowed reserves have to be refinanced periodically, and unexpected changes in international capital market conditions may lead to a sudden withdrawal of such finance. SDR holdings are also borrowed reserves in one sense, but since the allocations are permanent (except in the event of cancellation—which is theoretically possible), they are always available to the holder as an assured line of credit, irrespective of changes in financial market conditions. A reserve composition with a greater SDR component therefore represents a superior quality of reserves, providing a greater assurance of stability in the system.

It is important to emphasize that this concern with the quality of reserves is not merely theoretical. It is of immediate practical importance in a world in which international financial markets, though enormously important and capable of great flexibility, are also far from perfect. The experience with commercial bank lending to developing countries in the 1970s and 1980s amply demonstrated the extent of that imperfection as the markets first lent imprudently to developing countries, and then withdrew excessively, necessitating extensive official action to avoid destabilizing shocks and to restore reasonable flows of funds to these countries. The same problem surfaced again in the aftermath of the Mexican cnsis, when there were widespread fears of “contagion effects” on other developing countries. The fact that these effects were quickly contained only testifies to the effectiveness with which the Mexican crisis was handled, but it does not detract from the assessment of the underlying vulnerability of the system. This vulnerability would be reduced if a larger portion of the world’s reserves were to come from SDR creation.

The concern with the quality of reserves can only be heightened by the changes that have taken place in the relative size of resources available from official and bilateral credit arrangements that constitute a safety net for the system. The relatively lower level of reserves held by industrial countries compared with developing countries is partly a reflection of the availability of a more extensive safety net available to these countries. These arrangements include resources potentially available through the Fund, including the General Arrangements to Borrow and SDRs, the U.S. Federal Reserve’s reciprocal currency arrangements, and the European Monetary System’s support facilities. It is estimated that these resources aggregated SDR 110.7 billion in 1979, or about 40.5 percent of world nongold reserves. By 1990 the absolute size had increased to SDR 186.2 billion, but as a proportion of total reserves they had declined to 29.1 percent.

The Impact on Inflation

Avoidance of inflation is explicitly mentioned in the Articles of Agreement as a relevant consideration in deciding on SDR allocations. Any argument in favor of SDR allocations must therefore address this issue. The nature of the linkage between SDR allocations and inflation is well known. An allocation by itself cannot lead to any inflationary pressure unless it leads to additional spending by some of the recipients. It is possible to argue in principle that an SDR allocation leads to no additional spending, and the entire amount of the allocation is simply added to reserves. But this is unlikely to happen. In practice, one should expect that an SDR allocation will lead to some increase in total reserves, some switching from other borrowed reserves to SDRs, and some increase in spending. If SDRs are used to finance additional expenditure by some recipient countries, this leads to an accumulation of reserves in the country in which they are spent, and this reserve buildup could be associated with an increase in money supply and therefore in prices in that country. This, however, depends upon whether the monetary authorities sterilize the reserve accumulation by open market operations.

As the monetary authorities in the major industrial countries normally sterilize the impact of any reserve accumulation, it is reasonable to conclude that a modest SDR allocation would not lead to any inflationary pressure, especially when we consider the quantitative magnitudes involved. For example, a general allocation of SDR 36 billion over five years involves an average annual allocation of a little over SDR 7 billion a year. Of this amount, only about SDR 2.7 billion would accrue to developing countries and countries in transition. If we assume that half of this is spent, the additional expenditure would be only about SDR 1.4 billion a year, which would be reflected in an increase of this magnitude in reserves of different industrial countries. The monetary authorities of industrial countries should have no difficulty in sterilizing accretions to their reserves of this size.

The fear of inflation following a modest allocation of SDRs therefore appears grossly exaggerated. This conclusion is only reinforced by the fact that inflation in the industrial countries has been at a historic low of about 2.5 percent a year for the past three years, and there are no signs that it is about to accelerate. Even if we concede that the fear of inflation, which held back allocations of SDRs in the 1980s, had some justification in terms of inflation expectations not having been reduced, no such problem exists today.

The SDR as Principal Reserve Asset

Finally, we need to consider the case for an SDR allocation from the point of view of the objective of making the SDR the principal reserve asset of the system. Developing countries fully recognize that this objective should not be interpreted to mean continuing large allocations irrespective of the availability of liquidity from other sources. However, it must also be recognized that in the absence of an SDR allocation, with the expected growth in total reserves, the ratio of SDRs to nongold reserves will fall to less than 2 percent at the end of five years. This proportion was 3.8 percent when the second allocation was made, which raised the percentage to 6.5 percent in 1981. An allocation of SDR 36 billion would raise the proportion only to about 5 percent.

Industrial countries clearly believe, even if they do not always assert, that the objective of making the SDR the principal reserve asset of the system has lost its relevance. The earlier fears of exchange rate instability consequent upon a flight from dollar reserves have abated as the system of multiple currency reserves appears to have functioned reasonably well. However, industrial countries would not go so far as to say that the SDR has outlived its usefulness and should be abolished. There is probably a consensus that the SDR is an extremely useful instrument that can provide a multilaterally controlled safety net that could be quickly deployed in the event of an unforeseen strain on the financial system. Even this minimal concession is relevant in considering the case for an SDR allocation. The effectiveness of the SDR in being able to deal with unforeseen emergency situations depends upon the instrument not atrophying through lack of use. A fresh allocation at this stage, 15 years after the last allocation, is surely warranted for this reason alone.


We can now pull together the main strands of the argument developed in this paper. The Articles of Agreement can at best provide only broad directions about the considerations that must guide the allocation of SDRs. Interpretation of the Articles has to evolve over time, taking into account precedents and the current consensus about the world economic situation.

It is clear that the world economy has changed substantially from when the SDR was first invented, and even from the time of the Second Amendment. The rapid growth of international financial markets has created a situation in which industrial countries, and even many developing countries, can meet their growing demand for reserves from the market—at least in normal times. However, this is not true for a very large number of countries for whom the markets are either closed or too expensive. Nor can we say with any confidence that market perceptions of the creditworthiness of developing countries, which are reflected in the risk premiums that they have to pay, are especially accurate. For all their sophistication, international financial markets remain vulnerable to herd instincts and contagion effects, which can introduce a high degree of volatility into the access of individual developing countries.

In this situation a case can be made for periodic general allocations of SDRs as envisaged in the Articles, especially when inflationary pressure in the world economy is low. The definition of global need should be interpreted, as it was in 1978, with sufficient flexibility to allow a modest to moderate allocation of SDRs from time to time. Although such an allocation may seem to be needed primarily by a group of countries, a general allocation also improves the quality of aggregate reserves for the world as a whole. It also helps to maintain the SDR, if not as a principal reserve asset, then at least as a credible element in the international monetary system, with the potential to provide a much larger safety net if needed in an emergency.

No Future for the SDR

Horst Siebert 1

Michael Mussa has presented an inspiring paper making a case for moderate SDR allocations and providing an informative historical background. However, I come to a different conclusion.

SDRs were introduced (as Mussa points out) to solve the credibility problem that arose in the 1960s with central bank reserves in U.S. dollars increasing and with U.S. gold reserves declining: the so-called Triffin dilemma. Thus, SDRs were a solution to a specific problem. But the initial raison d’être of SDR allocations no longer exists. Do we still need them?

SDRs can be interpreted as a form of “paper gold” or “monetary gold,” that is, as an artificial international reserve. In a broad interpretation, they provide, free of charge, a credit option, or more precisely, an option to obtain foreign currency. When the option is used, a (privileged) interest rate is applied. The allocation of SDRs can thus be understood as a credit (or currency option) arrangement for international reserves where the credit option (currency option) is guaranteed by an international agreement that also specifies the conditions under which a credit (or currency) can be obtained. Starting from this interpretation, I want to make the following points.

It is true that we observe an increase in international reserves of central banks (in absolute terms) with an expansion in world trade. But such an increase in international reserves (documented by an import elasticity of reserves of, say, 0.85) is not a sufficient reason for SDR allocations, as Mussa himself points out. The ratio of reserve holdings to merchandise imports has remained surprisingly stable for the world as a whole and for industrial countries in the last 45 years, while increasing for the developing countries. This stability in the ratio of reserves to imports can be taken as an indication that the system has been capable of generating the reserves it needs.

Our specific question is not whether we need more reserves per unit of imports—which we apparently did not in the past—but whether we need more SDRs. In order to make the case for an additional SDR allocation, one would have to show (according to the First Amendment) “a long-term global need … to supplement existing reserve assets.”

Mussa does not succeed in proving that there is a global need: more SDRs are not a necessary condition for an increase in world trade. The share of SDRs in international (nongold) reserves declined from a maximum value of about 8 percent in 1972 to 2.3 percent at the end of 1995, but world trade has expanded substantially. Obviously, international trade was not restricted by the limited volume of SDRs. It is difficult to see how an expansion of SDRs would bring about an additional expansion of world trade.

Besides the problem of proving a “need,” the additional issue in Mussa’s paper is one of transforming the need into a quantitative yardstick of how much of the need should be supplied by SDRs, since SDRs should play only a supplementary role. Accepting the institutional arrangements of SDRs for the sake of the argument, economists should be able to have a basic model in which the optimal quantity of SDRs is determined by comparing the marginal costs and the marginal benefits of one additional unit of SDR, similar to the manner in which the optimal demand for international reserves can be specified. If marginal benefits exceed marginal costs, more SDRs should be provided. To address this issue requires a clear definition of the maximization problem; an important question is which agent’s preference function should be maximized. As we do not have such a model, the economic case for SDR allocation is weak.

Mussa’s answer is that some additional SDRs cannot hurt, especially as costs to net users are low. From a methodological point of view such a pragmatic answer is not convincing. It is also not convincing that the share of SDRs in international reserves given in a specific year in the past should remain constant. Why should it?

Thus, we do not have a clear quantitative yardstick with which to determine what the need is and which part of the need should be provided by SDRs.

The core question is not, of course, whether it is optimal to provide one additional unit of SDR within the given institutional arrangement but whether the system of SDRs contains institutional details that are not optimal and whether the institutional arrangement as such is meaningful.

Suboptimal institutional details distort behavior; such institutional details may be corrected, however, without jeopardizing the approach itself. Interest rates are such a case. Mussa mentions that the SDR interest rate is a market rate representing a weighted average of five countries whose currencies are in the SDR basket. This is only part of the picture. First, the SDR rate is a short-term rate, whereas SDR credits have increasingly changed into long-term credits. Since the reconstitution requirement was dropped in 1981, allocated SDRs can de facto be used for an unlimited time. Second, the SDR interest rate does not contain a risk premium, for instance, a premium for country risk. Consequently, distortions are implied (Schroder, 1990, p. 71). The short-term rate induces countries that have a deficit in their balance of payments (or that are threatened by depreciation of their currency) to use SDRs instead of paying the higher interest rate in the world capital market for long-term credits. This applies especially to countries facing a high-risk premium in the world capital market. For countries with a surplus, it is not attractive to hold SDRs relative to other reserve assets because of their relatively low short-term rates; in a way the IMF rules make it possible to force such countries to buy and hold SDRs by being designated.

To reduce these distortions, long-term rates should be used for SDR allocations and credits. This practice would bring the interest rates on SDR allocations partly into line with the market rates (with the exception of country risk premiums) and, as a consequence, partly establish the right incentives for deficit countries as well as for surplus countries. Thus, a reintroduction of the reconstitution requirement would not be quite so necessary.

A much more important issue is to what extent incentives of countries with deficits are systematically distorted by SDRs. This issue cannot be addressed adequately by suggesting (as in Mussa’s paper) that the scale of magnitude is negligible. Institutional arrangements have to be judged according to the criterion of incentive effects rather than according to the criterion of scale or of magnitude.

Apparently, there are incentive effects of SDRs. Consider two different types of countries, one willing to follow a stabilization policy with respect to its price level, for instance, by implementing a currency board approach to monetary and exchange rate policy, and the other being more reluctant to comply with stabilization constraints that allow a higher inflation rate and a depreciation of its currency. For this more inflation-minded country, a generous provision of SDRs is indeed not an incentive to steer national monetary and fiscal policies in such a way as to obtain stability.

The allocation of SDRs per se does not affect the monetary base of a country, as SDRs represent an asset as well as a position on the liability side of the balance sheet of the central bank (in a bookkeeping sense). The allocation of SDRs is therefore an enlargement of the balance sheet of the central bank. When SDRs are used to augment international reserves and when they only change the composition of the asset side of the central bank balance sheet, they do not affect the domestic monetary base, and the use of SDRs has no inflationary effect. In this case, the additional reserves are not used for interventions in the currency markets. However, if SDRs are used to cover an inflationary monetary expansion that is already going on and to finance a balance of payments deficit (arising, for instance, from domestic inflation) that is not sustainable in the long run, SDRs make a more inflationary policy possible. In this case, the additional reserves obtained from SDRs are used for interventions that defend an exchange rate for some time and camouflage the discrepancy between monetary policy and an exchange rate target.

Interestingly, Mussa admits that easy access to international reserves for countries with an acute shortage of reserves is insufficient to establish a global need. As a matter of fact, he points out “that the credibility of the position of many proponents of SDR allocations is not strengthened by the fact that they are persistent net users of SDRs.”

The incentive issue for deficit countries becomes apparent when conditional IMF credits and the unconditional use of SDR allocations are compared. Conditionality imposes conditions on the stabilization policy of a country that requires structural changes and institutional reforms. SDR allocations do not have a similar impact that allows deficit countries to finance balance of payments disequilibria and to postpone necessary adjustments. Thus, SDRs represent a moral hazard problem for the stabilization policy of the deficit country and give the wrong incentive. Inasmuch as SDR allocations induce even one country not to follow a stability-oriented policy, they establish the wrong incentives (Sachverständigenrat, 1995, p. 244). Institutional arrangements with wrong incentives are not good institutional arrangements. It seems to me that Mussa plays down these policy costs of instability.2

Between the lines, distributional considerations enter into Mussa’s argument in favor of providing access to liquidity at low (or in Mussa’s terms “zero”) cost (“There, but for the grace of God, go I”). Without being explicitly mentioned, the political demand for linking SDR allocations and economic development is nevertheless in the background. Here the problem arises as to what weight should be given to equity considerations in monetary institutional arrangements. In some domestic policy issues where “needs” are articulate, institutional safeguards—including constitutional restraints—have been implemented that let such overall restraints as a budget restraint for the government or for a country (in the form of the balance of payments) dominate other decisions. Such institutional arrangements make it harder to trade short-term (political) benefits for long-term (economic) costs. Examples are constitutional budget constraints and the institutional arrangement to depoliticize the money supply with the institution of an independent central bank. This example should remind us that equity considerations should not play any role in designing the institutional arrangement for international reserves.

The analogy to the domestic economy is telling. Poor people are short of money; this cannot be a reason to print more money. Therefore, “it makes no sense whatever to talk about the reserve adequacy problems of a subgroup of the international community” (Chrystal, 1990, p. 14). SDR allocations cannot substitute for the reduced willingness of industrial nations to finance a real resource transfer out of their government budgets, that is, out of savings.

The overriding issue of our debate on the future of the SDR is, of course, “whether the IMF should still be in the business of supplying reserves through the SDR mechanism,” as Mussa correctly puts it.

The international economy today is different from the time of Bretton Woods. Consider a scenario with n currencies and n central banks steering the national money supplies according to some criterion of price level stability. The central banks may directly target the price level, or they may steer monetary aggregates so as to keep the price level (relatively) constant. For the sake of argument, consider a case where all the countries increase their money supplies according to the growth of their production potential, correcting for a change in the velocity of money. If all countries move within these limits and if fiscal policy (and other economic policies) do not contradict monetary policy, exchange rates would remain stable.

If a country deviates from this international concert by allowing a higher national inflation rate, expectations for exchange rate variations will build up and movements of portfolio capital will affect exchange rates and force the currency to depreciate (as has recently been experienced in Europe). Expectations of exchange rate variations will also form in line with long-run changes in the supply side. Thus, the international monetary system can be understood as an interplay of the purchasing power parity (in a very broad interpretation) that forms the exchange rate expectations and of the interest rate parity that determines capital movements (especially of portfolio capital) and the short-term movements of the exchange rate. Exchange rates can overshoot, but, in principle, financial markets have a disciplinary function for national monetary policy and stabilization policy in general.

If national central banks need international reserves, they can obtain them in several ways. One way is for a country to run a current account surplus. A country gives up domestic goods and obtains an entitlement for foreign goods, either today or in the future, International reserves are thus obtained by having an excess of savings (private plus public sector) over investment. Another way is for a country to run a deficit in the capital account and thus obtain international reserves. This, however, implies repayment of debt later. It also depends on whether the country is creditworthy. Finally, a central bank can obtain foreign currency through portfolio shifts. For instance, it can buy foreign financial assets with domestic currency while at the same time foreigners buy domestic financial assets (for instance, shares of a firm) with foreign currency (Chrystal, 1990, p. 13).

In all these cases, countries incur costs of obtaining international reserves. They enjoy the benefit of having international reserves, which can buffer real adjustment problems.

In such a setting—which is characteristic of today’s world—central banks have ample options to obtain foreign currency. The concept of “liquidity shortage” as a systematic phenomenon for the world as a whole does not make sense. Of course, there are costs to obtaining reserves, but there is no supply restraint for reserves if they are not supposed to be obtained at a zero price or a low price. Now it is even being questioned whether a structural reserve adequacy problem existed in the 1960s (Chrystal, 1990, p. 14). This brings us back to the concept of global need. The term need is a moral or a political category, not an economic one. There are many needs in the world, and we are not able to satisfy all of them. For a need to become an effective demand, an institutional setting including prices (and costs) has to be specified under which the need can be revealed. Otherwise, a need cannot be defined.

According to this reasoning, monetary policy is inseparable from exchange rate policy; this, incidentally, requires that central banks, and not national governments, be in charge of exchange rate policy. It is questionable to talk of a “need for international liquidity” without looking at national money creation (Salin, 1990, p. 77). A country’s demand for international reserves clearly depends on its national monetary policy (and other stabilization policies) as well as on its exchange rate target. If there is an exchange rate target, international reserves represent a buffer allowing excessive money creation while still maintaining the exchange rate for a while. Consequently, the demand for reserves—or is it more appropriate to use the term need?—should increase with the importance of an exchange rate target or, more precisely, with the intensity of the inconsistency between monetary policy and the exchange rate target. When reserves decline, it is a signal that money creation was excessive (Salin, 1990, p. 78), leading to what Mussa calls “the potential embarrassment of devaluation.” Applying this philosophy, we have to define the demand for international reserves on a national level depending on the country’s willingness to accept inflation and devaluation.

To resolve the current issue of new members of the IMF not yet having received SDR allocations, the given volume of SDRs could be reallocated according to each country’s share in the sum of quotas. This would be consistent with the long-run target of reducing the role of SDRs rather than increasing it.

A political compromise may be to endow the new members of the IMF that have not yet received SDR allocations with additional SDRs. This action, however, would require a change in the statutes of the IMF, making the SDR allocation to new members a special case that does not represent a general expansion of SDRs. A general expansion would be a precedent that would open the door for other expansions of SDRs in the future, because it would represent staking a claim in institutional change in the wrong direction. A precedent for the future extension of SDRs should not be established.

So what is the conclusion? Let us ignore the issue of whether there was a structural reserve adequacy problem in the 1960s and whether the creation of SDRs “was a mistake” (Chrystal, 1990, p. 14). If the proposal were made to introduce the special drawing right today, one would not do it. One can leave it to individual countries to obtain their international reserves as they need them; each country should run its own monetary policy according to its own inflation aversion.

For a financial crisis, the SDR is not the right instrument anyway, and the IMF commands enough policy means to deal with such a case (Sachverständigenrat, 1995, p. 243). Insofar as SDRs have an organizational and bureaucratic justification for existing, their role should not be expanded. There is really no future for SDRs. The IMF may be well advised not to link its legitimacy to the expansion of SDRs. It would be better to concentrate on the problem of preventing national financial crises from spreading worldwide and on helping countries to gain access to financial markets.

With respect to the quotation from Adam Smith, I wonder whether Michael Mussa found the right quotation to support his case. The summary sentence for the paragraph quoted by Michael Mussa reads, “Paper is more convenient to the Americans for home trade,” and the summary sentence for the next paragraph in The Wealth of Nations is “while for their external trade they use as much gold and silver as is necessary.”

To conclude, I quote Adam Smith from this next paragraph—and if you substitute international currency reserves for gold and silver, you have the answer to our problem:

In the exterior commerce…, gold and silver are more or less employed, exactly in proportion as they are more or less necessary. Where those metals are not necessary, they seldom appear. Where they are necessary they are generally found (p. 894 in 1937 edition).

Interpreting “Long-Term Global Need” J. de Beaufort Wijnholds

The concept of the “long-term global need” to supplement existing reserves has confounded us for a long time. Just as the Fund struggled with the hard-to-operationalize concept of “fundamental disequilibrium” under the original Articles, it has since the Second Amendment been faced with the need for exegesis of Article XVIII.

Mussa seems to take up the role of theologian with relish, calling on certain countries to love their developing brethren a little more and to realize that, “There, but for the grace of God, go I.” While he is probably not entirely off the mark when suggesting that opposition to or reluctance about SDR allocations has to do with the low cost of holding reserves for industrial countries, I think there is more to it. I believe that, at least for some countries, there is also the concern about the proper functioning of the international monetary system and, closely connected to that, the feeling that considerations of a predominantly development nature should be limited in a monetary institution such as the Fund.

There is much in Mussa’s paper with which I can agree. I particularly agree with his view that the most convincing argument in favor of SDR allocations has become the cost consideration. Coming back to exegesis, however, one could question whether the argument of cost is strictly speaking a valid consideration when deciding upon the “long-term global need.” I am not sure that the architects of the amended Articles had this consideration clearly in mind.

This brings me to another matter of interpretation, namely, the fact—as mentioned by Mussa in his footnote 15—that the Fund’s lawyers believe that the objective of making the SDR the principal reserve asset does not constitute a criterion for allocation. And to digress for a moment, and to refer to his footnote 16, I believe that the only realistic chance to have made the SDR anything close to the principal reserve asset was at the time the substitution account was discussed. It was ultimately rejected on not very clear grounds. After 16 years I still find this regrettable and even wonder whether an attempt at reviving the notion of a substitution account might not be worthwhile. In this respect, I refer to recent reports concerning reserve diversification, particularly by Asian countries. Substituting SDRs for the dollars that central banks wish to switch into yen and deutsche mark could also be attractive from the point of view of preventing unnecessary pressure on the dollar exchange rate.

Returning to the main argument, although Mussa and I are in broad agreement on the desirability of keeping the SDR alive, I have some difficulty with the approach that he and the Fund staff have taken with respect to the quantification of the need for reserves. In my view this approach tends to produce results that overstate the need for reserves.

Let me return for a moment to the advent of widespread floating after the breakdown of the par value system. I agree that the reduction of the need for holding reserves was not as dramatic as was thought at the time, and in fact an article of mine that was published in 1974 argued that floating would generally mean managed floating and that therefore a considerable need for reserves would remain.1 I do not believe, however, that it is accurate to say that because the ratio of reserves to imports—analytically a rather imperfect indicator of reserve need—has remained roughly constant since the mid-1970s, the need for reserves has continued to grow as it did under the post-Bretton Woods system. I believe that quite a number of countries have over the years accumulated foreign exchange assets through interventions beyond what could be considered an optimal level but did not lose reserves in subsequent periods as they continued to have strong external positions. Industrial countries in weaker positions could generally borrow in the markets to build up reserves. Hence, the fall in the reserves to imports ratio (for industrial countries), which one could have expected, did not occur. Indeed, an analytical case has long been made for the existence of economies of scale in the holding of reserves—the argument going back to the work of Baumol on inventory theory.2 The Fund staff does take this into account to the extent that an import elasticity of reserves of 0.85 is applied. However, this figure could be on the high side, particularly for industrial countries.

Let me now comment briefly on indications that reserve levels for quite a number of countries, including some emerging market economies, are higher than optimal. First, the so-called symptomatic method for assessing reserve adequacy, which looks for qualitative signs of a shortage of international liquidity,3 does not point to widespread behavior that may be considered indicative of such shortages (that is, in particular a tendency to impose restrictions on capital outflows and on imports of goods and services, or a tendency to keep interest rates high to protect reserves). On the contrary, liberalization in industrial and many other countries has been triumphant. Also, some countries, such as Japan, have lowered interest rates considerably while still taking in large amounts of foreign exchange. Japan and some other Asian and European countries have accumulated reserves beyond levels that they would conceivably need for future interventions to counteract unwanted pressure on their currencies. Such accumulation has clearly been the outcome of considerations other than a perceived need to hoard such huge amounts of reserves. I would venture to say that quite a few central banks these days are partly acting as de facto investment funds for their governments. One indication of this development is that many central banks are known to have been aiming for higher yields on their reserves and in the process have put part of them in less than fully liquid assets. It could be an interesting exercise to survey individual central banks on their perceived reserve ease as well as on their investment behavior.

This of course does not mean that no countries are suffering from reserve shortages. They certainly exist among the developing countries and particularly among the transition economies. The case for an “equity allocation” of SDRs—providing SDRs to those countries that did not take part in earlier allocations—remains a compelling one in my view. However, this requires amendment of the Articles. And since Mussa’s paper deals specifically with allocations under the present Articles, I will not go into this matter now. The situation under the present Articles is that in comparison with quite a few industrial countries and some emerging market economies with more than adequate or clearly excessive reserve levels, there are developing and transition countries with reserve shortages that in purely quantitative terms are most likely to be relatively small in the global context. And, since under what I believe is a valid interpretation of the concept of global reserve need the existing maldistribution of reserves is not a criterion for allocation, creating new SDRs cannot be the solution to this problem under the present Articles.

Briefly looking into the future, I believe that the creation of a European Monetary Union (EMU) will contribute to a clear reduction in the need for reserves for the participating countries. It is also likely to be reflected in a more than marginal fall in the ratio of reserves to imports for industrial countries and the world as a whole. The Treaty of Maastricht requires countries participating in the EMU to pool their reserves in an amount of ECU 50 billion. (This requirement is based on 15 participating countries, which may not initially be the case. The amount is substantially less than is presently held by these countries.)

The following main effects on international reserves may be expected from the inception of EMU. (1) It should substantially reduce the need for reserves in Europe as intervention needs between participating countries will disappear. (2) Actual reserve holdings of the European central bank plus the participating national central banks will be considerably smaller than were held by those central banks prior to EMU, because holdings of deutsche mark will become assets denominated in euros and therefore domestic assets. This would erase a sizable portion of European reserve holdings without affecting reserve ease.

Although it is too early to say anything about the EMU with any precision, it seems useful to realize that once it is established it will have an important impact on the assessment of the need for international reserves.

Finally, I wish to make a few brief remarks on Mussa’s view that the real difference of opinion concerning the SDR is not about the quantification of the global reserve need but rather about whether the Fund should remain a supplier of unconditional liquidity. Although I take the view that important differences of substance in the area of quantification exist, as I have attempted to explain, I agree with Mussa that more fundamental doubts about the SDR by a number of Fund members may play an important role in determining their position with respect to allocations. I also agree with him that the argument of SDR allocations producing visible inflation is not convincing for the amounts that could be termed modest or moderate. The argument concerning the temptation to use SDRs to pursue inappropriate policies holds more sway, I believe, although here too the scale of allocations is relevant, as he observes. Although I agree that modest or even moderate allocations would not seriously undermine the international adjustment process, I feel more comfortable erring on the side of generosity in deciding upon an increase in conditional Fund liquidity (quotas) than I do in deciding upon SDR allocations.

Rationalizing SDR Allocation

John Williamson 1

Michael Mussa’s paper traces the causes of the long-standing stalemate over the propriety of further SDR allocations in terms with which it is difficult to disagree. The SDR was created in the 1960s as a primary reserve asset to supplement gold, for a world in which the stock of primary reserve assets was presumed to have long-run significance for the evolution of magnitudes like global inflation, and in which it was reasonable to assume that a shortage of primary reserve assets would lead to financial instability (in the form of the feared run on the dollar) or competitive payments objectives. The criteria to govern SDR allocations and cancellations were drafted for that world and enshrined in the First Amendment to the Fund’s Articles.

But then the Bretton Woods system collapsed, and the attempt to restore a modified version of it failed. As a result, the principal primary reserve asset—gold—lost any monetary role. The SDR was redefined as a basket of currencies, most of which were floating against one another. When the Articles were amended a second time to recognize most of the new realities, however, the Article establishing the criteria to determine allocation was left unchanged. It has proved a recipe for deadlock.

Mussa points out that, even in the new world of floating exchange rates, the global level of reserves continues to increase. This increase takes the form of holdings of reserve currencies; there is no longer any danger that their supply will fail to respond to an increase in the demand, or that exogenous, capricious changes in reserve supply will create systemic instability. Hence, none of the consequences of a shortage of primary reserve assets that motivated the creation of the SDR will materialize if no SDRs are created. In the sense that the term “need” was used by its creators, there is no need for an allocation: a country that does not want to see any SDR allocations can appeal to that fact, which is incontrovertible.

Many countries, however, continue to want SDR allocations. They reason that they are going to need to increase their reserves over time, and that it is less satisfactory to do this by borrowing than by receiving SDR allocations, partly because it is more expensive and partly because there is more risk of the reserves evaporating just when they are needed. The countries that can expect to gain most are those that find it most expensive (or, in extreme cases, impossible) to borrow on the international capital market, which tend to be poor countries; but all, except the suppliers of reserve currencies and countries that can borrow on terms as fine as they do, stand to gain. Mussa writes (p. 80 of this volume) that

… the vast majority of the Fund’s membership accounting for about half of Fund quotas2 faces net costs of holding reserves that are higher, in many cases significantly higher, than the true economic cost of creating reserves through SDR allocations. This, in my view, falls well into the range of a global need for SDR allocations to supplement other forms of reserve growth.

Now, as argued above, this is not the concept of global need that was in the minds of the creators of the SDR system. But the meaning of words can evolve to fit the times, and if there was a general will to resume allocations, Mussa’s reinterpretation could provide a linguistic veil that would avoid the need to go to the trouble of amending the Articles. The problem is that there is no such general will. And even if one agrees with Mussa that the phrase “global need” cannot reasonably be interpreted to mean that all countries, or that countries with more than 85 percent of Fund quotas, feel a need for an SDR allocation, the “general will” of the Fund is precisely that that can command an 85 percent majority.

That being so, it is necessary to examine why some Fund members resist allocations. Mussa lists three main concerns that have been cited by opponents:

the potential for adding to world inflation, the potential for allowing countries to pursue undesirable macroeconomic policies, and the potential for some countries to increase their absorption of goods and services at the expense of others.

He dismisses these on de minimis grounds. I have never found this a particularly compelling rationale: if something is bad, it is a comfort to know that it is only modestly bad, but that does not make it worth doing. Hence I will try and evaluate the three arguments on their merits.

It will help to have a model of the normal impact of an SDR allocation in our minds. My model is as follows. Think of a typical nonreserve center, or “peripheral,” country, P, that receives an allocation of 100. The authorities seek to maintain the reserve level constant3 by inducing a reduction in external borrowing, which requires either that they sell 100 of their foreign exchange reserves on the exchange market and allow the exchange rate to appreciate or that they expand domestic credit by 100 and hold the exchange rate constant as 100 of their foreign exchange reserves flow out. In either event the private sector will end up with 100 less of debts to the reserve center, R, and the central bank will end up with 100 less of R’s treasury bills. There will as a result be a savings to P of 100 (p—r), where p is the interest rate that R’s private sector charges to P’s borrowers and r is the interest rate that R pays on its treasury bills. Since the SDR interest rate, s, is set as the average treasury bill rate on the currencies in the SDR basket and interest arbitrage keeps these interest rates in line ex ante, s can be assumed equal to r, stochastic factors aside, and it would make no difference to P’s expected saving if its central bank converted some of its SDRs into R’s treasury bills.

With the above model in mind as the standard case, let us consider whether an SDR allocation4 would be inflationary. In the textbook case of perfect capital mobility, neither the sale of reserves nor the expansion of domestic credit would have any impact on interest rates or income in P, nor is there any obvious reason why the change in the asset composition of R’s financial sector, from making loans to P to holding more of R’s treasury bills, should influence income in R. In the real world case of imperfect capital mobility, a sale of reserves would be disinflationary while an increase in domestic credit would be inflationary in P; there is no reason for presuming a net inflationary impact in this world of generally flexible exchange rates. But the counterpart to a net disinflationary impact in P is a net cost-inflationary impact in R, while a net inflationary impact in P would be associated with a substitution of loans to P by more of R’s treasury bills in the portfolio of R’s financial sector (just as in the case of perfect capital mobility). Assuming that R’s central bank did not allow these portfolio changes to influence its money supply, there is no obvious reason for an inflationary impact on R.

One can create a somewhat better case for expecting a net inflationary impact by recognizing that the standard model laid out above is not the only possible reaction to an SDR allocation. A country might also increase its reserves or deliberately expand its absorption. In the first case there would be no inflationary impact in either P or R, but in the second case there might or might not be an inflationary impact in P, depending on the mix that was chosen between fiscal and monetary expansion and the resulting extent to which the exchange rate appreciated, but there would in either event be an inflationary impact in R. Whether that impact actually led to higher inflation would of course depend on whether R chose to offset it.

How do these conclusions compare with the view that an SDR allocation is inflationary if and only if it causes an increase in the money supply of the industrial countries (which may be interpreted as the R of my example)? There are two differences. First, I consider the impact on inflation in the periphery P, as well as that in the center R, to be a matter worthy of note. Second, I consider cost-inflationary impacts in R as well as those that might arise from monetary expansion (recall that I am specifically not dismissing effects because I would expect them to be de minimis). I would not quarrel with what I understand to be the position of the Fund staff, to the effect that any significant impact on inflation in the industrial countries depends on their being seduced into monetary expansion.

Consider next the argument that allocations would increase the scope for countries to pursue undesirable policies. There is indeed no guarantee that all countries would react as posited by my standard model above; some might instead expand demand, even if this would be inflationary, which is presumably what worries those who cite this reason to oppose allocations. But, if so, this is just another way of phrasing the previous objection. I thus find difficulty in understanding what additionality this argument is supposed to bring.

So far as the third argument is concerned, it is true that some countries will get the chance to increase their absorption of goods and services, and that this will be partly5 at the expense of others. Specifically, the countries that would be enabled to increase their absorption are those that would be able to acquire their reserves more cheaply than otherwise. The ones at whose expense they would increase their absorption are, as analyzed above, the reserve currency countries from whom the P countries borrow in the absence of an allocation.

But does the presumption of some net inflationary impact and of a transfer of resources constitute a case against SDR allocation? After all, the payment of, say, pensions also has an inflationary impact and it also causes a transfer of resources, which is, indeed, the object of the exercise. If the transfer of resources is desired, then one simply takes the attitude that the inflationary impact has to be offset. Typically, that transfer is desired in the case of pensions, perhaps because those from whom the income is transferred can think, “There, with the grace of God, will go I.”

Welfare, which depends instead on the sentiment “There, but for the grace of God, go I,” tends to have a more difficult political passage. And Mussa is right to say that, in the post-Bretton Woods world in which the consequences of an SDR allocation are essentially distributional rather than systemic, a decision to allocate has to be based on the reserve centers acting on that sentiment. Apparently countries deny that sordid considerations of financial gain have any impact on their positions on the desirability of SDR allocations, just like most proponents of welfare reform have convinced themselves that this would be in the best interest of the recipients (because after a transitional phase their recovery of the work ethic will leave them better off). It just happens, however, that a cynic could get a good prediction of which countries have supported and which have opposed allocations by ignoring their disclaimers and positing the same sort of self-interested behavior that we economists customarily impute to human beings.

I conclude that, so long as decisions on allocations are made by monetary technicians whose normal remit is to maximize national welfare and who are told their decisions should be based on an Article that has to be drastically reinterpreted to rationalize the case for an allocation, the probability of further SDR allocations ever being agreed under the present Articles and with present procedures is approximately zero.

This does not necessarily imply that the SDR is doomed. Perhaps the nonreserve centers will one day act on Mussa’s implied criticism of them for having been “persistent net users of SDRs that have already been allocated” and will try to switch into the SDR on a massive scale and then maintain such a wail of complaints about having to hold deutsche mark and dollars instead of SDRs that the reserve centers will be shamed into agreeing to a resumption of allocations. Or perhaps the reserve centers will come to see attractions in an ability to use newly created SDRs to supply conditional liquidity, which might permit the Fund to develop innovative roles such as acting as lender of last resort to the banking systems of countries that established currency boards and fixed to the SDR rather than to a national currency. Or perhaps some high-level political bargain will include an agreement that the present criteria for allocation be replaced by a provision that SDR allocations in a forthcoming basic period will be equal to x percent of the increase in the actual level of reserves in the preceding basic period. But, if the SDR is to have a future, it will somehow have to be liberated from the criterion of global need that was imposed on it in a world that has disappeared and that has become a straitjacket in the world of today.

General Discussion

Alec Chrystal questioned Michael Mussa’s calculation of the benefit of “owned” reserves created through an SDR allocation compared with the cost of borrowed reserves. Mussa was comparing the lower cost of allocating SDRs with the cost of acquiring and holding other reserves through sterilized intervention; but in Chrystal’s view it was more accurate to compare, as Max Corden had suggested, the difference between borrowing through an SDR facility and borrowing in the capital markets at market rates. The numbers would be different, although some recipients who had access to the cheaper borrowing facility would still derive a net benefit.

Chrystal also questioned the inference in Mussa’s paper that an SDR allocation constituted a free good. The SDR was not a free good; it could not be a free good, since it entailed some risk, which Mussa’s paper discounted entirely. The cost of providing reserves depended on an assessment of risk. The market—not the IMF—was the appropriate and professionally equipped risk assessor, and the market should thus determine the cost of providing reserves.

Mussa acknowledged that the creation of SDRs through an allocation could indeed entail a risk, which could be described as a general liquidity risk. In the absence of a complete Arrow-Debreu structure (decentralized, atomistic, markets with perfect information), the possibility existed that all countries would try to use the available liquid assets simultaneously, which would create major congestion—similar to everybody trying to drive on the same stretch of road at the same time. That possibility must certainly be taken into consideration in the creation of international liquidity, said Mussa; and, in fact, the designation mechanism in the SDR system (Article XIX, Section 5) provided a way of dealing with the congestion problem. However, since it had not been necessary to resort to designation for the last seven or eight years, Mussa did not think that liquidity risk posed a substantial problem.

Another risk was that countries might default on their obligations within the SDR system. But, in practice, the incidence of countries not meeting their obligations had been limited—currently representing less than one half, or even one fourth, of 1 percent—of total outstanding SDRs. Admittedly, this risk would be a matter of concern in the event of a large SDR allocation and a move to an enlarged SDR system.

So, although he would not deny that risks were attached to allocating SDRs, Mussa concluded that, for moderately sized SDR allocations, the magnitude of the risks was small compared with the very large disparities between the cost to many countries of acquiring and holding reserves versus the cost of creating them through an SDR allocation.

In measuring the demand for reserves to determine whether there was a need for supplementing the available supply, Christian de Boissieu countered that one must consider both the optimal level and the optimal composition of reserves. With regard to the optimal level, he said the IMF’s practice of relying on the relationship between reserves and imports to estimate the growth of demand for reserves was not wholly reliable. Estimates at the global level of the velocity of reserves—the inverse of the ratio of reserves to imports—also produced an unstable value. Had the IMF, he asked, attempted to estimate a stable velocity function at the global level or for some parts of the world?

The optimal composition of reserves involved three issues, de Boissieu continued: cost of reserves, conditionality, and “controllability.” In organizing a moderate allocation of SDRs, would the IMF, in effect, be conferring on the SDR the characteristics of high-powered money at the global level? An increment to the stock of SDRs, although it would increase the indirect control exerted by the IMF, could be problematic since the IMF was not a true central bank.

Horst Siebert, responding to the controllability issue, argued that having SDRs created an option, based on specific conditions, for a country to take certain actions; but the SDR was not an instrument that could control a country’s behavior.

Mussa, responding to de Boissieu, said that the SDR was essentially an unconditional line of credit and in present circumstances could only be regarded as very low-powered money.

Mussa acknowledged that the IMF had devoted considerable resources in the past to estimaring the demand for reserves—or its inverse, velocity—and had found that such estimates were subject to substantial errors. Nevertheless, in view of the modest level of SDR allocations that had been recommended over the years, he did not think the wide margins of error were a matter of concern. If, on the other hand, the IMF were to move in the direction of the Buira proposal (chap. 6), which involved trying to satisfy the projected growth of demand for reserves of many developing countries, then the IMF’s capacity to estimate this demand would be a matter of grave concern—in addition to the other concerns he had expressed earlier about large allocations.

Alexander Swoboda said that although there was major disagreement about the meaning of the term “need,” everyone agreed that it was an unfortunate choice of word, because it prompted people to ask the wrong question. The right question was not whether the “need” for reserves would be met, since it always would be in some fashion: if not by gold or dollars, then by deflation. In that narrow sense, there could never be a case for an SDR allocation. The appropriate question was whether an allocation was a good way of meeting that need. Therefore, the question that Mussa was asking—can SDRs be used to meet the demand for international reserves at a lower cost—was appropriate, legitimate, and in keeping with the IMF’s purpose of overseeing the welfare of the international monetary system.

Paul De Grauwe asked Mussa to discuss in greater detail his argument, as De Grauwe interpreted it, that small things did not matter if they are small enough; that is, although large SDR allocations produce inflation, small ones did not because of some nonlinearity in the money supply process. De Grauwe said he knew of no theory or model that produced nonlinearities in the money supply process at the national level and was interested to know what nonlinearities Mussa perceived at the international level.

Mussa said that, generally speaking, the main macroeconomic objective of monetary policy in the major industrial countries was stability. Even a fairly large allocation of SDRs was unlikely to induce these countries to change their monetary policy. The reserve inflows generated by other countries exercising their option to use their SDR allocation to buy currencies from the major industrial countries would expand the money supply. But these inflows would likely be sterilized. For the United States, for example, a modest allocation was not likely to induce a reserve inflow of more than 10 percent of the normal growth of the U.S. money base—about $20 billion a year. Inflows of SDR 1 or 2 billion would be sterilized and would not affect U.S. monetary policy.

However, the major industrial countries might not be prepared to sterilize completely the effects of really big allocations that induced yearly inflows of, say, SDR 8 to 10 billion of reserves, which would thereby pose inflationary risks. As the increments to reserve holdings grew larger relative to the monetary base of a major industrial county like the United States, monetary policy could very well be affected.

It was difficult to estimate precisely at what point the risk became significant that monetary policy in the major industrial countries would be pushed off track by the direct receipt of allocations or by reserve inflows from other countries. However, at low levels of allocation—on the order of past IMF proposals and as recommended in his paper—Mussa said the risk of inflation was effectively zero.

With regard to the inflationary effect on smaller countries, Mussa said there was some confusion between the effects of constraints and incentives. SDR allocations did alter the resource constraint somewhat, because they reduced the real resource cost of acquiring and holding reserves, but they did not necessarily alter the stability incentive by encouraging a more inflationary policy. A country that was otherwise constrained might use an allocation to expand somewhat more; but the expansion would result from a relaxation of the constraint rather than a change in the incentive.

There was another side of the argument as well. Some countries—transition economies, for example—might be so severely constrained that the only way they could finance any economic activity was by “resorting to the printing press.” Additionally, their budgets were subject to extreme pressure from the real resource cost of acquiring reserves. It was arguable that if the latter constraint could be relaxed through the less costly mechanism of an allocation, they would need to resort less to the printing press in accordance with their choice of policy options given the constraints under which they operated.

Barry Eichengreen asked Mussa whether the use of SDRs to fund a fast-disbursing, relatively large facility for countries in the midst of sudden financial crises would fall under the definition of “global need.”

Mussa responded that under the current Articles, IMF members could not lend SDRs directly to the IMF for the purpose of creating such a facility. An alternative approach would be a trust arrangement whereby leading IMF members agreed to relend their SDRs in some other form. A more relevant question—albeit difficult to answer—was why not create the facility through a quota increase rather than an SDR allocation.

Returning to the more general issue of long-term global need, Mussa said that the one clear conclusion emerging from the discussions so far was that the concept—whatever its meaning once was—was now ambiguous. But why, he asked, must this ambiguity always be resolved in opposition to SDR allocations? Why did it not afford some degree of flexibility? Unlimited flexibility was not desirable, of course; but the institutional arrangement whereby the mechanism can only be implemented with the consent of 85 percent of the voting power of the IMF provided enough discipline and control to permit some flexibility in interpretation of the concept.

Siebert took issue with Mussa’s assertion that the loosening of a constraint had no effect on incentives. Loosening the constraint once through an SDR allocation, allowing countries to follow a more expansionary policy than they otherwise would have, might set a precedent. The allocation would create an incentive for getting around the constraint the next time.


  • Byrne, William J., “Evolution of the SDR, 1974-1981,” in The Contemporary International Economy: A Reader, ed. by John Adams (New York: St. Martin’s Press, 2nd. ed., 1985), pp. 34050.

    • Search Google Scholar
    • Export Citation
  • de Vries, Margaret Garritsen, The International Monetary Fund, 1966-1971: The System Under Stress (Washington: International Monetary Fund, 1976).

    • Search Google Scholar
    • Export Citation
  • de Vries, Margaret Garritsen,, The International Monetary Fund, 1972-1978: Cooperation on Trial (Washington: International Monetary Fund, 1985).

    • Search Google Scholar
    • Export Citation
  • Fleming, J. Marcus, Toward Assessing the Need for International Reserves, Essays in International Finance, No. 58 (Princeton, New Jersey: Princeton University, Department of Economics, International Finance Section, February 1967).

    • Search Google Scholar
    • Export Citation
  • Hausmann, Ricardo, and others, “Overcoming Volatility in Latin America,” Report of the Office of the Chief Economist of the Inter-American Development Bank (Washington: IDB, August 1995).

    • Search Google Scholar
    • Export Citation
  • Heller, H. Robert, and Mohsin S. Khan, “The Demand for International Reserves Under Fixed and Floating Exchange Rates,” Staff Papers, International Monetary Fund, Vol. 25 (December 1978), pp. 62349.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, International Reserves: Needs and Availability (Washington: International Monetary Fund, 1970).

  • Lizondo, José Saúl, and Donald J. Mathieson, “The Stability of the Demand for International Reserves,” journal of International Money and Finance, Vol, 6 (September 1987), pp. 25182.

    • Search Google Scholar
    • Export Citation
  • Mussa, Michael, “The Evolving International Monetary System and Prospects for Monetary Reform,” Journal of Policy Modeling, Vol. 17 (October 1995), pp. 493512.

    • Search Google Scholar
    • Export Citation
  • Mussa, Michael, and others, Improving the International Monetary System: Constraints and Possibilities, IMF Occasional Paper No. 116 (Washington: International Monetary Fund, December 1994).

    • Search Google Scholar
    • Export Citation
  • Smith, Adam, An Inquiry Into the Nature and Causes of the Wealth of Nations (New York: Random House, Modern Library ed., 1937).

  • Buira, Ariel, and Roberto Marino, “Allocation of Special Drawing Rights; The Current Debate,” International Monetary and Financial Issues for the 1990s: Research Papers for the Group of Twenty-Four (New York: United Nations, April 1995) pp. 1122.

    • Search Google Scholar
    • Export Citation
  • Coats, Warren L„ Jr., Reinhard W, Furstenberg, and Peter Isard, The SDR System and the Issue of Resource Transfers, Essays in International Finance, No. 180 (Princeton, New Jersey: Princeton University, Department of Economics, International Finance Section, December 1990).

    • Search Google Scholar
    • Export Citation
  • de Vries, Margaret Garritsen, The International Monetary Fund 1972-1978: Cooperation on Trial (Washington: International Monetary Fund, 1985).

    • Search Google Scholar
    • Export Citation
  • Polak, Jacques J., “The Impasse Concerning the Role of the SDR,” in The Quest for National and Global Economic Stability, ed. by Wietze Eizenga, E. Frans Lumburg, and Jacques J. Polak (Dordrecht: Kluwer Academic, 1988), pp. 17589.

    • Search Google Scholar
    • Export Citation
  • Chrystal, K. Alec, “International Reserves and International Liquidity: A Solution in Search of a Problem,” in The International Financial Regime, ed. by Graham Bird (London: Surrey University Press in association with Academic Press, 1990), pp. 928.

    • Search Google Scholar
    • Export Citation
  • Sachverständigenrat zur Begutachtung der gesamtwirtschaftlichen Entwicklung, Im Standortwetlbewerb (Stuttgart: Metzler-Poeschel, 1995).

    • Search Google Scholar
    • Export Citation
  • Salin, Pascal, “Comment on Schröder” in International and European Monetary Systems, ed. by Emil-Maria Claassen (London: Praeger, 1990), pp. 7682.

    • Search Google Scholar
    • Export Citation
  • Schroder, Jurgen, “The Role of the SDRs in the International Monetary System,” in International and European Monetary Systems, ed. by Emil-Maria Claassen (London: Praeger, 1990), pp. 6576.

    • Search Google Scholar
    • Export Citation
  • Smith, Adaman Inquiry into the Nature and Causes of the Wealth of Nations (New York: Random House, Modern Library ed., 1937).


This paper has benefited from significant contributions by Peter Isard and from valuable suggestions from James Boughton, François Gianviti, Morris Goldstein, Donald J Mathieson, Jacques J. Polak, and David Williams. It draws extensively on past work of the staff of the International Monetary Fund. The views expressed, however, are solely the responsibility of the author and do not necessarily reflect the views of the Fund. Indeed, on a number of the issues examined in this paper, there are significant differences of opinion within the Fund.


Article XVIII, Section 1(a).


Gold holdings of the United States declined over the Bretton Woods period, and the official market for gold (with its fixed dollar price) was separated from the private gold market in 1968. There is little indication, however, that U.S. policy was “too expansionary” for the rest of the world before the late 1960s, After the Korean War bulge, inflation remained quite low in the United States and only rose persistently above 2 percent beginning in 1966. From 1945 through 1968, a number of countries devalued their currencies by significant amounts against the U.S. dollar, some of them more than once. Only two countries revalued their currencies against the U.S. dollar and by a comparatively modest amount.


If the demand for reserves is met partly by an allocation of SDRs, the presumption is that countries would not choose to meet this demand a second time by adding to their foreign currency reserves. Thus, an allocation of SDRs should normally be expected to substitute for growth of reserves in the form of foreign currencies. In principle, the situation with respect to gold is different if monetary authorities are committed to buy or sell it at a fixed price. When the private gold market was separated from the official market, it became predictable that additions to monetary gold holdings would be modest.


communiqué of the Ministers of the Group of Ten, The Hague, July 25–26, 1966.


See “A Report to the Board of Governors of the International Monetary Fund Containing the Managing Director’s Proposal on the Allocation of Special Drawing Rights for the First Basic Period,” as reproduced in de Vries (1976), Vol. II, p. 259.


See Fleming (1967).


See “The Need for Reserves: An Exploratory Paper,” written in 1966, and “Reserve Developments, 1951-68: An Analytic Review,” written in 1969, both of which are published in International Monetary Fund (1970).


See Fleming (1967), p. 85.


The basket consisted initially of the currencies of each of the 16 countries for which exports during 1968-72 exceeded 1 percent of world exports. The 16-currency basket was revised somewhat in 1978, based on applying the same criterion to the 1972-76 period, and in 1981 the composition of the basket was reduced to the currencies of the five countries with the largest shares in world exports. See de Vries (1985) and Byrne (1985).


Gold continued lobe held by many governments and central banks but was not used actively as a reserve asset.


The financial transactions and operations of the Fund are conducted through the General Department, the SDR Department, and several Administered Accounts. The SDR Department conducts all transactions and operations that affect members’ holdings of SDRs. All members of the Fund have chosen to be participants in the SDR Department.


Full use of these lines of credit became effectively available for unlimited duration when the reconstitution requirement was abrogated in 1981.


In Mussa (1995) and Mussa and others (1994), I have argued that a return to a system of pegged exchange rates among the world’s major currencies is not now feasible and is probably not desirable.


Thus, as interpreted by the Fund’s Legal Department, the objective of making the SDR the principal reserve asset does not constitute a criterion for SDR allocation.


On several occasions in the 1970s, serious consideration was given to the creation of a “substitution account” under the auspices of the Fund, which would have enabled members to exchange substantial amounts of their U.S. dollar reserves for reserves denominated in SDRs. An important objective of these proposed schemes was to add to the stability of the international monetary system by allowing members to diversify the currency of denomination of their reserves. At that time, U.S. dollars still constituted three-fourths of foreign currency reserves and other countries were somewhat reluctant to undertake the burdens of a reserve currency. One worry was that a sudden desire to shift out of dollars might depress the foreign exchange value of the dollar with disruptive effects on trade and with the consequence of reducing significantly the value of total world reserves. The SDRs that would have been created for the substitution account, however, would not have been created through allocations in the SDR Department, but rather through a parallel mechanism administered by the Fund.


To purge the data of possible trends associated with increases over time in the number of countries for which data on reserves were reported, Figure 1 relates only to countries that were members of the Fund at the end of 1952, which was the year in which Germany and Japan became members.


If gold reserves were valued at market prices (rather than at SDR 35 a troy ounce), there would be no apparent decline in the ratio of total reserves for industrial countries excluding the United States. For the United States, the level of reserves would be boosted by about SDR 70 billion, which would more than double measured total reserves as of 1995.


See “Proposal by the Managing Director of the International Monetary Fund” (October 25, 1978), as reproduced in de Vries (1985), Vol. III, pp. 275-9.


See, for example, the survey by Heller and Khan (1978) and the more recent study by Lizondo and Mathieson (1987).


Another reason for wide error bands in this context is the argument that the relationship between reserves and imports is likely to be less stable than the relationship between a measure of international Liquidity that is broader than reserves alone and a measure of international transactions that is broader than merchandise imports.


In principle, it would have been possible to generate an increase in the effective supply of monetary gold by raising its price. This would have raised the value of existing monetary gold hoards and stimulated increased production. However, there were worries that changing the gold price might add to instability, and, except on the occasion of the Smithsonian agreement, changing the gold price proved to be politically impossible. There were also concerns about the real resource cost of increasing monetary gold holdings—concerns that date back at least to Adam Smith.


The SDR interest rate is a weighted average of interest rates on three-month financial instruments denominated in the five component currencies of the SDR, The financial instruments are interbank deposits for Germany, certificates of deposit for Japan, and treasury bills for France, the United Kingdom, and the United States.


When the SDR system was created, it was recognized that there might be difficulties in getting countries to supply currencies in exchange for SDRs that other countries wished to use. To deal with this potential problem. Article XIX, Section 5, provides that “the Fund shall ensure that a participant will be able to use its special drawing rights by designating participants to provide currency for specified amounts of special drawing rights …” in accordance with certain general principles specified in the Articles. Since 1987, however, it has not been necessary to resort to this designation mechanism. Indeed, the reverse has generally been true; it has been difficult to find countries to agree to supply SDRs to countries that wish to acquire them.


The emphasis here on the scope of the problem of the high cost of reserves does not deny that qualitative considerations, such as the effect of SDR allocations in creating “owned reserves,” are also important in deciding about global need.


Raising the SDR/NGR ratio to 20 percent over the next five years would require large annual allocations of about SDR 40 billion—which would account for over 50 percent of the projected growth of demand for reserves. A more gradual approach could raise the SDR/NGR ratio with smaller annual allocations spread over a longer period.


The United States would receive an allocation of about SDR 1.8 billion (equivalent to about $2.7 billion), which would have no direct monetary effect. Based on past experience, other recipients of allocations might want to make use of about SDR 3 billion of those allocations, and one-fourth to one-third of this amount would be acquired by the United States. Unless sterilized, the resulting reserve inflow would add about $1.5 billion to the U.S. monetary base, in comparison with normal expected growth of the monetary base of about $20 billion. Obviously, there would be no difficulty in sterilizing a reserve inflow of this magnitude.


See Hausmann and others (1995).


The benefits of SDR allocations from meeting part of the growing demand for reserves at an economically appropriate cost and from ameliorating acute reserve stringencies should be expected to rise in proportion to the size of allocations. Benefits from increasing the portion of owned reserves might become significant if the ratio of SDRs to nongold reserves was somewhat higher. The potential risks and disadvantages plausibly follow a quadratic relationship; they remain very low for small through moderate allocations, and then rise increasingly rapidly as allocations become larger. The implication is that there might be an optimal scale of allocation somewhere in the moderate range.


Net creditors may perceive that their citizens would lose the opportunity to supply credit at attractive rates to countries that want to increase reserve holdings. However, private suppliers of credit probably see the rates they charge as necessary compensation for the risks that they undertake in individual credit transactions—even though, from a global perspective, it is known that countries that demand increased credit at high cost will, on average, use some part of it to finance increased holdings of reserves.


The views in this paper are those of the author and do not necessarily reflect the views of the Government of India.


The author appreciates comments from Jörg Kramer and Joachim Scheide.


The founding fathers of the SDR must have been concerned with these costs of expanding the volume of SDRs. Otherwise, it is difficult to understand why an institutional requirement of an 85 percent majority for an increase in SDRs exists.


J.A.H. de Beaufort Wijnholds, “The Need for Reserves Under Full and Limited Flexibility of Exchange Rates,” De Economist, Vol. 122, No. 3 (1974), pp. 225-43.


William J, Baumol, “The Transactions Demand for Cash: An Inventory Theoretic Approach,” Quarterly journal of Economics, Vol. 66 (November 1952), pp. 545-56.


J.A.H. de Beaufort Wijnholds, The Need for International Reserves and Credit Facilities (Leiden: Martinus Nijhoff, 1977), pp. 203-11.


The author is indebted to C. Fred Bergsten and Morris Goldstein for comments on a previous draft.


And, some might be inclined to add, containing an overwhelming majority of the population of Fund members.


In a world of growth, all these propositions need to be interpreted in a ceteris paribus way: the country will seek to maintain the same trajectory of reserve growth, it will buy 100 fewer reserves over the course of time, etc.


The analysis applies to allocations smaller than the size of reserves that the country would choose to hold in the absence of allocations. This is not an onerous restriction.


There would also be some saving in real resources, since, as Mussa points out, the SDR system is a cheaper way to manufacture reserves than doing so through the private markets.

  • de Vries, Margaret Garritsen, The International Monetary Fund 1966–71: The System Under Stress (Washington: International Monetary Fund, 1976).

    • Search Google Scholar
    • Export Citation
  • de Vries, Margaret Garritsen, The International Monetary Fund 1972–78: Cooperation on Trial (Washington: International Monetary Fund, 1985).

    • Search Google Scholar
    • Export Citation
  • de Vries, Margaret Garritsen, The IMF in a Changing World 1945–85 (Washington: International Monetary Fund, 1986).

  • Horsefield, Keith, The International Monetary Fund 1945–65: Twenty Years of International Monetary Cooperation (Washington: International Monetary Fund, 1969).

    • Search Google Scholar
    • Export Citation