The next session considered whether the role of the SDR could be enhanced by changing its characteristics. Three main questions were posed. First, how well has the SDR functioned as a unit of account, including as the unit of valuation for the general operations of the Fund? Second, should the purchasing power of the SDR be “hardened,” either vis-à-vis its component currencies or vis-à-vis representative baskets of goods and services? Third, to what extent is the potential for the SDR in the current system constrained by the absence of a well-developed market in which banks and other private market participants can trade SDRs? In that context, should private market participants be authorized to hold the SDRs allocated to member countries by the Fund? Four speakers addressed different aspects of this topic: Alec Chrystal, Holger Wolf, Jonathan Frimpong-Ansah, and Robert Heller.

The next session considered whether the role of the SDR could be enhanced by changing its characteristics. Three main questions were posed. First, how well has the SDR functioned as a unit of account, including as the unit of valuation for the general operations of the Fund? Second, should the purchasing power of the SDR be “hardened,” either vis-à-vis its component currencies or vis-à-vis representative baskets of goods and services? Third, to what extent is the potential for the SDR in the current system constrained by the absence of a well-developed market in which banks and other private market participants can trade SDRs? In that context, should private market participants be authorized to hold the SDRs allocated to member countries by the Fund? Four speakers addressed different aspects of this topic: Alec Chrystal, Holger Wolf, Jonathan Frimpong-Ansah, and Robert Heller.

The SDR and Its Characteristics as an Asset: An Appraisal

K. Alec Chrystal1

The SDR has been in existence for over a quarter of a century. Its creation was greeted with almost universal praise amid unbounded optimism for its future role in the international monetary system. However, the world looked very different in the 1960s from how it looks now. The SDR would certainly not be introduced today. The question that some might want to ask is: What can be done to improve the SDR to make it fulfill its intended role? However, others may take the view that it has clearly failed in its intended role, or that its intended role is no longer needed, and so it should be abolished.

As one of the first academics to be openly critical of the SDR (Chrystal, 1978), I might be expected to fall into the latter camp. However, I think it is now clear that the SDR does have a modest but useful role in the international monetary system and that this will continue, at least so long as the IMF itself survives. But this is not the grandiose role for which it was intended. Accordingly, the Articles of Agreement of the IMF must be modified to reflect this new reality so that not even lip service need be paid to the objective of turning the SDR into the “principal reserve asset” of the international monetary system. This was never going to happen from Day I and it is even less likely to happen now. It is neither a desirable nor a feasible goal, and its existence as an explicit objective is an obstacle to rational analysis.

My brief for the seminar is to discuss the characteristics of the SDR as an asset. This is hard to do independently of some analysis of the broader role of the SDR in its many potential dimensions. Hence, I have resisted the temptation to focus exclusively on the narrow issues of valuation and yield. These will inevitably be discussed in the context of a rather broad appraisal of the SDR. My general conclusion is that nothing should be done to change the characteristics of the SDR, as it is already appropriately structured to perform the limited function that it has. The only fundamental changes in characteristics of the SDR for the foreseeable future will be those required to accommodate the euro, if the European Union (EU) single currency ever becomes a reality.

I shall proceed by first giving a brief review of the arguments behind the invention of the SDR and where we stand on these arguments today. Then I shall attempt an assessment of the role of the SDR and draw conclusions for its constitution.

Whence Cometh the SDR?

Some History and Its Interpretation

The SDR was invented as a solution to the international liquidity problem of the 1960s. The conventional analysis, inspired by the work of Robert Triffin, was that the gold exchange standard of the Bretton Woods system was fundamentally flawed. The system allegedly depended for its credibility on the fact that the U.S. dollar was convertible into gold at $35 an ounce. Growing demand for reserves had to rely upon growing external dollar balances, yet these growing external dollar liabilities would reduce the U.S. reserve ratio—the ratio of U.S. gold reserves to external dollar liabilities. Hence, credibility of the system would eventually be threatened. In order to avoid potential collapse, the solution was to replace dollars in official reserves with SDRs.

Whatever the intention, it cannot be doubted that the SDR has not grown into anything remotely approaching “principal reserve asset” status. I shall proceed by pointing out the extent of decline in interest in the SDR outside IMF circles and I shall then discuss the available explanations of what went wrong.

The history of the SDR has been explained as follows:

In 1967, IMF members agreed to the creation of the Special Drawing Right (SDR), an artificial reserve asset similar to the IMF currency Triffin had envisioned. SDRs are used in transactions between central banks, but their creation had relatively little impact on the functioning of the international monetary system. Their impact was limited partly because by the late 1960s, the system of fixed exchange rates was beginning to show strains that would soon lead to its collapse. These strains were closely related to the special position of the United States. (Krugman and Obstfeld, 1994, p. 546.)

Few could argue with what is said in the above quotation. What is remarkable is that it is the only reference (save a single reference to an SDR-denominated bond issue) to the SDR in a modern standard international economics textbook of nearly 800 pages, over half of which is devoted to international macro and monetary issues and which was written by two of the world’s leading international economists. Even more remarkable, in a major survey of the evolution of the workings of the international monetary system, McKinnon (1993) does not mention the SDR even once! Clearly, the SDR project has not impressed the modern generation of economists as much as it impressed some of those who spent their formative years on the “international liquidity” debates of the 1960s, nor will it be high on the list of research topics for contemporary students of international monetary economics.

I asked a class of my own undergraduate students the following (slightly loaded) question: “What was invented in 1967 to save the world from depression?” To be helpful I gave a clue that it was commonly known by an acronym. The answers in order of popularity were the VCR, LSD, the CD, and the ATM. When I gave the correct answer none could explain what it was and one even suggested that it might be a contraceptive device! Only among my older MBA students could I find any name recognition at all. An ex-bond dealer (over 40) could vaguely remember SDR bond issues in the early 1980s but his younger colleagues regarded this as a curiosity from a bygone age.

This story has a serious point. It is that, outside the IMF, interest in and understanding of the SDR has declined sharply, and future generations of voters and politicians will be much more skeptical of the SDR’s relevance to the world economy than was the generation that invented it. If the latter cannot hand on a good reason for continuing the SDR facility, future generations may well abandon it. Thousands of doctoral dissertations are written each year on arcane topics in international finance but almost none of these is on the SDR. Fifteen years from now there will be few academics or officials left working who experienced the euphoria surrounding the birth of the SDR and even perhaps regarded it as a serious research topic.

If the SDR were a business product we would need to call for a relaunch (or shutdown). This seminar provides an ideal opportunity for just such a reassessment of the “value added” of the SDR facility. But the marketing blurb will have to be much more plausible and modest in scope than the original messianic advertising campaign.

A consensus view seems to have emerged in recent literature about what went wrong with the project to turn the SDR into the principal reserve asset. It runs in two parts as follows: first, the standard view of the problem as analyzed in the 1960s was basically correct. As mentioned above, this was based on the view that under the Bretton Woods fixed exchange rate regime, the world required a growing stock of international reserves. Triffin (1960) correctly diagnosed that this required growing dependence upon the U.S. dollar. As U.S. external liabilities inevitably grew, this would undermine credibility of the convertibility of dollars into gold. What was required to save the system was a new supranational reserve asset, which was to be the SDR. From then on reserve growth would be provided by SDRs, and the reserve role of the dollar would eventually be phased out.

The second part of the conventional story is that, although what happened in the 1960s was both a correct economic analysis and the implementation of a well-designed solution to the perceived problem, the world changed. Indeed, by the time the SDR came along the system was already on the verge of collapse. The move to floating combined with the emergence of globalized private capital markets meant that the international liquidity problem went away (see Williamson, 1994, and Crockett, 1994). Either one of these changes would have made it hard to argue that a liquidity constraint existed. Once exchange rates were floating, reserve use by monetary authorities became optional, so the concept of a “systemic shortage of liquidity” became meaningless. Equally, once there was an active and deep international capital market, creditworthy authorities could always borrow to augment liquid reserves. Therefore, from being exogenously restricted in supply (allegedly connected to the U.S. official settlements balance), reserves became endogenously available through borrowing in private markets. Also, because monetary authorities could choose whether to intervene in foreign exchange markets, it became impossible to determine whether a liquidity shortage existed. Indeed, some argued convincingly that there could be no such thing as a general liquidity shortage in a world of floating exchange rates. The reserve stock had become demand determined rather than supply constrained.

In short (according to the conventional view that will be challenged below), the SDR was a great idea that could have saved the Bretton Woods regime, but unfortunately it came too late.

There is less of a clear consensus about what the role of the SDR should be today. Some argue that there are still many liquidity constrained countries in the world, so significant SDR allocations should continue in order to meet their needs (Williamson, 1994). However, there is a danger that this argument is confusing the need for wealth redistribution with the issue of design of the monetary system. (In the domestic economy, we would not normally argue that the existence of poor people is an argument for expanding the money supply.) It is related to the question of the “aid link” to which I will return below. Others believe that because it is impossible to make a case on monetary grounds alone for any further SDR allocations, the facility should be allowed to wither on the vine. I shall also return to this issue later.

Some Questions About the Conventional View

My own critique of the SDR (Chrystal, 1978 and 1979) was based on a rather different way of looking at the reserve problem from that of the above consensus.

The international liquidity problem of the 1960s was perceived as a purely official sector problem that could be solved within the official sector. I approached the reserve holding question from the broader perspective of demand for “international media of exchange.” Trading nations had always required a balance of external currency and this had often been held in a centralized form (such as in the Bank of England of the nineteenth century). Karl Marx, among others, had understood that “just as every country needs a reserve of money for its home circulation so too it requires one for external circulation in the markets of the world” (Marx, 1886).

From this perspective the needs of central banks (or domestic treasuries) should not be viewed independently of the needs of the domestic financial sector that they serve and of which they are a part. Accordingly, I assessed the SDR by traditional monetary criteria in the potential roles of medium of exchange, unit of account, and store of value. On each of these counts the SDR scored very low marks. The SDR could not be used for intervention in financial markets (the primary function of foreign exchange reserves), it had undesirable characteristics as a unit of account, and it was a poor store of value. I concluded that “…there is little future for the SDR as presently constituted [and that]… it is unlikely to evolve into the major reserve asset” (Chrystal, 1978, p. 24).

My original criticism was aimed at the SDR of the mid-1970s, which was valued as a 16-currency basket and on which the interest rate was held at 80 percent of market rates. The 16-currency-basket SDR provided a poor unit of account because it was hard and costly to replicate and it included some currencies that were not widely traded. It was a poor store of value (asset) because it had a lower yield than substitute reserve assets. Perhaps in response to these criticisms, the valuation of the SDR was switched to its current 5-currency basket and the interest rate was raised to 100 percent of market rates. However, these improvements have done little to advance its cause, for a number of different reasons. The two most important are that because it is hard to make a case for further allocations, the proportion of SDRs in official reserves is small, and that the mere existence of the SDR as an index number adds nothing of value to the real world economy.

The most controversial argument I put forward in 1978 was that the creation of SDRs may increase rather than decrease desired holdings of foreign exchange reserves, and that mechanisms were available to turn this desired increase into reality. A substantial increase in foreign exchange holdings did indeed follow the introduction of the SDR, but the reason I thought it might happen was almost certainly wrong. My idea was that the existence of the SDR might encourage what used to be called “incest.” A central bank can print its own money but it cannot print foreign money. However, any two central banks can do a swap whereby they create reserves for each other. Such swaps did happen in the 1960s but only to a limited extent. They have monetary policy implications if one party then spends the reserves, and such swaps were normally only activated in a crisis. Through SDR transactions, incest could have become legitimate and could have been activated without any hint of a crisis. Suppose, for example, that the Bank of England uses its SDR allocation to acquire deutsche mark from the Bundesbank, which in turn uses the same SDRs to acquire dollars from the Federal Reserve, which in turn uses the same SDRs to acquire sterling from the Bank of England. The SDR allocation has passed around but the total stock of foreign exchange reserves has gone up threefold.

One factor that would have encouraged this activity in the 1970s was that the SDR bore a lower than market interest rate whereas the foreign exchange acquired could be invested at full market rates; hence central banks could seemingly increase their wealth2 by this game of pass the parcel. It did not happen much because the only countries that could play were those issuing the reserve currencies and by the late 1970s none of them was short of reserves. Indeed, once the SDR carried a full market level of interest rate, reserves could be borrowed at similar interest rates through the money markets. In other words, drawing on SDR allocations to acquire foreign exchange reserves from other central banks would have appeared attractive at the margin (as a pure portfolio operation) when the interest rate on the SDR was below the market rate, but not afterwards.

It is almost certainly true that for the major industrial countries, once the SDR came to bear a market rate of return, the level of foreign exchange reserves they chose to hold was entirely independent of the size of SDR allocations. This has to be true because the gift of the right to borrow at market rates to someone who can already borrow at similar rates leaves their real wealth unchanged and, thus, has no effect on any of their real wealth allocation decisions. Accordingly, allocations of SDRs would have no effect on the level of foreign exchange reserves that monetary authorities would choose to hold. In contrast, it has always been assumed in the past that further SDR allocations would reduce the level of foreign exchange reserves held. Thus, the general presumption that increased allocations of SDRs over time would reduce the level of foreign exchange reserves that monetary authorities would choose to hold is unlikely to be valid.

My final criticism of the conventional view on SDR creation is that, far from being a good idea that just came too late, the invention of the SDR may have contained the seeds of its own demise (Chrystal, 1990). The creation of the SDR was intended to let the United States out of its role as world banker and, hence, the United States was relieved of its obligation to follow the monetary and fiscal policies that this international role required, thereby becoming free to target only domestic objectives. Indeed, the Unites States may have been a willing party to being let off the hook in this regard, in the context of its internal war on poverty and external war in Vietnam. Equally, all the other countries in the system were persuaded that their reserve constraint was about to shift and, accordingly, would have been likely to follow more expansionary policies than would otherwise have been considered desirable. These factors contributed to the major inflationary surge in the world economy in the early 1970s and, in the process, changed the world monetary system (and macro policy agenda) forever. What would have happened if the SDR had not been invented can never be known, but it is at least possible that things could have worked out differently if the recommendation for new international liquidity creation had been more widely challenged. Indeed, it is hard to believe that there could have been agreement on the creation of a new reserve asset at any other time than during the brief interlude in history that a U.S. President could be reported as claiming that “We are all Keynesians now.”3

I am not alone in thinking that the creation of SDRs may have been a mistake from the beginning. Bergsten (1994, p. 44) has recently argued, though from a different perspective:

My conclusion is that the officials clearly focused on the wrong problem in spending a large amount of time on the negotiations of the Special Drawing Right (SDR). Polak considers that the time spent on the SDR system was not exorbitant. I think that it was not only exorbitant but largely wasted. The main problem was essentially ignored, and if only Ossola and Polak and all those very creative minds had devoted as much effort to reforming the adjustment system in the 1960s as they did to creating the SDR, we might have avoided the crises of the late 1960s and 1970s. May be, we could have avoided the breakdown of the Bretton Woods system.

An even more telling and seriously neglected line of reasoning dates back to the work in the 1960s of Milton Gilbert (1968) and is most cogently argued by Peter Oppenheimer (1987). According to this view, the distinction between “liquidity” and “adjustment” is a bogus distinction. In reality they are opposite sides of the same coin. There is only one problem and this is to achieve the appropriate equilibrating behavior on the part of the world’s key policymakers. International liquidity is an endogenously determined aggregate of no particular behavioral significance (even under the gold standard and the Bretton Woods system).

The volume and composition of liquidity in the international monetary system is the outcome of numerous and varied decisions by national sovereign authorities. The viability of an international monetary system depends on the behavior of those authorities, or at least on the behavior of the small number who really matter. What is important is that national authorities should be committed, in theory and practice, to the pursuit of equilibrating policies. This may sometimes involve conditionality of access to finance for disequilibria, and even arbitrary limits on such finance to the individual country. But to conclude from this that aggregate international liquidity can likewise be directly determined by some central decision-taking body is to commit a fallacy of composition. (Oppenheimer, 1987, p. 317.)

According to this view, the generation of “correct” levels of international liquidity should never have been the business of the IMF. The invention of the SDR could not have been the solution because “international liquidity” was not the real problem. In the terminology of modern microeconomics, the solution to whatever the international monetary problem might have been was to be found in the generation of “incentive efficient” behavioral commitments (rules of the game). If anything, the SDR facility created disincentives (or at least no new incentives) to appropriate equilibrating behavior.

W(h)ither the SDR?

From now on I will stop talking about the past and turn attention to the present and future, focusing on the SDR as it now is or might become. However, the above discussion is rather important for what follows because it establishes that, even under the most favorable interpretation of why the SDR was invented and what went wrong, there can be no serious argument for any attempt to impose more SDRs on the world system on grounds of a generalized shortage of liquidity. Nor should there be any case for attempting to reconstitute the SDR in any way merely to achieve for it some bogus goal—such as making it the principal reserve asset.

The SDR as Asset or Liability

My original brief was to discuss the characteristics of the SDR as an asset. However, in its modern form, the allocated amounts of SDRs are neither an asset nor a liability. They are simply the right to incur a liability, that is, to borrow at market interest rates (a weighted average of Group of Five treasury bill rates). A right to borrow would not normally be classified as an asset, though certainly borrowing rights are valuable to anyone who has restricted access to capital markets. The borrowing right can be thought of as an option, and it is the value of this option that is the true value of an SDR allocation. For this reason, I have always found it strange that SDR allocations are counted as international reserves at all, even though they do have some value to credit constrained governments. The addition to net worth resulting from an SDR allocation is much smaller than its face value. SDR holdings in excess of allocation are clearly an asset whereas disposals relative to allocation constitute an equivalent liability, so it is only holdings in excess of allocation that really constitute reserve assets.

However, it could be argued that standard accounting rules do not apply to official reserves. Another example is that borrowed reserves are reported gross, without the offsetting loan liability (which may appear somewhere else in official accounts but rarely in the reserves figures). Similarly, SDR allocations have been counted by many countries as a credit item in their balance of payments accounts when, in fact, they are nothing of the sort.

Nonetheless, the economics of all this is clear. For monetary authorities that can borrow at market rates, an allocation of SDRs gives them nothing because they can already borrow freely at almost identical interest rates—indeed, the SDR interest rate is calculated directly from the rates at which the U.S., U.K., German, French, and Japanese Governments can borrow in money markets. The only authorities that benefit from an SDR allocation are those that cannot borrow at market rates, that is, those that are least creditworthy. The more non-creditworthy a country is, the more it benefits from allocated SDRs. The realized benefit per period is the size of its allocation drawn down times the differential between what it would have to pay to borrow and the SDR interest rate.4

Some have advocated biasing SDR allocations toward developing countries—the so-called aid link. It is not commonly appreciated that the benefits of the SDR facility are already biased toward the least creditworthy countries, notwithstanding the fact that allocations have so far been linked to quota. The size of the allocation is not the true indicator of benefit accrued.

However, this important characteristic of the SDR facility does make it easy to understand why the major industrial economies have been reluctant to sanction any further SDR allocations, whereas the developing countries have continued to be enthusiastic advocates of significant allocations.

It was believed originally that SDRs would be highly desirable reserve assets, partly because the basket valuation gave them a “hedge” characteristic and partly because their guarantee by the international community gave them super-gilt-edged status. (They were of course, originally, and somewhat ludicrously, billed as paper gold.) Some authorities have clearly been happy to hold excess SDRs but others have not. How good a hedge they provide depends on what else is in the authorities’ portfolios.

Private use of the SDR as a numeraire has all but died out because no private firms have matching basket structures on the other side of their balance sheet. Exchange risks can be hedged exactly into local currency by the appropriate use of single-currency debt instruments combined with characteristic changing derivatives (dollar bond issues swapped into local currency by a non-U.S. issuer are common).5

For official reserve portfolios, which are not typically hedged, excess SDRs are likely to be close substitutes for the major currencies, especially given the comparative coherence of world interest rate movements. Ironically, any attempt to make the SDR more attractive to hold would also make it less attractive to use (that is, to draw down relative to allocation). Anything that increases its yield to excess holders would also raise the interest cost to borrowers. Gresham’s Law would then operate and fewer countries would wish to draw down their SDR facility and its circulation could thereby be restricted. This reasoning suggests that there is a very strong case for leaving the interest rate calculation on the SDR very much alone.

The only case in which the interest calculation should be changed is in the unlikely event that it was decided to move to a new valuation method. The most plausible of the potential moves available is to include an inflation adjustment into SDR valuation. It is to this that I now turn.

A “Hard” SDR?

The incorporation of an inflation adjustment in the valuation formula would create what is now called a “hard” SDR. Thakur (1994) has recently performed for us the outstanding service of assessing the technical feasibility of this suggestion. He concludes that “the most suitable method of valuation to establish a hard SDR is… one where the amounts of currencies in the SDR basket are adjusted periodically (say, monthly) to reflect the loss of purchasing power of each currency, as measured by its rate of inflation” (p. 480). His arguments supporting this conclusion are persuasive. However, many of the implications for moving in this direction are left (deliberately) to be drawn out by others.

If the SDR remains a purely official sector instrument, there is clearly no gain from creating a hard SDR. Any move that improves the return characteristics of the SDR as an asset will also discourage its use as a liability, and SDR accounts will freeze (just as gold reserves did once the private market was created).

Although the calculation of the basket weights is easy, the interest rate calculation would not be so straightforward. Thakur argues that “this should not present any technical difficulties because, by adjusting the weights of the SDR interest rate basket in line with the adjustment in the currency basket, the interest rate on the hard SDR could be calculated daily” (p. 485). It seems to me that it is more complicated. We should deduct the weighted inflation rate from the weighted interest rate (using his changing weights at each point in time) in order to arrive at the appropriate real SDR interest rate. However, the rate so calculated could be negative. If instead we calculate the weighted nominal rate, we would clearly be giving the SDR superior return characteristics (that is, inflation would tend to be compensated for twice over) and the Gresham’s Law problem would apply. If we were to go down the indexation route (which I am sure we will not), the real interest rate calculation would probably have to be based on long rates rather than three-month money market rates.

Why should the official central banking community, which has mainly nominal assets and liabilities, wish to introduce into its own intraofficial sector dealings an indexed asset? The answer is that it should not. The best hope for the SDR within the official sector is as something not too different from other currencies (in valuation and return characteristics).

The only serious case for creating the hard SDR is that it may encourage SDR use as a numeraire for private transactions. The argument for this echoes the long literature on the case for a commodity-backed reserve asset or for a gold base to the international monetary system. It would provide a peg of real value in the world economy—and the real SDR would better reflect the typical consumer basket than would any commodity currency. However, if private markets really wanted such an index number they could calculate it today at low cost. The reality is that long bond yields already fully reflect inflation expectations and the mere publication of some official index number would do nothing to increase the efficiency of these bond markets. Few borrowers or lenders have the exact exposure to inflation risk that the SDR basket would match; hence, its private market attraction would continue to be minimal.

The introduction of a hard SDR could preclude discussion of new allocations for all time, or, rather, it would effectively create the rule that the real stock of SDRs would remain constant. It would be hard to make a case for creating SDRs faster than this, though perhaps one could argue that these real SDRs should grow in volume at the same rate as the growth of real world trade. However, as SDRs have no direct role in world trade, it is far from clear that this makes any sense. Only if the SDR were made a genuine tradable currency could this connection be made.

In short, I believe that there is no strong case for creating a hard SDR, especially in the low-inflation environment of the mid-1990s. It could diminish rather than enhance the role of the SDR as an official settlement medium (because anything that enhances its yield characteristics makes it less attractive to draw down); its existence would do nothing to enhance the efficiency of private debt markets; and it would be unlikely to encourage more prudent domestic monetary policies than otherwise seem politically appropriate.

SDR as an IMF Numeraire and “Currency”

Although it is clear that the SDR would not be invented today if it did not already exist and it is equally clear that it will never be the principal reserve asset of the international monetary system, the existence of the SDR facility is sufficiently useful that it is not necessary to close it down. What the SDR facility has done is to provide a useful intraofficial sector payments system via (and including) the IMF that augments, albeit in a minor way, the standard currency payment channels. This is helpful in the making of payments among governments and central banks, and also between them, the IMF, and other international organizations (designated holders). However, such a payments mechanism could easily have been set up without the existence of the SDR per se, though the fact that this payments channel is there already is one of the main reasons why it is not necessary to close the SDR facility down.

To make sure that members that wished to draw down their SDR allocation could always do so, there is a potential process for the “designation” of recipients. However, designation has been rendered unnecessary by the dominance of “transactions by agreement.” To facilitate transactions by agreement, several major central banks stand ready on a continuous basis to buy and sell SDRs within established limits, typically in exchange for U.S. dollars. This “market making” activity is important because it increases the ease with which SDRs provide liquidity to the official sector, thereby increasing the acceptability of the SDR as an official sector settlement medium.

The use of the SDR as a neutral numeraire also has obvious attractions in that the SDR is not vulnerable to weakness in one of the dominant currencies and also has the appearance of symmetry inasmuch as one nation’s currency does not enjoy dominance. The symbolism here is probably important. The numeraire function does not require the existence of the SDR account. It is perfectly feasible to cancel SDR allocations and yet keep the SDR as the IMF accounting unit.

Seeing the SDR facility for what it is rather than what it was intended to be has two important implications for debates about further SDR allocation. First, SDR allocations have little to do with global reserve need. The calculation of the desirable level of SDR stocks should be related solely to the growth of intragovernmental transactions, since these are the only transactions in which the SDR has any role. Second, all new members of the IMF should automatically be given an SDR account with some initial balance related to quota size. There is no point in inviting someone to play Monopoly without giving them some Monopoly money to play with. IMF membership should automatically carry with it an SDR account and consequent membership of this bit of the official payments mechanism.

There would be little potential damage caused by modest SDR allocations. However, these should always be small and gradual. Proposals to more than double the existing stock are out of any reasonable ballpark figure. Allocations of up to at most 5 percent a year of the existing stock over time would both advance the role of the SDR payments mechanism and give a modest low-cost unconditional borrowing facility to the poorest countries. Such allocations should continue in relation to quota, once a catchup has been arranged for new members. They have no value for the rich countries anyway.

More Radical Reform?

Many more radical reforms could be envisaged for the SDR. It could, for example, become a private market currency if the IMF started to issue tradable SDRs in exchange for currency. However, this would require a far more fundamental reform than anybody dared to contemplate in the 1960s, when the tide of opinion was vaguely supportive of an international currency. Such support would not exist today and hence is not worth pursuing.

The one reform of the SDR that will have to be discussed is what to do about its valuation in the unlikely event that the euro becomes a reality at the end of the decade. As presently constituted, the currency weights are: U.S. dollar, 39 percent; deutsche mark, 21 percent; yen, 18 percent; French franc, 11 percent; and sterling, 11 percent.

It might seem that, if the three EU currencies in the SDR basket join (with others) to form the euro, the weight of the euro in the SDR would become 43 percent. However, this does not follow, as the reported structure of world trade would also change and the combined GDP of the EU would be much larger. A significant proportion of German, French, and U.K. trade is intra-EU; hence, the proportion of EU external trade in world trade could be quite different. As there is a danger that SDR weights may have to jump quite radically, this issue is not trivial.

One simple solution to the weighting problem might be to give the yen, dollar, and euro equal weights in a three-currency basket and to introduce it slowly over time. Perhaps the diplomatic procedure would be to continue to calculate weights as before—in relation to GDP and world trade—but this would inevitably involve some jump in valuation at the time (or soon after) currency reform. This discussion presupposes that the United Kingdom joins the EU single currency, which is far from a foregone conclusion even if the project proceeds. In any event, the SDR cannot remain unchanged if even a small subset of EU currencies (at least including the franc and the deutsche mark) is merged, so contingency plans will need to be made soon.


I have reached the unsensational, but hopefully sensible, conclusion that there is a strong case for leaving the SDR more or less alone, save for planning for its valuation after the advent of the euro. The SDR will never be the principal reserve asset of the international monetary system nor should it aim to be. However, the SDR account at the IMF does provide a modest but useful official sector payments mechanism, and therefore the facility should not be closed down The hassle associated with trying to get agreement to close it down would not be worth the effort—there is little to be lost by keeping it going. Any new allocations of SDRs should be modest and gradual and should reflect the needs of official sector transactions. The case for increasing SDR allocations cannot be made with reference to global reserve needs, as aggregate “liquidity” shortages cannot exist, and arguably never could. Nor should SDR allocations be attempted purely as a means of solving the financing needs of developing countries. Such abuse of the world’s monetary institutions will do nothing but discredit the SDR and make it of only sectional interest, though there are potential gains for all if a steady expansion of the SDR facility was to contribute to enhanced prospects of world trade growth.

The SDR is a bit like Esperanto. It was created out of great ambitions to save the world from economic (linguistic) divisions. It still has a small group of enthusiastic supporters, but reality has long been creeping in and its supporters have been diminishing greatly over time. The intended global role will never be fulfilled. Unlike Esperanto, the SDR will survive just because it is there and it would need a deliberate decision of the international community to kill it off. Such morbidity is unnecessary because the SDR facility is just about worth keeping, but only at a modest level. It is a useful neutral numeraire for IMF (and other international bodies’) affairs, and the SDR account is a convenient, but replaceable, transactions medium for the official sector. The SDR, however, has no serious role in the real international economy and will not have one without radical changes that are unlikely to be contemplated.

Currency Baskets as International Units of Account

Holger C. Wolf1

Monetary text books customarily define money by its three properties—as a means of transaction, a unit of account, and a store of value. In the vast preponderance of domestic exchanges, a single unit—the legal tender—serves to fulfill the first two functions: prices are quoted in units of the local currency and transactions are consummated using concrete or electronic tokens denominated in units of the same currency. It is important that this use of a single unit for both objectives reflects evolutionary convenience rather than logical necessity; the two functions can and readily have been performed by different units throughout history if circumstances warranted.

For domestic transactions, most instances of separate units of account can be traced to two broad causes—the presence of multiple mediums of exchange concurrently used and high inflation in terms of the legal tender. As both of these also feature—in modified form—in the case for separate international units of account, we begin with a brief review of domestic units of account before turning to the main focus of the paper: currency baskets as international units of account.

Domestic Units of Account

Multiple concurrently circulating units of payment create simplification incentives for unique units of account much akin to the simplification incentive for standard units of exchange in barter economies. The locally dominant monetary unit is often the most convenient choice, but separate units of account have also enjoyed extensive usage throughout history.

The second main motivation for using a separate unit of account is created by a volatile inflation rate of the dominant unit of exchange into goods and services, reducing its ability to serve as a stable reference point, in particular for transactions consummated over prolonged periods.2 Independent units of account—ranging from fictitious units to foreign currency—tend to readily emerge in these circumstances.

Separation of the domestic unit of account and the domestic unit of the means of transaction is thus quite readily achieved if circumstances warrant. The two causes of separation—multiple transaction mediums and high and variable inflation measured in terms of the dominant means of transaction—extend, with modifications, to transactions undertaken in the international realm, to which we turn next.

International Units of Account

The case for a separate unit of account for private international transactions rests on two pillars. First, one of the main arguments for the coincidence of the unit of account and the unit of the means of exchange—convenience—by definition does not fully apply to international transactions, since the international unit of account used to book the transaction will in general differ from the unit of the medium of exchange and the domestic unit of account for at least one of the trading partners. Second, although the choice of the domestic unit of account may be legally restricted—at least in transactions with the public sector and compliance with standards—firms are, with few exceptions, free to choose the international unit of account used in the transaction. As the unit of account determines the distribution of exchange risk between the trading partners, its selection becomes an important consideration in contracting. A related problem arises in the translation of nominal values in various currencies to a common reference magnitude. The use of volatile bilateral nominal exchange rates may distort relative rankings, motivating the selection of an alternative unit of account, be it units of the local consumption basket, the quantity of a reference commodity available in all relevant markets, or a currency basket.

The selection of an international unit of account is likewise of evident importance for the international transactions of governments and for the internal and external operations of international organizations. Indeed, the only currently widely used international units of account not also used as a national legal tender—the SDR and the ECU—are creations of the public rather than the private sector. To a significant extent, the objectives in specifying an optimal public international unit of account overlap with those characterizing private sector firms engaged in international trade or multinational production, be it the comparison of national performance indicators expressed in local currency, the need to construct balance sheets in a separate unit of account to filter out the effects of “noisy” bilateral exchange rate movements, or the wish to denominate financial instruments in a unit distinct from the unit of the means of payment in which principal and interest rate payments are undertaken.

An attractive unit of account for these purposes is defined by a high signal-to-noise ratio of changes in the exchange rate between national monies and the unit of account, reducing both the transaction risk and the potential distortions vis-à-vis the alternative of using a national money as unit of account. The first focus of our empirical work will be to examine the suitability of global currency baskets to serve as international units of accounts in this regard.

The second empirical problem discussed below is based on an intriguing link between “hard” units of account—units defined in terms of purchasing power over a specific bundle of goods—and optimal reserve baskets in a world of incomplete markets. To simplify the issue, it is assumed that the sole objective of reserve management is the safeguarding of the real purchasing power of reserves in terms of imports from the current trade partners, an assumption roughly consistent with rules of thumb expressing reserve adequacy in months of imports, although substitution of imports across countries is ruled out. The optimal unit of account for assessing the adequacy of reserves in this case is trivial. For a total reserve level expressed in terms of some reference currency (typically the U.S. dollar), it is simply the trade-weighted average of import partner export prices translated into dollars. In a complete market environment with positive real interest rates, the objective function could be trivially satisfied by matching the currency composition of import partners and reserves, eliminating exchange rate risk.

In practice, liquidity considerations restrict reserve holdings to a small subset of currencies. The resulting divergence of the trade and reserve basket implies that movements of the exchange rate between reserve currencies and the currencies of import partners imperil the maintenance of the real purchasing power of reserves in terms of imports. Optimal reserve management in this environment entails the selection of the reserve basket generating home currency returns exhibiting maximum positive correlation with changes in the trade-weighted home currency equivalent of import partner export prices. The problem of finding the basket that optimally fulfills this criterion links intriguingly to the distinction between “soft” currency baskets—defined in terms of units of currency—and “hard” currency baskets—defined in terms of constant purchasing power vis-à-vis some reference basket. Specifically, the home currency equivalent of the value of the ideal reserve basket stabilizing reserves in terms of imports is identical to the home currency equivalent of the hard unit of account defined in terms of imports.

Before taking a closer look at currency baskets as a particular subclass of units of account, it bears emphasizing that the solutions to both problems—the unit of account basket and the reserve basket—are both trivial and coincide in a first-best world. Specifically, in the joint presence of instantaneous purchasing power parity, uncovered interest rate parity and strictly positive returns to capital, the optimal composition for both baskets for a given country simply mirrors the trade composition.

Currency Baskets

The increased volatility of bilateral exchange rates in the aftermath of the collapse of the Bretton Woods system motivated substantial interest in the use of currency baskets as alternative units of account. To date, two such baskets, the ECU and the SDR, have attracted a significant following3 as units of account, while a number of other units have had more limited usage.4 A number of alternative basket specifications have attracted attention over time, with the major distinction being between fixed (“soft”) and flexible (“hard”) weight baskets.5

Fixed Weight Baskets

The fixed weight basket comprises a given number of units of each component currency. The exchange rate of the basket vis-à-vis any given reference currency is thus simply computed by converting the component currency units into the reference currency using bilateral nominal exchange rates and summing across currencies. Thus, denoting the units of currency q per unit of the basket as Uq and the units of currency q per unit of a reference currency j (that is, the bilateral nominal exchange rate) as ERiq, the price of one n-component basket in terms of currency j, Bj, is given by


For our second benchmark case, we assume that the reserve basket is held in interest rate bearing assets. Denoting the nominal interest rate (in percent) of an asset held in currency q as itq

article image
, the interest rate of the basket itB
article image
is likewise straightforwardly computed as the weighted average of the component interest rates, using the relative weight of the currencies in the basket:


The selection of a fixed nominal weight basket instead of a single currency as a unit of account is motivated by the possibility to cancel “noise” movements in the individual components: a well-constructed basket may exhibit a variability vis-à-vis some reference currency below the minimum of the variabilities of the basket components vis-à-vis the same reference currencies. However, the latter property is a special case dependent on the correlation structure of changes in the component exchange rates vis-à-vis the reference currency, not a general property.6

Variable Weight Baskets. The use of fixed weight baskets as international units of account can help in reducing the high noise-to-signal ratio widely thought to characterize relative common currency price movements measured based on bilateral nominal exchange rates. Fixed weight baskets are thus well suited to reduce costs associated with deviations of exchange rates from their “equilibrium” path. In some instances, for example, long-term contracts aiming to generate constant real transfers in terms of a specified basket of goods or international legal conventions (Gold, 1981), the variability of alternative units of account matters less than trend changes in the value of the unit of account vis-à-vis some reference point. Fixed weight baskets fail to provide good intertemporal units of account if the purchasing power of the basket components is subject to trend changes. For both the SDR and the ECU, this has been the case (Thakur, 1994).

Variable weight baskets provide a constant value standard by adjusting the number of units of each basket currency included at a particular point in time to maintain stable purchasing power in terms of some target variable, commonly the baskets of goods and services underlying domestic or foreign price indices. Denoting the target price index in country q by Pq, the weight given to country q at time t is given by


The number of currency units will thus be increasing for countries with inflation, but decreasing for countries with deflation. While fixed weight (“soft”) baskets maintain relative weights in nominal terms, variable weight (“hard”) baskets maintain relative weights in real terms vis-à-vis the target price index. The exchange rate vis-à-vis the basket becomes


Optimal Baskets

Despite the popularity of currency baskets, surprisingly little attention has been given to optimizing their composition—in terms of both components and weights—with respect to an explicit objective function. In the case of the SDR, the stated objectives are quite diverse, including a desire “to give the SDR stability in purchasing power as a reserve asset” (Polak, 1979, p. 635), a desire to obtain a “system of weighting … that is reasonably balanced in relation to the economic importance of the issuers of currencies in total commercial and financial transactions” (Gold, 1981, p. 29), as well as a desire—partly motivating this seminar—to improve the quality of the SDR as a private financial asset and not least as a unit of account. The relative importance of these criteria, the relative weight attached to the performance of the SDR along these dimensions vis-à-vis different national currencies, and the cost-benefit calculus underlying the selected cutoff points for inclusion remain largely unspecified, however.7

The scarcity of work basing the selection of the SDR basket on an explicit optimization function may partly reflect the multitude of tasks the SDR is charged to perform. The optimum basket defined for one particular task can by no means be assumed to be identical—or even reasonably close—to the optimal basket defined for another task. In particular, there exists an uneasy relationship between the desire to improve the SDR as a unit of account and as a reserve basket. The criterion relevant for choosing an optimal basket for the first objective—the correlation structure of exchange rate changes of the component currencies—is only of secondary importance for selecting an optimum reserve basket, where high liquidity is a precondition for any potential basket component. In consequence, there can be no presumption that both properties of a given basket can be improved simultaneously. Indeed, one might expect ex ante that the correlation of exchange rate movements between highly liquid, and typically highly integrated, markets is above average, implying a fairly substandard performance as units of accounts. For instance, the inclusion of both the French franc and the deutsche mark in the SDR basket is sensible from the point of view of creating a reserve basket with liquid components. Yet as the two currencies are linked through the exchange rate mechanism and hence largely move in tandem vis-à-vis all other currencies, inclusion of both makes little sense in terms of a unit of account trying to maximize offsets between its component members to enhance the relative stability of the basket vis-à-vis its components. Ultimately, a partial trade-off likely exists between improving the SDR—or any currency basket—as a unit of account and as a reserve asset. We resolve this problem by considering the two objectives separately, first determining the optimal composition of a currency basket unit of account without regard to its suitability as a reserve asset, then computing the optimal reserve basket without regard to its suitability as a unit of account.

The following subsections describe the objective functions more fully. Before turning to these descriptions, it may be worthwhile mentioning two issues beyond the scope of this paper but arguably deserving further attention. First, we consider only global baskets. An interesting further question is the possible improvement obtainable through regional baskets. Second, we constrain weights to be constant throughout the sample period. As fundamentals change, so do the optimal weights. Hence, starting from an optimal basket, the aggregate cost associated with a particular unit of account increases over time relative to those associated with a continuously re-optimized basket. Very frequent changes in weights reduce the user-friendliness of the unit of account, however. Whereas this consideration was apparently of some importance in the decision for a fixed but adjustable SDR basket, the fixed adjustment period selected, five years, does not appear to have been based on an explicit computation of costs and benefits. The approach adopted here permits a straightforward calculation of the benefits of alternative adjustment intervals—by allowing weights to change after prespecified time intervals—which might then be matched to the perceived informational costs of weight changes to derive an optimal adjustment lag.

Optimal Currency Basket Unit of Account. Our first objective is to find the currency basket minimizing the “noise” in the bilateral exchange rates vis-à-vis the basket. It is assumed that over the longer term, relative purchasing power parity holds as an approximation and that contractees have rational expectations regarding trend differentials in inflation. Consequently, their concern in selecting a unit of account is to minimize the likelihood of movements around the longer-term trend, or “noise,” a concept we capture by removing the mean and a (log-linear) trend from the exchange rate series. The challenge is then to find the basket Bk minimizing the weighted sum over all n countries of the variability of the detrended bilateral exchange rates vis-à-vis the basket


where Bki

article image
and B¯ki
article image
denote the actual and trend level of the exchange rate between currency i and the basket Bk. For the weighting schemes, ω, we employ the proportional IMF quota.8 By definition, the “optimum” basket thus defined depends on the exchange rate policies followed by the countries in the sample. For example, if all countries in the sample had successfully pursued an SDR peg, the SDR would trivially provide the solution. However, while a few currencies have attracted a wide following as pegs (in particular, the dollar (20 countries) and the French franc (14)), the diversity of exchange rate arrangements remains sufficient to make the search for an optimal basket satisfying the above condition a nontrivial exercise.

Optimal Reserve Basket. The second empirical problem addressed below involves the selection of a reserve basket preserving purchasing power in terms of imports or, put differently, a basket whose home currency returns are at least as large as the change in the ratio of the hard unit of account defined in terms of reserves to its initial period level. A number of simplifications are required to render this problem computationally feasible. First, we assume that the import partners and weights are fixed. Second, we restrict attention to the top three import partners for each country. Third, since import price indices are not available for many countries, we assume that the price of imports in trading partner currency rises at the rate of the trading partner consumer price index. Fourth, we assume that the objective of reserve management is solely to safeguard a given level of real purchasing power rather than to maximize returns beyond this given level. In consequence, all baskets at least achieving the minimal level are regarded as equally suited to the task at hand regardless of the magnitude of any excess returns achieved. The assumption reflects the common policy objective of a fixed purchasing power of reserves. Alternative specifications might be considered. However, with positive weights on the total return, the optimal solution typically involves a single currency rather than a basket. Fifth, we take the returns on the reserve assets as given. While our sample of potential reserve assets is limited to the 20 most liquid currencies, the assumption, although unavoidable, is somewhat unsatisfactory for the smaller of these asset markets.

Given these assumptions, and scaling initial reserves and imports to unity, the objective becomes to find the basket ZK maximizing


where k, i, and i denote the basket, time, and the country, respectively. Rk,i,t equals the value, in the home currency of country i, of the reserve basket. The value can change for two reasons. For given exchange rates, interest payments on the foreign assets—determined by equation (2)—raise the home currency value of reserves, while an appreciation (depreciation) of the home currency leads to a capital loss (gain) reducing (raising) the home currency value of reserves. Mi,t equals the value, in the home currency of country i, of a constant real volume of imports. Again, the value can change for two reasons. At given exchange rates, inflation abroad increases the amount of local currency required to purchase a constant bundle of imports, while appreciation (depreciation) of the home currency reduces (raises) the home currency cost of the bundle.

Methodology. A basket is defined by three features: the number of components, the components themselves, and the weights attached to each component, all of which are part of the search for an optimal solution. For a given objective function, the challenge thus involves selecting a given number of currencies out of either the entire set—for the first problem—or out of a subset of the sample currencies—the 20 reserve assets in the second case—and weighing them to satisfy the objective function defined over the entire set of 120 plus currencies. We use a genetic algorithm to search for the maximum in the quite large solution space.

Data. The data set includes all countries covered in the IMF’s International Financial Statistics (IFS) for which continuous quarterly data are available for the period 1985–94, 134 countries for the unit of account problem, and 125 countries for the reserve basket problem. The nominal exchange rates, consumer price indices, and short-term interest rates are taken from IFS, the import trade weights are taken from the IMF’s Direction of Trade Statistics.


We next turn to the results for the two problems, focusing on three features: the relation between the number of basket components and the overall basket performance; the performance of the current SDR relative to the best available unconstrained basket; and the optimal weighting of a basket consisting of the current SDR members.

Optimal Unit of Account

The suitability of individual currencies to serve as optimal units of account varies widely. The first two columns of Table 1 report the currencies with the smallest weighted exchange rate variabilities vis-à-vis the 134 sample currencies. Bahrain and Ethiopia share top honors, with Greece, Sierra Leone, and Norway rounding out the top five. The five top finishers have little in common, be it location, economic development, or exchange rate regime. Nor are they bunched in terms of inflation, either mean or variability.9 The high scores obtained by currencies rarely mentioned in discussions of international monetary reform underline again that the two tasks of improving the SDR as a unit of account and as a reserve basket are not necessarily consistent. Indeed, the top international reserve assets do not perform particularly well as single currency units of account: the United Kingdom makes the best showing in place 20, followed by France at 26, the United States at 49, Germany at 57, and Japan at 68.

Table 1.

Unit-of-Account Currency Baskets

article image

Scores represent quota-weighted sums over all countries of the standard deviations of detrended bilateral exchange rates vis-à-vis the basket during 1985-94; see condition (5).

Percentage share.

The middle column of Table 1 reports the optimal currency basket unit of account for three components, comprising Korea, Cyprus, and Sweden, with a score of 0.078, moderately below that of the best single currency unit of account. The basket components have little in common, a desirable property reducing the likelihood of shared shocks and hence increasing the chance of offsets in movements of the bilateral rates against reference currencies. The third column of the table reports the optimal five-currency basket chosen from the full set of 134 sample currencies. Since both St. Vincent and the Grenadines and Grenada have maintained a peg against the dollar throughout the period, a four-currency basket with the United States at the combined weight of St. Vincent and Grenada would have scored equally well.

Finally, the third column also reports the optimal weights if the current SDR basket were arranged to serve as a global unit of account. In such an optimal basket, the dollar and the pound would enjoy enhanced roles, while the importance of the deutsche mark and in particular of the yen would be attenuated. At 0.0786, this optimal SDR basket scores behind the optimal unconstrained basket, but not by much; a revised SDR with optimized weights thus provides a quite reasonable unit of account. The SDR in its current form scores quite poorly in comparison, however, at 0.08437, substantially higher than the optimal five-member basket. On the plus side, the SDR even in its current form does satisfy the strong criterion for currency baskets as units of account: averaged over all countries, the weighted exchange rate variance of the basket is less than the minimum variance of the five components.

Optimal Reserve Basket

In an ideal classical world with positive real interest rates, purchasing power, and uncovered interest rate parity, the problem posed for the optimal reserve basket—maintaining purchasing power in terms of imports—would be an easy hurdle to surmount. In practice, while home currency real returns on home assets are rarely negative, wide swings in nominal exchange rates occasionally lead to negative real and even nominal home currency returns on foreign assets. Depending on the co-movements of these returns, with deviations from purchasing power parity vis-à-vis import partners, the home currency value of reserves may thus fall short of the amount required to purchase a constant quantity of imports.

As it turns out, this possibility remains, at least for 1985–94, largely in the theoretical realm. The first column of Table 2 reports the performance of each candidate reserve currency under the assumption that reserves are held exclusively in that currency, with scores indicating the value of equation (6) using equal weights (ωi—1/125) for each currency. The maximum score is thus 5,000, the product of the number currencies and the number of calendar quarters. Out of the 20 candidate markets for holding interest-bearing reserves,10 17 generated returns sufficient to cover imports in most periods for most countries. The main exception has, alas, been the major reserve currency actually used: home currency returns on the U.S. dollar were negative for the sample period for a significant number of countries. The other two individual reserve currencies with ex post declines in purchasing power vis-à-vis imports for a significant number of countries are Canada and Australia, both fairly closely co-varying with the dollar.

Table 2.

Reserve Currency Baskets

article image

Value of equation (6) with weights equal to 1/125 for all countries. Maximum feasible score is 5,000.

5,000 minus the score as defined in footnote 1.

The second and third columns report the optimal composition of a 3-and a 5-currency reserve basket drawn from the group of 20 currencies. The relative performances of these multiple-currency baskets are characterized in terms of losses, shown at the bottoms of the columns, which are simply the amounts that their scores fall short of the maximum. While performance of the 5-member basket is better, the bulk of the optimal 5-currency basket consists of Ireland, Italy, and New Zealand, with a small weight on Sweden and France. Holding reserves in these 5 currencies would have enabled the 125 countries in the sample to maintain the purchasing power in terms of imports nearly perfectly, with a score on the objective function above 4,999 out of 5,000. As market depth in Ireland and New Zealand falls significantly short of the markets in the 5 current members of the SDR basket, the assumption of exogenously given returns might raise some eyebrows. But the very good individual performance of 17 of the 20 currencies suggests that well-performing baskets can also be found for a more restricted set of assets. Indeed, the best-performing current SDR member—France—boasts a performance that, while falling short of the optimal 3-membcr basket, is north of 4,999 and thus very close to the optimum in absolute terms. Imposing stricter conditions on the liquidity of candidate currencies thus will not materially alter results. One way of tightening is to restrict attention to the current SDR components. The optimal weights are presented in the fourth column of the table, revealing essentially a 2-component basket consisting of the French franc and sterling, with a minor role played by the deutsche mark and virtually no showing of the yen and the dollar. The performance of this optimal SDR basket, while significantly worse than the optimal unrestricted 5-component basket given in the third column, is again above 4,999 and thus absolutely very close to the maximum score.

There thus exists a wide variety of baskets scoring in the vicinity of the maximum, including baskets composed of the current SDR currencies. While performance deteriorates as the number of components are reduced, the deterioration is quite muted, with absolute scores remaining solidly north of the 4,999 level even for a single component. Is it, then, the case that the objective function is so easy to satisfy that virtually any currency basket scores well? Alas, no. In particular, the current SDR basket, at 4,766, scores very poorly, reflecting the high weight given to two relatively poor individual performers, the United Kingdom and the United States.


Increased volatility of nominal and real exchange rates following the collapse of the Bretton Woods system has raised interest in finding international units of account exhibiting reasonable stability vis-à-vis national monies, as well as reserve baskets with reasonably stable purchasing power in terms of imports. Currency baskets, notably the ECU and the SDR, have attracted significant interest in this regard. Surprisingly little attention has been given to optimizing these baskets vis-à-vis some explicitly stated objective function. It is hoped that the evidence in this paper presents some steps in this direction.

We computed the baskets maximizing two objective functions—a unit of account problem based on a low variability of exchange rates vis-à-vis individual currencies and a reserve basket problem based on maintaining real purchasing power in terms of imports. The analysis provided was strictly ex post. The optimal baskets defined above would have achieved the stated objectives during the sample period. However, little can be said about whether they will outperform alternative baskets in the future. The SDR in its current form proved to be fairly poorly suited to either problem relative to other baskets. However, the gap between the SDR and the best basket could be significantly narrowed by selecting optimal weights.

Beyond the specific results provided, the discussion pointed to a deeper problem in attempting to improve the functioning of the SDR as a unit of account and as a reserve asset: the qualities making a currency a good candidate differ across the two objectives. The primary membership criterion facing a candidate for membership in an optimal unit of account is a low correlation with the other members. Since no transaction need be undertaken with the unit of account, the liquidity characteristics of the candidate do not matter; consequently, we found relatively less known currencies in the optimal unit of account basket. In contrast, every member of an optimal reserve basket must be highly liquid; consequently, we restricted our search ex ante to the set of major international currencies. Fundamentally, there is thus no reason to expect—and ample reason not to expect—that the optimal unit of account basket will show any overlap with the optimal reserve basket. Nor will it necessarily be true that, starting from any arbitrary basket, altering weights can improve the quality of the basket along both dimensions—as borne out by our empirical results. The trade-off is thus inescapable—simultaneous improvement of the SDR as both a unit of account and a reserve basket is outside the realm of the feasible. In the end, a policy decision on the weighting to be given to the two criteria—if indeed they are both to be fulfilled by the SDR—must be taken.

Arguably, the reserve asset function is of greater promise as an area of improvement than the unit of account function. While both the perfect-markets optimal reserve basket and the optimal unit of account basket differ across nations, the real-world cost entailed in the loss of liquidity in a tailor-made reserve basket will, for most countries, provide a significant offset against the cost of holding reserves in a somewhat suboptimally weighted and composed but highly liquid standard SDR basket. The case for a uniform global reserve currency basket is thus fairly strong even if the strictly optimal composition of a tailor made basket differs across nations. The same argument does not, however, hold with equal strength for a global unit of account that is not simultaneously a liquid reserve basket. There is little apparent cost in adopting an individually optimal tailor-made unit of account instead of a globally optimal unit.

As a final caveat, it bears emphasizing again that the results presented above are strictly backward looking. We asked whether the SDR, in its current definition, was a good competitor relative to alternative baskets in 1985–94. We did not ask, and cannot answer, whether the SDR will be a good or a bad performer going forward. For instance, the dollar experienced a strong trend depreciation during the sample, resulting in a poor showing in terms of the reserve basket criterion. Looking forward, many observers believe that the long episode of dollar weakness will be replaced by a period of stability of even appreciation. If this should indeed be so, the performance of the dollar in the reserve basket will markedly improve. In like vein, there is little reason to believe that correlation patterns between exchange rates will stay constant over time. On the one hand, correlations of changes in the equilibrium rates depend on the difficult-to-forecast incidence of future supply and demand shocks. On the other hand, noise movements and peso events may significantly alter the correlation structure.

We thus do not view our results as indications of desirable changes in SDR weights. Rather, the hoped-for contribution of the paper is, by establishing that the SDR in its current form has ex post not been a particularly good unit of account or reserve basket, to stimulate discussion of the appropriate methodology for assessing the composition of the basket for future revisions.

Does the World of the SDR Still Exist? Jonathan Frimpong-Ansah

Alec Chrystal and Holger Wolf have presented to us well-researched papers on the characteristics of the SDR that should form the basis for a fruitful discussion this morning. Although the titles of the respective papers distinguish between the SDR as an international reserve asset and the SDR as an international unit of account, the two authors have researched both characteristics of the SDR in their respective papers. In doing so they have come up with useful relative analytical results that will contribute immensely to the evidence on the characteristics of the SDR that we need to answer this morning: whether the SDR has performed well enough as an international reserve asset and as an international unit of account to deserve to be preserved, or modified to serve some specific role in a future international monetary system.

The primary characteristics of the SDR, as envisaged in the early 1970s, were that it should become the principal international reserve asset and also a unit of account for international transactions. Implicit in these primary characteristics was an envisaged major reduction in the role of existing reserve currencies. It was expected that the U.S. dollar, in particular, would be gradually phased out as a reserve currency. Also implicit in the primary characteristics of the SDR was an enhanced role for the IMF as the managing institution of global liquidity. A third characteristic, though not well emphasized, was to establish a link between SDR allocation and development finance.

What has come out clearly in the two papers is that the two important enabling conditions for the SDR have not materialized after over two decades of its implementation. The IMF was not enhanced as a managing institution of global liquidity, the U.S. dollar and other leading currencies were not phased out, and therefore the SDR never became the principal reserve asset. There are a number of reasons:

  • The international monetary environment in the 1960s and early 1970s, when the SDR was conceived, was inflexible. In the advanced countries, exchange rate rigidity prevailed despite flexible market oriented trading conditions in commodities. In the large majority of developing countries, there were rigidities in both exchange rates and domestic pricing. The culture of economic rigidity that prevailed at the time suggested a centralized management of global liquidity. The situation has dramatically changed. The past two decades have witnessed generalized policy shifts in both advanced and less developed countries toward economic flexibility. In the resulting liberalized world exchange rate markets involving all major trading currencies, “the case for increasing SDR allocations cannot be made with reference to global reserve needs, as aggregate ‘liquidity’ shortages cannot exist,” to quote Alec Chrystal.

  • A second factor is that the rivalry between national sovereignty and multinational bureaucratic management has rarely been won by the latter, and it was perhaps naive on the part of the Committee of Twenty to hope that the SDR was going to be a special case to reverse the culture of national sovereignty. If the IMF itself and the Committee of Twenty had the benefit of today’s clearer perception, perhaps other approaches emphasizing adjustment rules for the reserve currencies might have been more successful. With the benefit of hindsight, it is easier to see that the events leading to the suspension of the work of the Committee of Twenty—particularly the floating of some major currencies in 1973—and the creation of the Interim Committee were important pointers that the SDR and the IMF’s managing role were not expected to be a total success,

  • With regard to the developing countries, the awakening to the logic of free markets and liberalized exchange rates in the past decade has resulted in better economic management. Today’s generally better economic management, promoted jointly by the IMF and the World Bank and the donor community, has resulted in greater inflows of external official and private capital. Exchange rate liberalization has also reduced the demand on central banks for external reserves. Some of the developing countries of the 1960s and 1970s have actually made the transition to become newly industrial countries. These trends have made the need for the SDK/aid link less topical, except in terms of the relative costs of borrowing and except for the few countries that still retain exchange rate rigidities.

Given this background, the conventional reasons (Williamson, 1994, and Crockett, 1994) for the demise of the SDR cannot be lightly dismissed. However, the focus of the two papers before us by Chrystal and Wolf on additional technical refinements offer fruitful areas for discussion and should be seen as important contributions to the debate on the future role of the SDR if the attempt is to be made to retain it and modify its role. Let us consider the more significant points in the two papers.

By focusing on total reserve need, that is, official and private needs, and making full use of the historical literature and analysis in certain cases, the authors have been able to introduce standard methods of verifying the performance and usefulness of the SDR as an international unit of account, medium of exchange, and a store of assets, in the past three decades. Chrystal concluded that on all points the SDR scores low marks.

Wolf goes further to show quite convincingly that the contents of a basket of currencies and the respective weighting not only make the difference in the basket’s operational usefulness but could be contradictory in simultaneously attempting to achieve a high-value basket and a stable unit of account. The questions then are:

  • If the SDR is to be retained, there is a selection exercise to make. Should it be used only as a store of assets or as a unit of account?

  • How would this be compatible with a possible limited role in official transactions within the IMF?

  • How should the weighting in the basket be revised in the different uses of the basket?

  • How do we find a solution for a future euro in the basket, as suggested by Chrystal?

  • And should there be a different system from the one that now exists for periodically revising the weighting in the basket?

These are all difficult areas that need further study, but may never be fully resolved.

Chrystal has raised an important fundamental question on whether the SDR did not contain the seeds of its own destruction. This is important in view of present trends of flexibility in the international monetary system. An impotent SDR may have created room for a relaxation in monetary and fiscal management in the countries issuing the major reserve currencies. The impression is created that the international monetary system might have been better off if the SDR had not been invented, or if there had been greater debate on it. I would like to seek an explanation in the more conventional argument and to ask whether an investigation into the factors that have induced exchange rate liberalization in the past two decades, SDR or no SDR, may not lead us to a more plausible explanation for the demise of the SDR. It is easier to see the hand of generalized economic flexibility in Oppenheimer’s (1987) argument that liquidity and adjustment are very much related. I would then agree with those who argue that the adjustment problem, which was overshadowed by the debate on the SDR in the 1960s and 1970s, was equally important.

But can we not say that the IMF’s subsequent promotion of liberalization and exchange rate flexibility repaired the damage and indeed reduced the envisaged role of the SDR? I see in these arguments some important lessons for the future of any global currency basket:

  • Trends in exchange rate flexibility will decidedly affect its fate;

  • It is better not to have a basket at all than to have an impotent one; and

  • We may have to think of simpler but more implementable solutions to future liquidity problems.

I also have a problem with the argument regarding the SDR/aid link in Chrystal’s paper. Let me start by saying that there is no misunderstanding concerning “the dangers of confusing the need for wealth redistribution with the issue of design of the monetary system.” The argument that the SDR has been beneficial to the less creditworthy countries also cannot be denied. But this is the case where access to SDRs has been easier and borrowing more expensive than the SDR rate. Unfortunately, access to SDRs has not been easy, and the borrower in a developing country has often had to pay a higher interest rate irrespective of his repayment record and irrespective of the greater effort being made with structural adjustment and better economic management.

The development effort is therefore inherently more expensive than it need be. The case that Chrystal makes, that the “benefits of the SDR facility are already biased toward the least creditworthy countries… notwithstanding the fact that allocations have so far been linked to quota,” is therefore a weak one, especially in the situation in which “the major industrial countries have been reluctant to sanction further SDR allocations.” We should all be happier to see the developing countries persevering in their present efforts at better economic management and consequently de-linking themselves from the aid syndrome. In this context, is the link idea not a red herring?

We now come to the most important conclusion arrived at in the two papers: that the SDR stands no chance of becoming an international unit of account or a major reserve asset, particularly in the unofficial markets. Chrystal, however, says it is not worthwhile getting rid of it because it already exists and it serves a useful purpose in official transactions and that the process is not worth the cost involved.

Wolf, on the other hand, says that it is not feasible simultaneously to improve the SDR as a unit of account and a reserve basket with a common weighting and composition. He goes on to say that there may be a good case for a uniform global reserve currency basket because most countries can create tailor-made reserve baskets with benefits that can offset the cost of holding a part of their reserves in a highly liquid standard SDR basket, even if it is suboptimally composed and weighted. In other words he suggests that the SDR may be retained and its weighting improved to serve as a partial reserve asset but not as a unit of account.

Both arguments are not convincing for the future of the SDR as an international reserve asset and as an international unit of account for the following reasons:

  • They show that in a real-world situation, the SDR can only be a second-best instrument as either a reserve asset or a unit of account.

  • They can only offer a weak case for the long-term retention of the SDR as a credible international reserve asset, even in the context of limited use.

We are therefore left with the unanswered question of whether the IMF, as a highly credible institution, should continue to promote and manage a reserve asset with suboptimal credibility and an uncertain future.

Given the fact that except for in a few poorly managed countries, there appears to be little enthusiasm for the SDR, one is forced to the conclusion that the two papers do not offer us a convincing case for the continued retention of the SDR with the characteristics originally intended. The suggestion that it should be retained for official transactions between the Fund and the Bank and member countries is perhaps the best that can be hoped for. But we remain uncertain about how long this limited use will last in an international monetary system that continues to be more flexible and in which the role of central banks continues to shrink.

Based on the contents of the papers I was mandated to discuss, I selected, as the title to this paper, “Does the World of the SDR Still Exist?” The two papers suggest that the world that exists today is not that in which the present SDR, in the present constitution of the Fund, can happily survive. But that does not mean that today’s world does not need another type of liquidity with its special characteristics. Today’s monetary world is more flexible and more globalized. But there are many economies in transition, and these include the East European countries, the Pacific Basin countries, and many of those formerly described as developing countries, going through the process of democratization in the rest of the world.

To think of the liquidity needs of all these countries in the way we thought of the link some twenty-five years ago is decidedly incorrect. Let me offer some suggestions, and I offer them in the framework of an undisputable fact: that the IMF is one of the world’s most credible multilateral finance institutions, which must be so preserved by endowing it only with responsibilities and provisions that are credible and implementable:

  • If all indications point to some form of a liquidity provision in the Fund, it is important for a quick decision to be reached on its characteristics and allocations made, however small, to break the fifteen years of inactivity.

  • The specific reference to the SDR/aid link should be dropped, because it does not do the image of the Fund any good: it is too controversial, it is diversionary, and it will never be implemented.

  • Steps should be taken to revise the Articles of Agreement of the Fund to remove the present characteristics of the SDR and replace them with characteristics that are acceptable and implementable. The estimation of a global liquidity need is a problem, and the allocation of SDRs on a global basis is also a problem; a global basket cannot be simultaneously a satisfactory reserve asset and a unit of account. We cannot produce an SDR that is a currency understood by a world of individuals. All these point to a smaller, national assessment of liquidity needs and regular ongoing allocations, rather than to global estimations and distribution.

  • Finally, we may be guided by the experience we have gained: any ideas on the creation and management of liquidity by the IMF that appear to be in conflict with the national sovereignty of member states, or smell of wealth redistribution, must be regarded as unworkable and unimplementable and no attempt should be made to obtain cosmetic agreement. If Keynes failed to make the IMF a world central bank, I doubt that any of us can do it.

The SDR as an International Means of Payment

H. Robert Heller

In this paper I will explore whether the SDR could be used as an international means of payment that would avail commercial and central banks with a means of making and settling international payments. The proposed payments function for the SDR is not thought of as a way to execute the majority of international payments. It is designed as a potential “backbone” for the international payments system that can be used to make payments in cases where assured finality on the books of a central bank or an official authority is of importance and during times of international financial stress.

Money is generally defined as an asset that fulfills three functions: it serves as a unit of account, as a store of value, and as a means of payment. Throughout this seminar, the emphasis has been on the unit of account and the store of value functions of the SDR. Not much attention has been paid to the SDR as a means of payment.

As the SDR has been less than a resounding success in international finance, we should ask ourselves why this has been so. Maybe the answer to this question is to be found in the absence of a true payments function for the SDR. Without such a payments function the SDR does not have all of the necessary attributes to make it a useful international monetary asset.

The SDR and the Need for an International Payments System

While the IMF is wondering why the SDR is not a resounding success, payments system working groups convened by the Bank for International Settlements (BIS) note with concern that there exists no truly global payment and settlement system that allows the users to obtain final settlement of cross-currency payments in an efficient and secure manner.

In a series of reports, including the Angell Report, the Lamfalussy Report, and the Noel Report, the various BIS working groups have analyzed the defects of the existing payment systems and catalogued ways in which the international payment system could be made more efficient and secure.

If we join the two questions, what can be done to make the SDR more successful? and what can be done to improve the functioning and the security of the international payment system? we just may be able to find one solution to both problems.

Lack of a Risk-Free International Payment System

At present, there exists no truly global payment system. Instead, we have a situation where various national payment systems are being used for both national and international purposes. Payments are executed in terms of national currencies, and ultimate settlement occurs with finality on the books of the central bank responsible for the issuance of that currency.

In addition to the official central bank payment system, various wholesale or retail payments may exist in a country. They may or may not have access to the facilities of the central bank.

Working groups constituted by both the private and the official sector have done much to study and clarify the risks involved in international transactions, and steps have been taken to reduce the risks inherent in making and settling payments. But risk reduction and risk management are not the same as risk elimination.

Risk elimination in payment systems is associated with payment finality. Payment finality is defined as the moment when a payment has become irrevocable and any future claim is extinguished. One way to provide payment finality is to effect an irrevocable transfer of assets on the books of the central bank.

The Missing Links in International Payments

A major source of risk in international payments is “principal” or “Herstatt” risk. It arises from the general inability to settle both sides of a foreign currency transaction at the same time with finality. When Bankhaus Herstatt failed in 1974, the deutsche mark side of substantial foreign currency transactions was settled with finality at the end of the German banking day, but the U.S. dollar side of the same transactions had not yet been paid. Thus, the entire principal amount of the foreign exchange transaction was found to be at risk.1

If one considers that the global foreign exchange volume now averages over $1 trillion dollars a day, it is clear that the risks involved are considerable.

Possible Solutions

The Noel Report of the BIS discusses possible solutions to the international payments dilemma. It discusses the following four possibilities to reduce payments risks:

  1. Make home currency payment and settlement systems more efficient and allow finality in intraday payments. That is, the Report finds that one risk-reducing action would be a move from once-a-day net settlement system to several settlements a day or to a real-time gross settlement system. Nevertheless, the Report recognizes that owing to the nonsynchronization of operating hours of the various payment systems, this will not permit the final delivery-versus-payment of all cross-currency transactions.

  2. Extend the operating hours of the various national payment systems so that there will be overlapping hours that will permit the simultaneous settlement of foreign currency transactions. Again, this solution would work only in the case of real-time gross settlement systems or systems with netting facilities operable during the time of operational overlap.

  3. Cross-border operational links between payment systems could give participating central banks the joint capability to execute transfers among the various currencies. Here the need for tight operational coordination is obvious and payments could again be executed only during hours of operational overlap.

  4. Central banks could operate multicurrency payment and settlement services that might be provided by a common agent acting on behalf of the central banks. In this system, payment and settlement facilities in the various national currencies would be provided by the common agent. Clearly, the common agent has to have sufficient liquid resources at its disposal to support this multicurrency payment and settlement system.

Clearly, there are potential problems with each one of these solutions. All of them have in common that they try to create an ever-expanding web—both over time and over geography—of central bank relationships. If one considers the implications of almost 200 countries around the globe participating actively in such a global payment and settlement system, one sees that the operational problems will eventually become overwhelming—long before all 200 countries are active in the system.

SDR-Based Payment Facility

Another way to provide for global payment settlements with finality around the clock would be for the IMF to operate what may be termed an “SDR payments facility.” Conceptually, this approach is similar to alternative (4) discussed in the Noel Report. But there is one decisive difference: instead of having a “common agent” operate a multitude of parallel payment systems in the various national currencies, the SDR payments facility would constitute a single central way station for cross-currency payments that would be operated by the IMF or a new IMF subsidiary on behalf of the central banks.

The SDR payments facility would be open to central banks and potentially also to commercial banks. Participants would maintain an SDR account at the IMF by depositing convertible currencies with the IMF. If desired for credit quality reasons, such an account could be funded only with central bank money.

Participants in this facility would maintain positive account balances. Thus, the IMF would come to hold a portfolio of convertible foreign currencies that could be used to make payments by holders of SDR balances.

All payments would take place in real time, so that the SDR payments facility would be a true real-time gross settlement system.

Operationally, a transaction from, say, Japan to the United States would look as follows:

The Japanese firm would instruct its bank to debit its yen account; the bank would pay the yen to the IMF (maybe through the intermediary of the Bank of Japan). The IMF would credit the Japanese bank’s SDR account and guarantee that the equivalent dollar amount would be credited to the U.S. bank’s account for the benefit of the ultimate U.S. recipient.

Transactions could be carried out over the existing SWIFT network, so that there would be virtually no additional costs involved in erecting a new payment system infrastructure. Only the relevant accounts would need to be opened at the IMF and funded by the participants.

Such an SDR payments facility would provide a payment system backed by one of the strongest financial institutions in the world. The IMF would also have access to additional credit facilities at central banks—thus assuring unquestioned access to liquid resources and providing for assured payment with finality. Central banks could issue the appropriate guarantees and make liquid funds available when needed.

Over time, banks and central banks would find it convenient to hold sufficient SDR balances at the payments facility so that payments out of the existing funds could be carried out at any time during the day or night. Thus, the SDR would become a central asset of the international financial system.

Some Concluding Thoughts

The world financial system is confronted with a significant problem: to provide for efficient and risk-free payments around the globe. Europe’s solution is to move toward a common currency that will obviate the cross-currency payment and settlement problem. Central and commercial banks around the world continue their efforts to reduce risk and to increase efficiency in cross-currency payments. But no one has yet developed the perfect global payment system that is risk free and works on a continuing basis at low cost.

At the same time the IMF finds that SDRs, which were initially conceived to become the central reserve asset of the international monetary system, are not used extensively. One reason for this lack of interest in the SDR is that while it functions as a unit of account and as a store of value, it does not have a true payments function. That is, the current SDR is not a fully functional international money, which would require that it have all three features.

Providing an international payments function through the establishment of an SDR payments facility at the IMF would contribute toward making the SDR a more usable asset. It would also solve a key problem for the international financial system by providing for essentially risk-free settlement of cross-currency transactions on a real-time gross settlement basis at any time of the day or night.

The implementation of such an SDR payments facility would not reduce national sovereignty in any significant way. National currencies issued by the respective central banks would still fulfill their national payment system roles. The SDR payments facility would only be used in international transactions that involve a switch into a foreign currency.

It is open to question whether a significant portion of international payments will actually flow over such an SDR facility. Making payments through the SDR payments facility involves a double conversion from the payer’s currency into SDRs and then into the payee’s currency. This is a more cumbersome and expensive procedure than is used by most payment systems—although even under current arrangements a significant portion of foreign exchange transactions involves a move through the dollar or another key currency. Thus, the SDR payments facility may be particularly attractive for payments from one minor currency into another minor currency.

The SDR payments facility would provide an official international linkage between the various national payment systems. It would be a “backbone” system that would lend increased stability to the entire international payment system. This would be particularly true in times of crisis, when doubts as to the reliability of private payment systems may become more pronounced.

The SDR payments mechanism would be particularly useful for transactions that need to be executed with certainty—and where slightly higher costs may not be an important factor.

Because the SDR payments facility would eliminate the risk of not being able to settle cross-currency transactions, the international financial system would become safer—a longtime goal of central banks and commercial banks alike.

General Discussion

Montek Singh Ahluwalia asked Holger Wolf whether the whole issue of an optimal basket for the SDR might not be a red herring. If such a basket were needed to remove from the decision making of individual participants what was actually a hedging decision, the markets would have produced an optimal basket. But they had not, presumably because different parties wanted very different baskets. The combination became important for the SDR because an exposure had to be taken or an asset accepted in a particular combination. But the IMF’s particular basket could always be offset in whatever hedging decision or optimum mix decision countries took in their other liabilities and assets. The present combination was almost not worth re-examining, and any kind of exercise based on the past, in terms of improving the basket, was of no relevance.

Wolf agreed in part. A main question in his paper had been how well the SDR basket had actually performed relative to other baskets over the last ten years. Also, in the next revision of basket weights, a decision might have to be made about what precisely the SDR was supposed to do. It should, for example, perhaps not be expected to serve widely as a unit of account outside the IMF.

In conceding Ahluwalia’s point on reserve assets, Wolf remarked that as the externality associated with the SDR being selected as a reserve asset would accrue to the currency components of the SDR and not to the SDR itself, there was really no case for trying to improve the SDR as a reserve basket at present. But the evidence did not suggest that virtually all international transactions could be easily hedged at this stage. Hedging was available only among a small subset of 20 international currencies and only for contracts of relatively short duration. Therefore, for any contracts or contingent liabilities of longer term, selecting a unit of account to minimize the residual exchange risk that could not be hedged was still relevant.

Onna Ruding, reacting to Heller’s proposal for handling payment and settlement risks, queried whether the IMF would be willing to serve as a clearing institution. He was puzzled that the central bank governors in the Group of Ten and the BIS, who were very much aware of this issue and were willing to cooperate, had, to his knowledge, never referred it to the IMF to resolve.

On the “link” proposal mentioned by Alec Chrystal and Jonathan Frimpong-Ansah, Ruding remarked that he had learned after many years of dealing in politics and elsewhere with SDR allocations that the most effective way to kill any proposal for SDR allocations was to advocate a link to development finance. He did not welcome this conclusion, but the link strengthened the arguments of those opposed to a new allocation because it ran counter to the criteria as stated in the Articles. If the link was to be promoted, the Articles would have to be changed. Therefore, he advised developing countries not to press too hard for SDR allocations as an indirect way of enhancing the flow of development assistance because this approach would be counterproductive, at least politically.

Chrystal responded that he had been trying to explain why this political impasse existed: Why did the industrial countries not want more allocations, and why did the developing countries want them? The problem for the SDR had been precisely that although there was no explicit aid link, there was a kind of implicit aid link, given the implicit interest subsidy associated with it, and that problem had to be overcome if the SDR was to be pushed forward. He did not think he was competent enough to come up with a solution, but the problem was there and had to be faced.

Fabrizio Saccomanni agreed with Ruding that the EC central bank governors at the BIS had been actively weighing the risks involved in the operation of an international payments system. They considered, however, that this matter should remain within the field of central bank cooperation because of its very important implications for the conduct of domestic monetary policy. The risk arose if a participant in the settlement system was unable to fulfill its obligations; there was also the question of whether the central bank should act as a sort of lender of last resort. Although this question had now been put into an international context, and the Noel Report had talked of setting up a common agent, the implication should not be that central banks were willing to distance themselves from the operation of such a common agent. In fact, the common agent should be an emanation of the central banks themselves.

Warren Coats queried Heller’s description of the clearing and settlement activities that he had proposed for the Fund. Heller did not seem to be proposing a clearing and settlement function such as the ECU clearing system; he seemed to propose having the Fund actually buy and sell currencies, thereby raising the same problem that the substitution account had faced. By bringing together both sides of the transaction, the Fund could ensure that both sides would be settled simultaneously, but it should not be involved in buying and selling currencies. One party should buy the SDRs by paying for an exchange to another party that had SDRs to sell.

Chrystal raised the question of why, even if a new payments mechanism was needed, it had to be in SDRs? Was it just because the IMF happened to have SDRs, even though all other payments were being made in currencies?

Robert Heller agreed with Coats that a clearing settlement facility in the IMF could basically operate as Coats had described it; the modalities could be arranged in various ways.

On Saccomanni’s intervention on an international payment system, Heller recognized that it was being discussed but queried whether five or six different payment settlement facilities were needed or whether a common agent would be preferable. The central banks had one common agent, which is what he was essentially suggesting, just as the IMF acted as the common agent. If the BIS was to be that common agent, that was also fine, but the BIS still kept its books in gold francs whereas the IMF used SDRs. The SDR was probably a little closer to current reality, he said, though some of his former colleagues at the Federal Reserve might prefer the gold franc. Those issues were for discussion, but he thought that everybody had agreed on the need to have some kind of a backbone to the international financial system, and he was attempting to merge two issues that were presently on separate tables.


  • Bergsten, Fred C, “Discussion,” in The International Monetary System, ed. by Peter B. Kenen, Francesco Papadia, and Fabrizio Saccomanni (Cambridge: Cambridge University Press, 1994), pp. 4350.

    • Search Google Scholar
    • Export Citation
  • Chrystal, K. Alec, International Money and the Future of the SDR, Essays in International Finance, No. 128 (Princeton, New Jersey: Princeton University, Department of Economics, International Finance Section, June 1978).

    • Search Google Scholar
    • Export Citation
  • Chrystal, K. Alec, “International Money and International Reserves in the World Economy,” in Trade and Payments Adjustment Under Flexible Exchange Rates, ed. by John P. Martin and Alasdair Smith (London: Macmillan, 1979).

    • 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; San Diego: Surrey University Press in association with Academic Press, 1990), pp. 928.

    • Search Google Scholar
    • Export Citation
  • Crockett, Andrew, “The Role of Market and Official Channels in the Supply of International Liquidity,” in The International Monetary System, ed. by Peter B. Kenen, Francesco Papadia, and Fabrizio Saccomanni (Cambridge: Cambridge University Press, 1994), pp. 82100.

    • Search Google Scholar
    • Export Citation
  • Gilbert, Milton, The Gold-Dollar System: Conditions of Equilibrium and the Price of Gold, Essays in International Finance, No. 70 (Princeton, New Jersey: Princeton University, Department of Economics, International Finance Section, October 1968).

    • Search Google Scholar
    • Export Citation
  • Krugman, Paul R., and Maurice Obstfeld, International Economics: Theory and Policy (New York: HarperCollins, 3rd. ed., 1994).

  • Marx, Karl, Das Kapital (1886).

  • McKinnon, Ronald I., “The Rules of the Game: International Money in Historical Perspective,” journal of Economic Literature, Vol. 31 (March 1993), pp. 144.

    • Search Google Scholar
    • Export Citation
  • Oppenheimer, Peter M., “The Endogeneity of International Liquidity,” in The Reconstruction of International Monctary Arrangements, ed. by Robert Z. Aliber (New York: St. Martin’s Press, 1987), pp. 30519.

    • Search Google Scholar
    • Export Citation
  • Thakur, Subash, “The ‘Hard’ SDR: An Exploratory Analysis,” Staff Papers, International Monetary Fund, Vol. 41 (September 1994), pp. 46087.

    • Search Google Scholar
    • Export Citation
  • Triffin, Robert, Gold and the Dollar Crisis: The Future of Convertibility (New Haven, Connecticut: Yale University Press, 1960).

  • Williamson, John, “The Rise and Fall of the Concept of International Liquidity” in The International Monetary System, ed. by Peter B. Kenen, Francesco Papadia, and Fabrizio Saccomanni (Cambridge: Cambridge University Press, 1994), pp. 5364.

    • Search Google Scholar
    • Export Citation


  • Coats, Warren L., Jr., “In Search of a Monetary Anchor: A ‘New’ Monetary Standard,” IMF Working Paper 89/82 (Washington: International Monetary Fund, October 1989).

    • Search Google Scholar
    • Export Citation
  • Fama, Eugene F., “Banking in the Theory of Finance,” journal of Monetary Economics, Vol 6 (January 1980), pp. 3957.

  • Fisher, Irving, “A Compensated Dollar,” Quarterly journal of Economics, Vol. 27 (1913), pp. 21335.

  • Gold, Joseph, “Development of the SDR as Reserve Asset, Unit of Account, and Denominator: A Survey,” journal of International Law and Economics, Vol. 16, No. 1 (1981).

    • Search Google Scholar
    • Export Citation
  • Greenfield, Robert L., and Leland B. Yeager, “A Laissez-Faire Approach to Monetary Stability,” journal of Money, Credit and Banking, Vol. 15 (August 1983), pp. 30215.

    • Search Google Scholar
    • Export Citation
  • Hall, Robert E., “Explorations in the Gold Standard and Related Policies for Stabilizing the Dollar,” in his Inflation: Causes and Effects (Chicago: University of Chicago Press, 1982), pp. 11122.

    • Search Google Scholar
    • Export Citation
  • McCallum, Bennett T., “Bank Deregulation, Accounting Systems of Exchange, and the Unit of Account: A Critical Review,” NBER Working Paper No. 1572 (Cambridge, Massachusetts: National Bureau of Economic Research, March 1985).

    • Search Google Scholar
    • Export Citation
  • Polak, Jacques, “The SDR as a Basket of Currencies,” Staff Papers, International Monetary Fund, Vol. 26 (December 1979), pp. 62753.

    • Search Google Scholar
    • Export Citation
  • Thakur, Subhash, “The Hard SDR,” Staff Papers, International Monetary Fund, Vol. 41 (September 1994), pp. 46087.

  • Williamson, John, “Valuation of the SDR: The Case for a Standard Five-Currency Logarithmic Basket,” in The International Monetary System and Its Reform, Part 2, ed. by Sidney Dell (Amsterdam: North-Holland, 1987), pp. 393406.

    • Search Google Scholar
    • Export Citation


  • Crockett, Andrew, “The Role of Market and Official Channels in the Supply of International Liquidity,” in The International Monetary System, ed. by Peter B. Kenen, Francesco Papadia, and Fabrizio Saccomanni (Cambridge: Cambridge University Press, 1994), pp. 82100.

    • Search Google Scholar
    • Export Citation
  • Oppenheimer, Peter M., “The Endogeneity of International Liquidity,” in The Reconstruction of International Monetary Arrangements, ed. by Robert Z. Aliber (London: Macmillan, 1987), pp. 30519.

    • Search Google Scholar
    • Export Citation
  • Williamson, John, “The Rise and Fall of the Concept of International Liquidity,” in The International Monetary System, ed. by Peter B. Kenen, Francesco Papadia, and Fabrizio Saccomanni (Cambridge: Cambridge University Press, 1994), pp. 5364.

    • Search Google Scholar
    • Export Citation

The author is grateful to Forrest Capie, Peter Oppenheimer, Geoffrey Wood, and D. Sykes Wilford for suggesting improvements to an earlier draft, and to Paul Goudie of Barclays Bank for information about activity in SDR-denominated deposits.


Of course, in reality there is no free lunch here. Foreign exchange reserves are expanding but so are domestic currency liabilities. However, the gross yield on the foreign exchange reserves would rise and the reserves constraint would be shifted. In effect, by this route, monetary authorities could print their own reserves.


Even more remarkably, I was reliably informed at this seminar that this statement attributed to President Nixon was originated by Milton Friedman.


In calculating the net benefit of SDR allocations, I have difficulty accepting the calculation put forward by Michael Mussa in his paper to this seminar (chap. 4). It is clear that there is a net benefit that arises from the difference between the SDR loan interest rate and the rate at which each country can borrow on world markets. However, there are two points of contention. First, Mussa implies that this net gain to users of SDRs is a free good. Surely, the international community is picking up this risk. It may be small but it is not zero, and it would increase with the scale of SDR allocations. An SDR allocation gives the less creditworthy countries an option that is valuable to them, but someone is writing that option. Second, Mussa compares the interest cost of drawing down an allocated SDR with the cost of acquiring reserves through sterilized intervention, where the domestic bonds issued bear a very high interest rate. This seems to me to be an illegitimate calculation. The nominal level of domestic currency interest rates is not the appropriate shadow price in this context and greatly exaggerates the benefit from SDR allocations.


Although private use of the SDR is minimal, it is not zero. Two U.K. banks, Barclays and Standard Chartered, and one Belgian bank, Credit Bank, take SDR-denominated deposits. In Barclays, these are all one-month or sixth-month deposits. The typical depositor is a monetary authority from the Middle East. The market is small—Bardays’ book is about SDR 100 million—and there is no SDR loan market. The banks therefore have to set up five transactions to cover the deposit, and this necessitates that deposit rates be low. Depositors could probably get a better yield on a hedged currency deposit. The only private sector interest in this market comes from communication companies that have some trade contracts in SDRs.


The author thanks Jonathan Frimpong-Ansah, Anne-Marie Guide, Robert Heller, Peter Isard, and Michael Mussa for helpful comments and suggestions and Kalamogo Coulibaly for research assistance.


The argument extends in principle to lower inflation. Building on Irving Fisher’s (1913) tabular standard, recent literature (Fama, 1980; Hall, 1982; Greenfield and Yeager, 1983; McCallum, 1985; and Coats, 1989, inter alia) examines a monetary system built around a “hard” unit of account defined in terms of a commodity bundle selected to co-vary one to one with the general price level.


Among others, the latter is or has been used by the African Development Bank, the African Development Fund, the Arab Monetary Fund, the Asian Clearing Union, the Development Bank of the Great Lakes Countries, the Economic Community of West African States, the European Conference of Postal and Telecommunications Administrations, the Inter-American Development Bank, the International Center for Settlement of Investment Disputes, the International Development Association, the International Fund for Agricultural Development, the International Monetary Fund, the Islamic Development Bank, and the Nordic Investment Bank, as well as by a number of private organizations, notably by St. Cobain for internal transfer pricing.


Other offidal baskets—partly still used, partly defunct, or superseded—include the European unit of account, the Arab-currency related units, and the Asian monetary unit. Private baskets—now largely dormant—include the European composite unit, Barclay Bank’s B-unit, and Credit Lyonnais’s international financial unit.


We abstract from intermediate measures, such as the asymmetric adjustable basket (Polak, 1979, and Williamson, 1987).


A trivial counterexample is a reference currency pegged to one of the basket components.


Williamson (1987) provides a partial exception in discussing the criteria relevant to the choice of the number of basket members and their relative weights, though his conclusions are again based on subjective evaluation.


A range of alternative weighting schemes—GDP, trade, and population—are also of some appeal. The computational complexity of the task at hand precluded a robustness search over these alternative schemes, but as relative CDP and relative trade weights figure in the calculation of the SDR quota, the results are thus unlikely to be radically different.


The standard deviations of quarterly inflation rates are relatively low for two of the five countries, low to moderate for another two, and moderately high for the fifth.


Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, taly, Japan, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States.


It is ironic that although it might be desirable to allow for an overlap of the operating hours of the various national payment systems so as to increase the possibility of settling both legs of a foreign exchange transaction at the same time, overlapping operating hours would also eliminate any window of opportunity to close a bank while all rele vant currency markets are closed. To complicate matters further, some countries require that a bank be closed when the courts are open—and clearly this requires the closing of a bank in the middle of the business day, when there are likely to be a lot of unfinished payments transactions in progress. Other countries’ legislation involves a so-called zero hour rule that states that all transactions during the day of bankruptcy are to be rolled back to the status that existed at the beginning of the day Nothing in the payments business is easy.

  • 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