The foreign currency is usually, but not always, the U.S. dollar. The term “dollarization” serves as a shorthand in this paper for the use of any foreign currency.

Of the 99 countries that have had IMF arrangements since 1986, 52 reported data on foreign currency deposits (FCD) to the IMF. In those countries reporting data, the median level of FCD over broad money in 1995 was 21.8 percent.

Several of the issues discussed in this paper (for example, foreign currency risks) arise also in other foreign exchange operations, even if the economy is not dollarized in the sense discussed above.

Other special circumstances may lead to dollarization. For instance, European countries that will not initially participate in the euro arrangements but that trade heavily with euro members are likely to see part of the demand for money shift to euros. In addition, very small economies in which tourism from abroad is an important sector are also likely to be dollarized.

This distinction is standard in the literature. See the useful surveys by Calvo and Végh (1996) and Giovannini and Turtelboom (1994). McKinnon (1996) terms the two motives direct currency substitution and indirect currency substitution.

While the discussion in this paper focuses chiefly on the assets of the public, dollarization of deposits has typically been accompanied by dollarization of lending.

Previous empirical studies on dollarization conducted in the IMF include those by Agénor and Khan (1996), Clements and Schwartz (1993), El-Erian (1988), Mueller (1994), Sahay and Végh (1996), Savastano (1992), and Savastano (1996).

Dollarization as asset substitution is also associated with other nonmonetary assets, such as foreign-currency-denominated stocks and bonds and borrowing in foreign currency. But these assets are more removed from monetary developments and are considered more tangentially in this study.

In Africa, (measured) dollarization is less widespread than in other areas because FCD are not allowed or are severely restricted in most cases. However, there are indications of extensive dollarization in banknotes.

Kamin and Ericsson (1993) estimated a money demand function with “ratchet” effects for Argentina for a period that includes a hyperinflation episode.

Calvo and Végh (1992, 1996) and Savastano (1992, 1996) also mention capital flows as a driving factor for dollarization.

While it is true that, according to available data, the decline in CBD is less than the increase in FCD in absolute terms, it should be noted that the CBD figures are likely to underestimate the actual stock of CBD. For legal and tax-related reasons, residents of one country may wish to transfer assets to a company based in a third country, which appears to be the holder of the deposit. (In fact, the three top countries of residence of nonbank depositors in the Western Hemisphere are Cayman Islands, Panama, and the Netherlands Antilles.) This means that the actual decline in CBD was probably larger than reported.

These data have been kindly made available by the U.S. Treasury Department (Financial Crime Enforcement Network); this is the first time this information has been released with the exception of the data for Argentina used by Kamin and Ericsson (1993).

For a discussion of this issue for the case of Argentina, see Garcia-Herrero (1997).

The issuance by the Mexican government of the dollar-indexed tesobonos in 1994 signaled a commitment to a fixed exchange rate and raised the risks involved in a potential reversal of capital flows. On balance, it does not appear that tesobonos had a positive effect on credibility. As suggested by Drazen and Masson (1994), one has to distinguish between credibility of policy makers and credibility of policies, and the latter was very low in this case.

The exchange rate instrument could still be effective to induce changes in relative prices if domestic prices are not rigid in dollar terms; that is, if the domestic currency is the predominant unit of account in the economy.

Moreover, unlike with loans in local currency, the burden of foreign currency loans (FCL) could not be reduced through devaluation.

As noted earlier, the problems described in this paragraph would also apply to banking systems exposed to foreign exchange risk because they are heavily engaged in foreign exchange transactions. However, such exposure is likely to be more pervasive in countries in which there is extensive currency or asset substitution.

The magnitude of the seignorage revenues obtained by the United States from the holding of its currency in foreign countries is relatively small. It is estimated that $200 to $250 billion circulated outside the United States in 1995 (Porter and Judson, 1996). If there was no such foreign demand for the U.S. currency, and the United States had to issue an equivalent amount of short-term treasury bills, the annual interest cost would be some $10 billion to $15 billion, less than ¼ of 1 percent of GDP.

For example, in several transition economies, a decline in the acceptability of $100 bills (owing to fears that they would be confiscated during the conversion to the new notes) appears to have briefly raised the demand for domestic currency.

This conclusion applies not only to exchange rate systems in which the rate is formally pegged but also to exchange-rate-based stabilizations more generally; that is, situations in which macroeconomic policies (monetary policy in particular) are geared to maintaining a stable exchange rate continuously, even though the exchange arrangement is flexible. In Croatia, the anchor was not an explicit, formal one, but ex post this episode can be considered as an instance of exchange-rate-based stabilization.

See Savastano (1996) for a useful review of this literature.

See, for example, Miles (1978) and Bordo and Choudri (1982).

A number of studies, including Estrella and Mishkin (1996), Friedman and Kuttner (1996), and Feldstein and Stock (1994), have used this type of technique to approach the analogous question of which monetary aggregate is the most appropriate focus of targeting.

The finding for Peru may be affected by institutional changes that took place during the sample period. The Central Bank of Peru has found a narrow, purely domestic currency aggregate to be the most useful intermediate target.

It may also be difficult to implement an intermediate target that includes foreign assets. The lack of reliable data on DCC would make it difficult to do so.

For similar reasons (a high degree of indexation), the Central Bank of Chile’s main monetary instrument is an indexed instrument.

For example, preliminary evidence suggests that the short-run offset coefficient for Uruguay is higher than that for Bolivia, which implies that foreign exchange monetary intervention would be a more effective instrument in Bolivia than in Uruguay. Offset coefficients measure the extent to which changes in the net foreign assets of the central bank resulting from open market operations are offset by market-originated opposite changes in those assets.

For a discussion of these issues see Monetary and Exchange Affairs Department (1995).

Thus, in Peru, central bank intervention in the dollar money market is unnecessary, owing to the very high reserve requirements on FCD, and wide limits on banks’ open foreign exchange positions that facilitate arbitrage between the domestic currency and foreign currency financial markets.

The use of dollar sight deposits expanded rapidly in Bolivia after these accounts were authorized in 1988, notwithstanding that interbank clearing and settlement of checks drawn on these accounts was, until 1991, effected on the basis of private arrangements outside the central bank (see Figure 3).

There is some evidence that Argentine branches of foreign banks cut their exposure in the interbank market in the aftermath of the Mexican crisis.

This issue is illustrated by the case of the Philippines, where payments dollarization has been encouraged by the fact that the foreign currency payments system—where dollar transactions are sent on an electronic payments system operated by Citibank in Manila—technically dominates the peso payments system—which depends on the physical delivery of checks.

For example, although primarily motivated by the restrictions imposed by the currency board arrangement rather than by dollarization, the capital adequacy ratio was raised to 11.5 percent in Argentina. Moreover, since Argentina applies more stringent criteria for weighting risky assets, that ratio is equivalent to about 16 percent under the Basle guidelines. In Bolivia, it will gradually be raised to 10 percent, reflecting concerns about dollarization as well as broader concerns about the soundness of the banking system. These compare with a Basle guideline of 8 percent.

These requirements need not exceed those used for nondollarized economies, since foreign exchange exposure depends on a bank’s net, rather than gross, position. Moreover, the risk on the bank’s gross foreign currency position accrues from credit risk, rather than foreign exchange risk. However, to avoid circumvention via derivatives, foreign exchange open positions may need to be assessed on the basis of risk accounting principles (see Garber, 1996).

However, other considerations—e.g., risk of capital flight—limit the scope for unremunerated reserve requirements on FCD.

When liquidity requirements are fulfilled with domestic assets—rather than foreign liquid assets—the central bank needs to have sufficient foreign reserves to maintain the liquidity of these assets in the case of a systemic crisis. Liquidity requirements can take the form of fixed coefficients that can be differentiated by maturity of deposits to reflect differences in volatility (as in Argentina) or of limits on maturity mismatches. While more difficult to monitor, the latter also limit interest rate risk. As countries strengthen their supervisory capacity and banks improve their internal risk controls, liquidity requirements need to be substituted by more flexible and comprehensive methods to assess risk. See Gulde, Nascimento, and Zamalloa (1997).

Mexico’s experience with tesobonos in 1994 suggests that government bonds denominated in foreign currency may be subject to speculative attacks, particularly when they have short maturities and inadequate foreign reserve backing.

Thus, in the wake of the Mexican crisis in 1994, a number of Asian central banks established a protocol for mutual liquidity support. Argentina established an arrangement that facilitates emergency borrowing from foreign commercial banks by local commercial banks. Although these examples are not specific to dollarized economies, similar arrangements can be tailored to them. See Baliño and others (1997).

The “mexdollar” system, which was widely used in Mexico until the forced conversion of deposits in August 1982, and the “Patzam” in Israel provide other illustrations.

Thus, in Bolivia, their demand is mostly limited to public enterprises, which are required by law to hold their deposits in local currency. Similarly, in Nicaragua, dollar-indexed deposits pay a premium over dollar deposits.

In Brazil, both price indexation and interest rate indexation were broadly used. In particular, the indexation of deposits to the overnight interest rate protected the purchasing power of LCD throughout the turbulent inflationary period of the 1980s. However, the need to limit risks incurred by financial intermediaries in the overnight market eventually undermined the monetary anchor, since it induced pervasive monetary accommodation and led to the de facto creation of indexed money. See Garcia (1996).

Price indexation has been facilitated by the introduction in 1967 of a unit of account, the UF, that is published by the central bank daily on the basis of the consumer price index.

Also, since “dedollarization” measures are very difficult to apply to cash, part of the flow out of FCD can move to DCC.

Other countries where financial liberalization may have contributed to limit dollarization include Armenia, Estonia, Lithuania, and Poland (see Sahay and Végh,1996).

A too tight monetary stance, if sustained, can have other adverse macroeconomic and financial implications, including an increase in the burden of public domestic debt.

If the demand for dollar credits is elastic, an increase in unremunerated reserve requirements on FCD should have similar effects. When the demand for dollar credit is inelastic, an increase in unremunerated reserve requirements on FCD would raise dollar deposit rates and induce a capital inflow. Thus, dollarization, measured in terms of deposits, could rise.

Reserve requirements were reduced from a uniform ratio of 50 percent to 30 percent and 20 percent, on demand and term deposits, respectively. While the share of FCL in total loans increased from 2.2 percent at end-1993 to 11 percent at end-April 1996, the share of FCD in total deposits increased from 3.8 percent at end-1993 to only 5.3 percent at end-April 1996.

High reserve requirements on FCD were intended to limit short-term capital inflows and the associated prudential risk of rapid credit growth. In principle, this should also have encouraged the demand for LCD, but so far, probably owing to the low interest elasticity of LCD, growth in FCD has significantly outpaced growth in LCD. Between 1990 and 1996, end-of-year FCD balances increased from the equivalent of almost 3 percent of GDP to close to 14 percent of GDP, while LCD only grew from almost 2 percent of GDP to less than 5 percent over the same period.

In Uruguay, which is an important regional offshore center, the authorities have also expressed the view that attempts to lower dollarization through high reserve requirements and similar measures would drive dollar deposits offshore.

After FCD were later authorized again, Bolivian banks were subject to several incipient runs. The spread over LIBOR (London interbank offered rate) of dollar deposit rates, which reached over 900 basis points in 1987, was still over 400 basis points in December 1996.

In Poland, for instance, although accounts in domestic currency are not convertible into foreign currency, they may be converted by withdrawing zlotys, converting them to dollars and redepositing them.

For example, in the Lao People’s Democratic Republic, the lack of high-denomination notes contributed to promoting the use of U.S. dollars and the Thai baht. Also, introduction of higher-denominated riel notes in March 1995 prompted a significant increase in the demand for domestic currency banknotes.

Many countries already restrict or do not allow the counting of vault cash in fulfillment of reserve requirements because of the difficulty in monitoring those balances and because those funds are not immediately available for interbank settlement purposes (see Monetary and Exchange Affairs Department, 1995). However, this could be too costly for banks if their required reserves against FCD must be held with an overseas bank.

The spread, of course, would have to be limited so as not to give rise to a multiple-currency practice.

Measures to restrict the use of foreign cash were already in place.

However, to limit disintermediation into cash or deposits abroad, such taxes would need to be small. In Peru, in the late 1980s, a 2 percent transaction tax on both domestic and foreign currency debits encouraged disintermediation because many households and small enterprises resorted to payments in cash.

IMF staff estimates of ratios of FCD to broad money for 48 countries were not available, and these are assumed to proxy for nondollarized economies.

Countries with currency boards (Argentina and Estonia in the sample) and countries that use another country’s currency (such as Panama, which is outside of the sample) have employed different frameworks.

Of course, as highlighted in Section IV, even measures of money including FCD fail to capture foreign currency cash (especially important under currency substitution) and offshore FCD (especially important under asset substitution).

The inflation experience of these four countries does not suggest that major problems were encountered by targeting the broader monetary aggregate.

These three countries have been able to bring down inflation through controlling the domestic component of money despite prima facie evidence of at least some currency substitution. The exclusion of FCD for Peru is consistent with the central bank’s view that domestic aggregates are more appropriate intermediate targets. As noted in Section IV, however, Berg, Borensztein, and Chen (1997) found aggregates including FCD to be better predictors of inflation.

Following Kochhar (1996), a distinction is not made between required and excess reserves, since excess reserves can provide a useful warning of an increase in liquidity and a buildup of potential pressures on inflation, NIR, and the exchange rate. Thus, FCR refers to the sum of required reserves against FCD plus any excess reserves on FCD.

The operating target is important not only where explicit indicative targets are set on reserve money, but more generally in the calculation of NDA ceilings at the level of the central bank, since these are derived from a projection of reserve money (together with the NIR target).

This treatment of required reserves is not, however, intended to impugn the virtue of a system of required reserves for FCD, which remains a valuable prudential tool.

An equivalent definition would be to include the foreign exchange assets obtained by the central bank as the counterpart to FCR within “other foreign assets, net” and therefore outside of the NIR concept typically employed for program monitoring purposes. NFA would thereby be unaffected, since they are the sum of NIR and other “foreign assets, net.”

Valuation is also relevant for NDA targets if the central bank—or banking system, if this is the level of the operational target—has net exposure in foreign exchange. With market price valuation, exchange rate depreciations will generate NDA expansion (via losses) if there are net foreign exchange liabilities. If there are NFA, on the other hand, it will generate NDA contractions (via profits).

Because of data limitations, the measured velocity includes FCD. But, given that it does not include CBD and DCC, dollarization may be a reason for its higher variability.