Abstract

In light of the principal recommendations from the preceding analysis (Table 5), this section briefly reviews the design of recent IMF-supported programs in countries with dollarized economies. The main findings from an analysis of a sample of IMF programs are that dollarization has had little impact on program design, and that programs have generally not included explicit measures to try to reduce dollarization—on the grounds that it is an inconvenience that will go away once its underlying causes have been corrected. Perhaps because of this, IMF programs have had little impact on the level of dollarization. At the same time, however, the performance of dollarized countries under IMF programs has not been noticeably worse than that of other countries.

In light of the principal recommendations from the preceding analysis (Table 5), this section briefly reviews the design of recent IMF-supported programs in countries with dollarized economies. The main findings from an analysis of a sample of IMF programs are that dollarization has had little impact on program design, and that programs have generally not included explicit measures to try to reduce dollarization—on the grounds that it is an inconvenience that will go away once its underlying causes have been corrected. Perhaps because of this, IMF programs have had little impact on the level of dollarization. At the same time, however, the performance of dollarized countries under IMF programs has not been noticeably worse than that of other countries.

Table 5.

Recommendations for the Design of IMF Programs in the Presence of Dollarization

article image

The review of past IMF programs in this section focuses on a sample of 12 countries (highlighted in Table 1) selected to provide diversity in regional distribution, stage of development, and initial economic problems. The sample includes (listed starting with the lowest level of FCD to broad money at end-1995, and ending with the highest) Estonia, Pakistan, Jordan, Poland, Hungary, Lao P.D.R., Argentina, Turkey, Cambodia, Peru, Uruguay, and Bolivia. The first five of these countries (where the ratio of FCD to broad money is below 30 percent) comprise moderately dollarized economies. The other seven countries are more highly dollarized economies.

Inconsistencies in programs with regard to the emphasis given to dollarization may, of course, arise for a variety of reasons. Every IMF-supported program is drawn up in the face of a large number of macroeconomic and structural problems, each of which may influence the design of the program, so that in any particular case other factors may have overridden considerations related to dollarization. Because there is no control for these other factors, the comparisons should be viewed as only descriptive.

General Program Design

There is—with the possible exception of Argentina—no evidence that dollarization has influenced the choice of a nominal exchange rate anchor in IMF-supported programs; other considerations (source of exogenous shocks, level of foreign exchange reserves, scope for fiscal adjustment) have instead been given more weight. In Argentina, there is some evidence that dollarization influenced the preference for an exchange rate anchor (Kiguel and Liviatan, 1994).

There is no compelling evidence that programs in dollarized economies have maintained higher levels of international reserves, to allow the central bank to extend its role as lender of last resort to cover banks’ dollar portfolios comfortably. There is little difference between average reserve levels of the seven highly dollarized economies and those of the five moderately dollarized economies (Table 6). Both are somewhat higher than the average of all the dollarized economies considered in this study (Table 7), but this larger sample also reveals little difference, either between highly and moderately dollarized economies or between dollarized and nondollarized economies.58

Table 6.

Some Features of Dollarized Economies

article image
Sources: IMF Staff Country Reports; and IFS.

In months of imports.

First observation for all variables is 1992.

First observations for FCD/broad money, velocity, and money multiplier are 1992; for inflation, 1993.

Table 7.

Gross Reserves in Months of Imports of Countries Reporting Information on FCD/Broad Money Ratios

article image
article image
Source: IMF, IFS. Among highly dollarized economies, data were unavailable for Azerbaijan, Belarus, Georgia, Sao Tome and Principe, and Tajikistan; among moderately dollarized economies, data were unavailable for Uzbekistan and Vietnam.

Classification based on observations for 1995; countries in bold are those selected for review.

Relatively few programs have included explicit measures to reduce dollarization. Such measures have included policies aimed at maintaining interest rate differentials in favor of LCD. Such a policy is especially evident in Poland, where it was the main consideration of interest rate policy. In most of the other countries in the sample, little concern is evident about the level of interest rates, even though in at least one—Turkey—the level of dollarization appears to display sensitivity—albeit weak—to the real rate of return on domestic currency deposits (Figure 8).

Figure 8.
Figure 8.

Turkey: FCD/M3 Ratio and Lira Deposit Rates

Sharply differentiated reserve requirements and discriminatory rates of remuneration on reserve requirements on FCD have been used in two of the countries in the sample—Bolivia and Peru—with the difference between reserve requirements on LCD and FCD reaching over 35 percentage points in Peru (Table 8). However, in other instances (Hungary, Pakistan, and Poland), reserve requirements have been used in a manner that encourages rather than discourages dollarization. In Poland, required reserves on LCD have been markedly higher than those on FCD, at least since 1990. In Hungary, whereas the reserve ratios on LCD and FCD were equal, the de facto treatment of FCD was more favorable because the basis for calculating mandatory reserves on FCD excluded until 1996 banks’ FCD at the central bank. In Pakistan, required reserves on FCD were effectively remunerated at higher levels than they were for LCD.

Table 8.

Selected Indicators of Measures Influencing Dollarization

article image
Sources: IMF Staff Country Reports; and IMF staff estimates.

Latest available information.

Argentina actually makes use of liquidity ratios rather than reserve requirements. Presently, these are 18 percent for deposits < 90 days; between 90 and 179 days, 13 percent; between 180 days and 365 days, 8 percent; and 0 percent for deposits > 365 days.

Reserve requirements for local currency site deposits indicated. For time HDC of 30–180 days, 14 percent and > 180 days, 12 percent. FRR for time deposit > 180 days drops to 15.5 percent.

RR and RR* have tended to be uniform for much of the 1990—95 period, and averaged just below 15 percent.

Reserve requirements for local currency demand deposits are higher at 20 percent. Before February 1996, rates were 17 percent on HCD and 2 percent on FCD.

In Lao P.D.R., a decree requiring all domestic transactions to be carried out in monetary assets denominated in local currency was issued. In Cambodia, a similar decree required all government transactions to be denominated in local currency. This has, however, had little effect on the prevalence of currency substitution in the two countries. In addition, the Lao program included measures to promote payments in local currency, with the explicit objective of reducing currency substitution. These have included improvements in payments, clearing, and settlement procedures; the promotion of the use of checks; and the issuance of large-denomination banknotes.

By contrast, some programs have on occasion accommodated measures that provided incentives for greater dollarization. In Argentina in 1992, for instance, in response to a short-lived run on the currency, the authorities allowed greater scope for dollarization by allowing foreign-currency-denominated checking accounts for domestic transactions, in order to enhance the credibility of the fixed exchange rate regime.

IMF conditionality has only rarely been attached to measures that can be expected to influence dollarization. Perhaps the strongest form of conditionality was in Poland, where failure to maintain a positive interest rate differential in favor of LCD triggered consultations with the IMF. In Lao P.D.R. in 1995, structural benchmarks were set on the removal of floors on the interest rates paid by banks. In Hungary, removal of ceilings on forint deposit interest rates was a structural benchmark under the 1991 IMF Extended Arrangement.

Design of Performance Criteria

For the most part, IMF programs with dollarized economies have adopted relatively standard program targets—performance criteria on the net international reserves (NIR) of the central bank and the net domestic assets (NDA) of either the central bank or the banking system, as well as in some cases indicative targets for reserve money, in addition to performance criteria for fiscal outcomes and external borrowing.59 The presence of FCD complicates the specification of these targets in a number of ways, which are summarized in the lower panel of Table 5 and are considered in more detail below. Table 9 summarizes how these complications were dealt with in the various countries in the sample.

Table 9.

Selection of Performance Criteria in Sample Countries

article image

SBA, Stand-By Arrangement; ESAF, Enhanced Structural Adjustment Facility; EFF, Extended Fund Facility.

M2X includes FCD. CC = currency in circulation, MI = CC + demand deposits in local currency.

BM = broad money; reserve money definition: CC = currency in circulation; RR required reserves; TR = total reserves; DD = demand deposits, excluding time deposits/CDs.

Performance criteria set on banking system reserves relative to liabilities—essentially prudential requirement.

Accounting rate applied just to items included in NFA; other foreign-currency-denominated components of balance sheet at actual rates.

Indicative target.

At the level of the banking system.

1996 SBA. SBAs in 1990 and 1992 included total reserves minus reserves of one large state-owned bank.

The first issue a financial program in a dollarized economy must address is whether the intermediate target should be a measure of money including or excluding FCD.60 The presumption would be that, where dollarization principally reflects asset substitution, the level of FCD is likely to be—barring wealth effects—unrelated to domestic money demand, so that a measure of broad money excluding FCD would constitute the appropriate intermediate target. Conversely, if dollarization largely reflects currency substitution, the level of FCD will be directly related to domestic demand, and a measure of money including FCD would be more appropriate. For the reasons discussed in Section IV, the extent of currency substitution is difficult to assess, and the relative stability of the different measures of broad money has to be determined on a case-by-case basis.

The programs reviewed did indeed tackle this issue on a case-by-case basis. Of the ten countries in the sample that used standard monitoring frameworks (that is, excluding those with currency boards), seven based their financial programs on a measure of broad money including FCD, implicitly assigning some role to currency substitution. In three of these countries (Cambodia, Lao P.D.R., and Turkey) there is ample anecdotal evidence of currency substitution. However, in the other four countries there would seem to be a lesser degree of currency substitution, but an intermediate target including FCDs was still chosen.61 The three remaining countries (Bolivia, Peru, and Uruguay) used a narrower measure of money excluding FCD, owing to a combination of the absence of a stable relationship between broader measures of money and the ultimate policy targets, the perceived inability of the authorities to control such broader aggregates, and an assessment that dollarization was principally caused by asset substitution.62

If the intermediate target is a broad aggregate, then a decision must be made whether to apply the operational target at this level—that is, at the level of the banking system—or at the level of the central bank. This decision relates largely to the issue of control and the availability of indirect monetary instruments, as would be the case in a nondollarized economy. Thus, in countries with more market-oriented financial systems (most of the countries in this sample), targets operated at the level of the central bank. In the three countries with financial systems more subject to state control, however, targets operated at the level of the banking system. In Cambodia and Lao P.D.R., credit limits were set at the level of the banking system largely because of the rudimentary stage of development of the banking sector. In Poland, however, the targeting of NDA of the banking system mainly reflected concern about continuing instability in the money multiplier, perhaps not unrelated to dollarization.

For countries setting monetary targets at the level of the central bank, the next question concerns the choice of monetary liabilities that will be subject to targeting. This question relates, in part, to the stability of the money multiplier (the selected measure of reserve money—including or excluding reserves against FCD (FCR)63—relative to the chosen intermediate target).64 Once again, this is largely an empirical question. Nonetheless, the usual presumption would be that, if FCD are included as part of the intermediate target, then banks’ FCR should be included in the central bank’s operating target. Reflecting this, all of the countries in the sample that based their financial programs on a measure of broad money including FCD, except Jordan, established operating targets on a measure of reserve money that included FCR. By contrast, those countries that based their financial programs on a measure of broad money excluding FCD established operating targets on a measure of reserve money that excluded FCR.

Reflecting the decision to treat FCR as a component of reserve money—as a domestic liability—10 of the 12 countries in the sample excluded FCR as relevant liabilities in the definition of NIR, although in one of these (Uruguay) this effect is undone by an adjuster. From a purely statistical perspective, FCR are liabilities to residents and should be treated as such in the monetary and balance of payments statistics. However, the operational concept of NIR should be defined differently from the purely statistical concept of net foreign assets (NFA) to provide a more meaningful indicator of the adequacy of a country’s reserves.65 As noted above, treating FCR as domestic rather than foreign liabilities and thus excluding them from the relevant measure of NIR has certain drawbacks. In particular, for the central bank, reserves garnered from FCR are not as secure as reserves gained from intervention or from medium- and long-term borrowing.66 The two countries that included FCR as a relevant liability for NIR were Jordan, on account of the possibility of the introduction of a new currency in the West Bank leading to substantial withdrawals of FCD and thereby of FCR, and Peru, where it was recognized that FCD were subject to the volatility of capital flows and could be withdrawn at any time. Similarly, the adjuster was used in the Uruguay program to ensure that the NIR target was not met simply through the attraction of capital inflows.

Finally, it is necessary to decide whether IMF programs in dollarized economies should value the foreign exchange components of balance sheets at pre-determined (accounting) or prevailing market exchange rates. This choice is relevant for money targets if there are FCD (for broad money targets) or FCR (for base money targets). If the balance sheet is valued at prevailing market exchange rates, exchange rate depreciations will generate (measured) money expansions if there are FCD or FCR.67 Valuation at predetermined exchange rates, by contrast, will render money unaffected by exchange rate movements.

The apparent generation of monetary expansions in the context of depreciations could be seen as desirable, in that it would require monetary tightening to ensure compliance with targets. It is not clear, however, that the degree of tightening required would be necessarily appropriate. Valuation at prevailing market rates also has the disadvantage that exchange rate volatility close to test dates could make it extremely difficult for the authorities to comply with program targets, despite their best efforts. This would undermine the principle that members should have reasonable assurances that IMF resources will be available if agreed policies are implemented. For these reasons, combined with the availability of reviews for modifying program targets, performance criteria have generally been defined using predetermined exchange rates. In the sample, excluding those countries operating fixed exchange rate regimes, only one country (Hungary) used prevailing market exchange rates in valuing the balance sheet.

Dollarization and Program Performance

The evidence on the effectiveness of inflation control in dollarized economies is mixed. While it seems that highly dollarized economies tend to be associated with higher rates of inflation than other economies (Table 10), the data in Table 6 (for the smaller sample) show little difference between highly and moderately dollarized economies. Moreover, some of the former had fairly low rates of inflation (10 percent or below). Money demand appears to have been more volatile in the highly dollarized economies in the sample, since the coefficient of variation of velocity is markedly higher than in the moderately and nondollarized samples (table 11).68 However, the deviation of inflation from program targets in the first program year, albeit positive, has been only marginally higher than the typical deviation found in the 1994 and 1997 reviews of IMF-supported programs (excluding those countries known to be heavily dollarized; Table 12). Finally, there is little evidence that a higher degree of dollarization implies greater variability in the money multiplier (Table 13), at least when FCD are included in money supply and FCR in reserve money, as is the case for most of the countries in the dollarized sample.

Table 10.

Inflation Country Reporting Information on FCD/Broad Money Ratios

article image
article image
Source: IMF, World Economic Outlook.

Classification based on observations for 1995; countries in bold are those selected for review.

Table 11.

Velocity in Countries Reporting Information on FCD/Broad Money Ratios

article image
article image
Sources: IMF, World Economic Outlook and IFS. Among highly dollarized economies, data were unavailable for Azerbaijan, Belarus, Croatia, Georgia, Mozambique, and Tajikistan; among moderately dollarized economies, data were unavailable for Albania, Guinea, Macedonia F.Y.R., Moldova, Uzbekistan, Vietnam, and Yemen.
Table 12.

Inflation Target Versus Outcome for the First Program Year in IMF Arrangements

article image
Source: IMF Staff Country Reports.

Outturnless target.

Excluding countries where the ratio of FCD/broad money is known to be > 30 percent. For the 4 ESAF countries in the sample (Bolivia, Cambodia, Lao P.D.R., and Pakistan), the average and median deviations were 4.7 and 4.8 percent, respectively. For the 8 SBA/EFF countries in the sample (Argentina, Estonia, Hungary, Jordan, Peru, Poland, Turkey, and Uruguay), the average and median deviations were 30.6 and 6.0, respectively.

Table 13.

Money Multiplier1

article image
Source: IMF, IFS. Among highly dollarized economies, data were unavailable for Belarus, Georgia, Latvia, Mozambique, Sāo Tomé and Principe, and Tajikistan; among moderately dollarized economies, data were unavailable for Albania, Macedonia F.Y.R., Trinidad and Tobago, Uganda, Ukraine, Uzbekistan, Vietnam, and Zambia.

Broad money/reserve money.