Macroeconomic and Structural Policies in Fund-Supported Programs - Review of Experience

A Fund-supported program is a package of envisaged policies which, combined with approved financing, is expected to achieve certain economic objectives such as fostering macroeconomic stability and orderly external adjustment, promoting growth and poverty reduction, and reducing vulnerability to future balance of payments problems or financial crises. This paper reviews experience with specific macroeconomic and structural policies intended to achieve these objectives.

Abstract

A Fund-supported program is a package of envisaged policies which, combined with approved financing, is expected to achieve certain economic objectives such as fostering macroeconomic stability and orderly external adjustment, promoting growth and poverty reduction, and reducing vulnerability to future balance of payments problems or financial crises. This paper reviews experience with specific macroeconomic and structural policies intended to achieve these objectives.

I. Introduction

1. A Fund-supported program is a package of envisaged policies which, combined with approved financing, is expected to achieve certain economic objectives such as fostering macroeconomic stability and orderly external adjustment, promoting growth and poverty reduction, and reducing vulnerability to future balance of payments problems or financial crises. This paper reviews experience with specific macroeconomic and structural policies intended to achieve these objectives.1

2. In designing their economic program, national authorities have at their disposal a number of instruments, including the exchange rate regime, the monetary stance, fiscal policies, and structural measures. Some of the considerations behind the setting of macroeconomic and structural policies are discussed in Policy Formulation, Analytical Frameworks, and the Design of Fund-supported Programs. This paper turns to experience, seeking to answer three broad questions for each policy instrument: Was use of the instrument geared towards achieving program objectives? Were the intended polices carried out? And what was the outcome?

3. Before turning to a summary of the main findings, four points are worth noting. First, by its very nature, cross-country analysis requires making generalizations—there are always exceptions, however, since individual programs must be tailored to the specific circumstances facing the member. Second, for expositional ease—and to complement the analysis of outcomes in Fund-supported Programs: Objectives and Outcomes—the discussion in this paper is organized around the role of each individual policy instrument.2 But these various policy elements are also intended to work together, and an important consideration in program design is the complementarity of instruments and their appropriate assignment to targets. Third, policy choices and their implementation reflect deep social and institutional determinants of macroeconomic stability which are not modeled here. More generally, caution is required in interpreting the empirical findings owing to possible omitted variable bias and difficulties in establishing counterfactuals.3 Fourth, during the period under review—1995–2000—Fund-supported programs in low-income countries underwent important changes with the shift in 1999/2000 from the Enhanced Structural Adjustment Facility (ESAF) to the Poverty Reduction and Growth Facility (PRGF); most of the experience of low-income countries reported in this paper pertains to ESAF-supported programs.

4. With these points in mind, the main findings based on aggregate, cross-country analysis are as follows. First, up-front devaluations or shifts in the exchange rate regime are the exceptions rather than the rule under Fund-supported programs—in less than 20 percent of all programs was the regime changed in the year the program was approved. Most regime shifts involved pegging the exchange rate in transition economies (as they embarked on disinflation programs) or moving to more flexible regimes in non-transition economies (as pegs were abandoned in the face of balance of payments difficulties). Among programs that explicitly targeted disinflation, GRA-supported programs typically used the exchange rate as a nominal anchor, while PRGF-supported programs tended to use money-based stabilizations. But rates of success did not differ markedly, making it difficult to generalize about which strategy is preferable. Rather, what appears to have been of greater importance in explaining success are the supporting policies—specifically, whether the targeted fiscal adjustment was achieved.

5. It is also worth examining whether external adjustment came at a lower output cost in countries with more flexible regimes (because of expenditure switching) and whether countries with pegged regimes prior to the program subsequently underwent greater external adjustment as balance sheet mismatches—built up because the guarantee implicit in the peg had encouraged excessive foreign currency exposure in the pre-crisis period—unwound. While there is evidence that countries with more flexible regimes achieve external adjustment at lower output cost, there is little empirical relationship between pegged regimes and the subsequent adjustment of the current account being greater than programmed.

6. Second, programs usually target at least some tightening of the monetary stance—in order to lower inflation, promote orderly external adjustment, and, especially in capital account crises, to help stem capital outflows. Empirically, the monetary stance is tightened, though usually by not as much as is programmed, leading to higher inflation than projected. Importantly, policies set in the context of Fund-supported programs appear to enjoy greater credibility, leading to higher money demand, and thus lower inflation for a given growth rate of broad money. The empirical evidence does not support the assertion that the monetary stance was set excessively tight in Fund-supported programs leading to lower output growth.4

7. Third, Fund-supported programs also target at least some fiscal consolidation to promote external adjustment, underpin macroeconomic stabilization, or put the public finances and debt dynamics on a more sustainable footing. In the event, however, there are typically large slippages in the fiscal adjustment targeted for the first program year, which widen in the following year, mainly because of primary (and, to a lesser degree, interest) expenditure overruns in GRA-supported programs and a combination of primary expenditure overruns and revenue shortfalls in PRGF-supported programs.5

8. The failure to maintain the programmed fiscal consolidation cannot be explained by the planned current account adjustment having been achieved—fiscal consolidation was not sustained even in cases where the external adjustment fell short of expectations. Fiscal slippages undermine disinflation efforts and result in significantly higher public debt ratios than programmed. (Below-the-line operations are, however, the most important source of public debt projection errors.) Fiscal adjustment does contribute to external adjustment—but cannot explain instances in which the country undergoes substantially more external adjustment than anticipated. Empirical evidence does not indicate that fiscal policies in Fund-supported programs have had negative consequences for growth.

9. Fourth, structural measures in Fund-supported programs can be classified according to their primary objectives—bolstering the management of aggregate demand, enhancing the flexibility of the economy and raising efficiency (both of which serve to strengthen a country’s growth prospects), and reducing vulnerabilities to future crises. Classifying structural measures into these three categories shows some alignment between structural measures and the broad objectives of economic programs. While it is difficult to establish the impact of individual structural reforms, the evidence suggests that fiscal structural measures have been useful in underpinning fiscal adjustment and that there is a positive correlation between growth-related structural measures in Fund-supported programs and medium-term growth performance.

10. The paper follows the structure of this summary. Concluding remarks are presented in Section VI.

II. The Choice of Exchange Rate Regime

11. Given the primacy of external adjustment and macroeconomic stability in Fund-supported programs, a natural starting point is the exchange rate regime. In particular, a flexible exchange rate can allow for more of the improvement in the current account balance to take place through expenditure switching rather than by monetary and fiscal restraint alone—though, in some circumstances, this can also be achieved through a discrete devaluation under an existing peg.6 Conversely, when disinflation is a primary objective, the use of the exchange rate as a nominal anchor can help induce policy discipline, engender confidence in the currency, and bring down inflationary expectations and real interest rates. The use of such “exchange-rate based stabilizations” is not uncontroversial, however (Box 1).

Exchange-Rate Based Stabilizations

Exchange rate based stabilizations (ERBS) are often advocated for countries starting from high and chronic inflation because the nominal exchange rate provides a highly visible anchor for private sector expectations. In particular, in countries with high dollarization and a high pass-through from the exchange rate to prices, the exchange rate can stabilize and coordinate expectations quickly, and may promote policy discipline.1 An exchange rate anchor could also be attractive to countries with high real interest rates, as an ERBS might reduce them more rapidly than a money-based-stabilization (MBS). Another benefit could be the relative ease of conducting monetary policy, in contrast to MBS, where the appropriate rate of money growth must be determined, often in situations of highly-unstable money demand. The transparency of ERBS may also enhance the credibility of the monetary authorities, thus reducing the costs of disinflation.

However, the debate on ERBS is still open. Some authors maintain that the costs of disinflations carried out with an exchange rate anchor are merely postponed. They point to some empirical regularities observed in ERBS (the so-called ERBS syndrome), namely a substantial real exchange rate appreciation and related deterioration in the external accounts, which often leads to a balance of payments crisis, and a boom-bust cycle in GDP, consumption, and investment.2 ERBS have been linked to financial crises as well.3 The ability of a predetermined exchange rate regime to impose discipline on other policies, notably fiscal policy, is also disputed.4 In addition, overvaluation under a pegged exchange rate regime may mask temporarily the extent of public indebtedness.

But other authors challenge empirical regularities that characterize the ERBS syndrome. For example, some authors do not find evidence that output dynamics differ based on the anchor used in the stabilization. Others find expansionary effects on output of ERBS from high inflation.5 Similarly, the claim that ERBS have a higher percentage of failures has been questioned (see Tables 1 and 2).6 In fact, these differences in findings may reflect the small samples used in some studies: for example, Calvo and Vegh (1999) examined 5 episodes of MBS compared to 12 ERBS. However, in studies where a large number of episodes are studied (typically identified by rules), the evidence of the ERBS syndrome is much weaker, if extant at all.7

Table 1.

Successful Stabilization Episodes

article image
Source: Hamann, 2001.

Criterion 1 defines success as inflation at t+2 and t+3 no higher than during the stabilization year. Criterion. 2 defines success as inflation at t+2 and t+3 no higher than 3/4 of the inflation rate prevailing the year before stabilization.

Table 2.

Disinflation Attempts under Alternative Exchange Rate Regimes

article image
Source: Ghosh, Gulde, and Wolf, 2003.

Schadler et al. (1995) in the 1994 Conditionality Review studied 16 countries (out of a total sample of 36) that adopted a monetary anchor—defined as either a money supply rule (1 country) or a predetermined exchange rate path (15 countries). They concluded that, while there is no substitute for tight financial policies and wage restraint, exchange rate anchors appeared to have sped up disinflation and helped keep inflation low. At the same time, they pointed to significant costs in terms of competitiveness, export growth, and possibly short-term output growth associated with the disinflation gains. Finally, they viewed the adoption of some nominal anchor as indispensable in reducing high or intermediate inflation, but underscored the key role of supporting policies.

Finally, when the exchange rate regime chosen for disinflation differs from the regime considered more suitable for the country from a longer-run perspective, issues of exit arise. For example, as discussed in Lessons from the Crisis in Argentina (SM/03/345), the currency board arrangement adopted by Argentina in 19991 was instrumental in bringing down inflation after decades of high inflation, but given extensive dollarization of the economy and turbulence in international capital markets, it was difficult to find an opportunity to exit the regime gracefully even as it became apparent that a lack of competitiveness was impeding growth, and that fiscal policy necessary to sustain the peg was not forthcoming. The Fund-supported program in Turkey (1999) pre-announced an explicit exit strategy (and timetable) for exiting the quasi-currency board arrangement adopted at the outset of the stabilization program. The pre-announcement does not appear to have undermined credibility of the regime, though in the event it collapsed for other reasons prior to the planned exit date.

1. See Calvo and Vegh (1999) and Hamann (2001) for surveys of the literature on ERBS.2. See for example Kiguel and Liviatan (1992), Vegh (1992), and Calvo and Vegh (1994).3. See Sobolev (2000) “Exchange Rate-Based Stabilization: A Model of Financial Fragility”, IMF Working Paper WP/00/122.4. Hamann (2001) does not find evidence of increased fiscal discipline of ERBS.5. Fischer, Sahay, and Vegh, 2002.6. See Easterly (1996), Ghosh, Gulde, Wolf (2002), Hamann (2001), Hamann and Prati (2002), and Santaella and Vela (1996).7. Fischer, Sahay, and Vegh, (2002) “Modern Hyper-and High Inflations,” IMF Working Paper WP02/97; and Hamann (2001).

12. This Chapter considers the role of the exchange rate regime in Fund-supported programs—Chapter III, below, takes up the related but distinct issue of the monetary stance. Section A sets the stage by considering the extent to which the exchange rate regime has been used as an explicit tool for achieving program objectives. Section B turns to outcomes, examining three questions: Did use of the exchange rate as a nominal anchor assist in disinflation? Did more flexible exchange rate regimes help achieve external adjustment at a lower cost in terms of output? And did countries with pegged regimes subsequently undergo greater external adjustment as balance sheet mismatches unwound?

A. Exchange Rate Regimes in Fund-Supported Programs

13. Table 1 reports the distribution of exchange rate regimes in the year prior to (year t-1), and the year of (year t), the approval of the arrangement. PRGF-supported members are split almost equally between pegged and flexible exchange rate regimes, whereas for GRA-supported members a larger proportion (60 percent) had pegged exchange rates.7 Transitions in the year that the Fund arrangement was approved occur in less than 20 percent of cases.8 When regime changes occur, these were frequently towards greater flexibility in non-transition GRA-supported countries (as pegs were abandoned9) or towards less flexible regimes in transition economies (to assist disinflation efforts).

Table 1.

Macroeconomic Performance Under Alternative Exchange Rate Regimes

article image
Sources: International Monetary Fund, AREAER and WEO; and IMF staff calculations.

Exchange rate regimes as classified by the Fund AREAER. "Pegged regimes" include exchange arrangements with no separate legal tender, currency boards, other conventional pegs, pegs with horizontal bands, crawling pegs, and crawling bands; "flexible regimes" include managed and independently floating regimes.

Classified by regime prevailing in year of program approval (t).

To reduce the influence of outliers, the inflation rate is mapped into the interval (-100, 100) percent.

14. To examine the determinants of regime choice in Fund-supported programs, Table 2 reports the results of estimating an ordered probit, where a higher score on the regime index indicates a more flexible regime. This analysis shows a great deal of persistence in regime choice—that is, consistent with the observation above, the exchange rate regime is seldom changed as part of a Fund-supported program. The exchange rate regime (whether or not it is changed at the time of program approval) can be explained by various explanatory variables for non-transition country programs. Specifically, in GRA-supported non-transition programs, a pegged (or less flexible) regime is more likely the larger the programmed decline in inflation. Moreover, though the programmed change in the current account balance is not statistically significant, a flexible regime is more likely the greater the estimated overvaluation of the real exchange rate. Other significant determinants are foreign exchange reserves (a higher level of reserves makes a peg more likely) and the output gap (a smaller gap makes a peg more likely). Overall, the probit explains 90 percent of the observations correctly. By contrast, in a similar analysis for PRGF-supported non-transition countries, only the lagged regime variable has the correct sign and is statistically significant—suggesting greater inertia in the choice of exchange rate regime for these countries.10 For transition economies, the fit of the equation is much worse, and only the lagged regime and the estimated degree of overvaluation are statistically significant.

Table 2.

Choice of Exchange Rate Regime: Results of Ordered Probit 1/2/

article image
Sources: International Monetary Fund, AREAER, IFS, INS, MONA, WEO; and IMF staff estimates.

A higher score indicates more flexible regimes on the eight-category scale of the IMF AREAER. The categories are: 1) no separate legal tender; 2) currency boards; 3) other conventional pegs; 4) pegged arrangements with horizontal bands; 5) crawling pegs; 6) crawling bands; 7) managed floats; and 8) independent floats.

Significant at:

1 percent

5 percent

10 percent levels.

Positive value indicates output below trend. Trend GDP is obtained from a Hodrick-Prescott filter.

Positive value indicates overvaluation. The trend real exchange rate is obtained from a Hodrick-Prescott filter.

B. Experience

15. A number of findings can be highlighted in terms of macroeconomic performance and exchange rate regime, though of course the regime choice may itself be endogenous to macroeconomic performance.11 Inflation for the full sample is lower under pegged exchange rates and inflation declines rapidly over the program period under both pegged and flexible regimes, though remaining higher in countries with flexible regimes (Table 1). The evidence on growth is less clear. Pegged regimes experienced modest variations in real growth while countries with flexible regimes saw an acceleration in real growth. For countries that switched regimes, pegging the exchange rate is associated with better inflation performance, though the sample of such countries is small and the results mostly driven by the experience of the transition economies. The growth experience of countries switching regimes is mixed: transition economies saw a sharp acceleration in growth under their exchange rate pegs (albeit after an initial collapse in output and a sharp depreciation of the real exchange rate), while non-transition economies that switched to flexible regimes fared better than those that switched to pegged regimes.

Disinflation Attempts

16. Most GRA-supported programs that started from high (above 20 percent per year) inflation rates and that targeted substantial disinflation used an exchange rate anchor. In 80 percent of the cases, the target was achieved, and in two-thirds inflation remained low up to three years later (Table 3, Figure 1). For GRA-supported programs starting with relatively low inflation rates, about half of the disinflation attempts were based on flexible regimes and these programs had higher rates of success than similar disinflations under pegged regimes. In PRGF-supported programs, by contrast, disinflation attempts starting from high inflation did not use the exchange rate as a nominal anchor. Success rates have been similar to GRA-supported programs: in about 80 percent of cases the initial disinflation was achieved, and in three-quarters inflation remained low up to three years later (Table 3, Figure 2).

Table 3.

Success Rates for Disinflation Attempts under Alternative Exchange Rate Regimes

article image
Sources: International Monetary Fund, AREAER, WEO, MONA; and IMF staff estimates.

Exchange rate regime at t+1.

Low inflation cases refers to end-of-period inflation of less than 20 percent and programmed change in inflation between t-1 and t+1 of less than -5 percent.

Success is defined as actual disinflation performance at least meeting the programmed disinflation target (i.e. 5%, 10% and 20%).

Success II refers to the cases within Success in which disinflation is maintained, as measured by the difference between the average of end-period inflation in t+2 and t+3 and inflation in t-1 at least meeting the programmed disinflation target (i.e. 5%, 10%, and 20%).

Moderate inflation cases refers to end-of-period inflation between 20 and 50 percent and programmed change in inflation between t-1 and t+1 of less than -10 percent.

High inflation cases refers to end-of-period inflation greater than 50 percent and programmed change in inflation between t-1 and t+1 of less than -20 percent.

Figure 1.
Figure 1.

Inflation and Growth in GRA Programs Under Alternative Disinflation Strategies 1/2/3/

(In percent; 1995-2000)

Citation: Policy Papers 2004, 047; 10.5089/9781498330008.007.A001

Sources: International Monetary Fund, WEO, MONA, and IMF staff estimates.1/ Inflation rates are transformed to be mapped into the interval (-100, 100) percent.2/ Exchange rate regime at t+1.3/ Definition of low, moderate, and high inflation available in preceding table.
Figure 2.
Figure 2.

Inflation and Growth in PRGF Programs Under Alternative Disinflation Strategies 1/2/3/

(In percent; period 1995-2000)

Citation: Policy Papers 2004, 047; 10.5089/9781498330008.007.A001

Sources: International Monetary Fund, WEO, MONA, and IMF staff estimates.1/ Inflation rates are transformed to be mapped into the interval (-100, 100) percent.2/ Exchange rate regime at t+1.3/ Definition of low, moderate, and high inflation available in preceding table.

17. While individual country circumstances—for instance, initial credibility of newly (re)-established central banks in transition economies—may suggest a particular (exchange rate-based versus money-based) disinflation strategy, the contrasting findings for disinflation attempts under GRA-and PRGF-supported programs do not allow for unequivocal conclusions about which strategy is more likely to succeed. Rather, a distinguishing feature between successful and failed stabilization efforts appears to be whether supporting policies were in place. Although the proximate reason that inflation targets are missed and disinflation attempts fail is often monetary overruns—as discussed in Chapter III, below—it is also worth considering some of the underlying causes of failures. To this end, Table 4 correlates success at disinflations—under both pegged and more flexible regimes—to fiscal performance under the Fund-supported program. From the Table, fiscal slippage is significantly greater in cases where disinflation was unsuccessful. Indeed, whereas the fiscal balance was marginally better than programmed in cases that succeeded in disinflation, it fell short by 2.3 percent of GDP in cases that failed—a difference that is statistically significant. Countries that succeeded in achieving and maintaining low inflation also managed to achieve their fiscal targets, compared to a fiscal slippage of 1.5 percent of GDP among those programs that failed to maintain low inflation.

Table 4.

Fiscal Adjustment and Success Rates for Disinflation Attempts

article image
Sources: International Monetary Fund, WEO anAMONA; and IMF staff estimates.

Average of projection errors for years t and t+1.

Fiscal balance refers to the change in fiscal balance in percent of actual GDP; inflation is end of period, in percent, per year.

Disinflation attempts refer to programs that envisaged disinflation between years t-1 and t+1 of: 1) over 5%, when initial inflation in t-1 is less than 20%; 2) over 10%, when initial inflation is between 20% and 50%; or 3) over 20%, when initial inflation is higher than 50%.

Success is defined as actual disinflation performance between years t-1 and t+1 at least meeting the disinflation target (i.e. 5%, 10%, and 20%).

Success II refers to the cases within Success in which disinflation is maintained, as measured by the difference between average inflation for t+2 and t+3 and inflation at t-1 at least meeting the disinflation target (i.e. 5%, 10%, and 20%).

Two-sided t-test for differences in mean. Significant at: 10% *; 5% **; 1% *** levels.

18. When the exchange rate is not pegged, the country needs some other monetary framework to conduct monetary policy (Box 2). In Fund-supported programs, countries with no exchange rate peg, used either a monetary target, or an inflation target, or had no explicit nominal anchor.12 Countries with monetary targeting aimed at more ambitious disinflations and achieved greater reductions in inflation than countries that had no explicit nominal anchor (Table 5).13 Even in those cases where the projected decline in inflation was smaller under monetary targeting than in countries without an explicit nominal anchor (GRA-supported programs in non-transition economies and PRGF-supported programs in transition economies), the actual decline in inflation was greater under monetary targeting. Only two countries in the sample had an inflation-targeting framework in the year the arrangement was approved. These countries sought to maintain inflation at about 5 percent; in the event, inflation turned out to be 6¼ percent per year.

Monetary Regimes when the Exchange Rate is not Pegged

When the country does not peg its exchange rate (thereby subordinating its monetary policy to maintaining the peg), it must have some other nominal anchor and monetary regime. Of countries without pegged regimes in the year of program approval, other monetary frameworks prevailed in 70 percent of the cases. In most cases, only a ceiling on net domestic assets (NDA) and a floor on net international reserves (NIR) were specified in a program context.1 A further 27 percent targeted monetary aggregates (e.g. reserve, base money, or a broader monetary aggregate such as M2), and only 4 percent had an inflation-targeting framework.

Countries that had only an NDA ceiling and an NIR floor as part of the program conditionality and no explicit monetary framework, arguably lacked a nominal anchor as the NDA/NIR configuration in Fund-supported programs is intended primarily to monitor progress towards external viability and safeguard Fund resources— not to act as a nominal anchor for inflation expectations or monetary policy.2

Predictable money demand is required for aggregates to serve as a useful nominal anchor. Partly because of unstable money demand functions, central banks in several emerging market countries have shifted to inflation targeting. Indeed, in the sample, inflation targeting became more prevalent subsequently to the year in which the arrangement was approved, with several emerging market countries adopting inflation targeting within three years of the approval of the Fund arrangement. In part, this may be because inflation targeting needs institutional and technical requirements—such as central bank operational autonomy, effective monetary policy instruments, a system of accountability for the central bank, and reliable models to forecast inflation and the impact of monetary policy actions on inflation—to make it an effective monetary framework.3 Some countries adopted inflation targeting “lite” before formally adopting this framework.4 (In some programs, some of the changes required to establish a formal inflation targeting framework are part of the measures included under the program). The time pattern of adopting inflation targeting frameworks suggests that such frameworks have been considered useful for reducing inflation from moderate or relatively low levels, rather than disinflating from high inflation (Figure).

No PRGF-supported country in the sample adopted inflation targeting.

uA01fig01

End-period Inflation in Inflation Targeting Countries (1995-2000) 1/

(in percent, per year)

Citation: Policy Papers 2004, 047; 10.5089/9781498330008.007.A001

Sources: International Monetary Fund, WEO; and IMF staff estimates.1/ Inflation rates are transformed to be mapped into the interval (-100, 100) percent.
1 Due to data availability, these statistics on nominal anchors are based on a smaller sample (78 arrangements) compared to the sample used in the rest of the paper: only the most recent arrangement for each country is considered in the same 1995-2000 period. The monetary regimes are classified following the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions into 5 categories: 1) exchange rate anchor; 2) monetary aggregate target; 3) inflation targeting framework; 4) Fund-supported or other monetary program; and 5) other.2 The Fund’s focus on NDA originated from the Polak model with a fixed exchange rate regime, where the overall money supply is endogenous and only its composition is under the authorities’ control. Even under a flexible exchange rate arrangement, limits on NDA can: (i) prevent sterilized intervention in the presence of capital outflows; (ii) allow an accommodation of money growth if capital outflows are quickly reversed; and (iii) act as a disciplining device on fiscal policy. However, NDA limits do not control the overall monetary expansion when the latter is generated by foreign inflows, and hence may not provide a nominal anchor.3 Truman E., (2003) Inflation Targeting and the International Financial System: Challenges and Opportunities (Washington: Institute for International Economics); Schaechter, A., M. Stone, and M. Zelmer (2000), “Adopting Inflation Targeting: Practical Issues for Emerging Market Countries,” IMF Occasional Paper 202.4 See Stone, M., (2003) “Inflation Targeting Lite,” IMF Working Paper 03/12.
Table 5.

Inflation Performance Under Alternative Monetary Regimes 1/2/

article image

The monetary regime classification is based on the regime prevailing in year of program approval (t). The sample used for this exercise is smaller than in the rest of the paper due to data availability. Only the last arrangement of each country is considered.

Inflation is end of period. To reduce the influence of outliers, the inflation rate was transformed to be mapped into the interval (-100,100) percent.

No explicit nominal anchor was in place, except - in most cases - for a ceiling on net domestic assets (NDA) of the central bank and a floor on net international reserves (NIR) under the Fund-supported program.

External Adjustment

19. Besides the inflation objective, the exchange rate regime may be important for external adjustment. Flexible exchange rate regimes should allow for more of the adjustment to take place through expenditure switching rather than by demand restraint alone, implying a smaller output cost of a given improvement in the current account balance. To examine this hypothesis, the change in real GDP growth (between years t-1 and t+1) is regressed on the change in the current account balance, where the latter is instrumented by the projected change (Table 6). A given improvement in the current account balance is associated with lower output growth under fixed exchange rate regimes (significantly so for GRA-supported programs in non-transition economies14), but the corresponding coefficient under floating regimes is not significantly different from zero. The hypothesis of equality of coefficients under fixed and flexible regimes is strongly rejected, suggesting that, for these countries, more flexible regimes facilitated external adjustment.

Table 6.

External Adjustment and Growth Under Alternative Exchange Rate Regimes: Regression Results 1/

article image
Sources: International Monetary Fund, WEO and MONA ; and IMF staff estimates.

*: significant at 10%; **: significant at 5%; ***: significant at 1%.

Changes between year t-1 and t+1; the current account balance (net of official transfers) is in percent of GDP at t-1; all regressors except for the constant term were transformed to be mapped into an interval (-100,100) to reduce the influence of outliers.

Exchange rate regimes are classified by AREAER; "fixed regimes" include no separate legal tender, currency board arrangement, other conventional pegs, pegs with horizontal bands, crawling pegs, and crawling bands; "flexible regimes" include managed and independent floats.

Instrumented by programmed change in the current account balance and the change in actual growth of the terms of trade.

t-statistics for the hypothesis of equality of coefficients between fixed and flexible regime.

20. A second hypothesis regarding the relationship between the exchange rate regime and external adjustment is that countries with pegged regimes are more susceptible to capital account shocks because the exchange rate guarantee implicit in the peg encourages unhedged foreign-currency denominated borrowing by the private sector.15 Although a number of capital account crises countries had de jure or de facto pegs prior to the crisis and much larger capital outflows than projected at the time the arrangement was approved, this does not hold more generally. In fact, countries with more flexible regimes were more likely subsequently to undergo greater external adjustment than programmed (Table 7).16

Table 7.

Exchange Rate Regime and External Adjustment 1/

article image
Sources: International Monetary Fund, AREAER, MONA, WEO; and IMF staff estimates.

The difference in proportions under pegged and flexible regimes is tested. Significant at the: 1% (***); 5% (**); and 10% (*) level.

Exchange rate regimes as classified by IMF AREAER ; "pegged regimes" include exchange arrangements with no separate legal tender, currency boards, other conventional pegs, pegs with horizontal bands, crawling pegs, and crawling bands; “flexible regimes” include managed and independent floats.

C. Summary

21. Empirically, countries are no more likely to alter their exchange rate regime at the outset of a Fund-supported program than otherwise. Successful disinflations have been undertaken both under pegged and under flexible exchange rate regimes, suggesting that the consistency of macroeconomic policies, and the fiscal adjustment achieved, may be of greater importance than the choice of the nominal anchor. At the same time, some nominal anchor seems needed for disinflation. The exchange rate regime also has implications for external adjustment—in some cases, countries with more flexible regimes achieved a given improvement in the current account at a lower output cost.

III. Monetary Policy

22. Beyond the exchange rate (and, under flexible exchange rates, the monetary) regime, the authorities must also specify the monetary stance under their Fund-supported economic programs. This Chapter therefore discusses the monetary stance in terms of the behavior of broad money. Since the monetary authorities typically control (or have influence over) narrow aggregates, a first question concerns the mapping between narrow and broad money—that is, behavior of the money multiplier. It turns out that this relationship is generally stable and well-predicted in programs, making it appropriate to consider the monetary stance in terms of broad money aggregates (Box 3). Section A therefore considers broad money growth rates and velocity targeted in Fund-supported programs and their relationship to program objectives. Section B examines the impact of monetary policy on inflation and growth. In particular, does a tighter monetary stance contribute to lower inflation? Do overruns of broad money growth account for cases where program inflation targets were missed; and does the composition of this overrun (between net domestic assets and net foreign assets) matter for inflation performance? Was the monetary stance associated with lower output growth?

Relationship between Program and Actual Money Multiplier

The text discusses the programmed monetary stance—and its impact on key macroeconomic targets—in terms of the behavior of broad money (and the velocity of broad money). Since national authorities typically control (or have more direct influence over) narrower monetary aggregates, this raises questions about the stability of the money multiplier, and whether errors in projecting the money multiplier are an important source of program slippages.

The Table below seeks to examine the behavior of money multiplier—defined as the ratio of broad money to reserve money—in Fund-supported programs that were arranged during 1995-2000. The actual money multiplier has remained remarkably stable around its historical average across all types of programs (top panel): the null hypothesis of a constant multiplier cannot be rejected by the data. In addition, according to the regression results (bottom panel), program multiplier appears to be a good predictor of the actual multiplier (none of the reported F-statistics are statistically significant), accounting for more than 80 percent of cross-country variation. As such, the link between narrow and broad money aggregates is relatively stable and predictable.

Table.

Money Multiplier: Program versus Actual 1/

article image
Source: International Monetary Fund WEO, IFS and MONA database; and staff estimates.

Money multiplier is defined as the ratio of broad money to reserve money; year t refers to the year of program approval; significant at:

10%

5%

1%.

F-statistics are reported.

mmA and mmP refer to actual and program multiplier, respectively. Due to limited data availability, the sample of GRA- and PRGF-supported programs includes both transition and non-transition country programs.

A. Programmed Monetary Stance

23. Across various Fund-supported programs, broad money growth rates are targeted to decline, as are inflation rates (Table 8). The higher the initial inflation rate and rate of monetary expansion, the greater the targeted deceleration. For countries whose initial inflation was below 20 percent per year, the targeted deceleration was modest—from 12 percent in the year prior to program approval to 10 percent in the year following program approval. For countries whose initial inflation exceeds 50 percent per year, the deceleration is more marked, declining to annual rates of 13 percent in the year following program approval.17

24. Nominal money growth provides one gauge of the intended monetary stance, but it does take account of the increase in money demand associated with either real growth or inflation projected under the program.18 A simple metric is the expected change in money velocity—with an increase relative to the historical trend indicating that a tighter monetary stance was envisaged (though a decrease need not indicate a monetary loosening if inflation is expected to decline).19 By this metric, programs in high inflation countries sought significant monetary tightening, especially in the GRA sample (Table 8).

Table 8.

Programmed Money Growth, Velocity and Inflation

(in percent)

article image
Sources: International Monetary Fund, WEO and MONA ; and IMF staff estimates.

Money growth and inflation are end-period figures and transformed to be mapped into (-100,100).

For non-transition economies, programmed velocity growth for year t and t+1 is relative to trend velocity growth as measured by the five-year historical average; velocity is defined as nominal GDP divided by (period-average) stock of broad money.

25. To examine more systematically the determinants of the programmed monetary stance (as captured by program velocity), Table 9 reports the results of a regression of the programmed change in broad money velocity. Higher initial inflation and a larger targeted improvement in the current account balance should call for a tighter monetary stance (an increase in programmed velocity), while a larger output gap, a flexible exchange rate, or a higher expected rate of remonetization of the economy (proxied by the targeted decline in inflation) would, ceteris paribus, argue for a looser stance. For GRA-supported programs, all variables have the expected signs and are statistically significant. Overall, the regression explains some 60 percent of the variation in velocity in GRA-supported non-transition programs. Among PRGF-supported programs and programs in transition economies, the most important determinants are the lagged inflation rate and the expected inflation decline, while the exchange rate regime is not statistically significant; nonetheless, the regression explains some 50 percent of the variation in transition economies but only about 35 percent of the variation in PRGF-supported programs.

Table 9.

Programmed Monetary Stance: Regression Results 1/

article image
Sources: International Monetary Fund, WEO and MONA ; and IMF staff estimates.

Significant at:

1 percent

5 percent

10 percent.

Monetary stance is measured as programmed velocity growth in year t relative to trend velocity growth (as measured by the five-year historical average).

The current account balance (net of official transfers) is in percent of GDP; inflation was transformed to be mapped into an interval (-100,100) to reduce the influence of outliers.

Output gap is defined as a percentage deviation of real GDP from its Hodrick-Prescott filtered trend; positive value implies current output below trend.

B. Experience

Inflation

26. Fund-supported programs generally succeed in reducing inflation—though by not as much as targeted. Slippages in the year of program approval were generally modest, about 1½ percent per year across GRA-supported programs and ½ percent per year in PRGF-supported programs, though as much as 5 percent per year for countries whose starting inflation rates were between 20 and 50 percent per year. In addition, for the following year, inflation was, on average, higher than programmed by about 4½ percent per year in GRA-supported programs and 2½ percent per year in PRGF-supported programs (Table 10). The decline in inflation was driven in part by lower money growth rates. Moreover, a given growth rate of the money supply in the context of a Fund-supported program is associated with lower inflation, possibly because greater credibility in the authorities’ policies engenders confidence in the currency and thus raises money demand (Box 4). In GRA-supported programs, this effect is both economically and statistically significant—ceteris paribus, inflation is 10 percentage points lower (in the year following program approval) under a Fund-supported program than it would be under similar money growth rates but without a program. The effect is weaker and not statistically significant among PRGF-supported countries.

Table 10.

Programmed and Actual Inflation, Money Growth, and NDA Contribution

article image
Sources: International Monetary Fund, WEO and MONA ; and IMF staff estimates.

Projection errors are calculated as actual minus program values after transformation.

Transformed to be mapped into (-100,100).

NDA contribution is defined as ΔNDA/ΔM where Δ indicates level difference.

Does Fund Support Engender Confidence in Disinflation Efforts?

Beyond the effects of slower money growth on inflation, if Fund support enhances the credibility of the authorities’ policies then this should be reflected in greater confidence in the currency and higher money demand. Higher money demand, in turn, should result in lower inflation for a given growth rate of money. To test this hypothesis, it is useful to consider a standard money demand function:
mp=αyν(1)

where m is broad money, p the consumer price index, y real GDP, and v is (residual) velocity. Inverting and taking first differences yields:

π=ΔmαΔy+Δν(2)

where the behavior of velocity is assumed to reflect the additional confidence that Fund support might impart.

Equation (2) is estimated for both low-and middle-income countries that had a Fund-supported program at some point during the period 1990-2000 and whose inflation is above 10 percent per year.

The results in the Table for upper-and lower-middle income countries suggest that, while Fund support has little immediate effect on confidence and inflation (in part because inflation in the current year may be largely determined), it has an economically and statistically significant impact by the following year; ceteris paribus, lowering inflation by as much as 10 percentage points.

The confidence effects of Fund support in low-income countries are much weaker (ceteris paribus, lowering inflation by 3 percentage points) and not statistically significant. Although broad money growth is highly significant in both regressions, the residual standard error is 15 percent per year in the low-income country regression compared to 7 percent per year in the middle-income country regression.

Table.

Inflation and Money Growth under Fund-supported Programs: Regression Results 1/

article image
Source: Staff estimates.

***: significant at 1%; **: significant at 5%; *: significant at 10%.

π, Δm and Δy represent inflation, broad money growth and real GDP growth, respectively; prog is a dummy variable indicating a Fund-supported program; Δm and Δy are instrumented with their own lags; annual dummies are also included in the regression (not reported).

27. While lower money growth contributed to the disinflation achieved, money growth tended to be higher than programmed—by about 4 percent in the year of program approval and 6 percent the following year in GRA-supported countries, and 1.6 percent and 0.4 percent respectively in PRGF-supported countries. Again, among countries starting with high inflation rates, the slippages are considerably greater—as much as 25 percent in the year following program approval in GRA-supported countries (Table 10). Table 11 seeks to explain some of the factors behind the slippage in broad money growth. In part, higher broad money growth reflects the effect of depreciation of the nominal exchange rate on foreign ansition economies. Beyond this effect, among non-transition economies, fiscal slippages are correlated with money growth slippages, while the performance of output growth appears to have little explanatory power (except for GRA-supported programs in year t+1). Finally, a larger programmed decline in inflation is associated with larger money growth slippages, presumably reflecting the difficulty of achieving ambitious disinflations. Overall, the regressions have greater difficulty in explaining slippages during the first program year, accounting for only 9 to 25 percent of the variation, but somewhat greater success in accounting for slippages in the subsequent year.

Table 11.

Determinants of Broad Money Growth: Regression Results 1/

article image
Sources: International Monetary Fund, WEO and MONA ; and IMF staff estimates.

***: significant at 1%; **: significant at 5%; *: significant at 10%.

ΔmER, ΔyER, ΔexrER and ΔfbalER represent projection error in broad money growth, real GDP growth, percentage change in the nominal exchange rate (national currency per US dollar) and fiscal balance in percent of GDP, respectively; Δπp refers to programmed change in inflation; projection errors are calculated as actual minus program values after transformation that maps underlying variables into an interval (-100,100).

Exchange rate regimes are classified by AREAR; “fixed regimes” include no separate legal tender, currency board arrangement, other conventional pegs, pegs with horizontal bands, crawling pegs, and crawling bands; “flexible regimes” include managed and independent floats.

Excludes capital account crisis programs.

28. In turn, Figure 3 correlates the slippage in broad money growth to the higher-than-programmed inflation rates. The relationship is statistically significant and—for the sub-sample in which money growth was higher than programmed—accounts for 30-60 percent of the variation of inflation projection error. The inflationary impact of monetary overruns naturally depend on whether there was a concomitant increase in money demand. A common hypothesis, in this regard, is that monetary expansions that reflect higher net foreign assets correspond to capital inflows responding to higher money demand, and should thus have a smaller inflationary impact; conversely, monetary overruns that reflect larger NDA growth than programmed should have a larger inflationary impact. Empirically, however, the source of the monetary overrun makes no difference to the inflationary impact (Table 12). This underscores the finding above that a NDA/NIR framework is not well suited to controlling inflation, which generally requires a more explicit nominal anchor. It also underscores the need to sterilize capital inflows or large donor support if the inflation target is to be achieved.20

Figure 3.
Figure 3.
Figure 3.

Projection Error in Inflation and Money Growth 1/

Citation: Policy Papers 2004, 047; 10.5089/9781498330008.007.A001

Source: International Monetary Fund, MONA, WEO, and staff estimates.1/ *: significant at 10%; **: significant at 5%; ***: significant at 1%.Sub-sample inlcudes observations with positive projection error in money growth only.
Table 12.

Projection Errors in Inflation and Money Growth: Regression Results 1/

article image
Sources: International Monetary Fund, WEO and MONA ; and IMF staff estimates.

Projection error in fiscal balance (in percent of GDP) was added to all regressions as a control; significant at:

10%,

5%,

1%.

INFER, MGER and fbalER refer to projection error in inflation (end-period), broad money growth and fiscal balance (in percent of GDP), respectively; D50 and D90 are dummy variables that equal 1 if ΔNDA/ΔM > 50% and ΔNDA/ΔM > 90% respectively, and 0 otherwise; NDAMG represents the contribution of NDA to broad money growth.

Exchange rate regimes are classified by AREAR; “fixed regimes” include no separate legal tender, currency board arrangement, other conventional pegs, pegs with horizontal bands, crawling pegs, and crawling bands; “flexible regimes” include managed and independent floats.