Chapter

CHAPTER 6 Inflation: The Case for a More Resolute Approach

Author(s):
Susan Schadler, and Hugh Bredenkamp
Published Date:
June 1999
Share
  • ShareShare
Show Summary Details
Author(s)
Steven Phillips

This chapter begins with a review of inflation performance and arrives at a mixed assessment: the record on ending high inflation is satisfactory, but the record on achieving low (single-digit) inflation is disappointing. Such an impression arises both from a long view over the past 15 years and from developments during SAF/ESAF-supported programs. For most of the countries under review, there was no tendency to reduce inflation during 1981–95. SAF/ESAF-supported programs tended to target a phased movement to single-digit inflation, but overshooting of program inflation targets was pervasive. Targets aside, some drop in inflation occurred during programs, but much of this only reversed a run-up just before programs began. Moreover, the gradual disinflation during many three-year programs generally did not continue into the following years. Together, these stylized facts account for the absence of a general movement, over time, into the low-inflation range.

This record raises the question of whether achieving low inflation is important, or whether somewhat higher, “intermediate” inflation might be satisfactory, if not ideal. After a brief review of performance, the chapter therefore assesses the likely benefits and costs of reducing inflation. The results indicate that both high and intermediate inflation rates are associated with slower growth than under conditions of low inflation. Furthermore, the disinflation process in programs is generally not associated with a slowing of growth. It is shown, however, that disinflation will often require that the public sector adjust itself to appreciably lower seigniorage income, highlighting the extent to which disinflation is a fiscal (hence distributional) issue.

Having established that the goal of low inflation is worth pursuing, the analysis next considers the determinants of program inflation performance and discusses how program design might be made more effective in delivering disinflation. It is found that slippages on policy commitments central to the disinflation process were frequent—though far from universal—and this no doubt contributed to weaknesses in inflation performance. However, even when key policy commitments were respected, inflation targets were still usually exceeded, implying that the design of programs needs to be bolstered in some way to be more consistent with inflation objectives. Overall, the evidence suggests that greater success in disinflation could be achieved through a complementary approach of stronger fiscal targets and wider use of nominal anchors. The final section summarizes the chapter’s findings and main policy conclusions.

The Inflation Record: Inside and Outside of Programs

Inflation performance can be considered from several viewpoints. A simple chronological view, disregarding the presence or absence of IMF-supported programs over the years, is the first approach used. Next, a program-centered perspective, looking at developments during SAF/ESAF-supported programs, is taken. Both perspectives are revealing, albeit with regard to somewhat different questions. The messages from each turn out to be similar.

The SAF/ESAF-supported programs under review have encompassed a wide range of inflation experiences, including severe inflationary conditions prior to the programs. In light of this diversity, it is useful to group countries (or programs) into three categories, according to the degree of initial inflation. First, “high” inflation is taken to be any annual rate above 40 percent. The distribution of inflation rates in the countries under review falls off sharply above this level, and such inflations tend to be episodic. Second, initial inflation rates below 10 percent are considered “low,” in the sense of being within striking distance of where most SAF/ESAF-supported programs aim to bring inflation over the medium term. Third, inflation rates of 10 to 40 percent are classified as “intermediate.”1

A Chronological Perspective: Inflation Trends from the Early 1980s

Although trends in inflation have varied widely among ESAF users over the past 10–15 years, some key patterns can be discerned. First, three country groups, shown in the top panel of Figure 6.1, exhibit broadly stable inflation over time. The eight CFA countries had the lowest inflation, reflecting the pegging of their common currency to the French franc. Average inflation in these countries was generally very low until the large devaluation of the CFA franc in 1994 and subsequently fell back, to about 7 percent in 1995 and 5 percent in 1996. Inflation in the four Asian countries was greater, often in the neighborhood of 10 percent, although inflation in Bangladesh recently averaged less than 5 percent.

Figure 6.1Consumer Price Inflation: Regional Median Values

(End of period, where available)1

Source: IMF staff estimates.

1When end-of-period data are not available, period average consumer price index figures are used. In a few cases, the GDP deflator is used.

2Excludes transition countries.

The 14 non-CFA African countries generally had higher inflation, although this is a diverse group. Most of these countries avoided lingering in the high-inflation range for more than a year or two, but sustained intermediate inflation was typical. Only Mauritania and (until recently) Burundi kept inflation usually in the low range. During 1981–95 there was no general tendency among the non-CFA African countries to reduce inflation. Although median inflation accelerated in the 1990s, finally exceeding 20 percent in 1995, preliminary data for 1996 show markedly slower inflation in many of these countries.

Two other groups (bottom panel of Figure 6.1) had much higher inflation, but they typically reduced it aggressively. Among the four Western Hemisphere countries, Bolivia and Nicaragua experienced extremely high inflation at times, but subsequently (in 1986 and 1992, respectively) they disinflated sharply; Guyana also had high inflation, followed by a moderation beginning in 1992. By contrast, Honduras had a record of low inflation until the 1990s, when its inflation rate accelerated into the intermediate range. By 1995, all of these countries but Honduras had inflation in the neighborhood of 10 percent. The six transition economies experienced high inflation early in the process of economic liberalization. All but Mongolia quickly brought inflation down to the intermediate range, and by 1995 Albania, Cambodia, and Vietnam had rates close to or below 10 percent.2

Figure 6.2 provides a different, but still chronological, perspective on inflation, comparing typical inflation rates in the period leading up to a country’s first SAF/ESAF-supported program and in recent years (1993–95). Outside the CFA countries, average inflation recently was below its level in the year before these programs began, with the implied disinflation being most pronounced in the Western Hemisphere and transition economy groups. However, this impression is qualified by the fact that inflation had often been on the rise during the preprogram period; such a pattern exists for the non-CFA African countries, the transition economies, and, to a lesser degree, the Asian economies. When recent inflation rates are compared with average levels over the three-year preprogram period, the absence of a downward trend for most non-CFA African and Asian ESAF users (apparent in Figure 6.1) is confirmed.

Figure 6.2Progress on Inflation from the Pre-SAF/ESAF Period to 1993–95

(Median values)

Source: IMF staff estimates.

Inflation performance was most mixed in non-CFA Africa, the largest country grouping considered here. In a handful of these 14 countries—The Gambia, Guinea, Lesotho, Mauritania, and Uganda—inflation was reduced considerably between the period before their first SAF/ESAF arrangements and 1993–95 (though sometimes much more slowly than originally envisaged). In half of the non-CFA African countries, however, inflation accelerated over this period.

Inflation During Programs: Initial Conditions, Targets, and Outcomes

Initial Conditions

Of the 67 cases of three-year SAF/ESAF arrangements approved during 1986–94, only 9 had high initial inflation, whereas 26 had intermediate initial inflation (Figure 6.3).3 The high-inflation cases comprised three transition economies (all with initial inflation over 200 percent), three countries in Africa, and two in the Western Hemisphere.4 Almost half of the sample had low inflation in year t - 1, split about equally between CFA countries and others. Among cases with intermediate or high initial inflation, on average there was a tendency for inflation to have been rising before the program. However, some programs were preceded by a significant slowing of inflation: in particular, nine cases had escaped from high inflation in their recent past, cutting inflation to intermediate or even low levels by year t - 1.

Figure 6.3Distribution of Initial Inflation Levels Among 67 SAF/ESAF-Supported Programs (Including CFA Cases)

Source: IMF staff estimates.

1This bar represents five cases in which initial inflation exceeded 100 percent.

Targets

Given the diversity of initial conditions, inflation objectives in SAF/ESAF-supported programs were themselves diverse. The most striking distinction was between the countries of the CFA franc zone and other ESAF users. For the CFA countries, inflation targets for the third program year were usually in the range of 2½ percent to 4 percent, consistent with their common exchange rate objective. Since initial inflation was usually already quite low in these countries, the objective was less often to reduce inflation than to keep it down. Outside the CFA countries, inflation objectives were somewhat less ambitious but usually called for inflation to reach 5–10 percent within the program period, approaching 5 percent in the long run. Inflation targets outside the CFA countries tended to be related to the degree of initial inflation, especially in the first program year, but there was a clear convergence of targeted inflation rates over time (Figure 6.4).

Figure 6.4Inflation Targets in SAF/ESAF-Supported Programs, by Degree of Initial Inflation1

(Group medians; excluding CFA African countries)

Source: IMF staff estimates.

1Targets set at the beginning of three-year SAF/ESAF supported programs.

Outcomes

For the 17 SAF/ESAF programs involving eight CFA countries, inflation was typically below program targets, in all three program years.5 Tending to start with very low inflation, these countries usually kept inflation low. The key exceptions were those outturns influenced by the large CFA franc devaluation of 1994. The three-year programs begun in 1992 and 1993 did not originally incorporate this devaluation into their inflation targets. However, targets in the programs begun in 1994 did so, anticipating cumulative inflation over the three-year period of about 50 percent on average. Actual inflation outturns in these three cases were fairly close to projections, averaging a bit less than 55 percent cumulatively.

Outside the CFA, overshooting of inflation targets was typical, in all three program years, although a minority of cases did perform well relative to target. Targets aside, many countries (about two-thirds) managed some reduction in inflation during programs. Figure 6.5 shows how the distribution of inflation among non-CFA countries shifts over “program time.” Overall, programs have been most successful in ending episodes of high inflation. By contrast, the record of transition to low inflation is disappointing: even by the third program year, the share of programs exhibiting low inflation (38 percent) is not much more than the proportion of cases that already showed low inflation in the preprogram year (30 percent).

Figure 6.5Distribution of Inflation Among 50 SAF/ESAF-Supported Programs (Excluding CFA Cases)

Source: IMF staff estimates.

1Represents cases in which inflation exceeded 100 percent.

2For the final program year, two observations are missing.

It is useful to examine these cases in more detail, according to the degree of initial inflation (Figure 6.6). In the 15 programs with low initial inflation, inflation on average exceeded its target by several percentage points. However, this central tendency conceals a considerable divergence within the group. In one-third of these cases, inflation moved upward into the intermediate range, with average inflation of 16 percent in the third program year, compared with targets in the vicinity of 5 percent. The rest of this group tended on average to reduce inflation slightly by the third program year and to come close to target.

Figure 6.6Inflation: Program Targets and Outturns1

(Median values; in percent)

Source: IMF staff estimates.

1Figures in parentheses above bars refer to the number of observations.

The 26 programs beginning with intermediate inflation (the largest group) also show a mixed record. Although some countries with intermediate inflation achieved their inflation targets, most fell considerably short: less than half achieved even one-half of the disinflation targeted for the first year. The picture is similar for the second year and only somewhat better for the third year. Nevertheless, many programs with intermediate inflation saw at least some disinflation: relative to inflation in year t - 1, many of the intermediate initial inflation cases brought inflation down in the first program year, and most did so by the second year. However, much of the disinflation achieved on average merely reversed a run-up of inflation during the preprogram period. This up-down pattern explains why for many countries there was no tendency to reduce inflation over the years, even though their programs were associated with some disinflation. Even by the third program year, only about one-fourth of the intermediate initial inflation cases reached low inflation.

In the nine programs with high initial inflation, inflation on average considerably overshot its target, in all program years. Still, these programs did tend to achieve a sharp reduction of inflation, making most of their progress in the first two program years. By the final program year, six of the nine cases managed to exit high inflation, although only Albania reached low inflation (Albania’s inflation rebounded to double digits in the following year).

That movement into the low-inflation range by the third and final program year was infrequent raises the question of what typically happened subsequently. Given that inflation did tend to decline (albeit often slowly) during each of the three program years, did inflation then typically continue to fall? In general, the answer is no—on average, inflation in the fourth and fifth years after a program was initiated stalled near the outcome of the third program year.

This finding has a brighter side, however: among the minority of three-year programs that were associated with significant disinflation, there was not usually a “rebound” of inflation in the following years. Looking at these cases two years further out (four and five years after a program was initiated) indicates a generally good record of preserving disinflation achievements. Among the six intermediate inflation cases with the strongest disinflation (those that halved inflation by the third program year), Guinea, Nepal, and The Gambia showed essentially no rebound over the medium term, and Madagascar also did fairly well. By contrast, Malawi and Mauritania suffered a rebound (the latter from a very low inflation rate, however). Among cases of high initial inflation, Bolivia, Guinea, and Guyana showed little or no reversal of strong disinflation.6

Conclusions and Interpretations

For many ESAF countries, the record on moving to low inflation, both inside and outside of programs, has been disappointing. The frequent overshooting of program targets, often by significant margins, raises issues of lost credibility, a problem that could undermine the effectiveness of future programs and disinflation efforts. It also raises questions about the degree of importance the IMF attaches to lowering inflation, especially from intermediate levels. It is noteworthy that programs that avoided interruption (as defined in Chapter 9)—where the IMF presumably judged policy results and compliance with policy commitments to be worthy of continued support—included some cases of very large inflation overshooting. Moreover, the programs avoiding interruption were not characterized by better inflation performance than in interrupted programs: on average, the former group actually had a greater excess over their three-year inflation targets.

To avoid sustaining a pattern of frequent overshooting of inflation targets, logic suggests two choices. Either programs should adopt less ambitious inflation targets, if inflation continuing in the intermediate range is considered acceptable; or, if low inflation is judged to be a worthwhile objective, greater efforts should be made to ensure that the objective is met. The next two sections consider the likely benefits and costs of achieving lower inflation, and whether the typical target of 5–10 percent inflation by the third year of ESAF-supported programs is appropriate for the countries considered.

Benefits of Low Inflation

One of the benefits of lower inflation, as discussed in Chapter 5, is that it may facilitate higher long-run growth. In this section, the implications of this relationship for setting inflation objectives in ESAF-supported programs are considered, as well as another possible benefit of disinflation, related to the distribution of income.

Assessing the Growth Benefits of Lower Inflation

The conventional wisdom about the association between inflation and growth makes a critical distinction between long-run and short-run relationships. In the long run, inflation may be expected to be negatively related to output growth, including through a causal effect from inflation, while short-run comovements may be negative or positive.7 This section focuses on the long-run relationship. Although few would doubt the desirability of ending high inflation, the advantages of ending intermediate inflation are not universally accepted—accordingly, attention is focused on this point, and on understanding why some but not all empirical approaches detect apparent growth benefits of low inflation.

On an informal level, low inflation is widely regarded as desirable, and indeed some empirical evidence now supports this preference. However, in the academic literature there is a certain amount of skepticism about a growth payoff from low inflation. Some studies find no significant role of inflation in determining growth (for example, Bosworth, Collins, and Chen, 1995); others acknowledge only an effect of high inflation on growth; and some present evidence that negative correlations between inflation and growth are not robust to excluding high-inflation outliers from the sample (for example, Bruno and Easterly, 1995; and Levine and Zervos, 1993). If only high inflation were injurious to growth, then the inflation record of SAF/ESAF-supported programs would be satisfactory. Our conclusion, however, is that by ignoring the highly nonlinear shape of the inflation-growth relationship some empirical approaches have missed or understated the costs of intermediate inflation.

Several factors complicate measurement of the growth benefits of lower inflation. One is that the relationship between the inflation rate and growth is unlikely to be linear. Such a relationship would rather implausibly imply that if low inflation is good, then zero is better, and rapid deflation better still.8 Furthermore, if prices are downwardly sticky, then zero or very low inflation might make an economy vulnerable to prolonged recessions; this suggests that inflation might be positively related to growth over some very low range. Beyond such a range, inflation is likely to be negatively related (or at least not positively related) to growth, but it is still not clear that one should expect a smooth relationship. For example, perhaps there is little or no injury to growth until inflation reaches some moderate threshold, say 10 percent. Beyond that, the effect on growth of moving from 10 percent to 20 percent might be greater than moving from 110 percent to 120 percent. On the other hand, many authors emphasize a distinction between intermediate and high inflation; perhaps serious harm to growth begins only after some higher threshold is crossed.

Other complications arise from the distribution of inflation. Its extreme right-skewedness can give great weight in empirical analyses to infrequent cases of very high inflation. Also, inflation data tend to be clustered below 10 percent, so that there are fewer observations than one would like (for analytical purposes) in the region of greatest policy interest for many ESAF users: the intermediate range.

These considerations suggest that linear regressions could be ill suited to examining the inflation-growth relationship. A similar impression emerges from a plot of inflation and growth, using a large panel of developing country data.9 Observations on per capita growth and inflation were first ranked according to inflation. The ordered sample of over 1,500 observations was divided into 15 equal-sized groups.10 For each group, mean inflation and mean per capita growth rates were computed, to make the basic patterns in the data readily apparent. The result was an inverted V-shaped relationship between inflation and per capita growth (Figure 6.7, top panel). Within most of the sample—that is, excluding the three groups with the lowest inflation—higher inflation is associated with lower per capita growth.11 However, looking at the data groups with the lowest inflation, a positive association is suggested. The turning point occurs at inflation of about 5 percent. At rates higher than this, growth tends to decline, but at a declining rate. It seems that the (marginal) effects of inflation become less rather than more severe as inflation moves into the high range.

Figure 6.7Association of Per Capita GDP Growth with Inflation and Logarithm of Inflation1

(Mean values within groups ranked by inflation)2

Source: IMF staff estimates.

1Based on 105 developing countries’ annual data, 1981–95; a total of 1,564 observations (11 missing observations).

2The 1,564 observations were ordered according to rate of inflation, then divided into 15 subgroups of 104 observations each (the highest inflation group contains 108 observations).

3The group with mean inflation of 1,047 percent (and mean growth of –2.5 percent) is not depicted in this panel.

4The group consisting mainly of negative inflation observations is not depicted in this panel.

The top panel of Figure 6.7 illustrates two reasons why a linear regression of growth against inflation can be biased toward finding only a weak negative relationship over the intermediate range. First, observations to the left of the relationship’s turning point will flatten the estimated slope of a regression line, a point emphasized by Sarel (1996). Clearly, there is a need for regression analysis to allow a “kink” somewhere in the range of low inflation (as was done in Chapter 5 using a spline technique). Second, beyond the turning point, the apparent convexity of the relationship—combined with the presence of observations with extremely high inflation—will also flatten the estimated slope away from the true slope over the intermediate range.12

The bottom panel of Figure 6.7 demonstrates that using logarithmic values of inflation handles the convexity problem quite well; beyond the turning point at low inflation, the negative relationship now seems close to linear.13 Moreover, once log values are added, there is no suggestion of a need for a regression to incorporate a second kink at some threshold between intermediate and high inflation. (In light of these points, the multivariate growth regression of Chapter 5 used a log specification of inflation together with a spline technique allowing a single kink in the low-inflation range.) Finally, the log specification allows even the most extreme high-inflation group to be included in Figure 6.7, although the impression of a negative relationship does not hinge on the presence of these observations (Box 6.1).

The results of this study (both the multivariate regression analysis in Chapter 5 and Figure 6.7) suggest that the inflation rate associated with maximum growth may be broadly similar to the typical long-run inflation target of most SAF/ ESAF-supported programs—5–10 percent by the third program year, and moving toward 5 percent in the long run. Growth may benefit from reducing inflation to single digits, but there seems to be no case for going below 5 percent.14

How important are these apparent growth benefits? The regression analysis of Chapter 5 found a statistically significant negative relationship between growth and inflation, but is this relationship economically significant? That is, if the relationship were entirely causal in nature, how large might be the implied growth benefits of moving to low inflation? The question is important because if the evidence suggested only small benefits, they might not justify the possible costs of disinflation. To address it, recall the multivariate regression analysis of Chapter 5, which controlled for a number of variables that might influence inflation and growth in opposite directions. Beyond the kink estimated to lie at 4½ percent inflation, the estimated coefficient on (log) inflation is close to –0.8. Such a value implies that, as the inflation rate is doubled (is halved), per capita GDP growth falls (rises) by more than half a percentage point. Although this estimate is sizable, it is not the largest in the recent literature using nonlinear techniques. For example, Sarel (1996) produced a far stronger estimate, one implying that a halving of the inflation rate is associated with a rise in growth of 1.7 percentage points.15De Gregorio’s (1993) result, for 12 Latin American economies, is close to the estimate here, suggesting that halving inflation might increase growth by 0.4 percentage points. A bit larger is the Judson and Orphanides (1996) estimate, with a reduction of inflation from 25 to 15 percent being associated with an increase in growth of about ½ of 1 percent.

Box 6.1The Inflation-Growth Association: Is It All Just High Inflation?

The robustness of a negative inflation-growth association has been the subject of some controversy. A specific charge is that such findings are driven by outliers; that is, the undue influence of a few countries with very high inflation. For example, Levine and Zervos (1993) demonstrated that merely dropping Nicaragua and Uganda from a large cross-section regression can cause the observed relationship to break down. A more general charge is that negative effects on growth begin only once inflation becomes “high” (by some definition).

By using techniques that allow the inflation-growth relationship to take a nonlinear form, it was found that the relationship does not depend on high inflation. The use of log inflation not only handles the relationship’s nonlinearity but also tends to reduce a regression’s sensitivity to high-inflation observations. On the more general charge, a graphical analysis of log values (Figure 6.7) gives no sign of a turning point where the (marginal) damage of inflation picks up once high inflation is reached. Moreover, Figure 6.7 suggests a clear negative relationship even when all high-inflation observations are disregarded.

The formal way to address these issues is to drop high-inllaiion observations from a multivariate regression analysis. In conducting this test, we chose to exclude not only the most extreme outliers, but all observations with inflation above either 80 or 40 percent (the thresholds suggested by Levine and Zervos). The results in the restricted samples were quite similar to the full-sample results: the estimated coefficient on log inflation (when inflation is above 4½ percent) is -0.79 in the full sample, -0.82 when the 80 percent inflation cutoff is used, and -1.03 when the cutoff is 40 percent. The coefficient’s standard error rises and its t-statistic falls somewhat as the sample range is narrowed (from a full-sample value of -6.46, to -4.08 and -4.14 for the 80 and 40 percent cutoffs, respectively), but its statistical significance is retained.

Comparisons with studies using a linear specification are not easily made. However, Fischer’s (1993) linear study included a nonlinear variant that allowed two kinks, set at 15 and 40 percent inflation. Even if the true relationship is a convex curve (and Fischer’s spline estimates are consistent with this), such multiple kinks may be a good enough approximation to make a comparison reasonable. Table 6.1 presents a comparison with the results of Fischer and Sarel; Fischer’s nonlinear findings also suggest a steeper negative relationship than found in this study.

Table 6.1Implied Reduction in Growth Rate when Inflation Exceeds 5 Percent(In percentage points, compared with growth associated with 5 percent inflation)
Inflation (percent)Fischer (1993) (nonlinear variant)Sarel (1996) (nonlinear)Present Study (nonlinear)Memorandum: Present Study (linear variant)
10-0.64-0.42-0.55-0.004
15-1.27-1.43-0.87-0.008
20-1.65-2.14-1.10-0.012
30-2.40-3.15-1.42-0.019
40-3.15-3.87-1.64-0.027

Although these estimates differ, the key point is that all the nonlinear estimates indicate that considerable losses are associated with inflation, even at the lower end of the intermediate inflation range. In contrast, a linear variant of the present study (also shown in Table 6.1) would suggest only small losses, even at the high end of the intermediate range.16

Thus, the evidence seems to suggest a considerable return to those countries avoiding not only high but also intermediate inflation. However, a question pertaining to these results is whether the inflation-growth association reflects only a direct causal link from (lower) inflation to growth. Some part of the observed association may reflect effects running from growth to lower inflation, or some third factor driving some countries to both faster growth and lower inflation. Of course, the multivariate regression analysis did control for certain policies (for example, the fiscal deficit) and supply shocks (terms of trade deterioration, drought) that could lead to both higher inflation and lower growth. However, it did not control for demand policy shocks that could move inflation and growth in the same direction in the short-run—the estimated negative coefficient might have been larger still if any such effects had been captured.

For policy purposes, it is not essential to know whether lower growth is caused only by inflation itself or also by other policies or policy characteristics—such as inefficiency of public expenditure—that are associated with inflation precisely because they are being financed through inflationary seigniorage. In this study, it is both understood and intended that the inflation variable may capture not only the direct influences of inflation on growth but also the effects from factors that are associated with inflation but are more difficult to measure. If lower inflation cannot be achieved without addressing these policy shortcomings, it is not misleading to refer informally to “the effects of inflation” on growth. As Fischer (1993) commented on his findings, even if inflation is seen merely as “a symptom of a government out of control” this does not contradict the view that eon-trolling inflation would help to restore growth.

A further issue is that of a “reverse” causality running from faster growth to lower inflation. Such an effect is intuitively plausible, and there is nothing in the evidence here to refute it. However, several studies suggest that reverse causality is not the whole story. For example, Fischer’s (1993) results suggest that inflation reduces growth by reducing both investment and productivity growth. Cukierman and others (1992) used an index of central bank independence as an instrument for the inflation rate; their results indicate that causation runs significantly, though not exclusively, from inflation to growth. Similarly, Barro’s (1995) results using different instrumental variables suggest that most of the negative association be found does not represent a reverse causation from growth to inflation.

In conclusion, one cannot be certain from the available evidence exactly how large the long-run growth benefits of disinflation might be for a typical—let alone any particular—country. It is questionable, however, whether doubts about the size of the effect, or the nature of the causality at work, would justify maintaining inflation at intermediate levels. The evidence suggests that a country that did so would risk forgoing sizable and durable gains in economic growth.

Implications of Low Inflation for the Distribution of Income

A different sort of potential benefit of low inflation relates to the distribution of income: to the extent that inflation adds to income inequality, disinflation may halt this effect. In principle, inflation can influence income distribution through numerous channels, not all working in the same direction, so that its overall effect will depend on various other country-specific factors.17 The standard presumption, however, is that inflation is a regressive tax, primarily because it transfers resources away from holders of currency and non-interest-bearing deposits, and the poorest elements of a society may be least able to find shelter from this effect. The effects of inflation on income distribution through this channel may be particularly strong in lower-income countries having a relatively unsophisticated financial sector (see Bulíř and Guide, 1995). Although there is little empirical work on the inflation-income distribution link, Bulíř and Guide suggest that inflation tends to have an overall regressive effect in lower-income countries.18 Thus, reducing inflation to low levels may limit this transfer of income away from those at the bottom of the distribution.19

Potential Costs of Disinflation

Two main considerations are usually advanced to explain why policymakers may be reluctant to move to low inflation. First, when wage and other rigidities exist, disinflation may entail short-term output costs. Second, the public sector will lose part of the seigniorage it collects through inflation, in general requiring a scaling back of public sector activities or a replacement of the lost revenue with more explicit (and hence, perhaps, more contentious) taxes.

Is Disinflation Associated with Short-Run Output Losses?

Although disinflation may yield an eventual payoff in terms of faster growth, the conventional wisdom of the Phillips curve is that disinflation may hurt growth in the short run. If policymakers have a high discount rate, they may thus decide that bringing inflation down is not worthwhile. Did growth in fact decline during episodes of disinflation in SAF/ESAF-supported programs?

Several authors (for example, Dornbusch and Fischer, 1993; and Ball, 1993) have provided evidence that short-run output losses from disinflation may be substantial; Dornbusch and Fischer argued that the prospect of such losses is an important reason why policymakers in some countries have maintained “moderate” inflation.20 However, a crack has recently appeared in the conventional wisdom of costly disinflation, at least for high inflation (see Easterly, 1996, who found evidence that disinflation from such levels is associated with short-run output gains). Even within the orthodox view, the extent and duration of a contractionary effect of disinflation may vary across countries depending on the structure of the economy or other initial conditions. For example, the “menu-cost” model of Ball, Mankiw, and Romer (1988) predicted that the size of the inflation-growth trade-off will vary negatively with the degree of initial inflation; they and others have provided evidence consistent with this view. Thus it is possible that those with greater need for disinflation may face the lowest short-term output costs of such disinflation.

Another factor that could diminish the contractionary impact of disinflation in ESAF countries is that the degree of wage inertia in economies with large informal sectors is likely to be less than that in more advanced economies. Ball found that the flexibility of wage-setting institutions is a determinant of the “sacrifice ratio” within his sample of moderate-inflation OECD countries.21 It is also possible that any contractionary effects of disinflation might be more than offset by positive influences on growth from other policy changes in SAF/ESAF-supported programs. Finally, some part of the long-run beneficial impact of lower inflation may be felt even in the short-run—this seems especially likely in cases of very high inflation.22

What does program experience suggest? For a first look at this question, Figure 6.8 shows that growth during SAF/ESAF-supported programs was often not only positive, but faster than preprogram growth, with the possible exception of the countries with low initial inflation. However, since the record of actual disinflation during these programs varied considerably within the program groupings, it is useful to look more closely, examining the comovements of growth and inflation during individual programs.

Figure 6.8Real GDP Growth During SAF/ESAF-Sup ported Programs1

(Median values; percentage change over previous year)

Source: IMF staff estimates.

1Figures in parentheses above bars refer to the number of observations.

In cases of high initial inflation, preprogram growth rates were typically very low, even negative. Among the countries exiting high inflation during their SAF/ESAF-supported programs, a sharp improvement in growth was observed (Table 6.2). Considering the size of the improvement: in growth associated with these cases of disinflation, it seems that any Phillips curve trade-offs, if they existed, were dwarfed by other, positive influences of the programs, perhaps including those arising from disinflation. Indeed, the short-run growth of countries disinflating from high initial levels—measured relative to preprogram growth—is the best among the SAF/ESAF-supported programs (refer again to Figure 6.8). These programs also experienced the sharpest disinflation, even when disinflation is measured using ratios to initial inflation. The Phillips curve question thus seems moot in cases of high initial inflation.

Table 6.2Growth Performance in Countries Exiting High Inflation During SAF/ESAF Programs
CountryFirst Program YearInitial Inflation (in percent)Program Year Exiting High InflationChange in Growth Rate Relative to Three-Year Preprogram Average (in percentage points)
Year tYear t + 1Year t + 2
Albania1993236.6First24.724.523.7
Bolivia198766.0First3.64.04.6
Guinea198771.8First
Guyana1990104.7Third-3.67.79.5
Kyrgyz Republic19941,363.0Second-6.714.618.9
Uganda1987237.0Third7.15.55.3
Source: IMF staff estimates.Note: For Guinea, data for preprogram growth are not available.
Source: IMF staff estimates.Note: For Guinea, data for preprogram growth are not available.

The Phillips curve question holds greatest interest as a potential explanation of why many countries with intermediate inflation have sustained it. To look for a possible short-term association of disinflation with diminished growth, we classify program cases with intermediate initial inflation according to three degrees of disinflation achieved over a specified time horizon. Disinflation is measured as the ratio of the inflation rate in a given program year to initial (year t - 1) inflation. If this measure equals or exceeds 0.9, a country is deemed to have experienced no (significant) disinflation through that time; if the measure is less than or equal to 0.5, disinflation is termed “greater,” and all other cases are termed “smaller.” For each group, growth rates are compared with preprogram growth; the exercise is repeated for time horizons of one, two, and three years (that is, for each program year).

The results of this procedure are presented in Figure 6.9. Two points stand out. First, regardless of whether disinflation occurs, and regardless of the degree of disinflation, there is no tendency for growth to slow. To the contrary, most of the calculations indicate a sizable increase in growth, whatever the inflation outcome. Second, there is no clear (bivariate) relationship between the extent of disinflation and changes in growth. Nor is there an apparent tendency for countries avoiding disinflation to do better than those experiencing some disinflation: changes in growth in the two groups are similar at one- and two-year horizons, while in the third program year those disinflating tended to do much better. Among the cases experiencing some disinflation, the evidence is mixed but leans toward those with smaller disinflations doing better on growth.23 Yet this is a rather mild form of Phillips curve: even for cases of stronger disinflation, growth performance is on average sustained above preprogram levels.

Figure 6.9Increase in Growth Rate During Programs with Intermediate Initial Inflation, by Degree of Disinflation Achieved1

Source: IMF staff estimates.

1Figures in parentheses above bars refer to the number of observations. The degree of disinflation is measured as ratio of inflation rate to initial inflation; cutoff points of 0.9 and 0.5 are used to distinguish between “some” and “none,” and “greater” and “smaller,” respectively. The change in growth rate is measured relative to average growth in the three years prior to program.

Of course, such a simple, bivariate analysis is not intended to address decisively the issue of output costs of disinflation.24 It is enough, however, to document the absence of a “stylized fact’’ that disinflation is usually associated with contraction of output—or even with a slowing of growth—in SAF/E SAP-supported programs.

Targeting Low Inflation: A Loss to the Public Sector of Significant Seigniorage?

The other traditional explanation for economies’ not reducing inflation to low levels involves the public sector’s loss of seigniorage, measured here as the increase in reserve money in a given year, scaled by either GDP or government revenues.25

Some seigniorage will of course arise from the growth of demand for real balances that accompanies output growth. Such seigniorage is non-inflationary, but it is likely to be fairly small—perhaps no more than ½ of 1 percent of GDP.26 In the year prior to SAF/ESAF-supported programs, seigniorage often extended (outside the CFA countries) much beyond its likely noninflationary component (Figure 6.10). Seigniorage exceeded 2 percent of GDP in almost half the cases. The importance to the public sector of seigniorage is perhaps better illustrated by considering seigniorage as a share of total government revenues (including seigniorage). In many cases, this share exceeded 10 percent; in a few cases it was nearly one-third.

Figure 6.10Distribution of Seigniorage Prior to SAF/ESAF-Supported Programs (Excluding CFA Cases)1

Source: IMF staff estimates.

1Increase in reserve money as percent of GDP (top panel) or government revenues including seigniorage (bottom panel) in the year preceding the SAF/ESAF arrangement. Cases in which reserve money contracted are included in the left-most grouping of each distribution.

How much seigniorage might policymakers lose by moving to low inflation? The answer depends not only on initial inflation, but also on how far the base for taxation through inflation has eroded as a result of inflation. In a cross-sectional examination of countries in the preprogram year (Table 6.3), seigniorage tends to be greatest as a share of GDP when inflation is toward the higher end of the intermediate range, around 30–40 percent annually. This suggests an incentive, from the perspective of the public sector, to disinflate when inflation is high: there is then the possibility of reducing inflation with little loss—and perhaps a gain—of seigniorage. In addition, disinflation from high levels may boost nonseigniorage revenue collection, through reversal of the Tanzi effect.

Table 6.3Seigniorage and Financial Depth Indicators Prior to SAF/ESAF Programs(Median values in last year prior to program; in percent of GDP unless otherwise noted)
Country GroupingInflation (percent)SeigniorageReserve Money (average)Broad Money (average)
Percent of GDPPercent of revenues
Low (non-CFA) inflation6.31.36.78.326.6
Intermediate inflation21.22.19.67.619.2
10–20 percent14.61.56.28.719.1
20–30 percent26.71.94.77.121.5
30–40 percent34.63.514.87.418.6
High inflation170.72.518.65.09.1
Sources: IMF, International Financial Statistics (various years); and IMF staff estimates.
Sources: IMF, International Financial Statistics (various years); and IMF staff estimates.

However, the picture is much different for cases of intermediate inflation. If the level of inflation that would yield maximum seigniorage is not much less than 40 percent, then ending intermediate inflation would imply a loss of seigniorage accruing to the public sector. Possible losses may be considerable: in this sample, inflation in the higher end of the intermediate range yields seigniorage with median values of 3½ percent of GDP and nearly 15 percent of public sector revenues. If inflation were brought down to 5 percent, seigniorage of perhaps no more than about 1 percent of GDP could reasonably be expected, implying a substantial adjustment for the public sector.

This analysis needs to be interpreted with care, in several respects. First, there is no necessary presumption that policymakers seek to maximize seigniorage—they are likely to be aware of the various costs and drawbacks of inflation. Nevertheless, seigniorage considerations may partially explain why moving from high to intermediate inflation may be easier—and therefore more likely—than moving from intermediate to low inflation. Second, it is not clear that lost seigniorage should be considered a “cost” of disinflation in the same sense that short-run losses of aggregate output would be; seigniorage is more of a distributional issue. The public sector’s loss of inflationary seigniorage represents a gain for holders of domestic currency who have paid an inflation tax. Of course, to give up seigniorage the public sector generally will have to cut its expenditure or increase other forms of taxation.27 Such offsetting measures will have their own distributional consequences—which may reinforce or run counter to the positive distributional effect of lower inflation—and they therefore may be difficult to implement. It is also conceivable that all feasible expenditure cuts or tax increases would have worse effects on growth than the inflation tax they are designed to replace, implying that the prevailing rate of seigniorage (and inflation) is, in that sense, optimal. It seems doubtful, however, that there is so little scope for efficient tax or expenditure reform in the countries under review that this would be an important constraint in practice.

Determinants of Disinflation in Programs

The evidence presented so far mounts a rather strong case for aiming for inflation within the single-digit range. Why then has the record of disinflation not been better in SAF/ESAF-supported programs? Slippages from program policy commitments are an obvious source of inflation overshooting. It is also possible that there have been more technical or strategic explanations of disappointing disinflation outcomes: that is, weaknesses in quantitative program design or in the strategies and tools used to combat inflation. This section considers the record in SAF/ESAF-supported programs with a view to finding the sources of pervasive inflation overshooting and drawing lessons for policy, No attempt is made to cover all possible determinants of inflation in all programs; instead, the focus is on the 35 programs that attempted to disinflate from intermediate or high inflation.28 The role of slippages from policy commitments is recognized, but emphasis is placed on a more fundamental question: where policy commitments were respected, was the design of programs likely to produce the disinflation targeted?

Slippages with respect to essential fiscal and credit policy commitments were indeed frequent. In addition, there was a strong tendency of money growth to exceed projections, but this was not a deviation from commitments, since only one program had a limit on money growth. Fortunately, policy commitments were respected in enough programs to assess whether these were usually adequate to achieve disinflation objectives, and the causes of pervasive money and inflation overshooting in these programs are investigated in what follows. Overall, the evidence suggests that programs targeted insufficient fiscal adjustment; moreover, where such underlying adjustment was adequate, inflation performance may have suffered from the absence of a nominal anchor.

Disinflation Strategies

To reduce inflation, two policy commitments received priority as clearly necessary conditions. First, programs sought to limit the contribution of credit or net domestic assets (NDA) to monetary growth. Second, fiscal restraint was planned that would reduce domestic demand and also help to contain the growth of NDA. Fiscal restraint should also have an effect on expectations—a credible fiscal adjustment strategy for the medium term should remove an important source of inflationary expectations.

A further consideration in disinflation strategy is the use of a nominal anchor. The 35 programs beginning from intermediate or high inflation almost universally made use of a credit (NDA) limit, but in only one case was there a formal limit (performance criterion) on a monetary aggregate.29 An NDA limit does not in itself constitute a nominal anchor, except in the case of a pure exchange rate float (a commitment to zero official intervention in the foreign exchange market, or to intervention aimed only at fulfilling predetermined targets for international reserves), no example of which is found among the countries under review.

Nor were the exchange rate or wage controls widely used as nominal anchors. There were only six programs that made use of an exchange rate anchor (Lesotho, 1988 and 1991; Nepal, 1987 and 1992; Nicaragua, 1994; and Zimbabwe, 1992) among the 35 programs surveyed. In seven other cases, exchange rate policy was set to act in a manner quite contrary to an anchor: in these programs, the authorities’ stated objective implied that they would adjust the nominal rate in response to actual inflation developments. Such pursuit of a targeted level of the real exchange rate is likely to make disinflation more difficult or even improbable. As for wage policy, one program (Albania 1993) involved an economy-wide attempt to use wage policy as an anchor.30

In principle, an inflation target itself could also serve as a nominal anchor. Although nearly all the programs surveyed noted a specific rate of inflation—and these figures are herein referred to as inflation “targets”—in practice these were not the sort of actively pursued, hard objectives that could constitute an anchor. Thus, initial inflation targets were systematically overshot throughout three-year programs, and there was no pattern of financial policy commitments being tightened in response to such overshooting (if anything, fiscal policy objectives tended to be relaxed as programs proceeded—see Chapter 4). Nor did inflation overshooting itself precipitate program interruptions: as noted earlier, programs that avoided interruption had inflation performance no better than that of other programs.

Thus, the large majority of programs (27 of 35) attempting disinflation from intermediate or high initial inflation relied only on an NDA limit and were without a nominal anchor; indeed, one in five programs explicitly had an exchange rate policy set to pursue a real exchange rate objective.31 However, there are signs of a (subtle) shift over time toward using anchors. Among the programs begun in the last three years of the survey period, there were no further (explicit) instances of real exchange rate targeting, and both cases of monetary and wage-based anchors began recently, in 1993 and 1994.

Why Were Inflation Targets Missed?

Within the general disinflation strategy that typically relied on demand restraint through fiscal and credit ceilings, usually without a nominal anchor, there are several possible explanations of the pervasive inflation overshooting.

  • Programs projected an undue expansion of real money demand.
  • Policy commitments were not implemented as programmed.
  • Programs placed insufficient emphasis on fiscal adjustment in their policy mix.
  • Policy commitments were not adequately defined nor sufficiently extensive.

This last possibility deserves elaboration. Standing by themselves, credit limits have no particular implication for either the price level or the stock of money, allowing a variety of policy behaviors and outcomes in which both money and inflation exceed program targets. Three stylized scenarios illustrate the possibilities. First, if NDA growth is slowed according to program policy commitments, room can be created for aggressive foreign exchange market intervention (that is, reserve accumulation substantially in excess of program targets) without accelerating inflation—but also without achieving the programmed disinflation. Second, a real exchange rate rule might be pursued, with obvious potential for inflation overshooting. Third, the authorities’ contribution to inflation overshooting might be more passive: if inflation fails to slow as targeted (reflecting inflation inertia and, possibly, low policy credibility), both the price level and nominal money demand will exceed program projections. While credit growth is limited to its original target, foreign exchange inflows may be pulled in to validate the inflation overshooting. In all three scenarios, program outcomes differ from program projections, even as program policy commitments are respected.

Did Programs Overestimate Real Money Demand?

The most obvious symptom of an overestimation of real money demand would be a tendency for inflation to exceed the target even when money targets were met. However, it is not feasible to examine this possibility for the cases of intermediate or high initial inflation, since broad money targets were nearly always exceeded, often substantially. Another symptom might be that when money targets are exceeded, inflation targets tend to be exceeded disproportionately, so that real money balances turn out to be smaller than projected (of course, such evidence would not be conclusive, since money growth in excess of program projections may itself weaken real money demand).32Figure 6.11 plots deviations from inflation targets against deviations from broad money targets for the first program year; it does not suggest a systematic tendency for inflation overshooting to exceed money overshooting. Since there is no prima facie evidence that frequent overshooting of inflation usually reflected overestimation of real money demand, it will be necessary to explain why and how money, as well as inflation, has tended to overshoot projections.

Figure 6.11Deviations from Target: Inflation Against Money Growth

(First program year; programs with intermediate and high initial inflation)1

Source: IMF staff estimates.

1Outlying observation at (169.6, 169.1) not shown.

Were Policy Commitments Implemented?

Program targets for both the fiscal balance and NDA were often exceeded, although slippages were far from universal. Thus, first-year program targets for the focal balance were missed in about half of the intermediate and high inflation cases. On average, the fiscal balance turned out to be about ½ of 1 percent of GDP weaker than programmed. NDA over-shooting was also common. Figure 6.12 indicates median NDA growth in all three program years; this was reduced significantly in high-inflation programs, but very little in intermediate-inflation cases.33 Although overshooting of NDA targets was far from universal, it did occur in 60 percent (21 of 35) of the first-year programs involving intermediate or high initial inflation. Among these 21 cases, the expansion of NDA in excess of target was usually large (the median excess being almost 15 percentage points), and it was associated with money overshooting in all but one case. As might be expected, there was some correspondence between NDA and fiscal policy performance relative to program projections. Thus, among the programs in which first-year NDA limits were respected, there was on average a slight overperformance with respect to the fiscal balance; when NDA limits were exceeded, the fiscal balance was on average weaker than projected, by about 1½ percent of GDP.

Figure 6.12Net Domestic Assets (NDA) Growth: Targets and Outturns1

(Median values; in percent of beginning-period broad money)

Source: IMF staff estimates.

1Figures in parentheses above bars refer to the number of observations.

2Median target not shown because of insufficient data.

As was evident from Figure 6.11, program projections for money growth were nearly always exceeded, often by large margins. Indeed, for the first program year, such projections were exceeded in 33 of the 35 cases of intermediate or high initial inflation; the median overshoot was about 15 percentage points.34Figure 6.13 provides a perspective on money outcomes in all three program years.35 For the cases with intermediate inflation, it is striking that there was on average no deceleration of money growth during the course of the programs, in contrast to program objectives. However, this pervasive monetary overshooting cannot itself be considered a deviation front program policy commitments, since these universally focused not on money but on NDA. About one-third of the cases of money overshooting reflected only NDA excesses, with net foreign assets (NFA) either at or below projection.

Figure 6.13Broad Money Growth Projections and Outturns1

(Median values; in percent)

Source: IMF staff estimates.

1Figures in parentheses above bars refer to the number of observations.

2Median target not shown because of insufficient data.

Adequacy of Program Design for Disinflation Targets

In the subsample of programs in which key financial policy commitments were respected during the first program year, it is convenient to focus first on the adequacy of NDA targets, and then assess the role of fiscal targets.

Were outcomes for money growth and inflation superior in programs where NDA ceilings were respected? The answer is generally no. Keeping NDA within program limits was strongly associated with faster-than-projected growth of NFA, and therefore was not associated with keeping money growth under projection: in the first program year, the money growth rate exceeded its target rate (which averaged about 20 percent) in 13 of 14 cases, by an average of 16 percentage points. This fact is illustrated in Figure 6.14, where almost all observations fail above the 45-degree line (money exceeds projections), including those to the left of the vertical (NDA is at or below target). Indeed, the figure shows a negative association between NDA and NFA relative to target—suggesting that, over a one-year horizon, NDA performance had little relationship to performance of money relative to target.36

Figure 6.14Deviations from Target: Net Foreign Assets (NFA) Against Net Domestic Assets (NDA)

(First program year; programs with intermediate and high initial inflation)1

Source: IMF staff estimates.

1Excluding three outlying observations.

Similarly, inflation targets were significantly exceeded in 8 of the 14 cases where NDA targets were respected, or about as frequently as when NDA ceilings were exceeded.37 Moreover, the magnitude of the average deviation from inflation targets was not minor (Table 6.4), and for the intermediate-inflation cases (which represent most of the sample) inflation performance was particularly weak. On the other hand, among the smaller sample of high-inflation cases where NDA targets were met, the degree of inflation overshooting was modest relative to the initial high inflation. Overall, the evidence does not support a strong positive short-run relationship between NDA and inflation performance.

Table 6.4Deviations from SAF/ESAF Program Targets, by Performance of Net Domestic Assets (NDA)(Cases of intermediate and high initial inflation, first program year)
Deviation from Target, as Percent of GDP
Deviation from Targeted Percentage ChangeCurrent account balanceNet private capital, and balance of payments errorsPublic external borrowing
InflationBroad moneyNDANFA1
NDA Growth Limit Respected
(14 cases)
Mean9.016.2-22.538.72.47.1-2.1
Median3.116.1-7.224.71.34.9-1.9
Of which
Intermediate inflation (9 cases)
Mean11.711.9-22.934.90.53.8-0.1
Median6.06.2-7.417.20.55.2-1.7
High inflation (5 cases)
Mean4.724.0-21.745.75.512.5-4.9
Median1.522.3-2.624.91.42.6-5.4
NDA Growth Limit Exceeded
(21 cases)
Mean15.326.953.8-26.91.11.8-0.8
Median4.914.614.5-6.50.60.0-0.7
Of which
Intermediate inflation (17 cases)
Mean4.413.130.4-17.31.02.8-0.2
Median4.413.69.5-5.41.30.2-0.6
High inflation (4 cases)
Mean61.585.7153.1-67.41.7-2.0
Median44.876.9152.1-57.5-0.7-3.6
Source: IMF staff estimates.

Net foreign assets.

Source: IMF staff estimates.

Net foreign assets.

Within the balance of payments, what accounted for the larger-than-projected NFA growth that was observed in nearly every case where NDA limits were respected? It did not generally reflect public sector external borrowing: on average, such borrowing turned out to be considerably less than projected (see Table 6.4).38 In contrast, the external current account seems to have played some role. In about half of the relevant cases, the current account turned out to be stronger than projected. In these instances, the overperformance was considerable, averaging about 5 percent of GDP. Similarly, private capital inflows appear to have been a factor. Because data are often incomplete and unreliable, the residual balance of payments category “net private capital inflows, including net errors and omissions” was constructed as a proxy. Among programs where NDA targets were respected, excesses over implicit projections for such inflows occurred in about three-fourths of such cases. The average deviation from projections was positive and quite sizable. Since this finding is based on a residual measurement, there can be no certainty that it represents shifts in net private capital movements (rather than shifts in errors in other balance of payments categories). Of course, neither current account nor private capital flows should be presumed to cause inflation overshooting. As likely is that such overshooting was driven by inflation inertia and faltering policy credibility, with foreign exchange inflows occurring only as an accommodating response after NDA were tightened.

That inflation typically overshot targets even when NDA limits and fiscal targets were respected suggests a weakness in program design. A strong possibility is that fiscal targets were insufficiently tight to achieve the projected disinflation. Although there is no evidence that monetary programs were set too loose, it may be that the fiscal -monetary policy mix gave too little emphasis to fiscal adjustment. It is difficult to know ex ante the amount of fiscal adjustment needed to achieve a given inflation objective. Fiscal deficits of varying sizes can fit inside a plausible financial program, and an ex ante assessment of fiscal adequacy would require information not readily available (such as the extent to which a given fiscal tightening will reduce absorption of nontraded goods). Ex post, any instance of inflation overshooting at least raises the suspicion that fiscal policy was insufficiently tight, insofar as more fiscal adjustment would likely have resulted in lower inflation.

Indeed, a direct look at program outturns, irrespective of program targets, confirms the association of greater fiscal adjustment with success in disinflation. Among three-year programs beginning from intermediate inflation, the distribution of changes in the overall fiscal balance was bi-modal, with a clear split between groups with and without a significant fiscal adjustment.39 Average disinflation in the programs with a significant fiscal tightening was much greater than in the other group (Figure 6.15).40 Moreover, all of the programs reducing intermediate inflation by half or more were in the group with significant fiscal adjustment. Still, not all programs making such adjustment were successful in reducing inflation. This pattern indicates that a strong fiscal tightening was a necessary but not sufficient condition for disinflation from intermediate levels.41 There is also some sign that fiscal adjustment leads to disinflation with a lag, highlighting the importance of a sustained effort.42

Figure 6.15Disinflation from intermediate Levels During Programs, by Degree of Fiscal Adjustment1

Source: IMF staff estimates.

1Figures in parentheses above or below bars refer to the number of observations.

2Degree of disinflation is measured by comparing inflation in the third program year to inflation in the preprogram year.

3Change in fiscal balance is measured using three-year averages (preprogram and program).

A second possible problem in program design is the infrequent use of a nominal anchor. Since only 1 of the 35 programs surveyed had a money limit rather than an NDA limit, the effectiveness of a money anchor cannot be judged from experience. However, comparisons can be made between programs that employed nominal anchors and those that did not. These should, however, be interpreted with caution because the anchor group has only eight programs for six countries, and there are inevitable questions of causality.43

Two basic observations can be made on relative inflation performance. First, there is some sign that the minority of program cases with specified nominal anchors had better inflation performance than others. Among the 35 cases of intermediate or high initial inflation, the 8 programs with such anchors in place at their outset tended to come closer to their inflation targets than did other programs; irrespective of program targets, the cases with anchors also tended to achieve a greater reduction in inflation (Table 6.5). This performance differential is particularly large during the first and second program years; the evidence is mixed for the third program year.

Table 6.5Inflation Performance, by Use or Nonuse of Nominal Exchange Rate Anchors(Ratios in percent; cases of intermediate and high initial inflation)
Cases With AnchorCases Without Anchor
tt+ 1t + 2tt + 1t + 2
Inflation outturn divided by target:
Mean111.9125.6172.8169.6206.0261.2
Median96.1127.2203.0135.8172.5192.1
Number of observations888262323
Inflation outturn divided by t - 1 inflation:
Mean72.651.952.3118.580.566.2
Median73.558.963.981.172.752.6
Number of observations888262625
Source: IMF staff estimates.
Source: IMF staff estimates.

Second, among the programs having no specified nominal anchor, those committing to a real exchange rate objective might be expected to have worse inflation performance than others.44 For the first and third program years, the evidence supports this view; the size of the inflation performance differential in the final year is especially striking. In the cases with real exchange rate targets, a slight disinflation was achieved in the first program year, and little if any progress was made thereafter (Table 6.6).

Table 6.6Inflation Performance Without Nominal Anchors, by Type of Exchange Rate Policy(Ratios in percent; cases of intermediate and high initial inflation)
Cases With Real Exchange Rate ObjectiveOther Cases Without Anchor
tt + 1t + 2tt + 1t + 2
Inflation outturn divided by target
Mean224.1182.1343.7149.5214.4232.1
Median137.9156.7318.1133.0175.3153.3
Number of observations766191717
Inflation outturn as share of t - 1 inflation
Mean88.875.684.2127.481.960.5
Median88.185.794.475.165.742.4
Number of observations666202019
Source: IMF staff estimates.
Source: IMF staff estimates.

An important consideration in an anchor strategy is the possible adverse effects on the external position. If the 6 cases with exchange rate anchors are compared with the 20 other intermediate inflation cases without anchors (Table 6.7), it does not appear that the external position suffered markedly from the constraints imposed by an exchange rate anchor, at least over a three-year horizon. In the first program year, cases with exchange rate anchors appear to be associated with a greater acceleration of import growth, a larger increase in the current account deficit, and smaller reserve accumulation. However, by the second and third years, reflecting mainly a deceleration of import growth in the anchor group, no clear difference in overall external performance is apparent: improvements in the current account and reserve coverage in “anchored” programs are comparable to those in the programs with flexible regimes. (The anchor cases tended to begin their programs from a stronger reserve position, and this may have made the early weakening of their external performance more acceptable for them.) Even anchored programs were associated with real depreciation, albeit considerably less than among the other group. Export volume growth in the anchor group does not compare unfavorably, in any of the program years.

Table 6.7Aspects of External Performance, by Use or Nonuse of Exchange Rate Anchors(Cases of intermediate initial inflation)
Cases With Exchange Rate AnchorCases Without Any Anchor
Level in t - 1Change from year t - 1Level in t - 1Change from year t - 1
tt + 1t + 2tt + 1t + 2
Reserves, months of imports
Mean3.1-0.40.20.82.10.20.60.4
Median2.5-0.50.01.41.60.40.50.6
Number of observations666520202017
Current account balance, percent of GDP
Mean-12.2-5.0-0.3-1.5-6.9-1.5-1.40.0
Median-7.8-4.91.51.0-4.9-1.1-2.2-0.6
Number of observations666520202017
Export volume growth, in percent
Mean5.018.72.44.82.57.30.99.7
Median2.616.56.720.80.95.01.08.4
Number of observations666516161615
Import volume growth, in percent
Mean3.510.5-1.84.04.62.13.8-1.1
Median3.97.8-5.51.75.74.54.12.7
Number of observations666516161615
Real effective exchange rate index, percent appreciation
Mean-3.4-5.4-6.5-5.8-8.8-13.8
Median-2.7-4.6-6.7-0.8-5.8-12.4
Number of observations666171715
Memorandum
Fiscal balance, percent of GDP
Mean-4.9-1.20.21.0-5.50.30.61.4
Median-5.8-0.41.0-0.1-4.40.61.41.1
Number of observations666620202018
Source: IMF staff estimates.
Source: IMF staff estimates.

Main Policy Conclusions

SAF/ESAF-supported programs have seen some countries succeed in moving from high to intermediate inflation, but the record of achieving low (single-digit) inflation is quite mixed. Many three-year programs (outside the CFA countries, almost two-thirds) have ended with intermediate or higher inflation, in contrast to their target of reaching inflation in the upper single-digits. Because the limited progress during programs is often preceded by a run-up of inflation, and because disinflation often stalls after the final program year, many countries under review still had intermediate or higher inflation even by the mid–1990s.

This chapter has argued that disinflation, ending not only high but also intermediate inflations, is indeed a worthwhile objective of ESAF-supported programs. The evidence supports a long-run negative association between inflation and growth that is robust once the relationship is allowed to take a (plausible) nonlinear form. A key implication of such nonlinearity is that moving from intermediate to low inflation may be just as important for growth as moving from high to intermediate inflation. Notwithstanding inevitable uncertainties about the nature and magnitude of the links involved, estimates of the effects of inflation are all large enough to create a reasonable presumption that ending intermediate inflations is important. Moreover, disinflation may result in a more equal distribution of income.

The experience from SAF/ESAF-supported programs also indicates that, contrary to popular perception, disinflation can be compatible with improved growth performance even in the short run. Even countries that cut their inflation rates by half or more during programs had growth that on average was not only positive but faster than preprogram growth. Furthermore, on average those programs experiencing some disinflation did as well as or better than countries avoiding disinflation. Apparently, any contractionary effects of disinflation tended to be offset by other, positive influences on growth.

For the countries under review, the loss of appreciable seigniorage received by the public sector may be a serious obstacle to disinflation. Particularly in countries with inflation in the intermediate range, seigniorage is often a considerable source of revenue for the public sector. Adjustment to low inflation requires making difficult decisions to scale back expenditure objectives or raise other forms of revenue. This being said, it is likely that the countries concerned have scope to reform the public finances in ways that improve resource allocation, and hence help rather than hinder economic growth.

Greater success in disinflation would be promoted by change on three fronts. First, and rather obviously, the frequency of slippages from financial policy commitments must decline. Second, the evidence suggests, albeit indirectly, that programs’ fiscal objectives may not always have been consistent with their inflation objectives, and therefore that greater fiscal adjustment may need to be targeted (and realized) for a given inflation objective. Third, an important additional contribution could come from expanding the use of nominal anchors when the fiscal adjustment envisaged is sufficiently strong. Most SAF/ESAF-supported programs attempting disinflation had no nominal anchor. In practice, when fiscal and credit targets were met, programs tended to see money growth and inflation exceed targets. The minority of programs using a nominal anchor (exchange rate pegs, mostly) in an effort to reduce high or intermediate inflation had superior inflation performance; in contrast, programs managing the exchange rate to pursue a competitiveness objective were associated with little disinflation even after three years.

Among anchors, nominal exchange rate targets are the logical first choice, since they can provide a transparent and unambiguous signal to support the disinflation effort. Many of the countries under review now have a level of official reserves that would not obviously disqualify’ them from such an approach. In the rather limited program experience with exchange rate anchors to reduce high or intermediate inflation, some aspects of external performance tended to be weaker than in other programs, but only early on. Moreover, export volume growth did not appear to suffer in these programs. The vulnerability of many ESAF countries to external shocks will often argue against exchange rate pegs, but they might still be considered as a transitional device. In such circumstances, the risk of introducing a potential rigidity would need to be carefully considered before introducing an exchange rate anchor, and flexibility in exiting the arrangement would be essential.

Wider use of monetary anchors, especially as a onetime device to support the transition to single-digit inflation, also deserves consideration, although it too would not be without drawbacks. As performance criteria in IMF-supported programs, money limits could give rise to operational problems when real money demand exceeded projections. Money anchors would preclude the sustained pursuit of a disequilibrium real exchange rate target, but they would also limit the scope for countering pressures for real exchange rate appreciation more generally. This problem underlines the importance of sufficiently strong and responsive supporting policies.

In situations in which sufficiently strong policies seem beyond reach, at least in the near term, program objectives, not only for inflation but probably also for growth, should be more modest than they have been in the period reviewed. A more realistic assessment of inflation and growth prospects would safeguard program credibility and provide guidance to policymakers about the extent of adjustment needed to place their country on a higher growth trajectory.

References

    BallLawrence1993“What Determines the Sacrifice Ratio?,”NBER Working Paper 4306 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    BallLawrenceN. GregoryMankiw and DavidRomer1988“The New Keynesian Economics and the Output-Inflation Tradeoff,”Brookings Papers on Economic Activity: I pp. 182.

    • Search Google Scholar
    • Export Citation

    BarroRobert1995“Inflation and Economic Growth,”Bank of England Quarterly BulletinVol. 35 (May) pp. 16676.

    BlejerMario I. and IsabelGuerrero1990“The Impact of Macroeconomic Policies on Income Distribution: An Empirical Study of the Philippines,”The Review of Economics and StatisticsVol. 72 (August) pp. 41423.

    • Search Google Scholar
    • Export Citation

    BosworthBarrySusanCollins and Yu-ChinChen1995“Accounting for Differences in Economic Growth,”Brookings Discussion Papers in International Economics No. 115 (Washington: Brookings InstitutionOctober).

    • Search Google Scholar
    • Export Citation

    BrunoMichael and WilliamEasterly1995“Inflation Crises and Long-Run Growth,”NBER Working Paper 5209 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    BulířAleš and Anne-MarieGuide1995“Inflation and Income Distribution: Further Evidence on Empirical Links,”Working Paper 95/86 (Washington: IMF).

    • Search Google Scholar
    • Export Citation

    CukiermanAlexP.KalaitzidakisL.Summers and S.Webb1992“Central Bank Independence, Growth, Investment, and Real Rates,”Carnegie-Rochester: Conference Series on Public PolicyVol. 39 (December) pp. 95140.

    • Search Google Scholar
    • Export Citation

    De GregorioJosé1993“Inflation, Taxation, and Long-Run Growth,”Journal of Monetary EconomiesVol.31 (June) pp. 27198.

    DornbuschRudiger and StanleyFischer1993“Moderate Inflation,”World Bank Economic ReviewVol. 7 (January) pp. 144.

    EasterlyWilliam1996“When Is Stabilization Expansionary? Evidence from High Inflation,”Economic PolicyVol. 22 (April) pp. 67107.

    • Search Google Scholar
    • Export Citation

    FischerStanley1982“Seigniorage and the Case for a National Money,”Journal of Political EconomyVol. 90 (April) pp. 295313.

    • Search Google Scholar
    • Export Citation

    FischerStanley1993“The Role of the Macroeconomic Factors in Growth,”Journal of Monetary EconomicsVol. 32 (December) pp. 485512.

    • Search Google Scholar
    • Export Citation

    JudsonRuth A. and AthanasiosOrphanides1996“Inflation, Volatility and Growth,”Board of Governors of the Federal Reserve System Finance and Economics Discussion Series No. 96–19 (Washington).

    • Search Google Scholar
    • Export Citation

    LevineRoss and SaraZervos1993“Looking at the Facts: What We Know about Policy and Growth from Cross-Country Analysis,”Policy Research Working Paper 1115 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    SarelMichael1996“Nonlinear Effects of Inflation on Economic Growth,”Staff PapersInternational Monetary FundVol. 43 (March) pp. 199215.

    • Search Google Scholar
    • Export Citation

    SchadlerSusanAdamBennettMariaCarkovicLouisDicks-MireauxMauroMecagniJamesMorsink and MiguelSavastano1995aIMF Conditionality: Experience Under Stand-By and Extended Arrangements Part I: Key Issues and Findings Occasional Paper 128 (Washington: IMF).

    • Search Google Scholar
    • Export Citation

    SchadlerSusanAdamBennettMariaCarkovicLouisDicks-MireauxMauroMecagniJamesMorsink and MiguelSavastano1995bIMF Conditionality: Experience Under Stand-By and Extended Arrangements Part II: Background Paper Occasional Paper 129 (Washington: IMF).

    • Search Google Scholar
    • Export Citation
1“Inflation” refers to the annual change of a country’s consumer price index (in a few cases, the GDP deflator is used either because of the absence of CPI data or concern over its quality). Where available, end-period data are used rather than a period-average measure. For low or stable rates of inflation, there is often little difference between such measures; in cases of higher inflation the end-period measure is often available.
2Both Albania and Cambodia suffered a partial rebound of inflation in 1996.
3“Initial inflation” refers to inflation in the last year before a program, which is denoted as “year t- 1.” The idea of measuring initial inflation as the average rate of inflation over the three years preceding a program was rejected because a number of countries with low or moderate inflation in the year or two before a program would be misleadingly classified as having high initial inflation.
4Here, and throughout this chapter, “case” refers to a program case, rather than a country.
5Inflation performance is here assessed over the three program years originally envisaged at the time of approval of the first annual arrangement; that is, any subsequent delays in approval of second and third annual arrangements are disregarded.
6For more recent successful disinflations, it is too early to assess medium-term performance.
7For example, in the short-run, supply shocks can make inflation and output move in opposite directions, whereas demand shocks can make them move together.
8This issue has practical significance—deflation occurred in some countries under review.
9The sample included annual data from 105 (nontransition) developing countries for 1981–95; roughly one-third of the data were from SAF/ESAF countries. The sample definition is thus similar to that analyzed in Chapter 5, but in this bivariate context only 11 annual observations had to be dropped because data were missing.
10This group size confines all observations with negative inflation rates to one group.
11For legibility, it is necessary to exclude from the data the point for the highest-inflation group (mean inflation over 1,000 percent, mean per capita growth of –2.5 percent).
12If a flattened but still statistically significant negative association nevertheless did emerge from such an analysis, its statistical significance might well depend on the presence of a few outlying cases of extreme inflation and low growth.
13In contrast, we found that log of (1 + inflation/100), the data transformation used in several studies, left the relationship still with a clearly convex form.
14The positive slope found for inflation below about 5 percent could suggest benefits of raising inflation, where it is very low, to such a level. However, this finding should not be taken at face value: it could be related to the circumstances of the CFA countries during the 1980–95 sample period. A significant positive slope was not a feature of Sarel’s (1996) results on a sample that included industrial economies and earlier data (and therefore more low-inflation observations).
15The reasons for Sarel’s result being stronger are not obvious, but a number of differences from the present study can be noted: a sample including advanced as well as developing economies; a sample period of 1970–90, with data averaged over four periods of five years; and various differences in specification of the growth regression equation.
16When the multivariate regression model of Chapter 5 was run without taking the log of inflation or using a spline, the estimated coefficient was only –0.00077, which would suggest that intermediate inflation is associated with only very slightly lower growth. Nevertheless, this coefficient was statistically significant at the 5 percent level (t-statistic of –2.65), a result that presumably hinges on the presence of the high-inflation, low-growth outliers.
17For instance, certain aspects of direct tax systems can create channels through which inflation has a progressive influence.
18Buiíř and Guide note that empirical work has focused on the relationship between growth and income distribution; furthermore, work on the link with inflation has focused on more advanced economies. Blejer and Guerrero’s (1990) work on the Philippines is an exception.
19Of course, the “net” effect of disinflation on income distribution may depend on how the disinflation is achieved. (See also the discussion of the role of seigniorage, below.)
20In their case studies of persistent moderate inflations, Dornbusch and Fischer (1993) concluded that the output costs of disinflation were usually a more significant deterrent to disinflation than loss of seigniorage. This result may reflect that their sample of “moderate” inflation was defined to exclude inflation above 30 percent. Also, their case studies featured countries more developed than SAF/ESAF-eligible countries, and more developed countries may be more vulnerable to output costs of disinflation.
21Dornbusch and Fischer (1993) suggested that a lower level of development may make disinflation less costly, noting that Indonesia, the least developed country in their study, was the exception to their finding of costly episodes of disinflation from the 15–30 percent range.
22The timing of such beneficial effects may depend on their mechanisms. Effects through investment, for example, might have some lag. Where very high inflation has crippled intra-economy trade, however, ending inflation might very quickly boost output.
23The program projections themselves embodied this expectation: in cases in which stronger disinflation (from intermediate inflation) was targeted, a smaller (but still positive) pickup in growth tended to be projected.
24One type of bias in this analysis would occur if countries facing smaller costs of disinflation were the ones that chose to disinflate. The significance of this bias in practice would depend on the degree to which disinflation costs vary among ESAF users as well as on policymakers’ abilities to predict such costs reliably.
25This is the measurement methodology used in Fischer (1982). Because seigniorage is defined in terms of total base money issuance, it is clear that this measure is not limited to monetization of the fiscal deficit; in particular, it can also capture quasi-fiscal activities not picked up by conventional fiscal deficit measures, as well as the issuance of reserve money associated with the purchase of foreign reserves.
26For example, with reserve money equal to one-tenth of annual GDP, growth in real output and real money demand of 5 percent would allow noninflationary seigniorage of ½ of 1 percent of GDP. Faster output growth would permit greater noninflationary seigniorage, but the potential for such additional seigniorage is sharply constrained by the typically small size (even in low-inflation economies) of base money in comparison to GDP.
27A third possibility would be to replace lost seigniorage with increased external financing, allowing increased absorption for the economy as a whole. However, this path may not always be available, nor generally intertemporally sound.
28These 35 programs (involving 28 countries) comprise 4 in Asia, 5 in the Western Hemisphere, 5 in transition economies, and 21 in Africa. No programs in the CFA are in this group, since all such programs begun before 1995 had low inflation in year t - 1.
29Cambodia established an “indicative” reserve money limit for its first program year; in Uganda, program documents stated that the authorities would attempt sterilization if balance of payments inflows were “excessive” but did not specify a particular monetary limit.
30Two other programs had a “benchmark” limiting government wages only. The desirability of wage controls is unclear, however. See Schadler and others (1995b) for a critical discussion of their use in transition economies.
31Among programs supported by Stand-By and Extended Arrangements, the use of nominal anchors has been much more frequent—see Schadler and others (1995a, b).
32A rigorous investigation of money demand issues would be more complex. Here the intent is merely to demonstrate the absence of the basic evidence that might have indicated a need to pursue such issues further and more rigorously.
33Throughout this study, “NDA growth” refers to the change in NDA as a percent of beginning period money. Growth of net foreign assets is defined analogously.
34The two cases avoiding monetary overshooting were among the minority of programs with nominal anchors.
35Full, three-year monetary projections were specified in fewer than one in four of all programs; this is the principal reason for the section’s focus on the first program year.
36Measurement difficulties could contribute to such a pattern: given the identity M ≡ NFA + NDA, any error in measuring NDA that is uncorrelated with errors in measuring M must be negatively correlated with errors in NFA measurement. A related potential qualification should also be noted: NDA and NFA targets are sometimes adjusted when foreign borrowing by the public sector deviates from program projections. The NDA and NFA data here are compared with original targets, and this may explain some part of the observed offset between NDA and NFA deviations. However, cases of NDA undershooting accompanied by NFA overshooting are not generally attributable to greater-than-programmed external borrowing by the public sector.
37In two other programs, Lesotho (1991) and Bolivia (1987), inflation excesses were slight.
38This suggests that the poor inflation performance cannot be explained by the possible failure of some programs to adjust NDA targets downward for excesses in foreign disbursements, nor by the fact that the present analysis is based on original rather than (possibly) adjusted targets.
39Among the former group, the minimum improvement of the fiscal balance was 1.7 percent of GDP, and the median change was 3.1 percent. Among the latter group, the maximum increase in the fiscal balance was only 0.5 percent of GDP, and the median change was –0.6.
40A similar result was found using the primary fiscal balance to measure fiscal adjustment. In principle, adjustment of this more exogenous component of the overall balance can initiate a decline in inflation, which in turn reduces the government’s interest burden and thus further improves the overall fiscal balance. In this sample, however, there was on average no decline in interest payments as a share of GDP, even in the group characterized by strong adjustment of the overall balance and considerable disinflation.
41Among the smaller group of programs with high initial inflation, fiscal adjustment was usually significant, but there were several programs that succeeded in reducing inflation despite no significant improvement in the overall fiscal balance. All such cases, however, did reduce domestic financing of the budget strongly (by at least 2 percent of GDP).
42A look at contemporaneous, one-year changes in the fiscal balance and inflation did not show a clear association, in contrast to Figure 6.15, which makes an allowance for lagged effects. Thus, for Figure 6.15 we measured the change in the fiscal balance using three-year averages, for both the program period and the preprogram period, while measuring disinflation using the third program year (t + 2) and final preprogram year (t - 1).
43This being said, the findings of the present study are consistent with those reported in Schadler and others (1995a, b), which were based on a larger sample.
44A problem is that some programs may have followed a real exchange rate objective, continuously or at times, without prior announcement—the comparison here is based only on what was made explicit at the outset of programs.

    Other Resources Citing This Publication