Chapter 5: Economic Recovery and Growth

International Monetary Fund. Independent Evaluation Office
Published Date:
November 2003
  • ShareShare
Show Summary Details

The pace of economic recovery under an IMF-supported program is an important determinant of the impact of a program on welfare and is, therefore, the focus of considerable attention. To the extent growth performance is significantly below program projections, it may signal shortcomings in program design, including the fiscal stance. As pointed out in the introduction, one of the criticisms of program design in IMF-supported programs is that they impose unduly tight fiscal policies leading to adverse effects upon economic recovery. In this chapter, we examine these issues using the large sample of programs studied for this evaluation.

Economic Recovery in the Program Period: Outcomes and Expectations

Table 5.1 summarizes the short-term growth experience in 159 IMF-supported programs.1 It presents average annual growth rates actually achieved during preprogram and program years for the whole sample as well as for the subgroups used for this study. We note that comparisons between preprogram and postprogram growth recovery should not be understood as indicating the impact of the program, but only as a description of what happened. The impact of IMF-supported programs on growth can only be determined by comparing actual outcomes with the counterfactual of what would have happened to economic performance without a program. There is now an extensive, albeit inconclusive, literature on the topic but this area goes beyond the limits of this evaluation.2

Table 5.1Experience with GDP Growth Prior to and During Program Periods: Annual GDP Growth(In percent)
Number of

Trend in the

Prior Decade
All programs1591.
SBA/EFF (transition countries)34−2.1−
SBA/EFF (nontransition countries)613.
Of which
Capital account crisis cases104.82.9−5.04.7
Other programs513.
Source: WEO database.
Source: WEO database.

The following are the main features that emerge from the data:

  • For all programs taken together, the average GDP growth rate achieved in the first program year T improved upon the level in T−1, and then improved further in T+1 when it actually surpassed the average of the preprogram decade.

  • The same pattern is discerned in the two subgroups consisting of ESAF/PRGF and SBA/ EFF transition cases. However, the SBA/EFF nontransition group shows a somewhat different behavior, with the average growth rate decelerating sharply from 2.4 percent in T–1 to 0.9 percent in T.

  • The SBA/EFF nontransition subgroup itself consists of two very different types of programs. There are 10 programs in this group which relate to so-called capital account crises while the others relate to more conventional balance of payments problems.3 The capital account crisis cases experienced a collapse in output with average GDP growth falling sharply to –5.0 percent in T compared with 2.9 percent in T–1. This was followed by a recovery in T+1, which almost offset the decline in the previous year. The other 51 programs in this category show only marginal deceleration, with growth decelerating from 2.3 percent in T–1 to 2.1 percent in T, followed by a respectable acceleration to 3.5 percent in T+1.

The average growth rates presented in Table 5.1 suggest that the perception that IMF-supported programs are associated with strongly negative effects on growth is not well founded, except in the case of capital account crises.4 However, averages can be misleading because of variations around the mean, so we have also examined the distribution of programs and identified the percentage of programs which show a deceleration in growth compared with T–1 and those that show negative growth. Two different time horizons are used for the program period, a one-year horizon T and a two-year horizon covering T and T+1. The two-year horizon is perhaps more relevant since many programs only commence in the middle of T. The results are presented in Table 5.2.

Table 5.2Programs Showing Deceleration or Negative Growth
One-Year HorizonTwo-Year Horizon1
Percentage of programs showingPercentage of programs showing
DecelerationNegative growthDecelerationNegative growth
All programs42363618
SBA/EFF (transition countries)20481040
SBA/EFF (nontransition countries)
Of which
Capital account crisis cases80908040
Other programs47234116
Source: WEO database.

Programs are classified as indicating deceleration or negative growth on the basis of the annual average growth rates in the two-year period T and T+1. A negative average growth rate over two years means GDP in T+1 was lower than in T−1.

Source: WEO database.

Programs are classified as indicating deceleration or negative growth on the basis of the annual average growth rates in the two-year period T and T+1. A negative average growth rate over two years means GDP in T+1 was lower than in T−1.

Although the average growth rate of all programs did not decelerate (see Table 5.1), it is clear that a substantial percentage of programs in all subgroups experienced a deceleration in growth not only over a one-year but also over a two-year horizon. The number experiencing negative growth is much smaller and this phenomenon is concentrated in the group of transition and the capital account crisis cases. In the transition cases, the negative growth is actually a continuation of negative growth in the preprogram period (see Table 5.1). In the other two groups, ESAF/PRGF and SBA nontransition others, negative growth over a two-year horizon was experienced by only a small proportion (3 percent and 16 percent, respectively) of cases.

The actual recovery and short-term growth performance in the postprogram period also needs to be compared with GDP growth projections in programs. This comparison is important because the public perception of the success of programs is often assessed not just in terms of the actual outcomes but in terms of achievement relative to the growth targets. Furthermore, large shortfalls in growth relative to projections can generate consequential problems because fiscal targets built into programs may become inappropriate.

To compare projections with actuals we use the cumulative growth over T and T+1 as the basis for comparison (Table 5.3). The main conclusions are the following:

  • Actual growth fell short of projected growth over the two-year period and the average shortfall for all programs amounted to 1.5 percentage points. Except for the subgroup of transition countries, where the actual two-year achievement is basically the same as projected, all the other groups show underperformance on average.

  • The shortfall in the case of ESAF/PRGF is 1.5 percentage points, the same as for all programs. The shortfall in the SBA/EFF nontransition cases in turn reflects divergent behavior in the two subdivisions within this group. There was a massive underperformance of 6.4 percent in the case of the 10 capital account crisis cases. The other programs in the subgroup show only a relatively modest shortfall of 1.6 percent of GDP, which is close to the average of 1.5 percent for ESAF/PRGF.5

Table 5.3Envisaged and Actual Two-Year Cumulative Growth Rates over T and T+1(In percent)
Projected Cumulative

Actual Cumulative


(Actual - Envisaged)
All programs7.76.2−1.51
SBA/EFF (transition countries)
Of which
Capital account crisis cases5.8−0.5−6.41
Other programs7.35.7−1.61
Source: MONA and WEO databases.

Difference statistically significant at the 95 percent or better confidence level.

Source: MONA and WEO databases.

Difference statistically significant at the 95 percent or better confidence level.

Since there is considerable variation around the means reported in Table 5.3, it is useful to look at the distribution of programs according to the differences between actual and envisaged cumulative growth (Figure 5.1). About 60 percent of the cases show a shortfall. In about 25 percent of programs, the shortfall in cumulative growth over the two-year horizon exceeds 4 percentage points.

Figure 5.1.Distribution of Programs According to Differences Between Actual and Envisaged Cumulative Growth over a Two-Year Period (T and T+1)

Sources: MONA and WEO databases.

Table 5.4 also suggests that the degree of optimism about growth in T+1 depends upon what has happened in T. For all programs, the growth rate projected for T+1 was too optimistic by 1 percentage point. However, for those programs where growth was negative in year T (one-quarter of the overall sample), the growth projected for T+1 was subject to greater overoptimism (double the actual growth). Program projections of growth tend to build in greater optimism about recovery when starting from an adverse situation, probably reflecting an under-standable expectation of reversal to normality.

Table 5.4Indicators of Growth Optimism for T+1
A. Growth rate at T+1

(In Percent)
All arrangements4.63.6
Arrangements where
growth was negative
in year T3.41.7
Source: MONA and WEO databases.
B. Frequency of casees

(In Percent)
Percentage of cases
where growth rates
are reduced from
T to T+117.839.5
Percentage of cases
where growth in
T+1 is negative1.312.6
Source: MONA and WEO databases.
Source: MONA and WEO databases.

Even more striking is that programs are reluctant to project slowdown in growth from T to T+1, let alone to project negative growth. Only 18 percent of programs projected a slowdown in growth, whereas this happened in almost 40 percent of cases. Programs seldom project negative growth, although in reality it happens in about 13 percent of cases. Programs tend to underpredict significantly more situations of adverse output developments than situations of favorable output developments.6 This tendency must be seen in the context of the fact that program projections are not just the outcome of technical analysis but are negotiated outcomes and there are strong compulsions to present as optimistic a picture as possible. Nevertheless, it does suggest that the reluctance of programs to “call a downturn” means that the appropriate fiscal stance in such circumstances is not addressed in the original program design.

Optimism in Projecting Private Demand and Investment

There are many reasons why growth outcomes during the recovery phase might differ from projections. These include (1) exogenous factors turning out to be different from what was expected; (2) policies on which the growth projection was based may not be implemented effectively; (3) the projections may have been based on an inadequate understanding of the determinants of short-term growth leading to an inadequate design of policies; and finally (4) acceptance of an overoptimistic projection as an outcome of the program negotiation process. These factors must have operated to different degrees in different programs and it is beyond the scope of this evaluation to go into all these issues. However, there is one factor which may explain some of the optimism about growth in many cases, which can be examined with the available data, and this relates to the tendency to be overoptimistic in projecting investment, especially private investment.

Crisis situations are typically disruptive and introduce uncertainty about economic outcomes that can be expected to have a temporary negative effect on private investment. The rate of economic recovery may depend significantly on the pace at which investment activity goes back to normal. Unfortunately, the MONA database does not contain data on projected private investment in programs. However, it contains information on projected total investment rates and this can be used to examine the extent of overoptimism regarding total investment and its possible relationship with growth shortfalls.

Earlier IMF staff studies have documented that IMF-supported programs typically overestimate the speed with which investment will recover.7Table 5.5, which presents available information on actual and projected investment rates for the large sample and for the individual subgroups, confirms that there was overoptimism on average for all programs and the extent of optimism increases from T to T+1.8

Table 5.5.Investment Projections and Actuals Under IMF-Supported Programs, 1993–2001(Annual investment in percent of GDP)
All programs20.621.020.7−0.322.021.2−0.8
SBA/EFF (nontransition countries)22.623.322.4−0.923.921.7−2.22
Of which
Noncapital account22.223.122.6−0.523.821.6−2.22
Capital account24.524.121.2−2.9324.622.5−2.1
SBA/EFF (transition countries)21.121.620.1−1.5323.020.7−2.32
Source: MONA and WEO databases.

Difference is actual minus projected.

Significant at the 95 percent confidence level.

Significant at the 90 percent confidence level.

Source: MONA and WEO databases.

Difference is actual minus projected.

Significant at the 95 percent confidence level.

Significant at the 90 percent confidence level.

There are interesting differences in investment behavior among the various subgroups and its relationship with growth outcomes.

  • In ESAF/PRGF programs actual investment rates are slightly higher than program projections. Thus, the shortfall in growth experienced by this subgroup cannot be attributed to investment shortfalls.

  • In SBA/EFF transition cases, actual investment rates fall short of projected levels by 1.5 percentage points in T and 2.3 percentage points in T+1, but the GDP growth rates achieved are very close to projections. This suggests that other positive factors, possibly the pace and impact of structural change in these countries, must have been stronger than expected and offset the negative impact of investment shortfalls.

  • The subgroup of SBA/EFF nontransition cases shows significant investment shortfalls and as seen in Table 5.3, this group also showed cumulative growth shortfalls.

As in other comparisons based on group averages, it is useful to look at the extent of variation. Figure 5.2 shows the distribution of the differences between actual and projected investment rates for T+1 for SBA/EFF programs. Investment rates were below projections in about two-thirds of programs. In about one-quarter of programs, investment rates were 5 percentage points of GDP or more below projections. The proportion of cases where programs predicted a decline in investment rates between T–1 and T+1 was also seriously underestimated. Programs projected a decline in 25 percent of cases, while in reality investment rates declined in 50 percent of cases.

Figure 5.2.SBA and EFF Programs According to Differences Between Actual and Envisaged Investment Rates for T+1

(In percent of GDP)

Sources: MONA and WEO databases.

To summarize, there is evidence of generalized optimism in programs regarding the prospects for investment. That optimism becomes significantly magnified in the context of projecting recoveries from adverse initial conditions, and/or projecting the possibility of deterioration in performance. The problem is especially important in the case of SBA/EFF nontransition cases, and within this group, the set of capital account crisis episodes appears to be worse affected. The specific assumptions that may be behind this optimism in private spending and investment projections and how these assumptions are linked to program instruments are critical to assess the appropriateness of the fiscal stance of programs. This is discussed next.

Is the Fiscal Stance in IMF-Supported Programs Unnecessarily Contractionary?

The issue of whether IMF-supported programs suffer from an unnecessarily contractionary fiscal stance has attracted special attention following some of the recent capital account crisis cases—notably Korea and Indonesia, both of which experienced large output declines and an increase in unemployment. Critics have argued that the fiscal adjustment proposed in those cases was inappropriate and may have even contributed to worsening the situation. The specific cases of Korea and Indonesia have already been examined in detail in an earlier IEO report and are not discussed individually in the present evaluation.9 In this report, we consider what light can be shed on this issue from the broader cross-country evidence studied for this evaluation. For this purpose, we focus on SBA/EFF programs in non-transition countries as this is the group where the problem of a contractionary effect is perhaps most relevant.

Table 5.6 presents some of the critical macroeconomic data distinguishing between the capital account crisis cases (Panel A) and noncapital account crisis cases (Panel B). The following features are relevant to our evaluation.

Table 5.6Macroeconomic Balances in Stand-By and EFF Arrangements in Nontransition Countries 1(In percent of GDP)
A. Capital account crisis cases
N = 10ActualEnvisagedActualEnvisagedActual
Current account−3.4−2.42.4−2.40.2
Government balance−3.3−1.8−4.3−1.6−3.7
Total investment24.524.121.224.622.5
Private sector balance−0.1−0.66.7−0.83.9
GDP growth (in percent)2.91.6−
Source: MONA and WEO databases.

All differences between envisaged and actual values are statistically significant (with the exception of growth in T+1 in capital account crisis cases; and growth in investment in T for noncapital account cases).

The average growth figures differ slightly from those in Table 5.3 because the Lesotho SBA programs (1994/1995/1997) and the Republic of Congo SBA program (1994) were excluded due to problems in the reliability of the current account data.

B. Noncapital account crisis countries2
N = 45ActualEnvisagedActualEnvisagedActual
Current account−3.1−3.1−2.2−3.4−2.2
Government balance−4.0−2.4−3.2−1.4−3.5
Total investment22.223.122.623.821.6
Private sector balance0.9−0.71.0−2.01.3
GDP growth (in percent)
Source: MONA and WEO databases.

All differences between envisaged and actual values are statistically significant (with the exception of growth in T+1 in capital account crisis cases; and growth in investment in T for noncapital account cases).

The average growth figures differ slightly from those in Table 5.3 because the Lesotho SBA programs (1994/1995/1997) and the Republic of Congo SBA program (1994) were excluded due to problems in the reliability of the current account data.

Source: MONA and WEO databases.

All differences between envisaged and actual values are statistically significant (with the exception of growth in T+1 in capital account crisis cases; and growth in investment in T for noncapital account cases).

The average growth figures differ slightly from those in Table 5.3 because the Lesotho SBA programs (1994/1995/1997) and the Republic of Congo SBA program (1994) were excluded due to problems in the reliability of the current account data.

  • The capital account crisis cases experienced a severe output contraction in year T, resulting in a massive underperformance in output relative to expectations. The noncapital account crisis cases do not show an output contraction on average, but they do show a shortfall in growth compared with projections, especially in T+1.

  • Both groups show an underperformance in investment rates relative to expectations with the phenomenon being more marked in the case of capital account crisis cases.

  • Both groups show an underperformance on the fiscal side with fiscal deficits significantly higher than program targets. Again, the phenomenon is more marked in the case of capital account crisis cases, reflecting the decline in GDP in these cases and the asymmetric response of revenues and expenditures.

  • Both groups also show overperformance on the external side, in that the current account deficit was reduced much more than programmed. This is particularly so in the capital account crisis cases where the current account adjustment on average was 4.8 percent of GDP higher than programmed in year T and 2.6 percent of GDP higher in T+1. The corresponding numbers for the noncapital account cases are 0.9 percent and 1.2 percent of GDP, respectively.

The experience of the noncapital account crisis cases appears to be a milder form of the experience of the crisis cases, with the problem surfacing not in a decline in output but in a shortfall in growth performance in T+1.

The fact that both output and investment were below programmed levels raises the possibility that these may be classic cases of Keynesian lack of effective demand, in which higher levels of output could have been achieved if fiscal policy in the short run had been less contractionary. This perception is reinforced by the fact that the current account deficit overcorrected compared to projections, even though the fiscal targets originally projected in the program were not achieved. This can be viewed as suggesting that the original fiscal deficit targets were excessively tight and a more relaxed fiscal stance might have allowed higher levels of output and employment. Of course, the current account deficit could be expected to widen in this situation, but since the data show overcorrection in this dimension, it can be argued that there was room for some deterioration while leaving the deficit within the financeable range.

The emphasis on tightening fiscal policy could be traced to unrealistic assumptions about the pace at which private investment demand will recover following the crisis. Programs typically assume rapid recovery, and therefore tend to push for greater fiscal adjustment to make room for private investment, whereas a more realistic recognition of the negative impact of crises on investor expectations would call for a more relaxed fiscal stance.

It can be argued that a more expansionary policy may not have been feasible if external financing was not available to finance the resulting increase in the current account deficit. However this does not seem to be the case in a number of the programs we have examined. One-fourth of the SBA/EFF programs in nontransition cases showed overperformance not only in the current account, but also in the buildup of reserves. External financing does not seem to have been a constraint in these cases and a less contractionary fiscal stance could have been more appropriate.

This essentially Keynesian argument focuses exclusively on the role of fiscal adjustment as a factor affecting aggregate demand. However, as pointed out in Chapter 2, “Relevant Considerations in Determining the Fiscal Stance,” this is only one of the factors relevant in determining the fiscal stance. Emerging market countries relying on international financial markets also have to consider the impact of the fiscal stance adopted in times of crises on market confidence and therefore the availability of external finance. Advocates of a tighter fiscal stance can legitimately argue that, in situations where debt sustainability is an issue, it may be necessary to accept a larger dose of fiscal adjustment to reassure markets and ensure revival of confidence, even though a more relaxed stance may be justifiable on countercyclical grounds. In this view, the benefits of countercyclical fiscal policy can only be enjoyed in circumstances where the underlying fiscal situation is sound and markets recognize that the relaxed fiscal stance reflects a temporary resort to automatic stabilizers, and not simply an unwillingness to take difficult decisions.

It is difficult to determine the extent to which the fiscal stance adopted in the various programs studied was the result of a conscious decision to send the right market signals and whether the scale of the adjustment proposed was appropriate under the circumstances. As pointed out above, our evaluation finds that program documents provide little analysis of the rationale for fiscal adjustment and its link with the recovery of private sector activity and growth. A clearer statement of the rationale would add to transparency by promoting better understanding of the different considerations involved in each case, with a fuller consideration of the underlying assumptions. It would also help to determine the degree of flexibility that must be shown at the time of program reviews. For example, there is a clear case for allowing flexibility in adjusting the fiscal deficit in the event that assumptions about investment demand prove overoptimistic. As pointed out above, IMF programs do show considerable flexibility in practice in revising fiscal targets, but the rationale for the revisions is often left unclear. This has the disadvantage that adjustments that are perfectly justified on grounds of automatic stabilization may be seen as a forced response to nonperformance, a perception that can undermine the very confidence which the program seeks to restore.

This evaluation recommends a series of practices in the design of future programs. Program documentation should explicitly discuss how the projected economic recovery is linked to assumptions on how private demand will respond to the impact of the program on confidence. This is critical for the discussion of the fiscal stance of the program. A tight fiscal stance is appropriate under the expectation of significant positive shifts in investment demand, thus creating room for this buoyant investment demand to be financed. However, the original fiscal stance may need to be modified if the same economic recovery is to be achieved with a less buoyant recovery of private demand. In this case, it may be appropriate to include a stronger countercyclical element. These discussions may be particularly critical when private demand has initially collapsed as a result of a crisis situation at the outset of a program. A more careful identification of these links will provide a more coherent framework for sensitivity analysis. It will help to identify the critical assumptions and alert the staff early in the process on what needs to be monitored as the program unfolds.10

When staff decides to loosen fiscal policy to compensate for shortfalls in private demand, it should clearly evaluate the costs and benefits of doing so via the expenditure or revenue side. The mere fact that revenues may be lagging should not be an argument to accept a shortfall as the preferred way of temporarily widening the deficit. There may be good equity and solid reason to allow certain social expenditures momentarily to increase.

This is the maximum number of arrangements for which a comparable series of data on projected and actual growth could be obtained from the MONA and WEO databases. This sample is larger than the sample of 133 arrangements used previously to compare developments in fiscal balances.

The literature follows the original study by Goldstein and Montiel (1985), which tried to isolate the impact of an IMF-supported program by the Generalized Evaluation Estimator (GEE), and the subsequent study by Khan (1990). The GEE attempts to provide a measure of the policies that would have prevailed in the absence of an IMF-supported program. Although some earlier studies have shown no impact or a negative impact of IMF-supported programs on growth, the results of this line of research have been rather sensitive to model specification and the choice of variables included in the analysis. For a general review of the literature on this topic, see Joyce (2002) and Haque and Khan (1998).

The distinction between capital account crises and other more conventional balance of payments crises that have their origin in the current account is now well established, though it is not always as sharp as it sometimes appears because even conventional current account crises may generate capital account feedback effects. The 10 programs identified for inclusion in this group are the eight IMF-supported programs identified in Ghosh and others (2002) (Argentina 1995; Brazil 1998; Indonesia 1997; Korea 1997; Mexico 1995; the Philippines 1997; Thailand 1997; Turkey 1994) plus Turkey 1999 and Argentina 2000.

Again, these results do not “prove,” in any sense, that IMF-supported programs are good or bad for the recovery of growth. For example, because of mean-reversion phenomena (i.e., the tendency of an economy to revert to normal growth rates after a shock) it could be argued that growth would be expected to be stronger in any event in years T and T+1, merely because the impact of the adverse shock that caused the country to seek IMF support would tend automatically to dissipate as time passes.

This finding is also consistent with a recent study by Musso and Phillips (2002). The study found a tendency toward growth optimism in programs involving large access to IMF resources, those usually associated with crisis situations and large capital flow reversals.

We found that programs forecast 1.3 percent of cases as having negative growth in T+1, while in reality this happens in 13 percent of cases. On the other hand, programs forecast 5 percent of cases to have growth larger than two times mean growth (a symmetrical deviation from the mean) while in reality this happened in 11 percent of cases. Thus, programs systematically underpredict negative output developments relative to favorable developments.

Goldsbrough and others (1996). Moreover, there is a large theoretical and empirical literature suggesting that a lagged response of private investment should be expected following a period of adjustment. See, for example, Dixit and Pindyck (1994) and Serven and Solimano (1994).

Regression results (not shown) also suggest a strong and statistically significant link between the projected acceleration of growth in programs and the projected increase in investment rates.

See IEO (2003).

In some instances, staff has also used independent output forecasts from academics or market analysts to complement program projections, for example, the recent staff reports of program reviews of the Brazil Stand-By Arrangement approved in 2002. This is a good practice that should be encouraged.

    Other Resources Citing This Publication