As noted above, maintaining inflation in the low single digits is an important feature of PRGF-supported programs in the mature stabilizers. In general, programs sought to keep inflation in the 4–6 percent range. The next two parts of this section consider the appropriateness of these targets and the monetary policy framework in which these targets have been pursued. A third part considers the nexus between private sector credit growth, fiscal policy, and economic activity.
Inflation and Economic Growth
The 15 countries considered in this review have kept inflation at relatively low levels for a sustained period, but the benefits of this achievement have been questioned. Critics have argued that such inflation has required high real interest rates and constrained potential seigniorage income. Although unanimity exists in the economic literature on the costs of very high inflation, consensus is lacking on the appropriate inflation range for low-income countries.1 More generally, the scope for monetary policy to impede growth far exceeds its ability to create it: high inflation—above, for example, 40 percent—is certainly inimical to growth, but keeping inflation low will not by itself induce a growth boom. Against this backdrop, this section considers the appropriate inflation range for low-income countries.
High inflation is harmful to economic growth, but the gains from continued moderate inflation on growth are ambiguous, reflecting the multitude of transmission channels.2 On the one hand, as inflation accelerates, a negative effect on growth is likely, given the associated increase in inflation uncertainty, which clouds price signals and limits both the quantity and quality of investments. On the other hand, some inflation could enhance real wage flexibility, or help avoid liquidity trap problems. In case of price stickiness, a low inflation target could render an economy vulnerable to prolonged downturns if adverse supply shocks were to occur (see below). Empirical analysis aimed at untangling these effects is hindered by the possibility of reverse causality—from growth to inflation and the endogeneity of both variables. For example, high inflation may signal poor institutions and policies more generally and be associated with lower growth as a result.3
Notwithstanding this complexity, considerable empirical support exists for policies that aim at keeping inflation between, say, 5 and 10 percent. The existence of a negative relationship between inflation and growth at higher rates of inflation is empirically well supported. By contrast, identifying the growth effects of moving from, say, 20 percent inflation to 5 percent has been challenging. According to Bruno and Easterly (1998), significant adverse growth effects can be found only for generally short-lived periods of high inflation, after which growth tends to return to its long-run path—classic money neutrality. However, several other studies indicate this may understate the adverse growth effects of moderate inflation:
-
Examinations of prolonged episodes of accelerated growth show that the onset of such periods is typically associated with lower inflation and that, moreover, the upfront improvement in macroeconomic indicators is more generalized for longer episodes of growth acceleration than for shorter ones.4
-
Quantifying the association between inflation and growth requires careful attention to the nonlinearities in the relationship between inflation and growth. This relationship appears to be convex: a given increase in inflation is associated with a larger negative growth effect, the lower the initial rate of inflation. But at low rates, higher inflation may have no effect on growth or its effect may even be positive. Since the work of Fischer (1993), several authors have tried to identify and locate such a “kink” in the relationship between inflation and economic growth associated with a maximum growth rate. Empirical studies using panels of countries have located this kink in the inflation-growth nexus at a level of inflation somewhere between 3 percent and 40 percent, with a cluster suggesting a level in the 5–10 percent range (see Table 3.1). Moreover, a positive growth effect of higher inflation was found at relatively low levels of inflation, suggesting that higher-than-single-digit inflation would not promote growth, while entailing risks of a negative effect. Notably, however, most of these studies have not focused on the kink in low-income countries as opposed to broader country groupings, and none of the studies covers the period since 2000.
Empirical Studies on Kinks in the Relationship between Inflation and Growth
Empirical Studies on Kinks in the Relationship between Inflation and Growth
| Inflation Threshold (percent) | Growth Effect of Higher Inflation Below the Threshold | Countries | Period | Inflation Measure | Remarks | |
|---|---|---|---|---|---|---|
| Fischer (1993) | 15 | Negative | 80 | 1960–89 | CPI | |
| Barro (1996) | 10–20 | Not significant | 117 | 1960–90 | 10-year average CPI | |
| Sarel (1996) | 8 | Positive | 87 | 1970–90 | 5-year average CPI | |
| Bruno and Easterly (1998) | 40 | Not significant | 97 | 1961–92 | CPI | |
| Ghosh and Phillips (1998) | >5 | Positive | 145 | 1960–96 | Average annual CPI | |
| Kochhar and Coorey (1999) | 5 | Positive | 84 (low- and middle-income countries only) | 1981–95 | Average annual CPI | |
| Khan and Senhadji (2001) | 11–12 (for developing countries only) | Not significant | 140 | 1960–98 | 5-year average CPI | 1–3 percent threshold for industrial countries. Controlled for investment and unemployment. |
| Burdekin and others (2000) | 3 (for developing countries only) | Positive | 51 | 1967–92 | 8 percent threshold for industrial countries. | |
| Gylfason and Herbertsson (2001) | 10–20 | Positive | 170 | 1960–92 | 5-year average GDP deflator |
Empirical Studies on Kinks in the Relationship between Inflation and Growth
| Inflation Threshold (percent) | Growth Effect of Higher Inflation Below the Threshold | Countries | Period | Inflation Measure | Remarks | |
|---|---|---|---|---|---|---|
| Fischer (1993) | 15 | Negative | 80 | 1960–89 | CPI | |
| Barro (1996) | 10–20 | Not significant | 117 | 1960–90 | 10-year average CPI | |
| Sarel (1996) | 8 | Positive | 87 | 1970–90 | 5-year average CPI | |
| Bruno and Easterly (1998) | 40 | Not significant | 97 | 1961–92 | CPI | |
| Ghosh and Phillips (1998) | >5 | Positive | 145 | 1960–96 | Average annual CPI | |
| Kochhar and Coorey (1999) | 5 | Positive | 84 (low- and middle-income countries only) | 1981–95 | Average annual CPI | |
| Khan and Senhadji (2001) | 11–12 (for developing countries only) | Not significant | 140 | 1960–98 | 5-year average CPI | 1–3 percent threshold for industrial countries. Controlled for investment and unemployment. |
| Burdekin and others (2000) | 3 (for developing countries only) | Positive | 51 | 1967–92 | 8 percent threshold for industrial countries. | |
| Gylfason and Herbertsson (2001) | 10–20 | Positive | 170 | 1960–92 | 5-year average GDP deflator |
Additional considerations in favor of determined inflation control relate to policy credibility and poverty alleviation. For countries that have only recently managed to achieve single-digit inflation—which includes almost all in this sample of PRGF countries—it may be wise to anchor successful disinflation by maintaining inflation in the single-digit range. Indeed, the gains from price stabilization may be realized in full only when low inflation is institutionalized and incorporated into expectations. Another consideration is that price stability may foster poverty reduction directly, in addition to its impact through economic growth. Inflation tends to hurt the poor disproportionally, because they often hold no financial assets that provide a hedge against inflation or they are dependent on state-determined income that is not fully indexed.5
Seigniorage Income
Increases in reserve money beyond the amount that accommodates changes in real money demand tend to be inflationary, and the associated part of seigniorage is aptly called the inflation tax.1 The reduction in inflation in PRGF countries has, indeed, entailed a decline in seigniorage revenue. Evidence suggests that seigniorage revenue (as measured below) is highest for inflation at 30–40 percent, far above the actual 6 percent average rate in this paper’s sample of PRGF countries.2 Seigniorage as a share of GDP can be expressed as3
s = rt–1*(1 + gr) – rt–1/[(1 + gp)*(1 + gy)],
where rt–1 is the previous period’s reserve money ratio (as a fraction of nominal GDP), and gr, gp, and gy are the growth rates of the reserve money ratio, the price level, and real GDP, respectively. For a constant reserve money ratio of 11 percent (the average in the country sample) and an average real growth rate of 5 percent, the noninflationary level of seigniorage would amount to 0.5 percent of GDP. With 6 percent inflation, seigniorage would amount to 1.1 percent of GDP. The actual 1999–2003 average in the sample was somewhat higher, at 1.4 percent of GDP, owing to a trend increase in reserve money as a share of GDP.
However, the revenue benefits of accomodating higher inflation rates would be very limited. For one, the inflation tax is relatively distortionary (especially at higher rates, see above) and regressive (given the impact of inflation on poverty), implying a strong case against the use of higher inflation as a means of raising government resources. A more compelling case for relying on seignorage income could be the relative insufficiency of most low-income countries’ tax systems. It is therefore possible that, on balance, a somewhat higher inflation tax would be less distortionary than the regular tax it could replace. But the revenue effect of raising inflation from 6 percent to, for example, 15 percent would be modest. With no change in the reserve money ratio, the revenue impact would be equivalent to 0.8 percent of GDP. But higher inflation would likely entail some decline in money demand and thereby in the reserve money ratio. Incorporating this effect could reduce the seigniorage bonus by half.4 Thus the seignorage that could be generated from allowing inflation to rise to 15 percent would only be around 0.5 percent of GDP, and it is highly questionable whether this gain would be worth the adverse effect that higher inflation tends to have on the poor.
1 Seigniorage revenue typically takes the form of a profit transfer from the central bank to the government. In cases where the central bank lends to the government at below-market interest rates, reducing the bank’s profits, the transfer may appear as lower interest payments. 2 Phillips (1999). 3 The formula does not take into account the remuneration of commercial bank reserves within reserve money. However, in the country sample, currency in circulation accounts for more than 60 percent of reserve money, on average, and remuneration is limited. 4 During 1994–2003, for disinflating nontransition PRGF-supported programs, a 1 percentage point increase in inflation was associated with a rise in velocity (and thereby in the reciprocal of the reserve money ratio) by about 0.4 percent (IMF, 2004d, Box 5).Some observers have argued that tight monetary policies have adversely affected poverty alleviation by constraining government spending. The main channel through which monetary policy can reduce available fiscal resources is by lowering seigniorage—that is, government revenue from an increase in reserve money (see Box 3.1).6 However, the revenue effect of raising inflation from 6 percent to, for example, 15 percent would be modest because seigniorage revenue would be raised by less than 0.4 percentage point of GDP.
One strong argument in support of higher inflation targets for developing countries is the need to accommodate exogenous shocks. Developing countries are more prone to supply shocks, such as adverse weather conditions, and sharp changes in the terms of trade. Moreover, the aggregate impact of such shocks may be more pronounced as a result of the limited diversification of production and consumption and the prevalence of de facto trade barriers (including, for example, lack of transportation and integrated distribution networks). Adverse supply shocks would tend to push up prices while limiting economic growth. In the presence of downward nominal rigidities, inflation targets that do not accommodate the supply shocks could serve to amplify their adverse growth effects. This consideration argues for inflation targets that accommodate the first-round effect of supply shocks (unless the shock is expected to be undone within a few months—that is, before monetary policy is effectively transmitted).
Cross-country evidence demonstrates the higher vulnerability of PRGF countries to external shocks. The year-on-year volatility of the terms of trade and export volumes (which may to a large extent be considered exogenous) and of real GDP growth (which also incorporates the policy response) is relatively large in PRGF countries, and in low-income countries more generally, in comparison with advanced and transition economies.7
Program design should address such volatility. In particular, programs can accommodate adverse supply shocks (within limits) while avoiding second-round effects, thus stabilizing inflation expectations.
-
Programs could be based on a target range for inflation around the operational (core) target, especially in countries where substantial exogenous shocks have occurred. For example, as illustrated in Figure 3.1, a band of 2 percentage points above and below the central target would capture more than 60 percent of realizations. Such change would not directly alter the monetary targets or the associated conditionality, but would formalize the inflation tolerance that is, in practice, often already exhibited in programs. The change may also help improve policy credibility by reducing the likelihood that inflation targets are missed. One of the risks associated with this approach is that, in practice, policies may simply be redirected toward the higher end of the inflation range, in which case the result would be an unwarranted policy relaxation with no increase in flexibility.8
-
In some cases, consideration could be given to focusing on measures of core inflation rather than the overall CPI in PRGF-supported programs.9 The monetary targets would continue to be based on nominal income and incorporate the first-round effects, while preventing second-round effects that would lift core inflation.
Inflation: One-Year-Ahead Projections and Outcomes
Sources: IMF, staff reports and World Economic Outlook database.On balance, these considerations support the use of single-digit inflation targets. Preserving the inflation gains already achieved by the countries in this sample is of vital importance. There is a need to further ingrain and institutionalize this achievement, enhancing the credibility, and thus the desirability, of the low-inflation strategy. However, pushing inflation too low—say, below 5 percent—may entail a loss of output and seigniorage revenue, suggesting a need for caution in setting very low inflation targets in low-income countries. These countries tend to be subject to larger output volatility and more pronounced price shocks, and program design should take these economic attributes properly into account. In particular, inflation targets should be set so as to help avoid risks of an unintended contractionary policy stance.
Financial Programming
This part of the section considers the effectiveness of reserve money targets and program conditionality as an anchor for monetary policy. There are several reasons for this assessment at this juncture. For one, as demonstrated in the 2004 review of the design of IMF-supported programs (IMF, 2004d), the commonly applied net domestic assets (NDA)–net international reserves (NIR) framework is not well suited for controlling inflation, which generally requires a more explicit monetary anchor. These papers also underscore the need to offset the monetary impact of donor support if inflation targets are to be met. Second, the large buildup of international reserves—beyond program targets—in the mature stabilizers could be indicative of the policy challenges related to reconciling exchange rate and monetary objectives.
Reserve money serves as an operational target for the 13 sampled countries that do not have a formal exchange rate peg (Box 3.2).10 The increase in reserve money projected to be in line with the inflation objective in turn provides the basis for the standard program ceiling on central bank NDA and the floor on central bank NIR or net foreign assets (NFA).11 Importantly, this framework does not constrain reserve money growth—that is, it does not provide a monetary anchor. In particular, the existence of a floor on central bank NFA provides scope for larger reserve accumulation than projected, with a corresponding rise in reserve money. This feature provides flexibility in accommodating increases in money demand through expanded interventions in the exchange market, thus avoiding below-target inflation.
Monetary Programming in IMF Program Design
Monetary policy targets for low-income countries are generally derived through financial programming. Financial programming involves the use of balance sheet relationships and a few simple behavioral relationships. It helps ensure the consistency of the design of monetary policy with (1) goals or assumptions regarding real GDP growth and inflation; (2) projected domestic financing of the budget and provision of sufficient credit to the private sector; and (3) the balance of payments (BoP) outlook as reflected in the net accumulation of foreign assets by commercial banks and the monetary authorities. Within financial programming, the exercise that directly relates to the central bank’s monetary program is also referred to as reserve money programming.1
The level of broad money (M) that would be consistent with the program targets for inflation (P, the GDP deflator) and real income (Y) is derived using the identity
where V is the velocity of circulation. In terms of growth rates:
Velocity depends on such factors as interest rates and changes in inflation, but it is typically—including in the mature stabilizers sample—assumed to remain constant over the program period. Provided that the assumption on velocity adequately reflects money demand, limiting the growth in money supply in line with equation (2) should help realize the targeted rate of inflation.
The increase in reserve money (RM) that would support this broad money target is generally derived from
or
where m is the reserve money multiplier, which reflects the ratio of currency in circulation to bank deposits and banks’ holdings of reserves at the central bank. Like velocity, the multiplier is often assumed constant over the program period.
Target values for the central bank’s NDA and NFA are derived using
Similarly, for the banking sector (B) as a whole,
with CG equal to net banking sector credit to the government, CP equal to net banking sector credit to the private sector, and BNW equal to net worth of the banking sector (often assumed constant over the program period). An important application of equation (6) allows verification if, given the targets and projections for the other variables, projected government borrowing is consistent with (does not crowd out) the scope for lending to the private sector, which, in turn, is important to sustaining private sector growth.
1 For a more general overview, see IMF (2004f, Appendix I).However, success in attaining monetary and inflation objectives may have been compromised by the simultaneous pursuit of exchange rate stability. In particular, if higher international reserves reflect higher supply of foreign capital—absorbed by the monetary authorities, for example, to avoid appreciation pressures—the increase could be inflationary. In fact, the resulting rise in reserve money (and thereby in commercial bank liquidity) could provide the basis for higher domestic bank credit and thus result in an even larger increase in broad money through the multiplier effect. Changes in reserve money supply can be minimized through offsetting domestic transactions, but such sterilization can be costly. Even countries labeled as independently floating (see Table 3.2) in practice often intervene to limit exchange rate variability. Moreover, the evidence is strongly suggestive of declining exchange rate flexibility in the mature stabilizers over time. This point is illustrated in Figure 3.2. For the 13 countries in this sample without a hard peg during 1990–94, 1995–99, and 2000–03, the dispersion of the monthly exchange rate changes is more highly concentrated around the mean (0.3 percent) in 2000–03 than earlier periods or relative to other PRGF countries not included in the mature stabilizer sample.12
De Facto Exchange Regimes, 2001
Fixed to the U.S. dollar, but with periodic adjustments, until mid-2003, when the currency was floated.
De Facto Exchange Regimes, 2001
| Currency union: Benin, Senegal |
| Conventional fixed peg to single currency: Bangladesh1 |
| Forward-looking crawling band: Honduras |
| Tightly managed floating: Azerbaijan, Guyana, Mongolia |
| Other managed floating: Ethiopia, Kyrgyz Republic, Rwanda |
| Independently floating: Albania, Madagascar, Mozambique, Tanzania, Uganda |
Fixed to the U.S. dollar, but with periodic adjustments, until mid-2003, when the currency was floated.
De Facto Exchange Regimes, 2001
| Currency union: Benin, Senegal |
| Conventional fixed peg to single currency: Bangladesh1 |
| Forward-looking crawling band: Honduras |
| Tightly managed floating: Azerbaijan, Guyana, Mongolia |
| Other managed floating: Ethiopia, Kyrgyz Republic, Rwanda |
| Independently floating: Albania, Madagascar, Mozambique, Tanzania, Uganda |
Fixed to the U.S. dollar, but with periodic adjustments, until mid-2003, when the currency was floated.
Dispersion of Monthly Exchange Rate Changes to the U.S. Dollar—Kernel Density Diagram
Source: IMF, International Financial Statistics.The recent experience of the mature stabilizers shows that the expansion of monetary aggregates has generally been higher than targeted, but without leading to a significant overshooting of inflation targets (Table 3.3).13 On average, actual broad money growth exceeded program projections by some 4–6 percentage points, about one-third as much as initially envisaged. The average deviation declines as the target approaches (except between t – 2 and t – 1), which points to increasing accuracy as the projection period shortens (see Appendix I for evidence of improved program accuracy in consecutive program updates). At the same time, there is little evidence that the higher-than-targeted broad money growth has translated into higher inflation: deviations of inflation from target are not statistically significant, and regression analysis reveals no statistically significant relation between the deviations from the previous years’ projections of money growth and inflation, as illustrated by Figure 3.3.14 It is important to recognize that these findings are fully consistent with the expected positive association between money growth and inflation—which is strongly supported by the data in our sample. Although there was scope to accommodate somewhat higher-than-projected money demand, a further increase in money supply would still have been inflationary.
Monetary Projections and Projection Deviations
(Averages, in percent)
The projections are for year t. The projections as of years t – 1 and t – 2 are those from the last staff reports in the previous year and the year before that, respectively. The projection as of t(SR1) is from the first staff report in year t, and the one at t(SR2) is from the final staff report in that year. If only one staff report was issued in year t, the last two observations coincide.
NDA (NFA) contribution is defined as the change in central bank NDA (NFA) divided by previous year’s reserve money.
Projection deviations are calculated as program values minus outcomes. Because the panel is unbalanced, deviations from the t – 1 and t – 2 projections do not correspond to the difference between the average projection for the sample subset and the average outcome for the full sample.
Significance at 10 percent level (*), 5 percent level (**), or 1 percent level (***).
The contribution of the deviation in NDA growth, to the deviation in reserve money growth, transformed to be mapped into (–100, 100),—comparable to the measure in the 2004 evaluation of program design (IMF 2004d, Table 10).
Monetary Projections and Projection Deviations
(Averages, in percent)
| Projection as of1 | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| t – 2 | t – 1 | t(SR1) | t(SR2) | Outcome | ||||||
| Projected | ||||||||||
| Inflation (end-year CPI) | 4.1 | 4.9 | 6.0 | 5.9 | 6.3 | |||||
| Inflation (GDP deflator) | 4.3 | 5.2 | 6.3 | 6.1 | 6.8 | |||||
| Real GDP growth | 6.0 | 5.9 | 4.9 | 4.9 | 5.1 | |||||
| Broad money growth | 12.0 | 12.3 | 12.2 | 13.3 | 17.5 | |||||
| Reserve money growth | … | 10.5 | 9.7 | 10.3 | 15.7 | |||||
| NDA contribution2 | … | –0.6 | –6.8 | –7.8 | –11.5 | |||||
| NFA contribution2 | … | 10.3 | 16.4 | 18.2 | 27.1 | |||||
| Velocity (% change) | … | –0.6 | –0.4 | –1.6 | –4.3 | |||||
| Money multiplier (% change) | … | 1.6 | 2.7 | 3.1 | 2.4 | |||||
| No. of observations | 29 | 43 | 54 | 54 | 54 | |||||
| Deviations from projections3,4 | ||||||||||
| Inflation (end-year CPI) | –0.7 | –0.5 | –0.3 | –0.4 | ||||||
| Inflation (GDP deflator) | –1.0 | –1.0 | –0.5 | –0.7 | ||||||
| Real GDP growth | 0.2 | 0.6 | –0.2 | –0.2 | ||||||
| Broad money growth | 4.5** | –5.9*** | –5.4*** | –4.3*** | ||||||
| Reserve money growth | … | –4.6* | –6.0*** | –5.4*** | ||||||
| NDA contribution2 | … | 6.9 | 5.1 | 3.7 | ||||||
| NFA contribution2 | … | –13.1 | –11.3** | –8.9* | ||||||
| NDA contribution5 | … | –4.1 | –6.1 | –0.7 | ||||||
| Velocity (% change) | … | 4.5*** | 3.8*** | 2.7*** | ||||||
| Money multiplier (% change) | … | –1.4 | 0.2 | 0.7 | ||||||
The projections are for year t. The projections as of years t – 1 and t – 2 are those from the last staff reports in the previous year and the year before that, respectively. The projection as of t(SR1) is from the first staff report in year t, and the one at t(SR2) is from the final staff report in that year. If only one staff report was issued in year t, the last two observations coincide.
NDA (NFA) contribution is defined as the change in central bank NDA (NFA) divided by previous year’s reserve money.
Projection deviations are calculated as program values minus outcomes. Because the panel is unbalanced, deviations from the t – 1 and t – 2 projections do not correspond to the difference between the average projection for the sample subset and the average outcome for the full sample.
Significance at 10 percent level (*), 5 percent level (**), or 1 percent level (***).
The contribution of the deviation in NDA growth, to the deviation in reserve money growth, transformed to be mapped into (–100, 100),—comparable to the measure in the 2004 evaluation of program design (IMF 2004d, Table 10).
Monetary Projections and Projection Deviations
(Averages, in percent)
| Projection as of1 | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| t – 2 | t – 1 | t(SR1) | t(SR2) | Outcome | ||||||
| Projected | ||||||||||
| Inflation (end-year CPI) | 4.1 | 4.9 | 6.0 | 5.9 | 6.3 | |||||
| Inflation (GDP deflator) | 4.3 | 5.2 | 6.3 | 6.1 | 6.8 | |||||
| Real GDP growth | 6.0 | 5.9 | 4.9 | 4.9 | 5.1 | |||||
| Broad money growth | 12.0 | 12.3 | 12.2 | 13.3 | 17.5 | |||||
| Reserve money growth | … | 10.5 | 9.7 | 10.3 | 15.7 | |||||
| NDA contribution2 | … | –0.6 | –6.8 | –7.8 | –11.5 | |||||
| NFA contribution2 | … | 10.3 | 16.4 | 18.2 | 27.1 | |||||
| Velocity (% change) | … | –0.6 | –0.4 | –1.6 | –4.3 | |||||
| Money multiplier (% change) | … | 1.6 | 2.7 | 3.1 | 2.4 | |||||
| No. of observations | 29 | 43 | 54 | 54 | 54 | |||||
| Deviations from projections3,4 | ||||||||||
| Inflation (end-year CPI) | –0.7 | –0.5 | –0.3 | –0.4 | ||||||
| Inflation (GDP deflator) | –1.0 | –1.0 | –0.5 | –0.7 | ||||||
| Real GDP growth | 0.2 | 0.6 | –0.2 | –0.2 | ||||||
| Broad money growth | 4.5** | –5.9*** | –5.4*** | –4.3*** | ||||||
| Reserve money growth | … | –4.6* | –6.0*** | –5.4*** | ||||||
| NDA contribution2 | … | 6.9 | 5.1 | 3.7 | ||||||
| NFA contribution2 | … | –13.1 | –11.3** | –8.9* | ||||||
| NDA contribution5 | … | –4.1 | –6.1 | –0.7 | ||||||
| Velocity (% change) | … | 4.5*** | 3.8*** | 2.7*** | ||||||
| Money multiplier (% change) | … | –1.4 | 0.2 | 0.7 | ||||||
The projections are for year t. The projections as of years t – 1 and t – 2 are those from the last staff reports in the previous year and the year before that, respectively. The projection as of t(SR1) is from the first staff report in year t, and the one at t(SR2) is from the final staff report in that year. If only one staff report was issued in year t, the last two observations coincide.
NDA (NFA) contribution is defined as the change in central bank NDA (NFA) divided by previous year’s reserve money.
Projection deviations are calculated as program values minus outcomes. Because the panel is unbalanced, deviations from the t – 1 and t – 2 projections do not correspond to the difference between the average projection for the sample subset and the average outcome for the full sample.
Significance at 10 percent level (*), 5 percent level (**), or 1 percent level (***).
The contribution of the deviation in NDA growth, to the deviation in reserve money growth, transformed to be mapped into (–100, 100),—comparable to the measure in the 2004 evaluation of program design (IMF 2004d, Table 10).
Deviations from Previous-Year Projections for CPI and Broad Money Growth1
(In percentage points)
Source: IMF staff reports.Note: See the note on Figure 2.6 for a list of the country names symbolized by each abbreviation.1 Regression conditioned on fixed country effects.The evidence that larger-than-programmed monetary growth has not triggered above-target inflation appears limited to the sample of mature stabilizers. Notably, the 2004 review of program design did find statistically significant correlation between slippage relative to the one-year-ahead projections in broad money growth and in inflation (IMF, 2004d, Section III.B).15 That analysis, however, covered all PRGF countries rather than a subset of mature stabilizers.16 Moreover, the study also showed that the larger overruns are recorded by high-inflation cases.
The above-target monetary expansion has accommodated a trend decline in velocity. The gradual decline in inflation (Section II) seems to have fostered an increase in broad money demand in real terms (that is, a decrease in velocity)—with lower inflation expectations, the opportunity costs of holding non-interest-bearing financial assets also fall.17 Financial sector development has likely also contributed to this trend. However, the baseline assumption in most programs has been for velocity to remain largely unchanged (see Figure 3.4), with only a minor average decline. As a result, velocity has on average been 3–5 percent lower than projected (Table 3.3).18 By contrast, the 2004 program design review found that the null hypothesis of equality of the relationship between money growth and inflation in the program and in reality could not be rejected.19 Findings concerning average velocity mask the pronounced variation in the changes in this parameter—which have a root mean squared error (RMSE) for velocity growth (in percent) of 7.0 for the final projection.20 This volatility imparts greater uncertainty to estimates of money demand, complicating the conduct of monetary policy.
Projected and Actual Percentage Change in Velocity
(In percent)
Sources: IMF staff reports and World Economic Outlook database.The overruns in broad money growth were primarily a reflection of higher-than-projected reserve money growth, the aggregate over which the monetary authorities have most direct control. Program targets for reserve and broad money were exceeded in approximately the same degree (Table 3.3), because the program assumptions for the reserve money multiplier—the parameter that measures the link between the two aggregates—were on average broadly accurate.21 The deviations in the multiplier from the program assumptions were not statistically significant. Averages, however, mask a great deal of volatility (Figure 3.5).22
Projected and Actual Percentage Change in the Reserve Money Multiplier
(In percent)
Sources: IMF staff reports and World Economic Outlook database.In turn, higher-than-targeted reserve money growth has mainly reflected larger-than-programmed accumulation of international reserves. As shown in Table 3.3 (upper panel), on average, programs correctly project a decline in central bank NDA and a rise in NFA. However, both changes tend to be even larger than projected, with the NFA overrun exceeding the NDA shortfall, resulting in a faster-than-projected expansion in reserve money.
This additional liquidity expansion is consistent with the architecture of the IMF-supported programs. Conditionality in IMF-supported programs generally requires NDA to be kept below and NFA to be kept above the targets set under the program.23 Recognizing the uncertainty that attends to estimates of money demand, IMF-supported programs generally do not impose firm limits on the aggregate increase in reserve money. The framework effectively allows for larger-than-programmed monetary expansion to take place as long as it reflects higher reserve accumulation.
Larger-than-programmed monetary and reserves expansion likely also reflects the objective of avoiding appreciation pressures in the face of external inflows. The systematic overruns in NFA growth seem related to the above-mentioned potential tension between reserve money objectives and (de facto) exchange rate targets. Had primacy been given to the money target, greater exchange rate flexibility would have been evident, along with less of a buildup in reserves. More than three-fourths of these NFA overruns have been sterilized through a lower central bank NDA (see Figure 3.6).24 Thus, monetary authorities have sought to keep foreign inflows from affecting the exchange rate while mitigating their impact on (reserve) money. The relative success of this strategy is illustrated in Figure 3.7: across country episodes, no association between exchange rate stability and overruns in NFA or money growth was apparent.25 Nonetheless, given the serious costs and limits of sterilization—including the adverse impact of higher real interest rates on private sector growth—for those countries in the mature stabilizer sample with the most heavily managed exchange rates, IMF staff has generally advocated a stronger focus on monetary objectives and a more market-based exchange rate.26
Deviations from the Programmed Levels of NDA and NFA1
Source: IMF staff reports.Note: See the note on Figure 2.6 for a list of country names symbolized by each abbreviation.1 Conditioned on fixed effects and deviations from projected government borrowing.Exchange Rate Variability and Deviations from Previous-Year Broad Money Projection
Source: IMF staff reports.Note: See the note on Figure 2.6 for a list of the country names symbolized by each abbreviation.These developments in the mature stabilizers point to the potential benefits of some changes in the design of monetary programs. First, there is scope for improving the programming assumptions regarding velocity, which are key to the conduct of monetary policy as well as financial programming. In particular, once inflationary pressures are at bay, monetary projections require a careful assessment of velocity trends and of the need to allow for a (further) decline. With this adjustment, the financial programming framework would continue to be a useful tool for guiding program design, as monetary management remains focused on monetary aggregates rather than interest rates. Second, the standard NDA-NFA framework for monetary policy in IMF-supported programs appears to provide a useful safety valve, allowing money growth to adjust to unforeseen changes in velocity. However, given the risks to inflation posed by an unwarranted expansion in reserves and money—evident in the larger set of PRGF programs more than in the subsample of mature stabilizers—the use of this channel merits close monitoring. In this context, indicative targets for reserve money growth could provide a useful check. Money growth in excess of the indicative ceiling should prompt further analysis to determine whether monetary policy merely accommodated higher money demand or entailed risk of inflationary pressure.
In the case of a managed floating exchange rate, interventions in the exchange market to resist durable real appreciation pressures may pose risks to the programmed inflation objective. The benefits of a managed float include that it can help absorb real internal and external shocks—which are prevalent in PRGF countries—through nominal exchange rate adjustment. However, the arrangement puts the responsibility for ensuring low and stable inflation firmly on the central bank. Resisting lasting appreciation pressure from such shocks through large-scale—and costly—interventions, building up NFA, and fueling inflation would negate the adjustment mechanism and undermine price stability.27
Is Private Sector Credit Being Crowded Out?
Concerns regarding the potential crowding out of private sector credit are an important reason why IMF-supported programs limit domestic financing of the public sector.28 The desire to foster the growth of credit to the private sector is influenced by findings that financial development, as measured by the volume of intermediated finance, is robustly associated with higher per capita income growth. At the same time, crowding out of credit to the private sector is generally difficult to establish. First, credit markets rarely clear through changes in interest rates alone. Because higher interest rates are likely to attract less creditworthy clients, banks often resort to credit rationing—a behavior not easily captured in the data. Second, financial repression remains prevalent in a number of the mature stabilizers (for example, Ethiopia), making interest rates sticky. Third, the poor health of the banking sector may inhibit the effectiveness of financial intermediation—limiting the impact of changes in government borrowing on private credit growth. And although there are a few studies (Christensen, 2004; Adam and Bevan, 2004) that find some evidence of crowding out in African countries, in particular, this may capture mainly the cases with severe macroeconomic imbalances. Adam and Bevan (2004) have also flagged the absence of crowding in of private sector credit after recourse to domestic financing by the government has been curtailed and broader macroeconomic stability achieved.
Against this backdrop, this section briefly considers, first, the magnitude of private sector credit growth (PSCG) accommodated in IMF-supported programs and, second, the extent to which crowding out, or lack of crowding in, has been problematic.
How Has PSCG Fared in the Mature Stabilizers?
Private sector credit growth in the countries in the sample has expanded at a rapid pace—albeit from a small base and remaining slightly lower than implicitly allowed for in their PRGF-supported programs. On average, during 2000–03, PRGF-supported programs envisaged private sector credit growth on the order of 19 percent—in real terms, about twice the pace of real GDP growth. Actual credit growth was marginally less (just under 18 percent, Table 3.4).29 More generally, private sector credit growth in the mature stabilizers compares favorably with that in other low-income countries. The median increase in the ratio of private sector credit to GDP in the sample over the 1995–2003 period was 5 percentage points of GDP, compared with 2½ percentage points in other low-income countries.
Credit to the Private Sector1
(Annual percent growth)
Credit growth in nominal terms during the first programmed year.
Credit to the Private Sector1
(Annual percent growth)
| Program | Actual | |
|---|---|---|
| Average | 19.1 | 17.7 |
| Median | 17.6 | 15.1 |
| Mode | 13.0 | 20.0 |
| Standard deviation | 12.4 | 24.5 |
| Sample size | 40 | 40 |
Credit growth in nominal terms during the first programmed year.
Credit to the Private Sector1
(Annual percent growth)
| Program | Actual | |
|---|---|---|
| Average | 19.1 | 17.7 |
| Median | 17.6 | 15.1 |
| Mode | 13.0 | 20.0 |
| Standard deviation | 12.4 | 24.5 |
| Sample size | 40 | 40 |
Credit growth in nominal terms during the first programmed year.
However, these averages mask important cross-country variations. Over the somewhat longer horizon of 1995–2003, credit to the nongovernment sector has grown significantly faster than income for most of the mature stabilizers, but particularly so for Bangladesh, Benin, Ethiopia, Honduras, and Senegal (see Figure 3.8). One possible explanation is in the deeper banking systems—the ratio of credit to GDP was about 18 percent in these 5 countries, compared with 11 percent in the other 10 mature stabilizer countries. Also, for some countries (Mozambique, Rwanda, and Tanzania), the association between GDP and credit growth is more linear.30 However, at least in the case of Rwanda, this likely reflects GDP recovery from extremely low levels more than a positive effect of private credit on real growth.
Cumulated Growth in Credit and GDP, 1995–2002
(In percent)
Sources: IMF, International Financial Statistics; and staff estimates.Note: See the note on Figure 2.6 for a list of the country names symbolized by each abbreviation.Crowding Out
There is limited evidence of government recourse to domestic financing crowding out private sector borrowing in the mature stabilizers:
-
First, in both the mature stabilizers and the broader grouping of low-income countries, credit to the private sector tends to be lower when levels of domestic debt are higher (Figure 3.9, top left panel). Further, the higher the stock of debt, the more pronounced this negative correlation tends to be (Figure 3.9, lower right panel compared with lower left panel). However, when one looks at annual changes in these variables instead of levels (Figure 3.9, top right panel), the correlation is still negative but no longer statistically significant.
-
Second, as noted above, over the somewhat longer period of 1995–2003, credit growth in the mature stabilizers has actually been higher than in other low-income countries, against the backdrop of a similar increase (about 2 percentage points of GDP) in domestic indebtedness in both groups.
-
Third, real interest rates (proxied by average treasury bill rates) have remained broadly unchanged in the 15 countries. Beyond this, there is no robust evidence of a negative association between real interest rates and changes in domestic government indebtedness for either the sample or the broader low-income countries grouping. Although, as noted above, crowding out could take place through channels other than interest rates (for example, credit rationing).
These considerations suggest that domestic financing should be relied upon only cautiously, especially where domestic and total debt stocks are high. The absence of convincing evidence notwithstanding, the risk of crowding out should not be discounted too readily—especially given the crucial role of private investment for sustaining growth. Caution in domestic borrowing is also warranted because domestic debt carries higher real interest rates than foreign debt; many projects may have an expected return exceeding the terms of concessional financing, but few projects are likely to have returns exceeding the interest rates on nonconcessional domestic borrowing. For countries that satisfy these conditions, one benchmark to help guide domestic borrowing decisions could be to ensure that annual recourse to such credits is consistent with stabilizing the debt stocks.
Given the close association between the level and the variability of inflation, their respective effects cannot easily be disentangled empirically. The positive relation also implies that the distinction may be of little relevance for policy purposes.
According to empirical work undertaken in the recent papers on the design of IMF-supported programs, there is no evidence that the monetary stance (measured by velocity) was set excessively tight in IMF-supported programs, leading to lower output growth (IMF, 2004f, Section III.B). Similarly, the Independent Evaluation Office (IEO) found no evidence suggesting “an excessive deflationary bias with respect to inflation targets” (IEO, 2004). For an overview of this and other effects of inflation on growth, see Temple (2000).
Easterly and Levine (2003) find no significant impact of macroeconomic policies (averaged over four decades) on the level of economic development after controlling for the impact of institutions, and Easterly (2004) conjectures that any impact of macroeconomic policies found in growth regressions may occur through its performance as a proxy for institutional quality. This view, however, contrasts with a range of evidence (see below) that is suggestive of an independent role of macroeconomic stabilization. It is likely that both macroeconomic stabilization and institutional improvements are important in triggering and sustaining higher rates of economic growth.
World Bank and IMF (2005, Box 2.6).
In cross-country panel studies, Li and Zou (2002) found a negative impact of higher inflation on income distribution, while Bulíř (2001) found evidence of such effect for a reduction in inflation to less than 40 percent. Fischer and Easterly (2001) used polling data to show that across countries the poor are more likely than the rich to report inflation as a personal problem.
In addition, tight monetary policy could raise real interest rates on government debt, lowering the room for spending. This channel is not considered in this paper. These effects of monetary policy on the government’s intertemporal budget constraint should be distinguished from controls over government borrowing, which affect the distribution over time of primary government spending.
The 1995–2003 average standard deviation of the annual growth rates of export volumes, the terms of trade, and real GDP deviated significantly (at a 1 percent level for the first two variables and at a 5 percent level for the latter) between low-income countries and advanced economies, and for the terms of trade also relative to transition economies (at a 10 percent level).
Another risk is that target bands may be less effective at helping shape expectations.
However, in low-income countries, volatile food prices often constitute a large share of the overall CPI, severely limiting the operational relevance of measures of core inflation.
It should be noted that more than one operating target may be employed simultaneously, reflecting multiple policy objectives (see below) as well as uncertainty regarding transmission channels. See, for example, Stone (2003).
As of mid-2004, 26 of the 38 PRGF arrangements incorporated this NDA-NFA framework. Under three programs, a ceiling was set on reserve money instead of central bank NDA (Guinea, Rwanda, and Uganda—of which the latter two are included in the sample of mature stabilizers); Vietnam has a ceiling on NDA of the banking system. For Nicaragua, which maintains a crawling exchange rate peg, and for the member countries of regional monetary unions, neither ceiling applies.
In the 20-odd countries with PRGF programs without a hard peg and excluded from the sample, the average monthly change during 2000–03 was 1.8 percent.
Equation (2) in Box 3.2 provides a useful framework for interpreting these data. Rewriting this equation, (projected) inflation can be approximated as a function of the growth rates in broad money (gM), velocity (gv), and GDP growth (gy):
gP = gM + gv – gy,
or substituting for broad money growth using the growth rates of reserve money (gRM) and the reserve money multiplier (gm):
gP = gRM + gm + gv – gy.
Deviations from the current-year projections of money growth and inflation also do not exhibit a significant positive correlation. The largest outliers relate to the Mozambique program. Following the floods in 2000, a monetary expansion pushed up broad money growth to 44 percent, after which (year-end) CPI inflation peaked at 22 percent in 2001.
The program design review study did not find a significant effect of deviations from projected money growth on inflation for the full sample of PRGF programs, but such effect was found in the subsample of cases in which money growth exceeded the projected level and also on inflation in the following year. By contrast, for the mature stabilizers, no such effect is found when the sample is limited to cases with an overrun in money growth.
A further difference with the current analysis is that the program design review was based on data up to 2000, and thus mainly reflected experience under Enhanced Structural Adjustment Facility (ESAF)-supported programs rather than PRGF-supported programs.
See also IMF (2004f, Box 5).
Further evidence is provided by tests for weak efficiency (Appendix II)—testing if a better projection could be made given a data set that includes just the projection itself—that reject this hypothesis for the projections of money growth and velocity, but not for inflation.
IMF (2004f, Section III.B).
This is compared with an average deviation of 2.7 percentage points. The RMSE is a measure of size of forecasting accuracy, giving the size of the typical deviation, based on a quadratic loss function, and thus weighting large deviations relatively heavily. When the RMSE significantly exceeds the average deviation, there are some deviations that differ significantly from the average.
This is in line with other recent studies (IMF 2004d, Box 3), which have found that the multiplier has remained fairly stable and predictable across all types of IMF-supported programs.
This is reflected in an RMSE of its percentage growth rate of 9.4 for the final staff report, compared with an average deviation of 0.7 percentage point.
Deviations from the projected year-end NDA or NFA level need not indicate that program conditions were missed. First, the deviation may be covered by a program adjuster. Second, generally, only two of the four end-quarter projections serve as performance criteria under the program, whereas the other two merely function as indicative targets.
For a given deviation from the targeted NFA level, most was offset through a divergence in NDA in the opposite direction, with an offset coefficient of about 0.8.
Interpretation is, however, complicated by the endogenous nature of exchange rate stability, which reflects not only the policy orientation but also the incidence of exogenous shocks.
This was discussed in the most recent IMF staff reports for Azerbaijan, Ethiopia, Guyana, and Tanzania. This advice is not inconsistent with the evidence on the importance of avoiding overvaluation, as discussed in IMF (2005d). The adverse effects of large-scale intervention and sterilization imply that if overvaluation (and “Dutch disease”) is an overriding concern, sharp increases in aid-based spending should be avoided.
See also IMF (2005c).
Similarly, an ambitious target for building up international reserves, or a combination of large-scale intervention and sterilization in response to high aid inflows, can also constrain the room for domestic borrowing and, given government borrowing, for the provision of credit to the private sector. The other important reason for limiting domestic financing is its high cost—particularly relative to the external financing on concessional terms, which is available to most of the mature stabilizers.
This implies an elasticity of PSCG with respect to GDP growth of close to 2 during 2000–03. Over the longer period of 1995–2003, this elasticity was lower (1.5) and below an estimated elasticity of 1.7 for some 146 developed and developing countries over 1995–2003.
In the case of the Kyrgyz Republic, the decline in credit is due to the liquidation of two loss-making state banks; PSCG was positive in real terms between 1997 and 2003. In Madagascar, it reflects poor and highly volatile growth performance.
