Chapter

IV. Revenue Growth and the Strength of Underlying Fiscal Positions

Author(s):
International Monetary Fund. Western Hemisphere Dept.
Published Date:
November 2007
Share
  • ShareShare
Show Summary Details

Fiscal balances in many Latin American countries improved steadily between 2002 and 2006. Initially, this reflected a reduction in expenditures as a share of GDP, which reached a low around 2004. Although spending picked up again in 2005 and 2006, fiscal balances continued to improve, reflecting an even greater increase in revenue growth. However, revenue ratios now appear to have stabilized, while spending growth continues unabated in many countries in the region. As a result, average fiscal balances are projected to weaken this year for the first time since 2002, with a further deterioration expected next year.

In light of continued relatively high debt levels, the region’s history of fiscal weakness, and the implications of weak fiscal positions for macroeconomic volatility, a return to primary deficits in Latin America would be a cause for significant concern. Whether or not this prospect is likely to materialize in the next few years depends on two considerations:

  • First, are current expenditure trends set to continue? If spending growth continues at the average rate of 8–10 percent (in real terms) experienced in the last two years, the region will probably return to primary fiscal deficits within 2–3 years, even if revenues remain at their current buoyant levels. To make room for higher capital expenditure and stabilize fiscal balances, growth of current expenditures will need to be curtailed and better targeted, particularly to social spending focused on poverty reduction.

  • Second, are relatively high revenues here to stay? Answering this question requires a careful analysis of whether recent changes in revenue ratios mostly reflect “structural” shifts such as changes in tax policy, tax administration, and commodity price changes that are likely to be permanent, or temporary factors such as cyclical tax buoyancy, or commodity price increases that may be reversed over the medium term.

Fiscal Developments, 2002-07

(In percent of GDP) 1/

200220032004200520062007
Commodity Producers 2/
Public sector revenue24.425.526.728.630.930.1
Commodity revenue4.56.06.58.19.89.0
Noncommodity revenue19.919.420.220.521.021.0
Public sector expenditures28.727.126.927.428.529.4
Current23.522.021.521.321.621.7
Interest4.83.63.22.62.52.5
Capital5.25.25.46.06.97.8
Public sector overall balance-4.3-1.7-0.11.22.40.6
Public sector primary balance0.52.03.13.94.93.1
Noncommodity Producers 2/
Public sector revenue24.224.524.425.226.226.3
Public sector expenditures28.228.227.026.927.127.7
Current23.623.422.422.322.722.8
Interest3.83.83.43.23.02.8
Capital4.74.84.64.64.44.9
Public sector overall balance-4.0-3.7-2.6-1.7-0.9-1.4
Public sector primary balance-0.30.20.81.52.21.4
Source: IMF staff estimates.

Unweighted averages.

Commodity producers: Argentina, Bolivia, Chile, Colombia, Ecuador, Mexico, Peru, Trinidad and Tobago, and Venezuela. Noncommodity producers: Brazil, Costa Rica, Guatemala, Honduras, Nicaragua, Panama, Paraguay, El Salvador, and Uruguay.

Source: IMF staff estimates.

Unweighted averages.

Commodity producers: Argentina, Bolivia, Chile, Colombia, Ecuador, Mexico, Peru, Trinidad and Tobago, and Venezuela. Noncommodity producers: Brazil, Costa Rica, Guatemala, Honduras, Nicaragua, Panama, Paraguay, El Salvador, and Uruguay.

The remainder of this chapter provides such an analysis of structural revenues and fiscal balances.12 Reflecting the special role of commodity-related revenues in the recent increase in revenue ratios in many Latin America countries, commodity revenues and noncommodity revenues are analyzed separately. To assess the medium-term prospects for commodity revenues, current revenues are adjusted in line with expected movements in relevant commodity prices. Noncommodity revenues are analyzed using econometric techniques, decomposing the observed revenue ratio into the level that would be expected, given the current state of the tax system, if the economy were in a “neutral” cyclical position; the portion that is attributable to the business cycle; and a residual, which could reflect other factors bearing on the tax ratio—such as changes in tax compliance or relative price movements—that may or may not be temporary. Finally, the results from these two steps are combined with data on noninterest expenditures to generate an overall view of the “structural” primary fiscal balance, that is, the primary balance if output were at a cyclically neutral position, commodity prices were at their medium-term expected prices, and both the tax system and expenditures were to remain unchanged. The structural primary balance hence gives a sense of the “underlying” strength of the fiscal position, abstracting from cyclical or temporary factors.

Public Sector Expenditures

(Percent of GDP) 1/

Source: IMF staff estimates.

1/ Unweighted averages for 17 countries.

The main results are as follows. First, for most countries analyzed, structural primary balances are currently weaker than actual (i.e., reported) primary balances. Hence, focusing on actual balances in Latin America somewhat exaggerates the strength of the underlying fiscal position. This is particularly true for nonfuel commodity exporters, since nonfuel commodity prices are projected to decline significantly in the medium term. Second, in most countries in which the primary balance is currently reported to be in surplus, the structural primary balance is also likely to be in surplus, albeit at a lower level. Since these are presently the majority of countries in the region, this is also true on average. Third, structural balance calculations in some countries are subject to a large margin of uncertainty, either because of uncertain commodity price projections, or because a significant portion of recent changes in noncommodity revenues as a share of GDP cannot be easily attributed either to cyclical conditions or to changes in the tax system. The assessment of the fiscal position in these countries hence depends on how this “residual” change in the revenue ratio is interpreted. Finally, the prospects for maintaining a structural primary surplus going forward will first and foremost depend on countries’ success in curbing the pace of spending growth. If spending were to continue to grow at its current pace, it is possible that the region’s structural surplus might disappear as early as next year.

Structural Commodity Revenues

Unlike tax revenues related to economic activity, such as income or consumption, commodity-related revenues depend on commodity prices and production volumes.13 The latter are usually viewed as “structural,” in the sense that they are determined by natural resource endowments and policy decisions. Hence, estimating structural commodity revenues is a matter of adjusting actual commodity revenues for medium-term expected changes in commodity prices. As a result, the estimation of structural commodity revenues is sensitive to the commodity price forecasts that enter in the adjustment.

To obtain an idea of this sensitivity, the analysis in this chapter uses two sources for price forecasts: IMF projections of commodity prices, which are primarily based on futures prices, and are available over a five-year period; and projections from the World Bank, which are based on an econometric model, and are available for a somewhat longer period, until 2015. Using data from these two sources, commodity price indices were constructed for nine major commodity producers in the LAC region—defined as countries with at least 2 percent of GDP commodity revenue.14 “Structural commodity revenue” was then defined as actual revenue divided by the ratio between current and five-year expected average commodity prices, for each country, according to either IMF or World Bank projections.

As is clear from the charts, the commodity price projections have a significant impact on the results. IMF projections for commodity prices suggest that structural commodity revenues have been rising in line with actual commodity revenues in most of the countries analyzed. (This is less true for Chile—as copper prices are projected to decline substantially over the medium term—and also for Peru and Bolivia, where commodity price declines are also expected to reduce revenue somewhat). In contrast, the World Bank projects lower medium-term prices for most commodities exported from Latin American countries, including energy. This translates into a generally more pessimistic view of underlying structural commodity revenues.

The difference across the two sets of projections matters particularly for oil producers. IMF projections, based on futures markets, forecast continued high oil prices in the medium term. But the model-based projections by the World Bank envisage a significant decline in oil prices (on the order of 25 percent) over the next five years, and an even bigger drop by 2015. As a result, structural commodity revenues estimated using these price projections are currently below actual commodity revenues by 3–4 percent of GDP in Venezuela and Trinidad and Tobago, and by around 2 percent in Ecuador and Mexico. In contrast, estimates based on the IMF price projections imply that current revenue levels will be sustained over the medium term, provided that production volumes are maintained.

Actual and Structural Commodity Revenues

(In percent of GDP) 1/

Sources: IMF staff calculations based on data from national authorities; World Bank Commodities Unit; UNCOMTRADE database; and WEO.

1/ Cyclical adjustments based on HP-filtered trend output with smoothing parameter = 6.25 (see Box 2).

2/ Actual revenue adjusted by the ratio of current to 5-year expected average commodity prices according to IMF (World Bank) projections.

Structural Noncommodity Revenues

The standard approach to estimating “structural” noncommodity revenues (see, for example, Hagemann, 1999) is to apply a cyclical adjustment to the reported revenue ratios. The extent of this cyclical adjustment will depend on two factors (see Appendix for details): the cyclical position—that is, whether the economy is deemed to be far away from a neutral cyclical state or not, see Box 2—and the “income elasticity of revenue,” which measures how much revenues tend to respond to changes in economic activity. If the income elasticity of revenue is 1—that is, revenues respond proportionally to changes in output—then the economic cycle has no impact on the revenue-to-GDP ratio, and the reported ratio will be deemed entirely “structural.” Similarly, if the income elasticity of revenue is different from 1 but actual output is close to potential output, then any cyclical adjustment to the reported revenue ratio will also be very small.

As it turns out, for the eight countries whose revenues are analyzed in this section15 the cyclical adjustment is very small for one or the other of these reasons. For three countries, estimated income elasticities were very close to 1 (a standard result). For four others, Costa Rica, Colombia, El Salvador, and Peru, the estimated elasticity was between 1.1 and 1.2; and for Panama it was 0.8 (see Appendix). However, the “output gaps” (deviations from neutral cyclical positions) for these countries are currently estimated to be small. As a result, any cyclical adjustment to revenue/GDP ratios is very minor. This methodology would therefore suggest that current noncommodity revenue ratios in these countries should be viewed as almost entirely “structural,” and hence permanent.

However, this conclusion is subject to an important caveat. In the standard cyclical adjustment approach, any change in revenue ratio that is not identifiably cyclical is assumed to be “structural,” whether or not it can be accounted for by changes in the tax system. This constitutes a potentially significant weakness in the methodology, as some of the supposedly noncyclical changes in revenue may well be due to one-off factors, which could be reversed in the future. The standard analysis was therefore extended by an additional step, which sought to quantify the impact of identifiable changes in the tax system on revenue, on the basis of either an econometric approach or direct estimates from country authorities or IMF staff (see Appendix).

In three of the eight countries analyzed here, identifiable changes in the tax system together with identifiable cyclical effects explain almost all of the recent changes in revenue ratios. But in the remaining countries, unexplained residuals are significant—about 1 percentage point of GDP or more. Argentina and Panama, in particular, show large positive residuals in 2006 and 2007—that is, reported revenue is higher than can be explained by identifiable cyclical or identifiable structural factors—while for Chile one finds a large negative residual. These residual revenues could be viewed as transitory, reflecting one-off increases or decreases in revenue that are likely to disappear, at least over the medium term. But they could also reflect unaccounted but nonetheless permanent structural changes, for example, unaccounted improvements in tax administration.

Changes in Noncommodity Revenue, 2002-06

(In percent of GDP) 1/

TotalIdentifiably cyclical 3/Identifiably structural 4/Residual
Argentina 2/2.870.000.422.46
Brazil0.280.000.68-0.40
Chile-2.580.000.04-2.63 5/
Colombia2.670.021.401.25
Costa Rica1.120.051.10-0.03
El Salvador2.240.002.180.06
Panama1.72-0.03-0.392.14
Peru0.020.020.95-0.96
Source: National authorities, and IMF staff calculations.

Refers to central government revenue.

2003-2006, as 2002 was a crisis year.

Based on estimation; see appendix.

Owing to changes in the tax system and tax progressivity.

Reflects increases in the GDP deflator driven by copper prices (see text).

Source: National authorities, and IMF staff calculations.

Refers to central government revenue.

2003-2006, as 2002 was a crisis year.

Based on estimation; see appendix.

Owing to changes in the tax system and tax progressivity.

Reflects increases in the GDP deflator driven by copper prices (see text).

Information about the specific source of revenue strength or weakness can sometimes provide clues. In the case of Chile, for example, an unexplained reduction in the ratio of noncommodity revenue to GDP appears to be driven not by lower revenues but by an increase in the GDP deflator, as copper price increases have outpaced the increase in consumer prices that drive VAT collection. For Argentina, the unexplained portion of the surge in revenues could be a result of improvements in tax administration or of faster economic growth in sectors with a higher tax burden. Improvements in tax administration could also play a role in explaining the surge in Panama.

Statistical criteria can also help decide whether unaccounted surges should be viewed as transitory. Intuitively, a rise in residual revenues is more likely to reflect a structural shift if it is large relative to past fluctuations in the revenue-to-GDP ratio, and if it has already been sustained for several years. Formal statistical tests suggest that for most countries analyzed, residual changes in revenue tend to disappear over time (they appear to be “stationary”; see Appendix for details). For the most part, therefore, it is wise to treat residual revenue buoyancy as temporary rather than structural. One significant exception is Argentina, for which statistical tests cannot reject that residuals persist over time.

To summarize, while standard cyclical adjustment suggests that current high levels of noncommodity revenues in many countries are largely structural, not all of these increases can be accounted for by changes in the tax system. Based on their statistical properties, these unaccounted “residuals” should generally be regarded as temporary. One exception is Argentina, which experienced a large shift in revenue several years ago—with little tendency to revert over time—which at this point is most plausibly regarded as structural.

Actual and Structural Noncommodity Revenues

(In percent of GDP) 1/

Sources: IMF staff calculations based on data from national authorities; World Bank Commodities Unit; UNCOMTRADE database; and WEO.

1/ Cyclical adjustments based on HP-filtered trend output with smoothing parameter = 6.25 (see Box 2).

2/ Structural revenues identified through regression analysis of a tax policy adjusted revenue series.

Expenditures

Unlike in industrial countries, expenditure commitments tied to the economic cycle (for example, unemployment benefits) do not play a major role in Latin American budgets. For this reason, the standard approach in the fiscal literature on Latin America is to assume that all government expenditure in these countries is “structural,” i.e., driven by policy, without an automatic countercyclical link to output and employment. In fact, expenditures in Latin America have in the past tended to be procyclical (Clements, Faircloth, and Verhoeven, 2007). Governments took advantage of buoyant revenue to expand expenditure in good times, and were forced to compress it in bad times owing to weak fiscal positions, and resultant borrowing constraints or high borrowing costs. Looking forward, we examine how strong underlying fiscal positions in the region would be now on the assumption that primary expenditures are independent of the cycle.

Even if expenditures are assumed to be noncyclical, computing the expenditure to GDP ratio when the economy is in its cyclically neutral position—in analogy to the structural revenue ratios shown in the figures of the previous sections—requires an adjustment to the observed expenditure to GDP ratio.16 However, for most countries in the region the gap between actual GDP and potential GDP has not been large in recent years. Consequently, spending ratios expressed as a share of potential output closely track reported spending ratios. Both concepts show a sharp rise in spending ratios in many countries in the region, particularly among the energy exporters, and the larger economies. Exceptions include Chile, Panama, and El Salvador.

Structural Primary Balances

Using the structural revenue estimates and noninterest expenditure data, it is possible to derive a set of tentative estimates of structural primary balances. These give a sense of the overall strength of the fiscal position in the countries analyzed. Two additional assumptions need to be made to overcome data difficulties. First, for commodity producers whose noncommodity structural revenues could not be analyzed in detail, the standard methodology for calculating cyclically adjusted revenues is applied using the average revenue elasticity estimated for the other countries in the region (about 1.1). Second, since commodity revenues often accrue to public enterprises, we focus here on overall public sector structural balances. This requires taking a view on structural, or permanent, noncommodity revenues outside the central government. For this purpose, we assume that all noncyclical changes in such revenues are structural.

Actual and Estimated Structural Primary Balances

(In percent of GDP) 1/

Sources: IMF staff calculations based on data from national authorities; World Bank Commodities Unit; UNCOMTRADE database; and WEO.

1/ Cyclical adjustments based on HP-filtered trend output with smoothing parameter = 6.25 (see Box 2).

2/ “Conventional” estimates view all noncyclical revenues as structural and use IMF commodity price projections; “alternative” excludes regression residual from definition of noncommodity structural revenues and uses World Bank projections.

The evolution of alternative structural balance estimates since 2002 allows some general conclusions:

  • With few exceptions, 2006 and projected 2007 structural primary balances remain in positive territory, although with large margins of uncertainty. In some cases, uncertain and conflicting commodity price projections, particularly for oil, cloud the picture.17 In others, including Argentina, Panama, and Colombia, there is uncertainty on how much of the large recent increases in noncommodity revenue should be viewed as permanent—although this uncertainty can in some cases be narrowed by examining the statistical properties of the estimated residuals, as discussed earlier.

  • However, a significant deterioration of both actual and structural fiscal balances is expected for this year, driven by sharply rising expenditure ratios.

Conclusions

Many Latin American countries made impressive improvements in fiscal positions between 2002 and 2006. This has been a crucial factor in reducing vulnerability to external shocks, as evidenced by the region’s relative resilience so far following the recent financial market turbulence. Following an initial period earlier in this decade of expenditure restraint, these fiscal improvements have recently come from the revenue side. These improvements have in part been driven by tax policy and administrative changes but also by rising commodity revenues, and in some cases by increases in noncommodity revenues that are difficult to explain either by cyclical conditions or explicit policy and administrative improvements. Statistical analysis suggests that—for the most part—these residual increases in revenue should be viewed as temporary.

Underlying structural primary balances still generally appear to be in surplus in the countries in which reported balances are in surplus. However, these surpluses are more modest than the unadjusted data indicate, particularly in commodity-exporting countries facing possible declines in prices over the medium term. Furthermore, primary surpluses will significantly shrink this year, as expenditure ratios continue to rise while revenue ratios stabilize. If expenditure growth is not curtailed, fiscal balances will quickly erode, and the region could soon return to primary deficits.

Actual and Structural Primary Balances, Public Sector(In percent of GDP)
20062007
ActualStructural 1/Pro-Structural 1/
MaxMinjectedMaxMin
Argentina4.03.70.03.32.82.1
Bolivia7.25.83.63.41.50.2
Brazil3.94.13.13.63.82.5
Chile8.55.73.28.86.33.7
Colombia3.43.11.13.63.40.7
Costa Rica3.43.22.81.41.20.3
Ecuador5.14.92.81.31.4-0.4
El Salvador-0.4-0.3-0.70.10.3-0.4
Mexico1.71.5-1.31.61.6-0.9
Panama4.94.83.43.93.92.5
Peru4.14.22.22.42.20.5
Trinidad & Tobago8.87.92.34.84.1-0.3
Venezuela0.60.0-5.6-5.2-3.7-9.9

Estimates. “Max” uses most favorable commodity price forecasts and assumes that any positive (negative) revenue “residuals” are permanent (transitory). “Min” uses least favorable commodity price forecasts and assumes that any positive (negative) revenue “residuals” are transitory(permanent) except for Argentina, where Min estimates assume that the (positive) residuals are permanent, based on the statistical properties of the residuals (see text and appendix table). If the residuals for Argentina are assumed to be transitory, the “Min” estimate for Argentina would be reduced to -1.4 from 2.1.

Estimates. “Max” uses most favorable commodity price forecasts and assumes that any positive (negative) revenue “residuals” are permanent (transitory). “Min” uses least favorable commodity price forecasts and assumes that any positive (negative) revenue “residuals” are transitory(permanent) except for Argentina, where Min estimates assume that the (positive) residuals are permanent, based on the statistical properties of the residuals (see text and appendix table). If the residuals for Argentina are assumed to be transitory, the “Min” estimate for Argentina would be reduced to -1.4 from 2.1.

In some countries, better control of expenditures may require institutional or structural reforms. International experience suggests that expenditure rules can be helpful but need to be supported by broad political consensus, consistent revenue policy, and in some cases expenditure reforms. Depending on country circumstances, such reforms could include reducing budgetary rigidities (Alier, forthcoming), increasing expenditure efficiency and flexibility, and strengthening public financial management systems.

Appendix. Estimation of Structural Revenue and Structural Balances

Methodology

Commodity Structural Revenue

Commodity revenues depend on commodity production or export volumes, prices, and the fiscal regime. Fiscal regimes and production/export volumes are taken to be part of the “structure” that in principle is under the control of the authorities. Hence, making the same functional form assumption as is commonly made in the literature on noncommodity structural revenue (see below), Rtc=βptγ, and Rs,tc=βpt*γ, where pt* is the long-run commodity price expected at time t, Rtc stands for commodity revenues, β and γ are parameters, and the subscript s is used to denote structural revenue. Substituting, β, it follows that

(1)

.

In the absence of reliable country-specific estimates for γ, this study follows the Chilean approach (Marcel and others, 2001) and assumes that γ = 1 for all countries. pt* is based on fiveyear-ahead commodity price forecasts published separately by the IMF (in the World Economic Outlook) and the World Bank—i.e., pt*=E[pt+5]t. In either case, export-share-weighted commodity price indices were created for each country, so that pt* is a weighted average of the expected prices of each commodity exported by that country.

Note that the above relationships do not explicitly recognize the role of the exchange rate in translating dollar commodity revenues into local currency revenues (implicitly, the exchange rate is subsumed in the parameter β ). This is admissible so long as the exchange rate is close to its equilibrium value, which is an acceptable assumption for most countries studied here. If exchange rates are not close to equilibrium, this would create an additional reason why structural commodity revenues could be different from actual revenues. An undervalued exchange rate implies that structural commodity revenues are lower than actual revenues, while an overvalued currency implies that they are higher.

Noncommodity Structural Revenue

The standard approach to estimating noncommodity structural revenue (see, in particular, Hagemann, 1999; see also Chalk, 2002) starts by assuming a constant elasticity relationship between revenue, R, and its tax base (for example, GDP or national income, denoted Y):

.

Using a star to denote potential output, it follows that Rs,tn,c=AYt*ɛ . Substituting the parameter A, structural revenue can hence be estimated by applying a simple cyclical correction to actual revenue:

where the ɛ is either an estimate of the revenue elasticity using time-series data for Rt and Yt or an assumed value (most studies indicate that ɛ is in a narrow range between about 1 and 1.25).

A potential problem with this approach arises from the fact that, for any specific ɛ and A, RtncAYtɛ ; that is, Rtnc/YtɛAt is not constant over time. However, it could still be the case that Rs,tnc=AYt*ɛ, i.e., that the parameters A andH define the (unobservable) relationship between structural balances and potential output. In that case, computing structural balances as

amounts to assuming that A = At—that is, that the “structural” parameter A shifts in every period in line with the actual realization of revenues in relation to GDP. In other words, in the standard approach, any change in revenues that cannot be explained by cyclical factors is considered structural.

This is implausible in many cases, as fluctuations in At may reflect one-off or other factors that are reversed over time, consistent with a long-run stable A. Indeed, it could be the case that there are no structural breaks in A other than these associated with identifiable policy actions, which can be accounted for using dummy variables or by adjusting the data. This hypothesis can be tested by testing for the existence of a long-run “co-integrating” relationship between the (adjusted) RtncYt series and Yt that is, testing the proposition that although Rtnc and Yt are “integrated”—i.e., follow a stochastic trend—the residual from a regression of one on the other is stationary. If this can be confirmed, it implies that any change of (tax-policy-adjusted) revenues that cannot be explained by cyclical factors should be considered temporary.

Structural revenues would then be given by the fitted values in the regression of Rtnc on Yt, evaluated at a cyclically neutral level,

(4)

plus any effect of tax policy changes that was previously removed from the series in order to estimate the parameters in the above equation.

This study uses both approaches, that is, Equations (3) and (4), to derive alternative estimates of the noncommodity structural revenues, based on the output gap estimates presented in Box 2, and the parameter estimates discussed below (see Vladkova-Hollar and Zettelmeyer, 2007). Underlying the figures in Chapter 2 is a set of potential output/output gaps based on the Hodrick-Prescott filter with the “smoothing parameter” λ set at 6.25. Output gaps according to the other concepts discussed in Box 2 were used for robustness checks. Final results are expressed as a share of GDP (see below).

Structural Balances

By definition, the structural balance equals structural revenue minus structural expenditures:

(5)

where B stands for balance and E for expenditure (or noninterest expenditure, if the focus is the structural primary balance), and the remaining notation is unchanged. In line with the literature on Latin America (see, for example, Alberola Ila and Montero, 2006), it is assumed that all expenditure is structural: Es, t = Et. Structural revenues are computed as described in the previous two subsections.

The structural balance as a share of GDP is obtained by dividing both sides by potential output Yt*Yt/g where g denotes the output gap (expressed as a ratio). Using lower-case letters to denote shares of current GDP and setting γ=1, this yields, based on Equation (3):

(6)

and based on Equation (4):

(7)

Estimation

Estimation of noncommodity structural revenues according to Equation (4) involved the following steps.

For countries considered commodity producers (which, in this context we define as having commodity-related revenues in excess of 2 percent of GDP), the central government revenue series was adjusted to exclude commodity revenues. For each country, the resulting noncommodity-related central government revenue series was then adjusted for the impact of changes in tax policy. The preferred methodology was to directly adjust the revenue series for the effect of changes in the tax structure through available impact estimates (IMF staff reports, working papers, country tax authorities’ estimates, authors’ own estimations). In cases in which a direct estimate could not be obtained, the effect of changes in the tax structure was controlled for through step dummies. Particular effort was made to avoid dummies in the latter part of the sample (2004–06), where a dummy would likely pick up some of cyclical improvement along with the effect of tax policy. This could not be avoided in the case of Costa Rica, where the introduction of automation in customs administration was launched in 2005.

The ordinary-least-squares estimates of the long-run policy-adjusted income elasticity of noncommodity central government tax revenues suggest that the estimated income elasticity is significantly different from unity only in Brazil and Costa Rica. In Panama, the lower point estimate of the income elasticity of tax revenues could perhaps be explained by the lack of full inclusion of the some dynamic sectors of the economy in the tax base. In most countries there is evidence of co-integration according to both the Johansen and the Engle-Granger tests. The main exceptions are Argentina, Chile, and Panama. However, for Chile and Panama the null hypothesis that the residuals in the relationship between GDP and revenue are stationary could not be rejected, whereas it could be rejected at the 5 percent level in the case of Argentina.

Long-Run Income Elasticity of Central Government Tax Revenues(OLS estimates, standard errors in parentheses)
Evidence for Cointegration
Estimated income elasticity(Number of CI vectors)ADF test (test statistic) 1/KPSS test (test statistic) 2/
Trace testMax. Eigenvalue
Argentina0.9910-2.360.26**
(0.01)
Brazil1.0311-1.960.18*
(0.01)
Chile1.0100-2.420.14
(0.03)
Colombia1.0911-2.110.21**
(0.05)
Costa Rica1.1111-4.92***0.17
(0.02)
El Salvador1.1611-2.6090.09*
(0.25)
Panama0.8000-1.950.11
(0.16)
Peru1.1511-3.670.23***
(0.07)

*, **, and *** denote rejection of the null hypothesis that the residuals are nonstationary at the 10%, 5%, and 1% significance level, respectively. Critical values for the Dickey-Fuller t-statistics when applied to residuals are taken from Hamilton (1994), Table B.9.

*, **, and *** denote rejection of the null hypothesis that the residuals are stationary at the 10%, 5%, and 1% significance level, respectively. Critical values for the KPSS LM-statistics when applied to residuals are taken from Shin (1994), Table 1.

*, **, and *** denote rejection of the null hypothesis that the residuals are nonstationary at the 10%, 5%, and 1% significance level, respectively. Critical values for the Dickey-Fuller t-statistics when applied to residuals are taken from Hamilton (1994), Table B.9.

*, **, and *** denote rejection of the null hypothesis that the residuals are stationary at the 10%, 5%, and 1% significance level, respectively. Critical values for the KPSS LM-statistics when applied to residuals are taken from Shin (1994), Table 1.

See also related work by Izquierdo, Ottonello, and Talvi (forthcoming).

“Commodity revenues” are defined as fiscal revenues that can be attributed to the activities of commodity-producing industries, whether from income and profit taxes, VAT, royalties, or export taxes. In the analysis that follows, commodity revenues are only analyzed for countries in which they exceed 2 percent of GDP.

Namely, Argentina, Bolivia, Chile, Colombia, Ecuador, Mexico, Peru, Trinidad and Tobago, and Venezuela.

Argentina, Brazil, Colombia, Costa Rica, Chile, El Salvador, Panama, and Peru. The sample was limited because the analysis requires extensive information on changes in the tax system and their revenue impact, which is available only for some countries.

When actual GDP is above trend (or “potential”) GDP, the reported spending-to-GDP ratio is reduced, as the denominator is larger than it would be at potential, and vice versa.

The calculations for Chile are broadly consistent with the government’s structural surplus target when similar commodity price projections are used (note that the government’s surplus target refers to the overall structural balance of the central government whereas the results here refer to the structural primary balance for the overall public sector).

    Other Resources Citing This Publication