Fiscal Positions in Latin America: Have They Really Improved?

The improvement of fiscal balances has been a cornerstone of Latin America’s macroeconomic stabilization and recovery following the crises of 1994–2002. After hovering around 5 percent in the last decade, the average fiscal deficit in Latin America declined steadily beginning in 2003, and reached a surplus in 2006. Initially, this reflected a reduction in government spending as a share of GDP, which reached a low around 2004. Beginning in 2005, spending picked up again, but was outpaced by an even larger increase in revenue growth. As a result, fiscal balances continued to strengthen (Table 1).

image002_400.png Are these improvements sustainable? Or do they merely reflect exceptionally favorable external economic conditions—including an unprecedented boom in commodity prices—and the strong cyclical recovery that Latin America has enjoyed in the last few years? The fact that Latin America’s recent fiscal improvements have come exclusively from the revenue side is a cause for concern (IMF, 2006, 2007; IADB, 2008; Izquierdo, Ottonello, and Talvi, forthcoming). In light of continued high debt levels, a return to the deficits of the 1990s could jeopardize the region’s newfound stability. The key question is hence how much of the recent revenue growth can be expected to be “permanent”—i.e. to survive a return to normal cyclical conditions—and how strong fiscal balances would be in these circumstances. This is the subject addressed in this paper.

We proceed in three steps. First, we analyze the sources of recent increases in the revenue to GDP ratio, distinguishing between revenues from commodity and noncommodity sources, and decomposing increases in the latter into three components: changes due to tax policy or tax administration; changes due to the economic cycle; and a residual. Based on this analysis as well as medium-term commodity price projections from two different sources, we compute “structural” revenue to GDP ratios separately for noncommodity and commodity revenues. Finally, we combine these with estimated structural expenditure ratios, under the assumption that expenditures in Latin America are not (automatically) linked to the economic cycle, to compute structural balance estimates for a number of countries in the region.

The approach used in this paper differs from standard structural balance methodology (Hagemann, 1999) in two ways. First, following Marcel et al. (2001), we distinguish between noncommodity and commodity revenues, and separately estimate the “structural” level for each. Second, when estimating noncommodity structural revenue, we consider the history of tax regime changes in each country in addition to the standard cyclical adjustment of observed revenues using the output gap. Conventional structural balance methodology implicitly assumes that all changes in the revenue ratio that are not identifiably cyclical— that is, cannot be statistically linked to the output fluctuations—are “structural,” whether or not they can be attributed to changes in the tax system. This approach might give too rosy a picture of the fiscal balance if revenue ratios are buoyant for temporary, but not identifiably cyclical reasons.

To address this problem, we adjust the observed tax revenue series, for each country, using the estimated revenue impact of all changes in the tax system that we are aware of, before regressing the adjusted series on changes GDP.2 The residual from this regression reflects the portion of revenue that is unexpected, given the state of both the tax system and the tax base (GDP). Structural revenue and balance estimates are computed both under the (conventional) assumption that this residual is structural, and under the alternative view that it is not.

We also take a position on which view is closer to the truth by examining the statistical properties of the residual. We are aware of four related recent studies of fiscal performance in Latin America. Alberola and Montero (2006) examine the relationship between the fiscal stance and the economic cycle in a paper that is primarily interested in debt sustainability. As an intermediate step, they estimate structural fiscal balances in nine Latin American up to 2004 using the standard assumption that all non-cyclical revenue changes are structural, and disregarding changes in tax structure in their regressions.3 Lozano and Toro (2007) compute structural balances for Colombia, based on the standard approach, and a revenue series that is adjusted for changes in the tax structure. Cubero and Sowerbutts (forthcoming) analyze structural revenue in Costa Rica using a very similar methodology as this paper, with consistent results. Finally Izquierdo, Ottonello, and Talvi (forthcoming; see also IADB, 2008, which is based on their analysis) calculate structural balances for a group of Latin American countries using a different methodology, which relies on statistical filtering of the observed fiscal data. The flavor of their results is different from those of this paper, in that they attribute a much larger portion of the recent revenue increase to cyclical factors.4

Our study has three main findings.

First, not surprisingly, commodity related revenues play an important role in the recent revenue boom of commodity producing countries. Whether or not these revenue increases should be viewed as permanent or not depends on the commodity. For fuel commodities, medium term projections by the IMF and World Bank envisage largely flat prices in the medium run. For non-fuel commodities, declines are envisaged, particularly for some metals. Furthermore, the assessment of whether non-fuel commodity price increases should be viewed as permanent or not turns out to depend on the forecast source. Model-based forecasts by the World Bank envisage greater declines over the medium term than IMF projections, which are largely based on futures markets data.

Second, revenue increases that are identifiably due to the business cycle play virtually no role in explaining the rise of the revenue-to-GDP ratio. The main reason is that estimated income elasticities of revenue are close to unity in most cases (they range between 0.8 and 1.35, with most elasticities clustered between 0.95 and 1.11). Hence, while noncommodity revenue levels are highly cyclical, revenue ratios should be quite insensitive to the cycle. Moreover, the estimated cyclical position of most Latin American countries is currently not very far from neutrality, namely in the order of 0–4 percent above “potential output.” Hence, a return to a cyclically neutral position would not have a big impact on revenue ratios.

Third, residual revenue changes that can be attributed neither to cyclical factors not to identifiable changes in the tax regime are quite large in a handful of countries, in the order of 1–3 percent of GDP. In these countries, structural balance estimates are sensitive to whether these residuals are interpreted as reflecting unobserved structural changes, or as temporary. Statistical tests indicate that for the most part they ought to be interpreted as temporary, but there are some exceptions.

In sum, there is little doubt that fiscal positions in Latin America have “really” improved in recent years. The business cycle cannot have played a significant direct role in raising revenue ratios. Improved fiscal positions seem to mostly reflect persistently higher commodity prices, as well as changes in taxation and tax administration. This said, structural balances in Latin America are weaker than reported balances, particularly in the case of nonfuel commodity exporters, which are projected to suffer significant price declines in the medium term. Furthermore, they are subject to a large margin of uncertainty, both because of uncertain commodity price projections, and because some of the recent changes in noncommodity revenues as a share of GDP are hard to attribute either to cyclical conditions or to changes in the tax system.



(2) This approach follows Swiston, Mühleisen and Mathay (2007). (3) Alberola and Montero (2006) do not take account of changes in revenue regimes when estimating elasiticies of revenue with respect to income and commodity prices. This may explain the fact that their estimated income elasticities of revenue are generally much higher than ours (see section II below). (4) Their methodology consists in applying a Hodrick-Prescott filter with a smoothing parameter calibrated to reproduce the degree of smoothing that is implicit in the structural fiscal balance estimates of the Chilean authorities. Because the commodity prices that drive the Chilean balance exhibit much less persistence than those of other commodities produced in Latin America (see appendix 3), this leads to a far larger adjustment than if structural balances are based on country-by-country properties of commodity prices, as in this study.

Paper published by the IMF and introduction reproduced here with the author’s permission.