Chapter

Chapter 7. Has Fiscal Policy Become Less Procyclical in Latin America?

Author(s):
Dora Iakova, Luis Cubeddu, Gustavo Adler, and Sebastian Sosa
Published Date:
December 2014
Share
  • ShareShare
Show Summary Details
Author(s)
Alexander Klemm

Fiscal policy in Latin America has been procyclical for many decades. The easy availability of funds during periods of economic expansion, against a backdrop of major social and infrastructure needs, has repeatedly prompted rapid increases in government expenditure. But spending often has had to be cut sharply later when economies have fallen into recession or faced a sudden stop of capital inflows. This procyclicality has been empirically documented in a growing literature that started in the late 1990s with Gavin and Perotti (1997). With very few exceptions, studies have found evidence of procyclical fiscal policy in developing and emerging market economies, and especially in Latin America.

However, given the improvements in macroeconomic performance and policy frameworks of many Latin American countries over the past decade, this chapter studies whether fiscal policy has become less procyclical in recent years. In assessing cyclicality, the analysis uses a broader measure of fiscal policy that gives credit for automatic stabilizers and controls for commodity prices. The measure is applied to a large sample of both emerging market and advanced economies. The panel data estimates also take account of the endogeneity of the output gap or growth using instrumental variable approaches, as these are affected by fiscal policy through the multiplier.

Consistent with previous studies, the analysis finds evidence suggesting countercyclical fiscal policy in advanced economies and, generally, procyclical policy in Latin America. For a broader sample of emerging market economies, the evidence is less clear. However, the results are very sensitive to specifications, making it hard to draw firm conclusions. A more detailed analysis of Latin American countries shows that, while many countries appear to have procyclical policy, country-specific empirical estimates typically lack statistical significance. However, the analysis finds that key countries in the region—Brazil, Chile, Colombia, El Salvador, and Mexico—have moved toward less procyclical policy since 2005.

While the focus of this chapter is on the cyclicality of fiscal policy, it should not be forgotten that this is just one aspect of fiscal policy. Other aspects—such as having to deal with a sudden financing constraint or having to reduce debt toward a sustainable level—may trump cyclicality concerns. This would happen even in the absence of uncertainty about the true output gap, which poses additional policy challenges. Hence finding that a country implemented procyclical policy does not mean that it did not take the best possible course of action, given its particular circumstances.

This chapter first covers methodological issues, including the definition of cyclical fiscal policy and the treatment of commodity and asset price shocks and changes in the composition of GDP. It then provides a brief review of existing studies, categorizing them by their explicit or implicit definition of procyclical fiscal policy. The chapter then describes the data used, including a new commodity price index, and presents the empirical results, including panel data regressions for advanced and emerging market economies and a more detailed analysis of Latin America based on country-by-country regressions. This is followed by a brief discussion of issues related to the quality of fiscal policy and then conclusions.

Methodology

The idea behind countercyclical fiscal policy is simple: fiscal policy should be tighter during booms and looser during recessions. To test for this empirically, previous studies have either looked at correlations between fiscal and macroeconomic variables or used a regression approach, which allows further controls. The typical regression relates the change in (a measure of) the fiscal balance to the output gap and a few additional variables:

where B is the fiscal balance, Y is nominal GDP, Y* is potential GDP, x is a vector of other control variables, fi is a country-fixed effect (which may be added in case of estimation on panel data), and ε is an error term. Variants in the literature include using real GDP growth instead of the output gap as a regressor, and focusing on government revenues or expenditures instead of a fiscal balance.

Table 7.1 presents a summary of key studies and their approaches. In these studies, the estimated coefficient on the output gap (β1) is the main indicator of the cyclicality of policy. A negative coefficient is evidence of procyclical policy, as it suggests that the fiscal stance is relaxed in a boom. Conversely, a positive coefficient implies countercyclical policy. With an insignificant coefficient, acyclical fiscal policy cannot be rejected against the alternative hypotheses of procyclical or countercyclical fiscal policies.

Table 7.1Literature Review: Definitions of Countercyclical Fiscal Policy
StudyMethodFinding1
Alesina, Campante, and Tabellini (2008)Regression of change in fiscal balance/spending on output gapOnly advanced Organisation for Economic Co-operation and Development (OECD) economies countercyclical
Catao and Sutton (2002)Regression of change in fiscal balance on output gapMost emerging markets procyclical
Céspedes and Velasco (2011)Regression of change in fiscal balance on output gap and cyclical component of commodity pricesDiversity across countries; some developing economies have become more countercyclical
Daude, Melguizo, and Neut (2011)Correlation between change in cyclically-adjusted primary balance and output gapMost of Latin America procyclical
Di Bella (2009)Regression of change in cyclically-adjusted primary balance on cyclically adjusted primary balance and debt rating during 2009 downturnEconomies with stronger fiscal positions and credit ratings more countercyclical
Egert (2012)Regression of overall balance/cyclically adjusted primary balance (CAPB) on output gap/growthEuropean countries countercyclical (overall balance) or acyclical (CAPB)
Frankel, Végh, and Vuletin (2013)Correlation between cyclical components of real government spending and GDPDeveloping economies more procyclical than advanced, but less than in the past
Gali and Perotti (2003)Regression of CAPB on output gap and debtSome European countries countercyclical, not less than before European Monetary Union
Gavin and Perotti (1997)Regression of change in fiscal balance/revenue/spending growth on GDP growthAdvanced economies countercyclical; Latin America procyclical
llzetzki and Végh (2008)Regression of real spending on real GDPDeveloping economies often procyclical
Jaimovich and Panizza (2007)Regression of fiscal balance or spending on growthAdvanced economies countercyclical; developing ones indeterminate
Kaminsky, Reinhart, and Végh (2005)Difference between spending growth in good and bad times; correlation between spending and growthMost non-OECD and half of OECD countries procyclical
Lane (2003)Regression of government spending on GDPProcyclical policies more likely in economies with volatile output and dispersed political power
Lledoand others (2011)Regression of government spending on GDP growthDeveloping economies, especially in sub-Saharan Africa, procyclical
Talvi and Végh (2005)Correlation between real output and government consumption/revenuesDeveloping economies procyclical
Végh and Vuletin (2012)Regression of tax rates on cyclical component of real GDPTax policy acyclical in advanced economies; procyclical in developing economies
Source: Compiled by IMF staff.

When estimating this type of regression, three main issues have to be addressed:

  • The estimation and endogeneity of the output gap

  • The definition of the cyclical stance

  • Other major influences on the fiscal balance, such as commodity-related revenues.

Output Gap

Potential GDP and the output gap are estimated using traditional Hodrick-Prescott filters to ensure systematic treatment of all countries, and to avoid arbitrary adjustments based on varying information sets. This approach, however, raises several issues that are relevant for the analysis of the fiscal stance. First, output gap estimates change whenever new data become available, even if past data are not revised. Hence, when assessing fiscal policy at some point in the past, it is necessary to distinguish carefully between the intended and the resulting cyclical stance.1 Although the main interest here is to analyze the actual fiscal outturn, the analysis is complemented by looking at real-time data to compare outturns and intentions (Box 7.1). Second, the volatility of trend growth in emerging market economies makes it hard to distinguish between the trend and cyclical components of growth (Aguiar and Gopinath, 2007), and as result to assess the cyclical stance of fiscal policy. To deal with this issue, we complement our analysis by using commonly used alternative specifications for the output gap, such as actual real GDP growth. Finally, since the output gap is partly the result of fiscal policy, and since estimating the equation (7.1) using ordinary least squares would produce biased results, we also estimate our regressions applying an instrumental variable approach.

Cyclical Stance

Previous studies have used two approaches to measure the cyclical stance. Some have considered only discretionary policy actions—such as tax cuts or budget revisions—to delimit the cyclical stance. In practice this means using changes in the cyclically adjusted primary balance (or a structural balance) as the dependent variable of the regression. Other studies have taken all actual changes in the fiscal balance, whether owing to discretionary action or occurring automatically (for instance, because of rising revenues) when the economy performs better than expected.

This study proposes an innovative third approach. Specifically, it includes as part of the cyclical response of fiscal policy the automatic stabilizers that are an inherent part of the economy’s tax and welfare system, such as the additional revenues gained during a boom owing to a rising average tax rate under a progressive tax system, or the reduction in welfare spending as the unemployment rate drops.2 However, the analysis does not consider as a policy response (1) the additional revenues from taxing deviations of GDP from potential at an unchanged average tax rate, and (2) declines in spending ratios that are only the result of GDP exceeding potential.

The reason for adopting this approach is that ignoring the contribution of systematic automatic stabilizers could be misleading in the analysis of policy. For example, when comparing the policy responses of two countries, noting a more active discretionary response in one of them but not reporting on the larger automatic stabilizers in the other would bias the assessment. Lesser reliance on discretionary measures could, in fact, be motivated by the presence of stronger automatic stabilizers, which reduces the need for policy action.

Box 7.1Comparing Actual and Intended Fiscal Policy

The intention of fiscal policy, based on output gap forecasts available at the time, could be different from the actual impact of fiscal policy. To check for this, regressions for this study were run on a historical output gap series using output gap estimates for a given year based on data available for the fall of the previous year. Past vintages of the IMF’s World Economic Outlook database were used to estimate the ex ante assumed output gap and intended fiscal stance. Key findings are summarized in Box Table 7.1.1.

Table 7.1.1Country-Specific Regressions: Coefficients on the Output Gap “Forecasts” Under Alternative Specifications
Δ Adjusted Primary Balance
Dependent VariableOLSIV-1IV-2
Argentina−0.38***−0.53−1.11
(0.12)(0.48)(0.79)
Belize−0.90−4.750.07
(0.77)(11.71)(1.96)
Bolivia2.062.503.09
(1.53)(4.12)(2.25)
Brazil1.00**−5.290.76
(0.39)(60.34)(0.94)
Chile0.636.75−1.79
(0.38)(15.28)(3.51)
Colombia0.35−1.140.60
(0.48)(1.03)(1.39)
Costa Rica1.51***1.521.99***
(0.33)(1.96)(0.41)
Ecuador−0.67−0.68−0.34
(0.39)(1.14)(1.11)
El Salvador0.45*1.05−0.75
(0.22)(0.68)(1.23)
Guatemala0.11−1.130.30
(0.48)(4.69)(0.39)
Guyana−0.512.240.50
(0.92)(3.80)(1.14)
Honduras0.901.501.11
(0.57)(0.85)(0.96)
Mexico0.030.151.15
(0.17)(0.46)(1.58)
Nicaragua−0.28−2.21−1.58
(0.64)(19.06)(1.86)
Paraguay0.08−3.04−2.11
(0.37)(4.03)(2.15)
Peru−0.42−3.860.56
(0.73)(3.78)(2.82)
Suriname−1.02−0.58−2.66
(1.30)(3.91)(2.00)
Uruguay−0.26***−0.47*−0.36
(0.05)(0.21)(0.31)
Venezuela−0.89*−1.02−0.62
(0.51)(1.06)(2.06)
Note: Robust standard errors in parentheses. IV-1 uses the lagged output gap as an instrument, IV-2 the U.S. 1-year Treasury bill and the export-weighted growth rate of trading partners. The coefficient on the output gap is shown. All regressions also include a constant, the lagged adjusted primary balance, and the commodity price index. OLS = ordinary least squares.

The analysis finds evidence that fiscal policy was intended to be procyclical in Argentina and Uruguay, although the result in the case of Argentina is insignificant when using instrumental-variable estimation.1 In both countries, the outcome of fiscal policy was also procyclical.

Costa Rica is the only country where the analysis finds significant evidence that fiscal policy was intended to be countercyclical, even though the result on final data was insignificant.2

1 Whether an instrumental-variable approach is needed in this case is debatable, as the output gap is a forecast that will only be endogenous to fiscal policy to the extent that such an impact is modeled, and fiscal policy does not deviate from the projected level.2 It is still possible that the intention of budgets was to be countercyclical and that the assessment was made based on a national definition of the output gap or that budget amendments or overruns ultimately led to procyclical policy.

The definition strikes a balance between ignoring automatic stabilizers and counting all temporary revenue gains as a policy response. Empirically, it is implemented by using changes in an adjusted primary balance as the dependent variable. Specifically, the adjusted balance is defined as:

where G is government spending. The year-to-year difference in this adjusted balance will rise if the average tax rate goes up and/or spending grows less than potential GDP.3

The Role of GDP Composition and Commodity and Asset Prices

Apart from the business cycle, tax revenues can also be strongly affected by the composition of GDP as well as commodity and asset prices.

The composition of GDP can play a role, as not all components are equally taxed. Exports tend to be lightly taxed, while consumption is relatively heavily taxed. It is therefore conceivable that an export-driven economic boom leads to a fall in the revenue ratio, even in the absence of a tax cut. Reflecting this, Kaminsky, Reinhart, and Végh (2004) note that a discretionary countercyclical policy can be accompanied by a rising, steady, or falling fiscal balance as a share of GDP. On the other hand, one could argue that a falling effective average tax rate, even if due to changes in the composition of GDP, is in effect procyclical, even if not a result of discretionary policy. A government trying to maintain a countercyclical or even acyclical policy would have to take measures to undo such a fall in effective average tax rates. This would also be consistent with treating a rise in average tax rates that results from progressive tax systems as part of automatic stabilizers, as we have done above.

Asset and commodity prices may also boost revenues beyond what can be explained by real GDP growth. In the case of asset prices this occurs through wealth and transaction taxes. A rise in commodity prices will increase profits of exporters, which can boost revenues. Even greater is the effect in countries exporting natural resources, particularly where profits are highly taxed or where the government owns and operates these enterprises. In economies where commodities play an important role, it is therefore important to consider the cyclical part of commodity revenues, although this can be difficult to ascertain in practice. For example, in the event of a commodity price boom, it may not be clear which part of the revenue increase is structural (say, due to China’s permanent rise in the world economy, or new oil extraction technology) and which part is temporary. A straightforward approach to separating cyclical and trend components would be to use filtering techniques, yet this has the strong drawback of assuming away any structural breaks in the series, and is therefore not used in our analysis.

In addition, even spending only the permanent revenue gains from structurally higher commodity prices is not necessarily acyclical. While there may not be an impediment from a fiscal sustainability perspective,4 spending the permanent revenue gains will still add to domestic demand, and would be procyclical if output is above potential. More generally, adjustment to permanent changes to commodity revenues may be countercyclical or procyclical, depending on the direction of the price change and the output gap.

Therefore, from the perspective of assessing the cyclicality of fiscal policy, the relevant question is not whether commodity-related changes in revenues are permanent or temporary, but whether their use increases or reduces economic cycles. To control for the effect of commodity prices on the fiscal balance, we propose including a country-specific commodity price index as a regressor, following Céspedes and Velasco (2011).5 The index is constructed as the change in commodity prices, weighted by the share of each exported commodity in GDP. Since the index is specific to each country it reflects the relevant dependence on commodities.

Estimation

The analysis also estimates country-specific and panel regressions. Given the ample evidence that fiscal policy affects economic growth and the output gap, we ensure that the estimation of equation (7.1) reflects the endogeneity of the output gap or growth rate. In the case of country-specific regressions, we use instrumental-variable regressions, with either the lagged output gap or the export-weighted growth rate of trading partners and the U.S. real interest rate as instruments.6 For panel regressions, we also report results using a system generalized method of moments (GMM) estimator proposed by Blundell and Bond (1998).

Data

The main data source is the IMF’s World Economic Outlook (WEO) database. In addition, we use U.S. Treasury bill rates from the Federal Reserve Board, and trade and commodity price data from the UN Comtrade database. For resource-related fiscal revenues in Chile and Mexico, we use the IDB Fiscal Resources data set. The data set covers the period from 1980 to 2012, although actual data availability varies by country and variable. For the regressions based on real-time data, WEO vintages from 1990 through 2012 are used.

For the primary balance, we use our own calculations adding back interest expenditure to overall government net lending. The WEO also reports a primary balance, using the more accurate approach of also deducting interest receipts. This, however, reduces the sample size, as not all countries report such receipts. As interest receipts are typically small, we chose this approximation to have a larger sample size.7

For commodity prices, we use a newly-calculated price index, based on time-varying weights, lagged by d periods:

where i denotes countries, j denotes commodities, t denotes time, Pj,t is the logarithm of the price in U.S. dollars, and xi,j,t-d are export values. This index allows us to take into account that the commodity export and import basket might change substantially over a long period, while ensuring that changes in the price index reflect changes in commodity prices rather than endogenous changes in export and import volumes in response to price fluctuations.

Findings

Panel Data Results

Panel data estimates were conducted for a group comprised of 19 Latin American economies, a broader group covering 134 developing and emerging market economies, and a group of 32 advanced economies, allowing for different intercepts for each country, but imposing the same slope within a region. The regressions address the endogeneity issue, control for the effect of commodity prices, and allow alternative specifications (output gap versus growth rates).

The results (Table 7.2) suggest that fiscal policy in Latin America has been procyclical, as the coefficient on the output gap is negative and statistically significant, both in a standard within-group regression and in a GMM regression that allows for endogeneity.8 In contrast, the results for advanced economies show a positive and statistically significant coefficient, implying countercyclical fiscal policy. For a broad set of emerging market and developing economies, the evidence is unclear, with mostly insignificant findings. Results, however, are sensitive to the specification used. For example, if growth rates are used rather than output gap, all regressions turn insignificant (see bottom part of Table 7.2).

Table 7.2The Adjusted Fiscal Balance in Various Panel Data Estimates
Dependent VariableΔ Adjusted Primary Fiscal Balance
Advanced EconomiesEmerging Market/Developing EconomiesLatin America
Estimation MethodWGWGGMMWGWGGMMWGWGGMM
(1)(2)(3)(4)(5)(6)(7)(8)(9)
Output gap0.13**0.14**0.30***−0.100.010.07−0.34**−0.34**−0.34**
(0.06)(0.06)(0.11)(0.12)(0.15)(0.13)(0.15)(0.14)(0.15)
Commodity price growth0.56***1.40**0.55***−0.450.39***0.41
(0.07)(0.64)(0.07)(1.17)(0.08)(0.26)
Adjusted deficitst-1−0.23***−0.23***0.04−0.57***−0.58***−0.74***−0.49***−0.44***−0.48***
(0.03)(0.03)(0.10)(0.11)(0.13)(0.14)(0.08)(0.08)(0.13)
Observations7917607602,5682,0362,036355333333
R-squared0.120.180.300.400.300.34
Number of countries333232146134134201919
AB AR(1)test00.110.08
AB AR(2) test0.410.380.49
Hansen p-value0.760.230.35
(1)’(2)’(3)’(4)’(5)’(6)’(7)’(8)’(9)’
Growth0.27***0.26***0.34***0.03*0.010.01−0.07−0.12−0.17
(0.04)(0.05)(0.09)(0.02)(0.03)(0.04)(0.10)(0.12)(0.23)
Commodity price growth0.49***−0.490.55***0.080.44***0.87
(0.12)(1.09)(0.07)(1.17)(0.13)(0.87)
Adjusted deficitst-1−0.28***−0.27***−0.11−0.58***−0.58***−0.72***−0.49***−0.42***−0.34*
(0.03)(0.04)(0.10)(0.11)(0.14)(0.15)(0.09)(0.10)(0.19)
Observations7917607602,5682,0362,036355333333
R-squared0.220.270.300.400.260.30
Number of countries333232146134134201919
AB AR(1)test0.010.070.05
AB AR(2) test0.290.150.27
Hansen p-value0.040.160.46
Note: WG = within-group regression. Robust errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. System generalized methods of moments (GMM) regressions estimated with xta-bond2. The outpug gap/growth and the lagged adjusted primary fiscal balance are treated as endogenous, using the first and second lag as instruments and the collapse option.

The coefficient on the commodity price is often significant and positive across all regions. This is in line with expectations, as the adjusted primary fiscal balance generally strengthens when commodity price growth is strong (unless countries spend more than their additional resource revenues). Despite the significance of the commodity price index, its introduction does not change the other coefficients very much, suggesting that any omitted variable bias from not including it could be limited.9

Country-Specific Results

Fiscal policy is likely to be run differently across countries, even within a region. Panel estimates, though common in the literature, can only give a first impression because they force the same cyclicality coefficient on all countries. Country-specific regressions were also estimated (using the same explanatory variables) for the 19 Latin American economies, applying an instrumental-variable approach with the lagged output gap as an instrument.

Table 7.3 presents results from country-by-country regressions of the adjusted fiscal balance for the period from 1990–2012. As before, these regressions include the lagged adjusted fiscal balance and the commodity price index, but to maintain readability only the coefficient on the output gap is reported. In addition to presenting ordinary least squares (OLS) results, we report two instrumental-variable specifications, as the system GMM approach cannot be used in pure time-series regressions. In the first of these instrumental-variable regressions the lagged output gap is used as an instrument. In the second approach the export-weighted GDP growth of trading partners and the U.S. one-year Treasury rate serve as instruments, as done in other papers in the literature. Finally, the table reports on regressions allowing a varying degree of cyclicality over time, showing coefficients for the period before 2005 and the change to the coefficient afterward. In these regressions the output gap and the commodity price index were interacted with the time indicator. The year 2005 was chosen because panel regressions suggest this to be the year with the most significant change in coefficient, while for comparability a single year for all countries seemed useful.

Table 7.3Country-Specific Regression: Coefficients on the Output Gap under Alternative Specifications
Δ Adjusted Primary Balance
OLS
Dependent VariableOLSIV-1IV-2Pre-2005Δ Since 2005
Argentina−0.32***−0.38**−0.24−0.26**−0.36
(0.08)(0.15)(0.19)(0.11)(0.67)
Belize0.222.400.200.54−0.53
(0.30)(1.76)(1.30)(0.35)(0.48)
Bolivia−0.45−0.21−0.14−0.410.01
(0.38)(1.00)(1.02)(0.43)(1.12)
Brazil0.325.550.01−0.160.74*
(0.19)(21.96)(1.66)(0.21)(0.33)
Chile0.27−1.091.37*0.000.90**
(0.26)(1.21)(0.69)(0.17)(0.34)
Colombia−0.14−0.500.14−0.310.68*
(0.16)(0.45)(0.33)(0.24)(0.38)
Costa Rica0.24−0.770.97*−0.350.75
(0.20)(1.45)(0.44)(0.33)(0.51)
Ecuador−0.48**0.600.97−0.280.13
(0.21)(2.45)(1.18)(0.23)(0.50)
El Salvador0.30−1.030.69***−0.080.61*
(0.19)(2.90)(0.22)(0.24)(0.32)
Guatemala0.14−0.750.58−0.420.62
(0.16)(1.69)(0.43)(1.24)(1.28)
Guyana1.12***−0.892.231.120.07
(0.23)(6.46)(4.86)(0.83)(0.83)
Honduras0.210.180.541.51*−1.42*
(0.14)(0.37)(0.57)(0.65)(0.69)
Mexico−0.210.430.53−0.32**0.46**
(0.14)(0.66)(1.07)(0.15)(0.19)
Nicaragua−0.33−0.71−0.24−0.81**0.84
(0.21)(0.61)(0.32)(0.29)(0.45)
Paraguay−0.074.980.10−0.130.19
(0.13)(35.75)(0.24)(0.25)(0.29)
Peru0.360.58*1.320.200.15
(0.29)(0.28)(0.99)(0.57)(0.53)
Suriname−1.10−9.85−2.81−1.260.79
(0.68)(10.88)(1.78)(0.77)(2.16)
Uruguay−0.45***−0.71***−0.27−0.46***0.13
(0.07)(0.18)(0.23)(0.08)(0.22)
Venezuela−0.60***0.280.36−0.54***0.26
(0.16)(0.83)(0.77)(0.15)(0.37)
Note: OLS = ordinary least squares. Robust standard errors in parentheses. IV-1 uses the lagged output gap as an instrument, IV-2 the U.S. 1-year Treasury bill and the export-weighted growth rate of trading partners. The coefficient on the output gap is shown. All regressions also include a constant, the lagged adjusted primary balance, and the commodity price index.

What stands out from the first three columns of Table 7.3 is the very small number of statistically significant coefficients. This is a common though rarely mentioned feature of studies on fiscal policy cyclicality, many of which do not report tests of significance. Still, the coefficients for some countries show evidence of procyclical policy under both the OLS and instrumental-variable regressions. In other countries, such as Ecuador and Venezuela, the evidence for procyclical fiscal policy does not hold up in the instrumental-variable estimates. The coefficient is neither positive nor consistently statistically significant in any of the 19 countries. In other words, there is no significant evidence for countercyclical policy in any Latin American country. In summary, for most countries, acyclical policy cannot be rejected, although a mildly cyclical (with a coefficient close to zero) or erratic (with large standard errors) fiscal policy is also consistent with the evidence.10

The time-varying regressions offer further insights. Fiscal policy in Brazil,11 Chile, Colombia, El Salvador, and Mexico appears to have become less procyclical since 2005. Only Honduras appears to have become more procyclical. As the post-2004 period includes the global financial crisis and related fiscal stimulus, the next boom period will provide a test of whether more countercyclical policies will prevail. In most countries, the intended and actual impact of fiscal policy broadly coincided (for a fuller discussion see Box 7.1).

The Quality of Fiscal Policy

As noted earlier, the cyclical stance of fiscal policy is only one dimension of its quality. A given fiscal stance could be achieved with many different underlying tax and expenditure policies. Hence, a move toward more countercyclical policy could be problematic if certain risks are not addressed.

Fiscal Sustainability

A countercyclical policy response—and in particular deficit-increasing policy during recessions—must not go so far as to put medium-term finances at risk. In Latin America, public debt remains very high, on average, having stopped declining in 2007 (Figure 7.1). Going forward, the evolution and relatively high levels of debt since the global financial crisis may constrain countercyclical policy action during downturns. However, the situation differs greatly across countries, as some (for example, Chile, Paraguay, and Peru) have very low debt stocks.

Figure 7.1Public Debt in Latin America

(Percent of GDP)

Source: IMF, World Economic Outlook database.

1 For details on Argentina’s GDP see Appendix 2.1 of the IMF’s April 2014 Regional Economic Outlook: Western Hemisphere (IMF, 2014).

Fiscal expansions in downturns are meant to address a demand shortfall and thus should be reversible or limited in time. If the higher expenditures are structural in nature, it will be harder to readjust the stance when the economy improves. Chile has tried to reduce this risk by linking increases in structural spending to permanent revenues (for example, a recent tax reform to finance education spending). Most other countries in Latin America, however, do not make this distinction.

Fiscal Institutions

Many countries in Latin America have adopted reforms to strengthen fiscal institutions, including fiscal rules. Provided that such rules are well designed, they can support sustainable fiscal policy while avoiding procyclicality. Indeed, countries such as Chile, Colombia, and Mexico have managed to move toward more countercyclical policy while following a fiscal rule. Fiscal transparency is also important, for both policymakers and the public. Some recent examples of nontransparent policies include the use of one-off transactions to reduce reported deficits or the increased use of deficit-neutral operations such as policy lending, which may still increase fiscal liabilities.

Conclusions

This chapter has considered the cyclicality of fiscal policy in Latin America using a new methodology that gives full credit to automatic stabilizers, while controlling for commodity prices and allowing for endogeneity. The evidence suggests that fiscal policy in Latin America has been procyclical, on average, rather than acyclical or countercyclical as in most advanced economies. Country-specific estimations, however, yield mostly insignificant results, as is common—but often unacknowl-edged—in comparable studies. In more recent years, Brazil, Chile, Colombia, El Salvador, and Mexico appear to have moved toward less procyclical or more countercyclical fiscal policy. It remains to be seen whether this development will prevail during times of closed or positive output gaps, when previous fiscal stimulus measures should be unwound. More generally, countries need to rebuild their buffers, not least to be prepared for any future negative economic shock.

Apart from cyclical considerations, fiscal policy should be sustainable and transparent. Fiscal institutional measures such as well-designed fiscal rules can support sustainability without leading to procyclical fiscal policy. Fiscal transparency has improved in many countries over the past decade, but recent examples indicate the reappearance of problematic behavior such as the use of one-off transactions and operations that are chosen to avoid increasing reported deficits.

References

    Aguiar, M., and G.Gopinath,2007, “Emerging Market Business Cycles: The Cycle Is the Trend,Journal of Political Economy115(1): 69102.

    Alesina, Alberto, FilipeCampante, and GuidoTabellini,2008, “Why Is Fiscal Policy Often Procyclical?Journal of the European Economic Association6(5): 1006036.

    Barnett, S., and R.Ossowski,2003, “Operational Aspects of Fiscal Policy in Oil-Producing Countries.” In Fiscal Policy Implementation in Oil Producing Countries, edited by J.M.Davis, R.Ossowski, and A.Fedelino. Washington, DC: International Monetary Fund.

    Beetsma, Roel, and MassimoGiuliodori,2010, “Fiscal Adjustment to Cyclical Developments in the OECD: An Empirical Analysis Based on Real-Time Data, Oxford Economic Papers62: 41941.

    Bernoth, Kerstin, AndrewHughes Hallett, and JohnLewis,2008, “Did Fiscal Policy Makers Know What They Were Doing? Reassessing Fiscal Policy with Real-Time Data,DNB Working Paper No. 169. Amsterdam: De Nederlandsche Bank NV.

    Blundell, Richard, and StephenBond,1998, “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,Journal of Econometrics87(1): 11543.

    Catão, Luis, and BennettSutton,2002, “Sovereign Defaults: the Role of Volatility,Working Paper No. 02/149, International Monetary Fund, Washington, DC.

    Céspedes, Luis Felipe, and AndresVelasco,2011, “Was this Time Different? Fiscal Policy in Commodity Republics,BIS Working Paper No. 365. Basel: Bank for International Settlements.

    Cimadomo, Jacopo,2012, “Fiscal Policy in Real Time,Scandinavian Journal of Economics114(2): 44065.

    Daude, Christian, AngelMelguizo, and AlejandroNeut,2011, “Fiscal Policy in Latin America: Countercyclical and Sustainable?Economics: The Open Access, Open-Assessment E-Journal5.

    Di Bella, Gabriel,2009, “Fiscal Policy Responses to the Crisis: How Much Space for Countercyclical Policy?” In Regional Economic Outlook: Western Hemisphere, World Economic and Financial Surveys, International Monetary Fund, Washington, DC.

    Egert, Balazs,2012, “Fiscal Policy Reaction to the Cycle in the OECD: Pro- or CounterCyclical?CESifo Working Paper No. 3777.

    Forni, Lorenzo, and SandroMomigiano,2004, “Cyclical Sensitivity of Fiscal Policies Based on Real-time Data,Applied Economics Quarterly50(3): 299326.

    Frankel, Jeffrey, CarlosVégh, and GuillermoVuletin,2013, “On Graduation from Fiscal Procyclicality,Journal of Development Economics100(1): 3247.

    Gali, Jordi, and RobertoPerotti,2003, “Fiscal Policy and Monetary Integration in Europe,Economic Policy18(37): 53372.

    Gavin, Michael, and RobertoPerotti,1997, “Fiscal Policy in Latin America,NBER Macroeconomics Annual, Vol. 12. Cambridge, Massachusetts: National Bureau of Economic Research.

    Ilzetzki, Ethan, and CarlosVégh,2008, “Procyclical Fiscal Policy in Developing Countries: Truth or Fiction?NBER Working Paper No. 14191. Cambridge, Massachusetts: National Bureau of Economic Research.

    International Monetary Fund (IMF), 2014, Regional Economic Outlook: Western Hemisphere, World Economic and Financial Surveys, Washington, DC, April.

    Jaimovich, Dany, and UgoPanizza,2007, “Procyclicality or Reverse Causality?IDB Research Department Working Paper No. 599. Washington, DC: Inter-American Development Bank.

    Kaminsky, Graciela, CarmenReinhart, and CarlosVégh,2004, “When It Rains It Pours: Procyclical Capital Flows and Macroeconomic Policies,NBER Macroeconomics Annual, Vol. 19. Cambridge, Massachusetts: National Bureau of Economic Research.

    Klemm, Alexander,2014, “Fiscal Policy in Latin America over the Cycle,Working Paper No. 14/59, International Monetary Fund, Washington, DC.

    Lane, Philipp,2003, “The Cyclical Behaviour of Fiscal Policy: Evidence from the OECD,Journal of Public Economics87(12): 2661675.

    Talvi, Ernesto, and CarlosVégh,2005, “Tax Base Variability and Procyclical Fiscal Policy in Developing Countries,Journal of Development Economics78(1): 15690.

    Végh, Carlos, and GuillermoVuletin,2012, “How Is Tax Policy Conducted over the Business Cycle?NBER Working Paper No. 17753. Cambridge, Massachusetts: National Bureau of Economic Research.

See Beetsma and Giuliodori (2010), Bernoth and others (2008), Cimadomo (2012), and Forni and Momigiano (2004), who tend to find that intended policies have been less procyclical than actual ones.

Alternative definitions of automatic stabilizers are possible. These would include any increase in revenue, even at constant or falling tax rates as a stabilizer, for example if the starting assumption is an economy of lump-sum taxes. We use the more demanding definition of an automatic stabilizer that requires an increase in the revenue-to-GDP share, as would happen under a progressive tax schedule.

Klemm (2014) also looks at primary expenditures (excluding transfer payments) as an alternative specification of the cyclical stance: Δ(GY*)=β0+β1(YY*Y*)+β2(GY*)t1+γx+fi+ɛ. While this measure has the advantage of avoiding the question of cyclical adjustments, it is only a valid measure to the extent that there is no policy change in the revenue side. The results were similar to, though somewhat weaker than, those obtained for the fiscal balance.

Natural resources can also be seen as a national asset, the sales of which should be counted as a capital transaction, in which case only the real return on that asset should be thought of as revenue (Barnett and Ossowski, 2003).

One approach would be to use a fiscal balance excluding commodity-related revenues. Unfortunately, this is available only for a few countries. Another simple alternative is to analyze spending rather than balances, which will be valid unless there are simultaneous structural revenue reforms.

An alternative approach is to use a vector autoregression (as do Ilzetzki and Végh, 2008), but this is more applicable for quarterly (or longer) data sets. Quarterly data are difficult to obtain and interpret, however, given that budgets are usually annual, and given the different choices by countries in terms of the extent to which fiscal accounts are prepared on an accrual versus a cash basis.

In the case of Brazil, where interest receipts are particularly high, we instead use net interest expenditure. Indeed, in the case of Brazil, the WEO reports net rather than gross interest spending.

GMM estimates are presented with the standard specification tests: the Arrellano-Bond AR(1) test, which is expected to be rejected, and the AR(2) test and the test of over-identifying restrictions (Sergeant/Hansen test), which should both not be rejected. The specification tests of system GMM regressions are passed in all cases, except regression (3’), where the test of over-identifying restrictions rejects the validity of instruments with a p-value of 4 percent. However, as the result is in line with all other results on advanced economies, the rejection of this specification does not affect the overall interpretation.

Alternative specifications of commodity price indicators—commodity prices weighted by total commodity exports instead of GDP, and commodity prices applied to net rather than gross commodity exports—yielded similar results (Klemm, 2014).

Similar country-by-country analysis using the adjusted spending ratio shows even fewer statistically significant results (see Klemm, 2014).

In the case of Brazil, policy lending, which is not part of the fiscal balance but may affect the fiscal stance, has grown in importance. If net policy lending is added to the adjusted fiscal balance, the coefficient on the output gap interacted with the post-2005 dummy becomes larger, but as the standard error rises even more, turns insignificant.

    Other Resources Citing This Publication