The Macroeconomic Effects of Fiscal Consolidation in Emerging Economies: Evidence from Latin America

We estimate the short-term effects of fiscal consolidation on economic activity in 14 countries in Latin America and the Caribbean. We examine contemporaneous policy documents to identify changes in fiscal policy motivated by a desire to reduce the budget deficit and not by responding to prospective economic conditions. Based on this narrative dataset, our estimates suggest that fiscal consolidation has contractionary effects on GDP, consistent with a multiplier of 0.9. We find these effects to be close to those in OECD countries based on a similarly constructed dataset (Devries and others, 2011). We also find similar estimation results for the two groups of economies for the effect of fiscal consolidation on the external current account balance, providing support for the twin deficits hypothesis.

Abstract

We estimate the short-term effects of fiscal consolidation on economic activity in 14 countries in Latin America and the Caribbean. We examine contemporaneous policy documents to identify changes in fiscal policy motivated by a desire to reduce the budget deficit and not by responding to prospective economic conditions. Based on this narrative dataset, our estimates suggest that fiscal consolidation has contractionary effects on GDP, consistent with a multiplier of 0.9. We find these effects to be close to those in OECD countries based on a similarly constructed dataset (Devries and others, 2011). We also find similar estimation results for the two groups of economies for the effect of fiscal consolidation on the external current account balance, providing support for the twin deficits hypothesis.

I. Introduction

Public debt has risen in many emerging market and developing economies (EMDEs), precipitating the need for fiscal consolidation. There is, however, no consensus in the literature regarding the size and even the sign of the macroeconomic effects of fiscal consolidation in EMDEs. There is also little agreement on how the effects compare with those in advanced economies (AEs).

Theory provides an ambiguous guide to the relative size of fiscal multipliers in EMDEs and AEs. Some factors imply larger multipliers in EMDEs, such as tighter liquidity constraints facing households and firms that imply a sharper reduction in private consumption and investment during periods of fiscal consolidation. But other factors imply smaller multipliers in EMDEs. For example, they typically feature higher perceived sovereign default risk, implying more scope for confidence effects that could partly offset the direct effects of fiscal consolidation (Blanchard, 1990, for example).

Existing empirical studies generally suggest that fiscal multipliers are smaller in EMDEs than in AEs. Our review of 133 recent estimates of fiscal multipliers finds that they are on average 50 percent larger for AEs than for EMDEs (Figure 1).2 However, methodological differences across the studies, as well as differences in data quality, may play a strong role in explaining this difference. Several recent studies for AEs use narrative methods, which draw on policy documents to identify the timing and intention of fiscal policy changes, with the goal of more precisely estimating causal effects (Romer and Romer 2010, for example). Such narrative-based studies often yield larger multiplier estimates than do the more conventional approaches—such as those based on cyclically adjusted fiscal data or the estimation of structural vector auto regressions (SVARs)—on which most studies for EMDEs rely.3 There is little available evidence based on such narrative approaches for EMDEs.

Figure 1.
Figure 1.

Distribution of Empirical Multiplier Estimates by Country Group

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

This paper attempts to fill this gap in the literature and provide fresh evidence on the macroeconomic effects of fiscal consolidation in EMDEs by constructing a new narrative dataset of fiscal consolidation episodes for 14 countries in Latin America and the Caribbean (LAC) for 1989–2016. Following the approach of narrative studies conducted for AEs, we examine the behavior of economic activity in LAC economies following discretionary changes in fiscal policy that, historical sources suggest, are not correlated with the short-term economic outlook. As we explain in Section II, we consult contemporaneous policy documents to identify fiscal actions motivated by a desire to reduce the budget deficit and ensure long-term public financial sustainability, rather than to prevent abnormal growth or respond to cyclical pressures.4 Our approach is closest to that of Devries and others (2011), who construct such a dataset for 17 OECD economies.

Based on our new dataset, we estimate the effects of fiscal consolidation for LAC economies. As Section III explains, our baseline specification uses the local projection method (LP) proposed by Jordà (2005). The estimation results, reported in Section IV, suggest that fiscal consolidation in our sample of LAC economies are typically associated with contractions in output, consistent with an average fiscal multiplier of about 0.9. This estimate is about three times larger than the average of multiplier estimates across existing empirical studies available for LAC economies (as reported in the Annex). Interestingly, our estimation results for LAC economies are very similar to those we obtain for AEs using a comparable narrative dataset (Devries and others, 2011). The results hold up to a battery of robustness checks.

To explore the channels at work and shed light on the factors that shape fiscal multipliers, we extend the analysis in three directions. First, we investigate the effects of fiscal consolidation on the unemployment rate, private consumption and private investment, the external current account balance, and the real exchange rate. Second, we assess whether the effects of fiscal consolidation depend on the initial level of sovereign default risk and the state of the business cycle. Finally, we investigate how the output effects of fiscal consolidation depend on the composition of the consolidation package across spending and tax measures. Section IV concludes the paper.

II. Identifying Fiscal Consolidation: A New Narrative Dataset

We construct a new dataset of fiscal consolidation measures taken by the governments of 14 LAC economies to reduce budget deficits during 1989-2016. The approach builds on earlier work by Devries and others (2011) that identified such fiscal consolidation measures for 17 OECD economies. Additional documentation on each fiscal policy change can be found in our companion paper, David and Leigh (2018), in which we provide detailed citations for each data point to show how we determine the motivation and estimated budgetary effects from the historical record.

Following the approach of Romer and Romer (2010) and Devries and others (2011), we examine contemporaneous policy documents to assess the motivation, expected size, and timing of discretionary policy actions, including changes in both taxes or spending. We focus on policy actions that are not driven by a desire to respond to current or prospective economic conditions, but that are instead motivated by considerations such as reducing an inherited budget deficit and ensuring long-term public financial sustainability.5 There are some cases of fiscal actions that imply expansions in the budget deficit that are motivated by such long-term objectives. Whenever such fiscal actions occur, we record them in the dataset with a negative sign. The measures of the magnitude of fiscal policy changes included in the database rely on estimates of the revenue or expenditure impact of the given policy action at the time of implementation (expressed in annual terms) and at the prevailing level of GDP.

A possible concern regarding our narrative approach, highlighted by Jordà and Taylor (2015), is that such fiscal consolidation episodes reflect past economic developments and could thus have a forecastable component. We address this concern in Section III by applying the Jordà and Taylor (2015) augmented inverse propensity score weighting procedure.

We only include in the dataset fiscal measures that were implemented, according to historical policy documents. If measures were announced but did not come into effect, we do not include them in the database. For example, the government of Costa Rica announced a consolidation package in 1990 that in addition to revenue measures amounting to 1.5 percent of GDP, also envisaged 2 percent of GDP in expenditure cuts. As the latter were not implemented, we only include the revenue measures in the dataset. Occasionally, announced measures are only partly implemented or are implemented with delays and the coding of the budgetary impacts in the dataset takes these features into account.

The narrative approach we use is motivated by reducing estimation biases associated with conventional approaches for identifying the causal effects of fiscal policy. As a number of studies explain, measuring changes in fiscal policy based on changes in the cyclically-adjusted primary balance (CAPB)– a conventional approach–can be problematic. The CAPB includes shifts in fiscal variables unrelated to policy decisions–including those driven by swings in asset or commodity prices, which also affect economic activity. A commodity price boom may, for example, stimulate private investment while also boosting government revenue. Another shortcoming is that changes in the CAPB may reflect policy responses to current macroeconomic conditions, such as a loosening of fiscal policy motivated by the onset of a recession.

A related literature identifies fiscal policy shocks using structural VARs. The pioneering work of Blanchard and Perotti (2002) applies this approach to quarterly data for the United States. As Romer and Romer (2010) point out, even this more refined approach still assumes that, after controlling for lags of output growth, changes in government revenue and spending are uncorrelated with other short-term developments affecting output. It therefore ignores the issue of nonpolicy changes in cyclically-adjusted fiscal data. This concern is particularly relevant when using annual data, as is typically the case for studies focusing on EMDEs.6

The dataset contains 76 observations with non-zero fiscal actions over the period of 1989–2016. Each fiscal action is expressed in terms of its budgetary impact as a share of GDP, as listed in the Annex, and is described in detail in David and Leigh (2018). In a number of cases, these actions enter the database with a negative sign, as in the case of temporary tax increases that are subsequently reversed. Figure 2 reports the density function for the size of the identified fiscal actions. The average budgetary impact (conditional on an action occurring) is 0.9 percent of GDP., with a standard deviation of about 1 percent of GDP.

Figure 2.
Figure 2.

Distribution of Budgetary Impact of Narrative Fiscal Shocks

(Percent of GDP)

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

We further classify the consolidation episodes as “tax-based” or “expenditure-based” depending on whether tax hikes or expenditure cuts account for most of the budgetary impact of the consolidation (Guajardo, Leigh and Pescatori, 2014).7 There are 55 observations for tax-based consolidations and 18 for expenditure-based episodes. Expenditure-based consolidations are somewhat larger in size at about 1.2 percent of GDP, which compares to 0.8 percent of GDP for tax-based consolidations.

An important feature of this narrative approach is the coverage of fiscal actions across both revenue and spending instruments, which allows us to control for offsetting measures that may be deployed as part of a package. In an important recent contribution, Gunter and others (2017) construct a comprehensive dataset of changes in VAT rates across numerous economies. VAT rate changes might, however, be accompanied by changes in other taxes or government spending that are not included in the study.

To shed light on how our narrative approach for measuring fiscal changes compares with more conventional measures, we compare our narrative episodes with the change in the cyclically-adjusted primary balance (CAPB) reported in the October 2017 IMF World Economic Outlook (WEO) database. As a series for the CAPB is not available in the WEO database for some of the14 LAC economies in our sample, we construct a measure of the CAPB based on conventional procedures, as we explain in the Annex.

Figure 3 plots the episodes identified through the narrative analysis against the contemporaneous change in the CAPB. The correlation between the two measures is strong. However, there are numerous cases in which the CAPB approach and our narrative approach come to different conclusions regarding the presence and size of fiscal consolidation. Our inspection of the 16 largest discrepancies—detailed in the Annex—suggests that our narrative approach more accurately identifies the size of deficit-driven fiscal consolidation. The discrepancies are generally driven by specific economic or budgetary developments that cause the conventional CAPB-based measure to inaccurately identify the size of deficit-driven fiscal consolidation. These developments include, in a number of cases, responses to current or prospective economic conditions, including the onset of economic crisis.

Figure 3.
Figure 3.

Two Measures of Fiscal Consolidation: Changes in CAPB versus Narrative Fiscal shocks

(Percent of GDP)

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Notes: Labels indicate cases where either the CAPB or the narrative approach identify fiscal consolidation and the discrepancy between the two measures exceeds 2.5 percent of GDP. Labels indicate three-letter ISO country codes. The diagonal line indicates points along which the series are equal (45° line).

Orthogonality to News Regarding the State of the Economy

The narrative approach aims to identify fiscal policy changes that are exogenous to current and prospective news about the state of the economy. To ensure that this is the case, we follow Guajardo, Leigh and Pescatori (2014) and test whether the narrative fiscal shocks are correlated with unexpected movements in output. We construct a measure of economic news based on real-time revisions to forecasts of real GDP published in the IMF’s World Economic Outlook database. These are defined as the revision to the forecast for current-year GDP made in the fall of year t relative to the forecast made in the fall of the previous year (t–1). We then regress our series of narrative fiscal shocks on this measure of economic news.

The results of this test are reported in Table 1. We find that the narrative fiscal consolidations are not significantly related to contemporaneous unexpected movements in output, both in an economic and a statistical sense. This conclusion is valid for emerging as well as for advanced economy samples. It is worth emphasizing that the finding that the narrative fiscal shocks are orthogonal to current economic developments does not imply that they are unrelated to past developments. Indeed, since the fiscal consolidations we identify are motivated by reducing inherited budget deficits, they are likely to be correlated with past developments and should thus be predictable to some extent. In the robustness section, we estimate fiscal multipliers using an average treatment effect that takes the predictability of the shocks into account.

Table 1.

Regressions of fiscal actions on economic news

article image
Notes: Regressions are estimated for 1989 - 2016. The table reports point estimates and heteroskedasticity-robust standard errors. See the text for description of the News variable. * Significant at 10%; ** significant at 5%; *** significant at 1%.

III. Empirical Strategy

We estimate the macroeconomic effects of fiscal consolidations using the local projection method (LP) proposed by Jordà (2005). One of the advantages of this procedure is that it does not constrain the shape of the impulse response functions, and is therefore less sensitive to misspecification than estimates obtained from VAR models. Another advantage is its flexibility in estimating state-dependent impulse responses. Importantly for the extensions we will undertake in Section IV, there is no need to make assumptions about transition probabilities across states and about the feedback between shocks and states, as LP estimates already incorporate average transition characteristics from the data (Ramey and Zubairy, 2018).

The benchmark specification for different horizons (h=0, 1, 2) in years is as follows:

yi,t+hyi,t1=αih+γth+βhΣs=tt+hFCi,s+δXi,t+εi,t+h,(1)

where y is the macroeconomic variable of interest (the log of real GDP, the current account balance as a share of GDP, or the log of the real effective exchange rate, among other variables); FC denotes our measure of fiscal consolidation in percent of GDP; and Xt is a set of control variables that includes two lags of real GDP growth and two lags of the fiscal shocks. We include time ((γth)) and country (αih) fixed effects to capture common shocks and time-invariant features of fiscal policy and growth, respectively. The vector of controls Xt also includes the contemporaneous growth rate of the commodity export value and its lags, which is an important driver of business cycles and fiscal policy in EMDEs (Céspedes and Velasco, 2014; Fernández, González and Rodríguez, 2018).

The coefficient βh corresponds to our multiplier estimate. Following Ramey and Zubairy (2018), we define fiscal multipliers as the response of the level of real GDP—or the relevant macroeconomic variable of interest–relative to the cumulative fiscal shock over a given horizon.8 Alternative definitions of fiscal multipliers have been frequently considered in the literature, such as the ratio of the peak of the output level response to an initial government spending/tax shock (Blanchard and Perotti, 2002) or the ratio of the average output response to an initial fiscal policy shock (Auerbach and Gorodnichenko, 2012). However, it is important to calculate multipliers by comparing the integral of output responses to the integral of fiscal shocks (rather than impact effects) because the effects of fiscal policy can either build or be reverted over time.

The regressions are estimated by ordinary least squares with the narrative shocks included directly in the panel models. The annex presents the definitions and sources for the variables used in the analysis.

IV. Results

A. Fiscal Multipliers in Emerging Economies

Figure 4 reports impulse responses obtained from estimation of Equation 1. The shaded regions indicate 90 percent confidence intervals based on Driscoll-Kraay standard errors that are robust to autocorrelation and cross-sectional dependence. The figure displays the response of the level of real GDP following a fiscal shock of 1 percent of GDP, and thus corresponds to the output multiplier estimate at each horizon. We find that fiscal consolidations in LAC economies lead to an output contraction of 0.5 percent on impact and of 0.9 percent after two years, with a 90 percent confidence interval of 0.6-1.2 percent.

Figure 4.
Figure 4.

Real GDP: Estimated Effect of a 1 Percent of GDP Fiscal Consolidation

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Shading indicates 90 percent confidence interval.

We then compare these responses to a group of advanced economies, using the same methodology. To do so, we re-estimate Equation 1 using the narrative fiscal shocks constructed by Devries and others (2011) for 17 OECD economies covering the period 1978-2009, which we extend through 2014 based on the narrative dataset of Alesina and others (2018). In this comparator group, fiscal consolidations are estimated to lead to a fall in output of 0.3 percent on impact, rising to 0.7 percent after two years.

In contrast to the literature, we conclude that the average output responses for these two groups of emerging market and advanced economies are similar, with point estimates within one standard error of each other at horizons of 1 and 2 years. This result may reflect the fact that we are using a common narrative identification strategy and common estimation specification across groups, whereas the literature has tended to use different strategies across samples.

B. Robustness Checks

Table 2 reports the sensitivity of the baseline results to a series of robustness checks, focusing on the multiplier estimates on impact (h=0) and after two years (h=2).

Table 2.

Estimation Results: Effect of a 1 Percent of GDP Fiscal Consolidation in year t + h (Percent).

article image
Notes: Country and time fixed effects included in all regressions. Driscoll-Kraay standard errors in parentheses. For the AIPW estimates empirical sandwich standard errors (clustered by country) are reported. *Significant at 10%; **significant at 5%; ***significant at 1%.

Sensitivity to outliers

First, we investigate the sensitivity of the results to outliers. While especially large or small fiscal consolidations are worth considering, it is natural to ask how important they are for the results. Investigating whether unusually large fluctuations in output are driving the results is also warranted. We therefore re-estimate the baseline equation after dropping the largest and smallest 5 percent of the fiscal policy changes and real GDP growth rates from the sample. As Table 2 reports, the multiplier estimates are similar based on this trimmed sample, implying that outliers are not driving the results.

The Average Treatment Effect of Consolidation based on a Matching Estimator

One possible concern regarding consolidations identified with the narrative approach is that these episodes might be predicted by past variables. To address this concern, Jordà and Taylor (2015) propose the use of an augmented inverse propensity score weighting (AIPW) estimator. The weights used are based on the predicted component of the narrative episodes obtained from a probit model of the probability of treatment. The intuition behind this estimation strategy is that less weight is given to consolidations that are better predicted by a vector of control variables.

To assess whether the results obtained with the matching estimator differ from our baseline, we convert the narrative fiscal shocks into a binary variable Dt (referred to as the “fiscal treatment”) that takes a value of 1 when a fiscal consolidation occurred and 0 otherwise. It is important to note that this transformation has implications for the identification of the effects of fiscal actions, since the narrative approach relies not only on the timing of fiscal adjustments but also on their size. Indeed, the loss of the latter dimension is a significant drawback of the AIPW estimator.

The first-stage probit model includes the following determinants of the probability of treatment, pt : past fiscal consolidations (two lags); two lags of GDP growth; the change in the commodity export price and two of its lags; the lagged debt-to-GDP ratio; lagged inflation; the lagged current account to GDP balance, lagged changes in perceived sovereign default risk; and country fixed effects.9 In an ideal randomized control trial, the distributions for the predicted probability of treatment between treatment and control groups should be identical. Annex Figure A1 presents smooth kernel density estimates of the distribution of the propensity score (p^t) for treated and control units. The overlap between the distributions appears to be much larger in the sample of emerging economies relative to advanced economies. Moreover, in the sample of emerging economies, the estimated probability of treatment shows relatively little mass at values close to unity.

The point estimates reported in Table 2 for the fiscal multiplier in emerging economies change little with respect to the baseline OLS results. Multipliers are slightly larger than our baseline estimates on impact (around 0.6), and reach 0.67 after two years, which is statistically indistinguishable from the baseline results. While the multiplier point estimate for advanced economies rises somewhat, the difference between the two groups remains small and not statistically different from zero.

V. Extensions

A. The Effects of Fiscal Consolidation on Unemployment and Domestic Demand

In this section, we examine the effects of fiscal consolidations on the unemployment rate and on private domestic demand (private investment and private consumption), employing the same local projections specification used previously.10 Quantifying the effects of consolidations on private demand is particularly interesting, as the size of multipliers depends on private sector reactions to fiscal policy.

As shown in Figure 5, fiscal consolidations lead to a significant increase in the unemployment rate. Over a two-year horizon a fiscal consolidation of 1 percent of GDP leads to an increase in the unemployment rate of 0.3 percentage points in LACs, confirming the contractionary effects of consolidations. These effects are somewhat smaller than what is observed for advanced economies, where the unemployment rate multiplier reaches 0.5 at year two. The mitigated impact on unemployment in LAC may reflect the presence of a large informal sector in many countries, which offers an alternative margin of labor market adjustment following a demand shock.

Figure 5.
Figure 5.

Impact of Fiscal Consolidations on Unemployment

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Dark gray shading represents the 90 percent confidence interval.

Figure 6 reports the response of private consumption and investment following fiscal consolidations. These lead to a decline in private consumption in both country groups, with multipliers of around 1 after two years. The effects of consolidations on private investment are also similar across both groups of countries, with a multiplier reaching about 0.5 after two years. Overall, there does not seem to be any evidence of crowding-in effects of austerity measures on private demand. On the contrary, consolidations also linked to falls in private consumption and investment.

Figure 6.
Figure 6.

Impact of Fiscal Consolidations on Private Demand

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Dark gray shading represents the 90 percent confidence interval.

B. The Effect of Consolidations on the External Current Account Balance

An implication of many open-economy macro models with non-Ricardian features is that a fiscal consolidation would lead to a real exchange rate depreciation and would be accompanied by an improvement in the current account balance. Conversely, fiscal expansions lead to real exchange rate appreciation and a deterioration in the current account balance. Such predictions are in line with the so-called twin deficits hypothesis, which posits that fiscal consolidation can reduce external imbalances. In this section, we re-examine this issue for emerging economies by using our narrative dataset covering 14 LAC countries.

Figure 7 presents results for the effects of consolidations on the current account to GDP ratio using the same specification used to assess the impact of consolidations on GDP and adding the changes in the commodity export price and its lagged values as control variables. We find that consolidations improve the current account balance in line with the results obtained for AEs by Bluedorn and Leigh (2011) and supporting the “twin deficits” hypothesis.

Figure 7.
Figure 7.

Impact of Fiscal Consolidations on the Current Account

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Dark gray shading represents the 90 percent confidence interval.

Nevertheless, the magnitude of the effects of consolidations on the current account is somewhat larger in the LAC sample relative to advanced economies. In LAC, a one percent of GDP fiscal consolidation leads to an improvement in the current account balance of around 0.8 percent of GDP after two years. For advanced economies we find an impact of around 0.5 after two years using the narrative episodes already mentioned.11

Standard models suggest that the exchange rate is a mechanism for current account adjustment in response to fiscal policy.12 Figure 8 depicts the impact of fiscal shocks on the real effective exchange rate and support that this mechanism is at play. These specifications are similar to the baseline, but the log of the real effective exchange rate is the macroeconomic variable of interest.

Figure 8.
Figure 8.

Impact of Fiscal Consolidations on the Real Effective Exchange Rate

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Dark gray shading represents the 90 percent confidence interval.

We find that fiscal consolidations lead to a depreciation of the real exchange rate. The magnitude of the estimated effects is larger in the LAC sample, with an estimated depreciation of around 3 percent within two years compared to a depreciation of around 1.3 percent for AEs over the same period.

C. Differentiating between Booms and Slumps

Based on traditional Keynesian models, periods of economic slack are times where there is excess capacity in the economy, and therefore less crowding out of fiscal actions, which would imply larger fiscal multipliers. Furthermore, the proportion of credit constrained agents is also likely to be higher during recessions, which would also imply higher multipliers (IMF, 2017). In addition, monetary policy reaction to fiscal actions could also be different during recessions if central banks are more likely to accommodate the impact of additional spending on demand.

There is a burgeoning empirical literature attempting to estimate how multipliers vary depending on the state of the business cycle with inconclusive results. While several studies confirm the prior that multipliers are higher if fiscal actions are preceded by recession periods (Baum, Poplawaski Ribeiro and Weber, 2012; Jordà and Taylor, 2015; Auerbach and Gorodnichenko, 2012, among others); other studies do not find such differential effects (Alesina and others, 2018; Ramey and Zubairy, 2018).

In this sub-section, we examine whether the output effects of consolidations differ depending on the state of the business cycle. We carry out the estimation of the previous models on two bins of data depending on whether the economy is experiencing a boom (denoted by the b subscript) or a slump (denoted by r):

yi,t+hyi,t1=Si,t1[αbih+γbth+βbhΣs=tt+hFCi,s+δbXi,t]+(1Si,t1)[αrih+γrth+βrhΣs=tt+hFCi,s+δrXi,t]+εi,t+h.(2)

The state indicator variable Si,t–1 takes the values 0 or 1 depending on the sign of the output gap (obtained using the HP filter with a smoothing parameter of 6.25).13 The vector Xi,t contains the same control variables as in previous sections.

Figure 9 shows that differentiating between booms and slumps does not seem to influence the effects of fiscal consolidations in emerging economies, with consolidations being contractionary in either case. Multiplier estimates exceed 0.5 at the two-year horizon in both states (reaching 0.7 in slumps), and remain statistically significant at the 5 percent level. Results for advanced economies also fail to uncover stark differences in multipliers across states of the business cycle, with estimates of similar magnitudes to those obtained for emerging economies.

Figure 9.
Figure 9.

Impact of Fiscal Consolidation on Output: Booms vs. Slumps

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Dark gray shading represents the 90 percent confidence interval.

D. The Role of Perceived Sovereign Default Risk

The literature suggests that fiscal consolidations could be expansionary if they help to reduce borrowing costs by dissipating doubts about the financial solvency of the government (Guajardo, Leigh and Pescatori, 2014). Therefore, one would expect that consolidations that were preceded by periods of high perceived sovereign risk could lead to smaller output losses.

To test this hypothesis, we allow our multiplier estimates to vary across two states of perceived sovereign default risk. As a proxy, we use the Institutional Investor Ratings (IIR) index that is based on assessments of sovereign risk by private sector analysts on a scale of zero to 100 (with 100 assigned to the lowest perceived sovereign default probability). We split the samples into high (low) risk if the IIR index is below (above) the median for the sample of Latin American or advanced economies.

We then re-estimate the baseline equation 1 depending on whether the consolidation was preceded by a high-risk period or a low-risk period. Figure 10 presents the results. There is no evidence that fiscal consolidations are expansionary in emerging economies. The magnitude of the impact of fiscal consolidations on GDP is somewhat larger for episodes preceded by low perceived sovereign default risk within two years (1.2 compared to around 0.9 for periods of high risk).

Figure 10.
Figure 10.

Impact of Fiscal Consolidation on Output: High vs. Low Perceived Sovereign Risk

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Dark gray shading represents the 90 percent confidence interval.

For the sample of advanced economies, the impact of consolidations on output is not statistically significant when they are preceded by periods of high sovereign default risk (point estimates reach 0.2 after two years), but are highly significant for periods of low risk (reaching 0.9 after two years). Overall, the finding that consolidations preceded by high sovereign risk periods entail somewhat smaller estimated output losses over the medium-term is in line with the findings of Guajardo, Leigh and Pescatori (2014).

E. Differentiating Tax-Based from Spending-Based Consolidations

A number of studies based on data for advanced economies suggest that fiscal consolidations have smaller contractionary effects when implemented primarily through cutting government spending rather than by raising taxes. In this sub-section, we investigate this possibility for LAC economies.

We assess the differential impact of spending-based versus tax-based consolidations by estimating the following equation, where TBi,t is an indicator variable that is equal to 1 when a tax-based consolidation is underway, and zero otherwise.14 The vector of control variables includes the commodity export price and its lags, as well as two lags of real GDP growth and two lags of the narrative consolidation shocks (irrespective of whether they were tax-based or spending-based):

yi,t+hyi,t1=αih+γth+(βeh+βthTBi,t)[Σs=tt+hFCi,s]+δXi,t+εi,t+h.(3)

Figure 11 depicts the multiplier impact of tax and spending-based consolidations on GDP. For both types of packages and country groups, consolidations are contractionary, reinforcing the results presented in previous sections. For the sample of advanced economies, we confirm that spending-based consolidations are on average less contractionary than those that are tax based. Estimated multipliers reach 1.7 after two years in the case of tax-based consolidations, but only amount to about 0.4 for spending-based consolidations over the same horizon.

Figure 11.
Figure 11.

Impact of Fiscal Consolidations on Output: Tax- vs. Expenditure-Based Packages

Citation: IMF Working Papers 2018, 142; 10.5089/9781484361696.001.A001

Note: Dark gray shading represents the 90 percent confidence interval.

For LAC economies, however, there is no significant difference between the impact of spending- and tax-based consolidations. If anything, point estimates suggest that spending-based consolidations lead to somewhat larger declines in output after two years, with multipliers of 1.6 compared to 0.8 for tax-based consolidations. However, inference regarding the comparative size of multipliers across consolidation packages is inconclusive, since the impact of spending-based consolidations is imprecisely estimated in the sample of emerging economies. It is important to note that the average size of spending-based consolidations in the LAC sample is 1.2 percent of GDP, which is larger than the average size of tax-based consolidations (0.8 percent of GDP). The frequency of tax-based consolidations is also larger than spending-based consolidations in the sample of emerging economies.

Alesina and others (2017) argue that the differences in multipliers between spending and tax-based consolidations can reflect the persistence of the fiscal consolidation. Based on a New Keynesian model, they show that the size of the multiplier for government spending is decreasing in the persistence of fiscal shocks, while the multiplier for taxes is increasing in the persistence of the measures. Income taxes are modeled as an increase in labor income taxes, which distort labor market decisions by introducing a wedge between net wages and the marginal product of labor. A persistent increase in taxes makes the negative effect on labor supply more permanent, increasing the tax multiplier.

To inform the interpretation of the similarity of our estimation results for tax and spending shocks in our LAC sample, we therefore investigate the persistence of both types of shocks. We estimate autoregressive panel models of the following form:

FCi,t=ci+ρ1FCi,t1+ρ2FCi,t2+εi,t.(4)

For advanced economies, we confirm the result of Alesina and others (2017) that spending-based consolidation packages are more persistent than those that are tax-based. The sum of estimated coefficients ρ1 and ρ2 is 0.55 for spending-based packages (p-value of 0.00) and 0.12 for tax-based packages. For our LAC sample, however, spending-based shocks are no more persistent than tax-based ones. The sum of the coefficients ρ1 and ρ2 is −0.05 and 0.05, respectively. Based on the hypothesis of Alesina and others (2017), this similarity in persistence could account for our finding of an insignificant difference between multipliers across different types of consolidation packages in LAC economies.

VI. Conclusions

Based on a new narrative dataset of fiscal actions, this paper concludes that fiscal consolidation is typically contractionary in the near term in LAC economies. Our estimation results are consistent with an average multiplier of 0.9 after two years for LACs, implying larger output effects than suggested by existing studies based on more conventional identification approaches. When we compare these results with near-term multipliers for advanced economies, which we estimate using comparable methods and existing narrative datasets, we find a remarkable similarity. The effects of fiscal consolidation on economic activity in advanced and emerging market economies may thus be more similar than typically assumed.

We also find that fiscal consolidation leads to a rise in the external current account balance in LAC economies, in line with a strong “twin deficits” link. The results are broadly comparable to those that we obtain for AEs. A significant depreciation in the real effective exchange rate typically accompanies the external adjustment process. In addition, we find little evidence of crowding-in effects. Private consumption and investment typically decline following fiscal consolidation, and the unemployment rate rises. However, the rise in the unemployment rate is typically smaller than for AEs, which may reflect the presence of a larger informal sector.

We find that fiscal consolidations undertaken in periods of economic booms and slumps do not have significantly different estimated effects on economic activity in LAC economies. Their effects are contractionary in both cases. We also find little evidence of differences in multipliers according to whether the composition of consolidation packages is primarily based on spending- or tax-side measures in LACs. In contrast, we do find some evidence that consolidations that are preceded by lower perceived sovereign default risk result in larger output costs.

The Macroeconomic Effects of Fiscal Consolidation in Emerging Economies: Evidence from Latin America
Author: Mr. Yan Carriere-Swallow, Mr. Antonio David, and Mr. Daniel Leigh
  • View in gallery

    Distribution of Empirical Multiplier Estimates by Country Group

  • View in gallery

    Distribution of Budgetary Impact of Narrative Fiscal Shocks

    (Percent of GDP)

  • View in gallery

    Two Measures of Fiscal Consolidation: Changes in CAPB versus Narrative Fiscal shocks

    (Percent of GDP)

  • View in gallery

    Real GDP: Estimated Effect of a 1 Percent of GDP Fiscal Consolidation

  • View in gallery

    Impact of Fiscal Consolidations on Unemployment

  • View in gallery

    Impact of Fiscal Consolidations on Private Demand

  • View in gallery

    Impact of Fiscal Consolidations on the Current Account

  • View in gallery

    Impact of Fiscal Consolidations on the Real Effective Exchange Rate

  • View in gallery

    Impact of Fiscal Consolidation on Output: Booms vs. Slumps

  • View in gallery

    Impact of Fiscal Consolidation on Output: High vs. Low Perceived Sovereign Risk

  • View in gallery

    Impact of Fiscal Consolidations on Output: Tax- vs. Expenditure-Based Packages