## wp18142 - 2016. Each fiscal action is expressed in terms of its budgetary impact as a share of GDP, as

## Source details

**Canonical URL:** [wp18142 - 2016. Each fiscal action is expressed in terms of its budgetary impact as a share of GDP, as](https://www.imf.org/-/media/files/publications/wp/2018/wp18142.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18142.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18142.pdf.json)

---

### Fiscal actions dataset and summary statistics
- 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).
- Some actions enter the database with a negative sign (e.g., temporary tax increases subsequently reversed).
- Average budgetary impact (conditional on an action occurring): 0.9 percent of GDP.
- Standard deviation of budgetary impacts: about 1 percent of GDP.
- Figure 2 reports the density function for the size of the identified fiscal actions.

### Identification issues and alternative approaches
- Structural VAR-based identification: Riera-Crichton, Végh, and Vuletin (2016) show various tax series respond to output fluctuations, implying presumed exogenous fiscal changes (one quarter horizon) may be anticipated.
- Bias sign depends on cyclicality of fiscal policy, which varies across emerging and advanced economies.
- Real-time forecast errors in fiscal variables (Auerbach and Gorodnichenko, 2013; Furceri and Li, 2017) are an alternative but can introduce measurement error due to changes in non-policy factors.

### Classification of consolidation episodes
- Episodes classified as “tax-based” or “expenditure-based” depending on whether tax hikes or expenditure cuts account for most of the budgetary impact (Guajardo, Leigh and Pescatori, 2014).
- Observations: 55 tax-based consolidations; 18 expenditure-based episodes.
- Expenditure-based consolidations average size: about 1.2 percent of GDP.
- Tax-based consolidations average size reported elsewhere in the text: 0.8 percent of GDP for tax-based consolidations.

### Narrative approach and comparison with CAPB
- Narrative approach covers fiscal actions across revenue and spending instruments, allowing control for offsetting measures.
- VAT rate changes may be accompanied by other tax or spending changes not captured in some studies.
- CAPB comparison (October 2017 IMF WEO database):
  - CAPB series missing for some of the 14 LAC economies; a CAPB measure was constructed following conventional procedures (Annex).
  - In three cases (Costa Rica 2016; Jamaica 2003 and 2004) packages are balanced between tax hikes and spending cuts and are not classified.
  - Correlation between narrative episodes and contemporaneous change in CAPB is strong, but numerous discrepancies exist.
  - Inspection of the 16 largest discrepancies (Annex) suggests the narrative approach more accurately identifies the size of deficit-driven fiscal consolidation.
  - Discrepancies generally driven by specific economic or budgetary developments, including responses to current or prospective economic conditions such as onset of economic crisis.

### Orthogonality to news regarding the state of the economy
- Narrative approach aims to identify fiscal policy changes exogenous to current and prospective economic news.
- Economic news measure: revision to the forecast for current-year real GDP made in fall of year t relative to fall of year t–1, based on real-time WEO forecast revisions.
- Regression of narrative fiscal shocks on economic news (1989 - 2016):
  - Latin America: coefficient -0.04 (Std. Err. (0.03)), R2 0.16, N 364.
  - Advanced economies: coefficient -0.08 (Std. Err. (0.05)), R2 0.34, N 403.
- Conclusion:
  - Narrative fiscal consolidations are not significantly related to contemporaneous unexpected movements in output.
  - Orthogonality does not imply absence of correlation with past developments; consolidations motivated by reducing inherited deficits may be predictable.
  - Robustness section estimates multipliers using an average treatment effect that accounts for predictability.

### Empirical strategy
- Method: local projection (LP) method (Jordà 2005).
  - Advantages: no imposed shape on impulse responses; flexible for state-dependent responses; incorporates average transition characteristics from data.
- Benchmark specification horizons: h = 0, 1, 2 years.
  - Dependent variables y: log real GDP, current account balance as share of GDP, log real effective exchange rate, among others.
  - FC: measure of fiscal consolidation in percent of GDP.
  - X: controls including two lags of real GDP growth, two lags of fiscal shocks, contemporaneous growth rate of commodity export value and its lags.
  - Time and country fixed effects included.
- Multiplier interpretation:
  - Coefficient βh corresponds to multiplier estimate.
  - Fiscal multipliers defined as response of level of real GDP (or macro variable) relative to cumulative fiscal shock over a given horizon (“integral multipliers”).
- Estimation: OLS with narrative shocks directly in panel models. Annex presents variable definitions and sources.

### Results — Fiscal multipliers in emerging economies
- Impulse-response summary for a 1 percent of GDP fiscal consolidation:
  - LAC economies:
    - Output contraction on impact (h=0): 0.5 percent.
    - Output contraction after two years (h=2): 0.9 percent.
    - 90 percent confidence interval for h=2 stated: 0.6-1.2 percent.
  - Comparator (advanced economies, narrative shocks extended to 2014):
    - Fall in output on impact: 0.3 percent.
    - Fall in output after two years: 0.7 percent.
- Conclusion: average output responses for LAC and advanced economies are similar; point estimates within one standard error of each other at horizons of 1 and 2 years.

### Robustness checks (selected estimates from Table 2)
- Baseline estimates:
  - LAC h=0: -0.50*** (0.14)
  - Advanced h=0: -0.34*** (0.11)
  - LAC h=2: -0.88*** (0.15)
  - Advanced h=2: -0.71*** (0.18)
- Trimmed sample (drop largest/smallest 5 percent of fiscal changes and real GDP growth rates):
  - LAC h=0: -0.50* (0.24)
  - Advanced h=0: -0.47*** (0.12)
  - LAC h=2: -0.84*** (0.13)
  - Advanced h=2: -0.89*** (0.17)
- AIPW matching estimator (accounts for predictability):
  - LAC h=0: -0.60*** (0.24)
  - Advanced h=0: -0.28*** (0.09)
  - LAC h=2: -0.67* (0.44)
  - Advanced h=2: -0.94** (0.28)
  - First-stage probit controls: past consolidations (two lags), two lags of GDP growth, commodity export price changes and lags, lagged debt-to-GDP, lagged inflation, lagged current account to GDP, lagged changes in perceived sovereign default risk, country fixed effects.
- Interpretation:
  - Point estimates for emerging economies remain similar to baseline; multipliers slightly larger on impact (around 0.6) and reach 0.67 after two years.
  - Difference between emerging and advanced economies remains small and not statistically different from zero.

### Extensions — Unemployment and private domestic demand
- Unemployment (Figure 5):
  - Over two years, a 1 percent of GDP fiscal consolidation leads to increase in unemployment rate of 0.3 percentage points in LACs.
  - Advanced economies: unemployment rate multiplier reaches 0.5 at year two.
  - Interpretation: mitigated unemployment impact in LAC may reflect large informal sectors.
- Private consumption and investment (Figure 6):
  - Private consumption: declines in both groups, multipliers around 1 after two years.
  - Private investment: similar across groups, multiplier about 0.5 after two years.
  - Conclusion: no evidence of crowding-in; consolidations associated with falls in private consumption and investment.

### Extensions — External current account balance and real exchange rate
- Current account (Figure 7):
  - Consolidations improve current account balance.
  - Magnitudes after two years:
    - LAC: improvement around 0.8 percent of GDP.
    - Advanced economies: impact around 0.5 percent of GDP.
- Real effective exchange rate (Figure 8):
  - Consolidations lead to depreciation.
  - Magnitudes within two years:
    - LAC: estimated depreciation around 3 percent.
    - Advanced economies: estimated depreciation around 1.3 percent.

### Extensions — Booms vs. slumps
- State-dependent estimation using output gap (HP filter, smoothing parameter 6.25); boom (b) vs slump (r) bins.
- Results:
  - Differentiating booms and slumps does not materially influence effects in emerging economies: consolidations are contractionary in either state.
  - Multiplier estimates exceed 0.5 at two years in both states, reaching 0.7 in slumps, and remain statistically significant at the 5 percent level.
  - Average size of consolidations: slumps 1 percent of GDP; booms 0.7 percent of GDP.

### Extensions — Role of perceived sovereign default risk
- Hypothesis: consolidations preceded by high perceived sovereign risk could be less contractionary if they reduce borrowing costs.
- Proxy: Institutional Investor Ratings (IIR) index (scale 0 to 100; 100 = lowest perceived default probability).
- Sample split by IIR median: high (IIR below median) vs low (IIR above median) perceived risk.
- Results (Figure 10 summary):
  - No evidence that fiscal consolidations are expansionary in emerging economies.
  - Magnitude of GDP impact somewhat larger for episodes preceded by low perceived sovereign default risk within two years (text cutoff prevents full numeric reporting).
- Additional reported points:
  - For advanced economies:
    - Impact of consolidations on output not statistically significant when preceded by periods of high sovereign default risk (point estimates reach 0.2 after two years).
    - Consolidations preceded by low sovereign default risk are highly significant, reaching 0.9 after two years.
  - Overall: consolidations preceded by high sovereign risk periods entail somewhat smaller estimated output losses over the medium-term.

### Differential effects of tax-based versus spending-based consolidations
- Estimation uses indicator S_it = 1 for tax-based consolidation, 0 otherwise; controls include commodity export price and lags, two lags of real GDP growth, two lags of narrative consolidation shocks.
- Advanced economies:
  - Spending-based consolidations less contractionary than tax-based.
  - Estimated multipliers after two years:
    - Tax-based: 1.7
    - Spending-based: about 0.4
- LAC economies:
  - No significant difference between spending- and tax-based consolidations.
  - Point estimates after two years:
    - Spending-based: 1.6
    - Tax-based: 0.8
  - Inference is inconclusive due to imprecise estimates for spending-based consolidations.
  - Average sizes in LAC sample:
    - Spending-based consolidations: 1.2 percent of GDP.
    - Tax-based consolidations: 0.8 percent of GDP.
  - Frequency: tax-based consolidations more frequent than spending-based in the emerging economies sample.

### Persistence of consolidation shocks and interpretation (Alesina et al. (2017) hypothesis)
- Theory: in a New Keynesian model, government spending multiplier decreases with persistence; tax multiplier increases with persistence.
- Empirical autoregressive panel model estimates:
  - Advanced economies:
    - Sum of coefficients (ρ1 + ρ2) for spending-based packages: 0.55 (p-value 0.00).
    - Sum for tax-based packages: 0.12.
    - Interpretation: spending-based packages more persistent than tax-based in AEs.
  - LAC sample:
    - Sum for spending-based shocks: -0.05.
    - Sum for tax-based shocks: 0.05.
    - Interpretation: spending-based shocks are no more persistent than tax-based in LAC; similar persistence may explain insignificant difference in multipliers by consolidation type.

### Broader findings and macroeconomic responses in LAC economies
- Estimated average multiplier for LAC economies: 0.9 after two years.
- Comparison: near-term multipliers between AEs and LACs show remarkable similarity using comparable methods.
- External sector:
  - Fiscal consolidation raises external current account balance in LAC economies (strong “twin deficits” link).
  - Significant depreciation in real effective exchange rate typically accompanies external adjustment.
- Domestic demand and labor market:
  - Little evidence of crowding-in effects.
  - Private consumption and investment typically decline after consolidation.
  - Unemployment rate rises, but typically less than in AEs (may reflect larger informal sector in LACs).
- Business cycle phase:
  - Consolidations in booms and slumps are contractionary in LAC; no significant difference across phases.
- Composition:
  - Little evidence of multiplier differences by spending- vs tax-based packages in LACs.
- Sovereign risk:
  - Some evidence that consolidations preceded by lower perceived sovereign default risk result in larger output costs.

### Annex — Comparison of Narrative Deficit-driven Shocks and CAPB (cases where discrepancy > 2.5 percent of GDP)
- Overview:
  - 16 cases examined where discrepancy between narrative approach and CAPB exceeds 2.5 percent of GDP.
  - In all but two cases (Jamaica (1992) and Chile (2008)), CAPB increase is larger than narrative deficit-driven consolidation.
  - For missing WEO CAPB entries, authors construct CAPB per Escolano (2010); constructed CAPB highly correlated (correlation coefficient 0.9) with WEO CAPB where available.
- Selected country findings (discrepancy drivers emphasized):
  - Argentina (2002): CAPB improved ~3.7 percent of GDP; narrative shows little action-based consolidation; crisis responses and large non-cash expenditures dominate.
  - Bolivia (1994): CAPB improved close to 3 percent of GDP; improvement driven by revenue increases (foreign grants 1 percent of GDP; tax revenues 0.6 percent of GDP) not expenditure cuts.
  - Bolivia (2006): CAPB improved over 6 percent of GDP due to higher hydrocarbon revenues (royalties increased 3 percent of GDP) and favorable prices.
  - Chile (2008): policy record indicates fiscal easing of 0.5 percent of GDP while CAPB fell by 4 percent of GDP (discrepancy 3.5 percent); CAPB does not accurately identify easing size.
  - Costa Rica (1994): narrative easing 0.5 percent of GDP; CAPB falls by 3.1 percent of GDP (discrepancy 2.6 percent) due to sales tax reduction and import fall and transfers.
  - Dominican Republic (2015): CAPB rose 2.8 percent of GDP; no action-based consolidation—rise mainly reflects capital grant from Petrocaribe debt restructuring (3 percent of GDP).
  - Ecuador (1999, 2000): CAPB improvements driven by oil revenue recoveries and other endogenous factors; narrative consolidation smaller.
  - Jamaica (1992): policy record suggests consolidation 2.1 percent of GDP; constructed CAPB falls by 2 percent of GDP; CAPB estimates sensitive to method.
  - Mexico (1995): CAPB improved 3 percent of GDP; measures were crisis response under IMF program (VAT increase estimated impact 1.2 percent of GDP; cuts in primary expenditures 1.4 percent of GDP).
  - Paraguay (1989, 1990, 1998): CAPB improvements reflect both policy measures and large non-policy windfalls (Itaipú royalties); non-policy factors explain substantial portions of discrepancies.
- Drivers of discrepancies:
  - Specific financial operations (e.g., Bank of Jamaica losses).
  - One-off revenue realizations (bauxite, cellular license fees).
  - Windfall non-tax revenues (Itaipú royalties) driven by weather/external conditions.
  - Policy responses to current or prospective economic conditions, including crisis responses.
  - Tax policy timing and tax administration improvements causing CAPB misattribution.
- Overall conclusion from 16 largest discrepancies:
  - Narrative approach more accurately identifies size of deficit-driven fiscal consolidation for examined cases.
  - Discrepancies generally driven by specific economic or budgetary developments causing CAPB to misidentify deficit-driven consolidation.

*Source: wp18142 - 0.6 percent of GDP.*

### 2016. Each fiscal action is expressed in terms of its budgetary impact as a share of GDP, as

### wp18142 - 2016. Each fiscal action is expressed in terms of its budgetary impact as a share of GDP, as

### Fiscal actions dataset and summary statistics
- 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.
- The average budgetary impact (conditional on an action occurring) is 0.9 percent of GDP.
- The standard deviation of budgetary impacts is about 1 percent of GDP.
- Figure 2 reports the density function for the size of the identified fiscal actions.

### Identification issues and alternative approaches
- Riera-Crichton, Végh, and Vuletin (2016) show that when using identification strategies based on structural VARs, various tax series respond to output fluctuations, suggesting that presumed exogenous changes in fiscal policy (at a one quarter horizon) may in fact be anticipated by economic agents.
- They argue that the sign of the bias introduced in multiplier estimates using this approach depends on the cyclicality of fiscal policy, which varies across emerging and advanced economies.
- The use of real-time forecast errors in fiscal variables is another alternative approach (Auerbach and Gorodnichenko, 2013; Furceri and Li, 2017), although this too can lead to the introduction of measurement error due to changes in non-policy factors.

### Classification of consolidation episodes
- Consolidation episodes are classified 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).
- 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.

*Source: wp18142 - 2016. Each fiscal action is expressed in terms of its budgetary impact as a share of GDP, as*

### 0.8 percent of GDP for tax-based consolidations.

### 0.8 percent of GDP for tax-based consolidations.

### Narrative approach and comparison with CAPB
- The narrative approach covers fiscal actions across both revenue and spending instruments, allowing control for offsetting measures in packages.
- VAT rate changes might be accompanied by changes in other taxes or government spending not captured in some studies.
- Comparison with the change in the cyclically-adjusted primary balance (CAPB) from the October 2017 IMF WEO database:
  - CAPB series missing for some of the 14 LAC economies in the sample; a CAPB measure was constructed following conventional procedures (Annex).
  - In three cases (Costa Rica 2016; Jamaica 2003 and 2004) packages are balanced between tax hikes and spending cuts and are not classified in either category.
  - The correlation between narrative episodes and contemporaneous change in the CAPB is strong, but numerous discrepancies exist.
  - Inspection of the 16 largest discrepancies (Annex) suggests the narrative approach more accurately identifies the size of deficit-driven fiscal consolidation.
  - Discrepancies are generally driven by specific economic or budgetary developments, including responses to current or prospective economic conditions such as onset of economic crisis.

### Orthogonality to news regarding the state of the economy
- The narrative approach aims to identify fiscal policy changes exogenous to current and prospective news about the economy.
- Economic news measure:
  - Defined as the revision to the forecast for current-year real GDP made in the fall of year t relative to the forecast made in the fall of year t–1, based on real-time WEO forecast revisions.
- Regression of narrative fiscal shocks on this economic news measure (Table 1) — regressions estimated for 1989 - 2016:
  - Latin America: coefficient -0.04 (Std. Err. (0.03)), R2 0.16, N 364.
  - Advanced economies: coefficient -0.08 (Std. Err. (0.05)), R2 0.34, N 403.
- Conclusion:
  - Narrative fiscal consolidations are not significantly related to contemporaneous unexpected movements in output, economically or statistically.
  - This orthogonality does not imply absence of correlation with past developments; fiscal consolidations motivated by reducing inherited deficits are likely correlated with past developments and thus somewhat predictable.
  - Robustness section estimates multipliers using an average treatment effect that accounts for predictability.

### Empirical strategy
- Method: local projection (LP) method (Jordà 2005).
  - Advantages: does not constrain impulse response shape; flexible for state-dependent impulse responses; incorporates average transition characteristics from the data.
- Benchmark specification for horizons h = 0, 1, 2 years:
  - Dependent variable y: log real GDP, current account balance as share of GDP, log real effective exchange rate, among others.
  - FC: measure of fiscal consolidation in percent of GDP.
  - X: controls including two lags of real GDP growth, two lags of fiscal shocks, contemporaneous growth rate of commodity export value and its lags.
  - Time and country fixed effects included.
- Multiplier interpretation:
  - Coefficient βh corresponds to multiplier estimate.
  - Fiscal multipliers defined as response of level of real GDP (or macro variable) relative to cumulative fiscal shock over a given horizon (“integral multipliers”).
- Estimation:
  - Regressions estimated by ordinary least squares with narrative shocks included directly in panel models.
  - Annex presents variable definitions and sources.

### Results — Fiscal multipliers in emerging economies
- Figure 4 (impulse responses for a 1 percent of GDP fiscal consolidation):
  - LAC economies: output contraction of 0.5 percent on impact (h=0) and 0.9 percent after two years (h=2), with a 90 percent confidence interval of 0.6-1.2 percent (note: interval stated in text).
  - Comparator group (advanced economies, narrative shocks from Devries et al. 2011 extended to 2014):
    - Fall in output of 0.3 percent on impact, rising to 0.7 percent after two years.
  - Conclusion: average output responses for emerging (LAC) and advanced economies are similar; point estimates within one standard error of each other at horizons of 1 and 2 years.

### Robustness checks
- Table 2 reports sensitivity of multiplier estimates at h=0 and h=2 across specifications:
  - Baseline:
    - LAC h=0: -0.50*** (0.14)
    - Advanced h=0: -0.34*** (0.11)
    - LAC h=2: -0.88*** (0.15)
    - Advanced h=2: -0.71*** (0.18)
  - Trimmed sample (dropping largest and smallest 5 percent of fiscal changes and real GDP growth rates):
    - LAC h=0: -0.50* (0.24)
    - Advanced h=0: -0.47*** (0.12)
    - LAC h=2: -0.84*** (0.13)
    - Advanced h=2: -0.89*** (0.17)
    - Implication: outliers not driving results.
  - AIPW matching estimator (addresses predictability of consolidations):
    - LAC h=0: -0.60*** (0.24)
    - Advanced h=0: -0.28*** (0.09)
    - LAC h=2: -0.67* (0.44)
    - Advanced h=2: -0.94** (0.28)
    - Note: converting narrative shocks into a binary treatment loses size dimension; first-stage probit includes past consolidations (two lags), two lags of GDP growth, commodity export price changes and lags, lagged debt-to-GDP, lagged inflation, lagged current account to GDP, lagged changes in perceived sovereign default risk, country fixed effects.
- Interpretation:
  - Point estimates for emerging economies change little relative to baseline; multipliers slightly larger on impact (around 0.6) and reach 0.67 after two years.
  - Difference between emerging and advanced economies remains small and not statistically different from zero.

### Extensions — Unemployment and private domestic demand
- Unemployment (Figure 5):
  - Over a two-year horizon, a 1 percent of GDP fiscal consolidation leads to an increase in the unemployment rate of 0.3 percentage points in LACs.
  - For advanced economies, the unemployment rate multiplier reaches 0.5 at year two.
  - Interpretation: mitigated impact on unemployment in LAC may reflect large informal sectors offering alternative labor adjustment margins.
- Private consumption and investment (Figure 6):
  - Fiscal consolidations lead to declines in private consumption in both groups, with multipliers of around 1 after two years.
  - Effects on private investment similar across groups, with multiplier reaching about 0.5 after two years.
  - Conclusion: no evidence of crowding-in effects; consolidations are associated with falls in private consumption and investment.

### Extensions — External current account balance and real exchange rate
- Current account (Figure 7):
  - Consolidations improve the current account balance.
  - Magnitude:
    - LAC: a 1 percent of GDP fiscal consolidation leads to an improvement in the current account balance of around 0.8 percent of GDP after two years.
    - Advanced economies: impact of around 0.5 after two years.
- Real effective exchange rate (Figure 8):
  - Fiscal consolidations lead to depreciation of the real exchange rate.
  - Magnitude:
    - LAC: estimated depreciation of around 3 percent within two years.
    - Advanced economies: estimated depreciation of around 1.3 percent within two years.

### Extensions — Booms vs. slumps
- State-dependent estimation using output gap (HP filter, smoothing parameter 6.25); boom (b) vs slump (r) bins.
- Results (Figure 9):
  - Differentiating between booms and slumps does not materially influence effects in emerging economies: consolidations are 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 similarly show no stark differences across states.
- Note: average size of consolidations in slumps is higher (1 percent of GDP) compared to booms (0.7 percent of GDP).

### Extensions — Role of perceived sovereign default risk
- Hypothesis: consolidations preceded by high perceived sovereign risk could be less contractionary if they reduce borrowing costs.
- Perceived sovereign default risk proxy: Institutional Investor Ratings (IIR) index (scale 0 to 100; 100 = lowest perceived default probability).
- Sample split: high (IIR below median) vs low (IIR above median) perceived risk.
- Results (Figure 10 summary):
  - No evidence that fiscal consolidations are expansionary in emerging economies.
  - Magnitude of impact on GDP is somewhat larger for episodes preceded by low perceived sovereign default risk within two years (1.2 compared to around ... [text cuts off here]).

_Italic: Source: wp18142 - 0.8 percent of GDP for tax-based consolidations._

### 0.9 for periods of high risk).

### wp18142 - 0.9 for periods of high risk)

### Impact of consolidations under different perceived sovereign risk
- For advanced economies: the impact of consolidations on output is not statistically significant when preceded by periods of high sovereign default risk (point estimates reach 0.2 after two years).
- For advanced economies: consolidations preceded by periods of low sovereign default risk are highly significant, reaching 0.9 after two years.
- Overall interpretation: consolidations preceded by high sovereign risk periods entail somewhat smaller estimated output losses over the medium-term, in line with Guajardo, Leigh and Pescatori (2014).

### Differential effects of tax-based versus spending-based consolidations
- Estimation strategy: use an indicator variable S_it equal to 1 when a tax-based consolidation is underway and 0 otherwise, controlling for the commodity export price and its lags, two lags of real GDP growth, and two lags of the narrative consolidation shocks (irrespective of type).
- Advanced economies:
  - Spending-based consolidations are on average less contractionary than tax-based consolidations.
  - Estimated multipliers reach 1.7 after two years for tax-based consolidations.
  - Estimated multipliers reach about 0.4 after two years for spending-based consolidations.
- LAC economies:
  - No significant difference between the impact of spending- and tax-based consolidations.
  - Point estimates suggest spending-based consolidations lead to somewhat larger declines in output after two years: multipliers of 1.6 for spending-based vs. 0.8 for tax-based consolidations.
  - Inference is inconclusive because the impact of spending-based consolidations is imprecisely estimated in the sample of emerging economies.
  - Average size of spending-based consolidations in the LAC sample is 1.2 percent of GDP.
  - Average size of tax-based consolidations in the LAC sample is 0.8 percent of GDP.
  - The frequency of tax-based consolidations is larger than spending-based consolidations in the sample of emerging economies.

### Persistence of consolidation shocks and interpretation (Alesina et al. (2017) hypothesis)
- Theory: in a New Keynesian model, the multiplier for government spending decreases in the persistence of fiscal shocks, while the multiplier for taxes increases in persistence. Persistent income tax increases can amplify the tax multiplier via more permanent labor-supply distortions.
- Empirical autoregressive panel model estimated to assess persistence:
  - For advanced economies:
    - Sum of estimated coefficients (ρ1 + ρ2) is 0.55 for spending-based packages (p-value of 0.00).
    - Sum of estimated coefficients (ρ1 + ρ2) is 0.12 for tax-based packages.
    - Interpretation: spending-based consolidation packages are more persistent than tax-based ones, consistent with Alesina and others (2017).
  - For LAC sample:
    - Sum of estimated coefficients (ρ1 + ρ2) is -0.05 for spending-based shocks.
    - Sum of estimated coefficients (ρ1 + ρ2) is 0.05 for tax-based shocks.
    - Interpretation: spending-based shocks are no more persistent than tax-based ones in LAC; this similarity in persistence could account for the insignificant difference between multipliers across consolidation types in LAC economies.

### Broader findings and macroeconomic responses in LAC economies
- Estimated average multiplier for LAC economies: 0.9 after two years.
- Comparison with advanced economies: near-term multipliers estimated using comparable methods show remarkable similarity between AEs and LACs.
- External sector:
  - Fiscal consolidation leads to a rise in the external current account balance in LAC economies (strong “twin deficits” link).
  - A significant depreciation in the real effective exchange rate typically accompanies the external adjustment.
- Domestic demand and labor market:
  - Little evidence of crowding-in effects.
  - Private consumption and investment typically decline following fiscal consolidation.
  - The unemployment rate rises, but the rise is typically smaller than for AEs (may reflect a larger informal sector in LACs).
- Business cycle phase:
  - Consolidations undertaken in periods of economic booms and slumps do not have significantly different estimated effects on economic activity in LAC economies; effects are contractionary in both cases.
- Composition:
  - Little evidence of differences in multipliers according to whether the consolidation package is primarily spending- or tax-based in LACs.
- Sovereign risk:
  - Some evidence that consolidations preceded by lower perceived sovereign default risk result in larger output costs.

*Source: wp18142 - 0.9 for periods of high risk).*

### Chapter 4 of World Economic Outlook, October (Washington, DC: International

### Chapter 4 — World Economic Outlook, October (Selected Annexes and Tables)

### Literature on Fiscal Multipliers (Selected citations)
- Jooste, C. 2012. “Analyzing the Effects of Fiscal Policy Shocks in the South African Economy,” Department of Economics working paper 2012-06 (Pretoria: University of Pretoria).
- Jordà, Ò. 2005. “Estimation and Inference of Impulse Responses by Local Projections,” American Economic Review 95(1): 161–82.
- Jordà, Ò. and A. Taylor. 2015. “The Time for Austerity: Estimating the Average Treatment Effect of Fiscal Policy,” Economic Journal 126: 219–55.
- Kraay, A. 2012. “How Large is the Government Spending Multiplier? Evidence from World Bank Lending,” Quarterly Journal of Economics 127(2): 829–87.
- Mertens, K. and M.O. Ravn. 2013. “The Dynamic Effects of Personal and Corporate Income Taxes in the United States,” American Economic Review 103: 1212–47.
- Mountford, A. and H. Uhlig. 2009. “What are the Effects of Fiscal Policy Shocks?” Journal of Applied Econometrics 24: 960-992.
- Nakamura, E. and J. Steinsson. 2014. “Fiscal Stimulus in a Monetary Union: Evidence from U.S. Regions,” American Economic Review 104(3): 753–92.
- Ramey, V. 2011. “Identifying Government Spending Shocks: It’s All in the Timing,” Quarterly Journal of Economics CXXVI: 1–50.
- Ramey, V. and S. Zubairy. 2018. “Government Spending Multipliers in Good Times and in Bad: Evidence from U.S. Historical Data,” Journal of Political Economy, forthcoming.
- Numerous IMF working papers and regional studies cited (examples include Matheson and Pereira 2016; Muir and Weber 2013; Narita 2014).

### Table A1 — Empirical Fiscal Multiplier Estimates (Selected country multipliers)
- LAC countries: G 0.3, T 0.2
- Argentina: G 0.0, T 0.0 (Anós‐Casero and others (2010), SVAR)
- Brazil: G 0.5, T 0.5 (Matheson and Pereira (2016), SVAR)
- Caribbean (14 countries): G 0.5, T 0.6 (Narita (2014), SVAR)
- Caribbean (7 lower debt countries): G 0.8, T .
- Caribbean (4 higher debt countries): G 0.0, T .
- Chile: G 0.6, T . (Fornero, Guerra‐Salas, Perez (2017), SVAR)
- Mexico: G 0.7, T . (Valencia (2016), State‐level panel data)
- Paraguay: G 0.8, T 0.1 (David (2017), SVAR, LPM. Average of estimates)
- Peru: G 0.3, T 0.2 (Vtyurina and Leal (2016), TVAR, downturn)
- Non‐LAC EMDEs: G 0.6, T 0.5 (aggregate row)
- China: G 2.7, T . (Wang and Wen (2013), Government consumption)
- Malaysia (downturn): G 2.7, T 0.1 (Rafiq and Zeufack (2012)*)
- Malaysia (upturn): G 2.0, T 0.2 (Rafiq and Zeufack (2012)*)
- United States (selected estimates):
  - G 0.8 (Owyang, Ramey, and Zubairy (2013), Narrative method, high unemp.)
  - G 0.7 (Owyang, Ramey, and Zubairy (2013), Narrative method, low unemp.)
  - G 0.6 (Auerbach and Gorodnichenko (2012), SVAR, expansion)
  - G 2.5 (Auerbach and Gorodnichenko (2012), SVAR, recession)
  - G 0.7, T 1.1 (Barro and Redlick (2011), Defense spending)
  - G 0.9, T 0.9 (Blanchard and Perotti (2002), SVAR)
  - G .1.8 (Mertens and Ravn (2013), Narrative method, PIT)
  - G .0.6 (Mertens and Ravn (2013), Narrative method, CIT)
  - G 1.5 (Nakamura and Steinsson (2014), Military procurement)
  - G 0.9 (Ramey (2011), Narrative method)
  - G .3.1 (Romer and Romer (2010), Narrative method)
- AEs (aggregate): G 1.0, T 0.7 (top-of-table aggregate)
- OECD (Guanardo, Leigh, and Pescatori (2014)): G 1.0, T 3.1 (Narrative method, 17 OECD)
- OECD (Alesina and others (2017)): G 0.3, T 1.1 (Narrative method, 17 OECD)

(Note: Table A1 reports second-year multipliers or, when these are unavailable, impact multipliers. “G” denotes government spending, and “T” taxes. Missing data entries are reported as in the source.)

### Table A2 — Budgetary Impact of Narrative Fiscal Shocks (Percent of GDP) (Selected episodes)
- Argentina 1996: Total 0.25, Tax 0.25, Spend 0.00
- Argentina 1997: Total 0.75, Tax 0.75, Spend 0.00
- Bolivia 1995: Total 0.90, Tax 0.90, Spend 0.00
- Bolivia 2004: Total 2.00, Tax 2.00, Spend 0.00
- Bolivia 2005: Total 4.10, Tax 4.10, Spend 0.00
- Brazil 2015: Total 0.80, Tax 0.30, Spend 0.50
- Chile 1990: Total 0.50, Tax 0.50, Spend 0.00
- Chile 1991: Total 0.17, Tax 0.17, Spend 0.00
- Chile 2003: Total 0.60, Tax 0.20, Spend 0.40
- Chile 2004: Total 0.40, Tax 0.40, Spend 0.00
- Chile 2008: Total -0.50, Tax 0.00, Spend -0.50
- Colombia 2000: Total 0.90, Tax 0.00, Spend 0.90
- Costa Rica 1990: Total 1.50, Tax 1.50, Spend 0.00
- Costa Rica 1991: Total 3.10, Tax 3.10, Spend 0.00
- Dominican Republic 2004: Total 1.70, Tax 0.50, Spend 1.20
- Dominican Republic 2013: Total 3.80, Tax 1.80, Spend 2.00
- Guatemala 1995: Total 0.80, Tax 0.80, Spend 0.00
- Mexico 1989: Total 0.90, Tax 0.90, Spend 0.00
- Paraguay 1989: Total 2.60, Tax 2.00, Spend 0.60
- Paraguay 2001: Total 1.80, Tax 0.50, Spend 1.30
- Peru 1992: Total 1.00, Tax 1.00, Spend 0.00
- Uruguay 2002: Total 3.27, Tax 1.58, Spend 1.70
- Jamaica 1992: Total 2.10, Tax 2.10, Spend 0.00
(Selections shown; full Table A2 provides a larger set of country-year Total/Tax/Spend entries in percent of GDP.)

### Table A3 — Variable Definitions and Sources (selected)
- Real GDP: Log of real GDP in local currency. Source: WEO database (October 2017 published vintage).
- Real Private Investment and Consumption: Log of real values in local currency. Source: WEO database (October 2017 published vintage).
- Current Account to GDP Ratio: WEO database (October 2017 published vintage).
- Unemployment rate: WEO database (October 2017 published vintage).
- Narrative Fiscal Consolidations: Elaborated by the authors based on contemporaneous reports and budget documents. Source: David and Leigh (2018).
- Cyclically Adjusted Primary Balance: Cyclical adjustment based on WEO data follows standard IMF approach (see annex). Authors’ calculations based on WEO database.
- Commodity Export Value: Calculated using trade data on 33 commodities. Source: Gruss (2014).
- Real Effective Exchange Rae: Change in the Log of the real effective exchange rate index (2010=100) x 100. Source: WEO database (October 2017 published vintage).
- Debt to GDP Ratio: IMF FAD Historical Debt Database.
- Output Gap: Cyclical component of real GDP using the HP filter with smoothing parameter of 6.25. Authors’ calculations based on WEO database.
- Perceived Sovereign Default Risk: Institutional Investor Ratings index. Source: Institutional Investor LLC.

### Annex — Comparison of Narrative Deficit-driven Shocks and CAPB (Cases where discrepancy > 2.5 percent of GDP)
- Overview:
  - The annex examines 16 cases where the discrepancy between the narrative approach and changes in the CAPB exceeds 2.5 percent of GDP.
  - In all but two cases (Jamaica (1992) and Chile (2008)), the increase in CAPB is larger than the deficit-driven fiscal consolidation identified by the narrative record.
  - CAPB source: October 2017 IMF World Economic Outlook (WEO) database. For missing WEO CAPB entries, the authors construct CAPB following conventional procedures (Escolano, 2010). The constructed CAPB measure is highly correlated (correlation coefficient 0.9) with WEO CAPB where available.

- Argentina (2002):
  - CAPB improved by about 3.7 percent of GDP.
  - Narrative record shows little evidence of action-based fiscal consolidation motivated by deficit reduction; policy actions were responses to a sharp financial crisis and recession (currency board abandoned January 2002, moratorium on public debt service, bank deposit freeze, exchange controls, export surrender requirements).
  - Large expenditures not captured in cash balance (e.g., issuance of Arg$36 billion of bonds; interest capitalization Arg$26.4 billion); comprehensive measure would bring augmented primary and overall deficits of consolidated public sector in 2002 to 11 percent and 25 percent of GDP, respectively.

- Bolivia (1994):
  - CAPB improved by close to 3 percent of GDP.
  - Fiscal actions were aimed at reducing inflation and rebuilding international reserves; narrowing of the primary deficit mainly reflected increases in revenues (foreign grants 1 percent of GDP; tax revenues improvement 0.6 percent of GDP) rather than expenditure cuts.
  - Classified as not deficit-driven consolidation.

- Bolivia (2006):
  - CAPB improved by over 6 percent of GDP.
  - Improvement largely reflected higher hydrocarbon revenues (royalties increased by 3 percent of GDP in 2006) and favorable international prices, not action-based consolidation.

- Chile (2008):
  - Policy record indicates fiscal easing of 0.5 percent of GDP.
  - CAPB fell by 4 percent of GDP based on authors’ calculations, implying a discrepancy of 3.5 percent of GDP.
  - Motivation for reduction in structural target was long-run considerations; country desk calculations also point to a decrease in CAPB of over 4 percent of GDP. Authors conclude CAPB does not accurately identify the size of fiscal easing for this episode.

- Chile (2010):
  - CAPB improved by about 3.8 percent of GDP.
  - Improvement reflected the end of temporary stimulus measures adopted in 2009 and temporary taxes to finance reconstruction after the February 2010 earthquake. Authors consider the discrepancy due to CAPB picking up effects of fiscal actions responding to the economic outlook.

- Costa Rica (1994):
  - Narrative record indicates fiscal easing of 0.5 percent of GDP; CAPB falls by 3.1 percent of GDP (discrepancy 2.6 percent of GDP).
  - Sales tax rate reduced from 13 percent to 10 percent in January 1994, estimated revenue loss of 1.5 percent of GDP over three years (assigned 0.5 percent to calendar year 1994). Revenues also fell due to a fall in imports and increased transfers to the electricity company; these responses to economic conditions are excluded from the narrative shocks.

- Dominican Republic (2015):
  - CAPB rose by 2.8 percent of GDP.
  - No evidence of action-based fiscal consolidation; rise in CAPB mostly reflects a capital grant from restructuring of Petrocaribe debt recorded as an above-the-line capital grant in 2015 (3 percent of GDP).

- Ecuador (1999):
  - CAPB improved by 4.5 percent of GDP.
  - 1999 marked by domestic financial crisis; improvement driven by recovery in oil revenues (petroleum revenue increased by 2.9 percent of GDP in 1999) and tax measures (import tariff surcharge and a financial transactions tax). Many revenue gains and tax measures were motivated by endogenous shocks and thus excluded from narrative consolidation measure.

- Ecuador (2000):
  - CAPB improved by 3.5 percent of GDP.
  - Policy record suggests consolidation of 0.5 percent of GDP (discrepancy 3.0 percent of GDP). Revenue measures of 1.3 percent of GDP partly offset by expenditure increases of 0.8 percent of GDP; oil revenues increased by 3 percent of GDP in 2000.

- Jamaica (1992):
  - Policy record suggests consolidation of 2.1 percent of GDP.
  - Baseline constructed CAPB falls by 2 percent of GDP; discrepancy noted as 4.1 percent of GDP. CAPB estimates for Jamaica 1992 are sensitive to method used; WEO does not report CAPB for Jamaica in 1992.

- Methodological note on CAPB construction:
  - The authors’ conventional CAPB measure adjusts government revenues using an elasticity of 1 for the output gap (estimated with an HP filter smoothing parameter 6.25), uses the GDP deflator to express adjusted series in nominal terms, then computes adjusted primary balance by subtracting interest revenue and primary expenditure and dividing by nominal GDP.

*Source: Chapter 4, World Economic Outlook, October (Washington, DC: International Monetary Fund).*

### 0.6 percent of GDP.

### 0.6 percent of GDP.

### Summary of the discrepancy between CAPB and narrative measures
- Almost half of the reported discrepancy—1.6 percent of GDP (2-0.6)—is due to differences in the CAPB versions.
- Of the remaining 2.5 percentage point discrepancy (4.1-1.6), most is explained by Bank of Jamaica-related operations.
- Specific financial operations and non-policy factors often cause the CAPB-based measure to inaccurately identify the size of deficit-driven fiscal consolidation.

### Country case studies and key quantitative findings
- Jamaica (1991/92–1992/93)
  - The deficit of the Central Government (including Bank of Jamaica losses) remained broadly constant in 1991/92 and 1992/93; modest improvement if such losses are excluded.
  - Fiscal surplus of the rest of the public sector increased from 0.7 percent of GDP in 1991/92 to 3.3 percent of GDP in 1992/93.
  - Conclusion: specific financial operations explain why the CAPB-based measure misstates the size of consolidation.

- Jamaica (1998)
  - The CAPB rose by 3.7 percent of GDP.
  - Authorities framed FY 1998 program aiming at output growth of 4-5 percent by FY 2000, reducing annual inflation to 3 percent by that year, and keeping international reserves at around the equivalent of 14 weeks of imports.
  - Central government balance (excluding costs related to support of the financial system) would turn to a surplus of about 2.5 of GDP.
  - Much of the CAPB increase reflected deliberate tightening to reduce inflation; not considered primarily deficit-reduction motivated.

- Jamaica (1999)
  - The CAPB improved by 3.7 percent of GDP.
  - Policy record indicates an action-based fiscal consolidation of 0.7 percent of GDP, implying a discrepancy of 3 percent of GDP.
  - Narrative: expenditure cuts of 0.7 percent of GDP on a net basis; much of observed revenue improvement (~2 percent of GDP) linked to one-off factors (bauxite and cellular license fees).
  - Conclusion: CAPB overstates the size of fiscal policy change.

- Mexico (1995)
  - The CAPB improved by 3 percent of GDP.
  - Measures were a policy response to a financial crisis under an IMF-supported adjustment program.
  - Key measures: increase in general VAT from 10 to 15 percent effective May 1995; estimated impact of VAT increase of 1.2 percent of GDP.
  - Cuts in primary expenditures of 1.4 percent of GDP, including reductions in the wage bill (0.7 percent of GDP) and capital expenditures (0.6 percent of GDP).
  - Conclusion: CAPB improvement reflects crisis-response and prospective conditions rather than a conventional deficit-driven consolidation.

- Paraguay (1989)
  - The CAPB improved by 5.2 percent of GDP, but action-based consolidation amounted to 2.6 percent of GDP (discrepancy of 2.6 percent of GDP).
  - Tax revenue increased by about 2 percent of GDP in 1989 due to policy measures.
  - Non-tax revenues (royalties from Itaipú) increased by about 1.4 percent of GDP in 1989, mostly due to favorable weather.
  - Conclusion: non-policy factors not adjusted for in CAPB explain a large share of the discrepancy.

- Paraguay (1990)
  - The CAPB improved by 3.7 percent of GDP.
  - Improvement reflected policy actions to reduce inflation and stabilize reserves—cutbacks in current and capital spending by the public sector.
  - Conclusion: CAPB reflects a policy response to prospective economic conditions.

- Paraguay (1998)
  - The CAPB improved by 2.9 percent of GDP.
  - Incoming administration (August 1998) implemented measures including excise duty increases, reduction of some tax exemptions, and tightening tax administration.
  - An indirect estimate of the impact of measures was in the order of 2 percent of GDP; deficit for 1998 as a whole estimated to be 1% percent of GDP, just above that of 1997.
  - Windfall gains from Itaipú royalties increased to 2.8 percent of GDP in 1998, 0.9 percent of GDP higher than in 1996.
  - Conclusion: gap between CAPB and narrative measure reflects both policy response to economic conditions and non-policy factors.

### Drivers of discrepancies identified
- Specific financial operations (e.g., Bank of Jamaica losses) that are captured differently across CAPB versions.
- One-off revenue realizations (e.g., bauxite and cellular license fees).
- Windfall non-tax revenues (e.g., Itaipú royalties) driven by weather or external conditions.
- Policy responses to current or prospective economic conditions, including responses to financial or balance-of-payments crises.
- Tax policy changes and tax administration improvements whose timing or nature lead CAPB to misattribute revenue changes to cyclicality rather than policy.

### Overall conclusion from the 16 largest discrepancies
- The narrative approach more accurately identifies the size of deficit-driven fiscal consolidation for the examined cases.
- 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, including policy responses to economic crises and non-policy windfalls.

### Figure note
- Figure A1: Overlap Check — Empirical Distributions of the Treatment Propensity Score (Control units vs Treated units; separate panels for LAC and Advanced).

*Source: wp18142 - 0.6 percent of GDP.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18142.pdf_