## _wp1459

## Source details

**Canonical URL:** [_wp1459](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1459.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1459.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1459.pdf.json)

---

### Introduction and contribution
- Context:
  - Prior literature (starting with Gavin and Perotti (1997)) finds fiscal policy more procyclical in emerging and developing economies than in advanced economies.
  - Studies differ in definitions of cyclicality and treatment of automatic stabilizers and special revenue factors (commodity or asset price booms, compositional GDP changes).
- Contribution:
  - Systematic overview of possible definitions of cyclical fiscal policy and categorization of existing studies.
  - Empirical analysis on data covering 1980 to 2012 using a new operational definition of procyclical policy that: (i) gives credit to automatic stabilizers that are part of tax or benefit systems; (ii) does not give credit to automatic revenue increases that stem from taxing additional GDP at stable tax rates.
  - Uses instrumental-variable approaches and system-GMM to address endogeneity of the output gap/growth.

### New definition of procyclical fiscal policy and measurement implications
- Key conceptual choices:
  - Variables: R = revenue, G = primary government spending, B (=R-G) = primary fiscal balance, Y = GDP, Y* = potential output, Δ indicates difference between actual outcome and hypothetical outcome if output were at potential.
  - Acyclical revenue: average tax rate constant so revenues remain constant as a share of GDP (revenue rises at the rate of nominal growth).
  - Acyclical primary spending: primary spending constant as a share of potential GDP (so spending ratio falls in a boom).
  - Regression implication: under perfectly acyclical policy the coefficient in a simple regression relating the fiscal balance to the output gap equals the expenditure ratio; coefficients above that imply countercyclical policy and below imply procyclical policy.
- Three definitions of countercyclical policy:
  - Definition A (strongest): discretionary fiscal policy must tighten; cyclically adjusted primary balance must improve.
  - Definition B (middle, preferred): tightening may be due to discretionary measures or automatic stabilizers; primary balance must improve by more than the temporary revenue gain; gives credit to automatic stabilizers but not to revenue gains at stable average tax rates.
  - Definition C (weakest): tightening may be due to discretionary policy, automatic stabilizers, or additional revenue at stable average tax rates; primary balance must improve.
- Empirical strategy:
  - Typical regressions: fiscal variables (fiscal balance or spending) on output gap or growth.
  - Endogeneity addressed via instruments (export-weighted growth of trading partners, US real interest rate) and system-GMM (Blundell and Bond (1998)), treating output gap/growth and lagged fiscal variables as endogenous.

### Data and measurement specifics
- Main data sources:
  - World Economic Outlook database (Fall 2013).
  - US treasury bill rates from the Federal Reserve Board.
  - Trade and commodity price data from the UN Comtrade database.
  - Resource-related fiscal revenues in Chile and Mexico from the IDB Fiscal Resources data set.
- Sample coverage: 1980 to 2012 (actual availability varies by country and variable).
- Real-time regressions: WEO vintages from 1990 through 2012 used for historical output-gap series.
- Primary balance construction: own calculations adding back interest expenditure to overall government net lending (exception: Brazil uses net interest expenditure).
- Commodity price index: newly-calculated, time-varying weights, lagged by d periods, constructed from log prices in US dollars and export values; two alternatives (commodity export shares and commodity terms of trade) produced very similar results.

### Panel-data empirical findings (within-group and GMM estimates)
- General observation: results are sensitive to specification and estimation technique.
- Within-group estimates (selected coefficients on the output gap; see Table 4):
  - Advanced economies (columns 1–3):
    - Output gap: 0.434***, 0.132**, 0.037
    - (standard errors): (0.068), (0.061), (0.132)
    - Lagged fiscal balance: -0.296***, -0.234***, -0.318***
    - (standard errors): (0.028), (0.025), (0.061)
    - Observations: 791, 791, 576
    - R-squared: 0.207, 0.121, 0.180
    - Number of countries: 33, 33, 31
  - Emerging/developing economies (columns 4–6):
    - Output gap: 0.285, -0.102, -0.240**
    - (standard errors): (0.200), (0.116), (0.106)
    - Lagged fiscal balance: -0.587***, -0.572***, -0.217
    - (standard errors): (0.083), (0.107), (0.128)
    - Observations: 2,568, 2,568, 311
    - R-squared: 0.312, 0.302, 0.153
    - Number of countries: 146, 146, 29
  - Latin America (columns 7–9):
    - Output gap: -0.122, -0.335**, -0.346*
    - (standard errors): (0.139), (0.149), (0.183)
    - Lagged fiscal balance: -0.490***, -0.495***, -0.040
    - (standard errors): (0.093), (0.082), (0.111)
    - Observations: 355, 355, 103
    - R-squared: 0.248, 0.299, 0.111
    - Number of countries: 20, 20, 9
- Interpretation of Table 4:
  - Advanced economies: evidence consistent with countercyclical fiscal policy (primary balance tightens in stronger activity); CAPB result insignificant, suggesting tightening generally not discretionary.
  - Emerging/developing: mostly insignificant; CAPB shows negative and significant coefficient on smaller sample (procyclical).
  - Latin America: evidence of procyclical policy; significance persists with CAPB for the subset with CAPB data.
  - Adjusted primary balance coefficients lie between CAPB and primary balance coefficients.
- Table 5 (commodity-price control and system-GMM):
  - Specification splits: each region estimated with WG, WG + commodity-price control, and GMM; commodity-price coefficient often significant and positive when significant.
  - Advanced economies: patterns point to countercyclical policy though sensitivity exists; some GMM results indicate countercyclicality.
  - Emerging/developing: generally unclear and mostly insignificant.
  - Latin America: output-gap based results including system-GMM indicate procyclicality; growth-rate based results are insignificant.
  - System-GMM diagnostics: generally passed except regression (3’) where Sargan/Hansen rejects overidentifying restrictions with p-value 4 percent (does not alter overall interpretation for advanced economies).

### Spending-based panel estimates (Table 6)
- Rationale: adjusted spending ratio (spending as share of potential GDP) avoids some fiscal-balance complications; positive coefficient indicates procyclicality (interpretation reversed relative to balance regressions).
- Selected output-gap specification coefficients:
  - Advanced economies (columns 1–3):
    - Output gap: 0.132**, 0.138**, 0.302***
    - (standard errors): (0.061), (0.057), (0.106)
    - Lagged adjusted prim. spending: -0.234***, -0.228***, 0.042
    - (standard errors): (0.025), (0.025), (0.103)
    - Observations: 791, 760, 760
    - R-squared: 0.121, 0.184, 0.302
    - Number of countries: 33, 32, 32
  - Emerging/developing economies (columns 4–6):
    - Output gap: -0.102, 0.010, 0.066
    - (standard errors): (0.116), (0.149), (0.131)
    - Lagged adjusted prim. spending: -0.572***, -0.585***, -0.741***
    - (standard errors): (0.107), (0.134), (0.145)
    - Observations: 2,568, 2,036, 2,036
    - R-squared: 0.398, 0.299, 0.343
    - Number of countries: 146, 134, 134
  - Latin America (columns 7–9):
    - Output gap: -0.335**, -0.343**, -0.337**
    - (standard errors): (0.149), (0.142), (0.150)
    - Lagged adjusted prim. spending: -0.495***, -0.441***, -0.479***
    - (standard errors): (0.082), (0.082), (0.125)
    - Observations: 355, 333, 333
    - R-squared: 0.248, 0.299, 0.343
    - Number of countries: 20, 19, 19
- Growth-rate specifications (selected advanced economy results): Growth 0.267***, 0.260***, 0.335*** in columns (1)'-(3)'.
- Interpretation:
  - Advanced economies: spending-based estimates indicate countercyclical policy, significant mainly in growth specifications.
  - Emerging/developing: mostly insignificant.
  - Latin America: spending-based evidence suggests procyclicality.
- Robustness: alternative commodity-price indicators (different weighting or net vs gross exports) produced broadly similar results.

### Country-specific regressions, time variation, and real-time vs ex-post measures
- Motivation: allow heterogeneity across countries; regressions run 1990–2012 with lagged adjusted fiscal balance and commodity price index.
- Instrumental-variable specifications:
  - IV-1: lagged output gap as instrument.
  - IV-2: export-weighted GDP growth of trading partners and the US 1-year treasury rate as instruments.
- Time-varying regressions: coefficients estimated pre-2005 and change since 2005 (2005 chosen based on Appendix Table A4).
- Key empirical patterns:
  - Very small number of statistically significant country-specific results.
  - Two countries initially identified as procyclical in OLS: Ecuador and Venezuela; OLS procyclicality for these does not hold up in IV estimates for Ecuador and Venezuela.
  - Guyana: positive and significant in OLS but not supported by IV.
  - No country-specific results provide strong evidence of countercyclical policy.
  - Instruments often weak: many first-stage regressions have extremely low R-squared; weak-instrument concerns yield implausible IV coefficients in some cases (example: IV-1 in Brazil).
- Time-varying findings:
  - Fiscal policy appears to have become more countercyclical since 2005 in Brazil, Chile, Colombia, El Salvador, and Mexico.
  - Honduras notable: pre-2005 Δ coefficient 1.510* (0.654); since 2005 coefficient -1.422* (0.685) — summary statement: “Only Honduras appears to have become more procyclical.”
- Real-time (historical output gap) regressions (Table 8) to assess policymakers’ intentions using WEO vintages:
  - Costa Rica (ex ante coefficients on historical output gap):
    - OLS: 1.512*** (0.329)
    - IV-1: 1.519 (1.962)
    - IV-2: 1.992*** (0.413)
  - Honduras (ex ante):
    - OLS: 0.900 (0.567)
    - IV-1: 1.498 (0.847)
    - IV-2: 1.115 (0.962)
  - Interpretation: Costa Rica’s budgets appear to have been aiming to be countercyclical based on data available at the time; two countries exhibited procyclical fiscal policy even using historical data.
- Caveats:
  - Output gap estimates change with new data; ex-post assessments may differ from what policymakers intended in real time.
  - For historical output gap regressions the need for IV is debatable because the forecasted gap may be less endogenous.
  - Brazil: increasing policy lending (not in fiscal balance) affects measured cyclicality; adding net policy lending increases the post-2005 coefficient but standard error rises and result becomes insignificant.

### Commodity and asset price effects
- Commodity-exporting countries: commodity price movements can boost revenues beyond real GDP growth via wealth/transaction taxes and commodity profits; separating cyclical vs structural components is difficult.
- Table 3 logic on apparent cyclicality when positive commodity revenue occurs (qualitative):
  - If additional commodity revenue is spent and output gap is positive: appears acyclical but is procyclical.
  - If additional commodity revenue is saved and output gap is positive: appears counter-cyclical but is acyclical.
  - If additional commodity revenue is spent and output gap is negative: appears acyclical but is counter-cyclical.
  - If additional commodity revenue is saved and output gap is negative: appears pro-cyclical but is acyclical.
- Approaches used:
  - Exclude commodity-related revenues (data sparse).
  - Analyze spending rather than balances.
  - Control for commodity price developments via commodity price index; commodity-price coefficient often significant and always positive when significant, indicating adjusted primary balance strengthens with commodity price growth unless countries spend more than additional revenues.

### Overall conclusions
- Using a methodology that gives full credit to automatic stabilizers, controls for commodity prices, and allows for endogeneity:
  - Panel-data evidence broadly confirms countercyclical fiscal policy in advanced economies (sensitive to specification and estimator).
  - Fiscal policy has been procyclical on average in Latin America.
  - Evidence for the broader sample of emerging and developing countries is inconclusive.
- Country-level evidence is frequently insignificant; acyclical fiscal policy cannot be ruled out for most countries.
- Two countries show significant evidence of procyclicality at the country level (noted but findings weaken under IV).
- Real-time data indicate Costa Rica intended countercyclical policy.
- Recent movements toward more countercyclical policy in Brazil, Chile, Colombia, El Salvador, and Mexico.
- Honduras is the notable outlier toward greater procyclicality.

*Source: _wp1459 (References, conclusions, appendix and tables inventory as provided in the source content).*

### References .............................................................................................................

### _wp1459 - References

### Tables and Appendix Tables (inventory)
- Tables
  - 1: Possible definitions of countercyclical policy in a boom ...................................................................................................... 8
  - 2: Definitions of counter-cyclical fiscal policy in the literature ............................................................................................ 10
  - 3: Apparent cyclicality of fiscal policy from overall fiscal balance in the presence of positive commodity price shocks .......................................................................................................... 12
  - 4: Within-group estimates of the cyclicality of fiscal policy for various fiscal balances. .................................................. 15
  - 5: The adjusted fiscal balance in various panel data estimates. ........................................................................................... 17
  - 6: Adjusted primary spending ratio in panel data estimates. .............................................................................................. 18
  - 7: Coefficients on the output gap from country-by-country regressions ................................................................. 21
  - 8: Ex ante cyclicality: coefficients on the historical output gap from country-by-country regressions ................................ 22

- Appendix Tables
  - A1: Overview of specifications in papers estimating the cyclicality of the fiscal balance ...... 24
  - A2: Overview of specifications in papers estimating the cyclicality of spending ................... 25
  - A3: Overview of specifications in papers estimating the cyclicality of revenues ................. 26
  - A4: Within group regressions of the adjusted fiscal balance in Latin America, with time interactions ................................ 27
  - A5: Coefficients on the output gap from country-by-country regressions of the adjusted spending ratio ................................ 28
  - A6: Country-by-country regressions using adjusted balance net of resource revenues .......... 29

### Introduction and Main Findings — key points
- Context and methodological variation
  - A number of papers starting with Gavin and Perotti (1997) have shown that fiscal policy has tended to be more procyclical in emerging and developing economies than in advanced ones.
  - Studies differ substantially in methodological approaches, notably in the definition of cyclicality of fiscal policy and in the treatment of automatic stabilizers and special revenue factors (for example, commodity or asset price booms or compositional changes in GDP).
  - The choice to include or exclude automatic stabilizers is often not explicit or justified across studies.

- Contribution of this paper
  - Provides a systematic overview of possible definitions of cyclical fiscal policy and categorizes existing studies according to those definitions.
  - Contains an empirical analysis on the latest data from a broad sample of developing and advanced economies using a new definition of procyclical policy.

- New definition of procyclical fiscal policy (conceptual)
  - Gives credit to automatic stabilizers that are part of the tax or benefit systems.
  - Does not give credit to automatic revenue increases that stem from taxing additional GDP at stable tax rates.
  - In practice, this new definition lies between the common approaches of (i) counting only discretionary measures and (ii) recording any revenue change.

- Empirical approach
  - Typical estimation approach: regress fiscal variables (fiscal balance or spending) on indicators of the economic cycle (output gap or growth).
  - Recognizes endogeneity of the output gap or growth because these are affected by fiscal policy through the multiplier.
  - Uses instrumental-variable approaches, including a system-GMM estimator, to address endogeneity.

- Main empirical findings
  - Panel data methods provide evidence suggesting countercyclical fiscal policy in advanced economies.
  - Evidence suggests generally procyclical policy in Latin America.
  - For a broader sample of emerging markets, evidence is less clear.
  - These results are broadly in line with the majority of existing literature.
  - Results are very sensitive to specifications, making it hard to draw firm conclusions.

*Source: _wp1459 - References (pages and table listings as provided in the source content).*

### conclusions.

### _wp1459 - conclusions.

### Main empirical conclusions
- Many Latin American countries "appear to have procyclical policy," but "country-specific empirical estimates typically lack statistical significance."
- "The evidence for the presence of procyclical policy is consistently significant only in two countries."
- "A few countries (Brazil, Chile, Colombia, El Salvador, and Mexico) appear to have moved toward less procyclical policy since 2005."
- "Commodity exports are important in many emerging economies, including in Latin America." The paper "controls for the impact of commodity prices on fiscal variables, which is found to be strong and significant."
- Caveat: "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," because "having to deal with a sudden financing constraint or having to reduce debt toward a sustainable level may trump cyclicality concerns."

### Methodological issues and definitions
- Output gap measurement:
  - Potential GDP is hard to measure; options include filtering techniques (e.g., HP filter) or production function approaches.
  - If potential GDP is estimated using a filtering technique, "estimates will change whenever new data become available," so distinguish "intended and the resulting cyclical stance." The intended stance should use data available when decisions were taken; the resulting stance should use final/latest data.
  - Emerging markets can have "very volatile trend growth" (Aguiar and Gopinath (2007)), complicating trend/cycle separation.
  - The paper "generally use[s]" an HP filter but complements output-gap regressions with specifications replacing it by the real growth rate; "basic results will be calculated using the latest available data" and some regressions use real-time data.
- Cyclical stance definition:
  - The paper includes both "automatic stabilizers and discretionary policy, but not the automatic revenue gains that would be obtained under constant tax rates due to a growing economy."
  - Variables: R = revenue, G = primary government spending, B (=R-G) = primary fiscal balance, Y = GDP, Y* = potential output, Δ indicates difference between actual outcome and hypothetical outcome if output were at potential.
  - Acyclical revenue: average tax rate remains constant so revenues remain constant as a share of GDP (revenue rises at the rate of nominal growth).
  - Acyclical primary spending: primary spending constant as a share of potential GDP, so spending ratio falls in a boom.
  - Regression implication: for a simple regression relating the fiscal balance to the output gap, "the coefficient that would be obtained under a perfectly acyclical policy... is the expenditure ratio." Any estimated coefficient above the expenditure ratio = countercyclical; below it, even if above zero, = procyclical.
  - The paper suggests running a regression on an "adjusted fiscal balance, so that a coefficient of zero becomes the expected outcome under acyclical policy."
- Three definitions of countercyclical policy (labels A to C):
  - Definition A (strongest): "Discretionary fiscal policy must tighten." Implication: "The cyclically adjusted primary balance must improve." For spending: "Expenditure net of transfers... must be cut." For revenue: "Taxes or other levies must be increased."
  - Definition B (middle): "Tightening may be due to discretionary measures or automatic stabilizers." Implication: "The primary balance must improve by more than the temporary revenue gain." Spending: "Total (or primary) expenditures... must fall." Revenue: "Tax revenues must rise as a share of GDP."
  - Definition C (weakest): "Tightening may be due to discretionary policy, automatic stabilizers or additional revenue at stable average tax rates." Implication: "The primary balance must improve." Spending: "Not applicable." Revenue: "Revenues rise in real terms."
  - The paper's preferred definition is the middle ground (its adjusted primary balance), giving credit to automatic stabilizers but not to revenue gains at stable average tax rates.
- Comparability with literature:
  - Many past studies use different definitions (A, B, or C) or focus on expenditures or revenues only; this affects results and comparability.
  - Most papers find "signs of fiscal procyclicality, especially in developing and emerging economies, including in Latin America," while "among advanced economies, there are many which employ counter or acyclical policy."
  - Endogeneity concerns are noted; some studies addressing endogeneity (e.g., Ilzetzki and Vegh, 2008) still find similar results; one exception is Jaimovich and Panizza (2007).
- Commodity and asset price effects and composition:
  - "The composition of GDP can play a role" because "not all components are equally taxed."
  - Asset and commodity prices can boost revenues beyond real GDP growth (wealth/transaction taxes; commodity profits). Separating cyclical vs structural components of commodity revenues is difficult.
  - Filtering techniques (e.g., HP filter) may be used but have limitations for commodity prices.
  - Spending even permanent commodity revenue gains can be procyclical if "output is above potential."
  - For assessing cyclicality, the key is whether commodity-related revenue use "increases or reduces economic cycles," not whether revenue changes are permanent or temporary.
  - Table 3 logic on apparent cyclicality when positive commodity revenue occurs:
    - If additional commodity revenue is spent and output gap is positive: "Appears Acyclical" but "But is Pro-cyclical."
    - If additional commodity revenue is saved and output gap is positive: "Appears Counter-cyclical" but "But is Acyclical."
    - If additional commodity revenue is spent and output gap is negative: "Appears Acyclical" but "But is Counter-cyclical."
    - If additional commodity revenue is saved and output gap is negative: "Appears Pro-cyclical" but "But is Acyclical."
  - Approaches to handle commodity effects: use fiscal balance excluding commodity-related revenues (data sparse); analyze spending rather than balances; or "control for commodity price developments" by adding commodity prices as a control variable.
- Specification and estimation:
  - Common regression form (difference specification) is used; regressions run on both output gap and growth rate.
  - Primary balance used is the "primary balance adjusted as discussed above (equation (6))." Commodity-price controls are often included.
  - Alternative specification: based on the primary spending ratio.
  - Endogeneity handling: instrumental-variable regressions (common instruments: export-weighted growth rate of trading partners and the US real interest rate).
  - Panel estimators: system-GMM (Blundell and Bond (1998)) to address Nickell (1981) bias on lagged dependent variable and endogeneity.
  - Country-specific regressions: instrumental-variable regressions using either lagged output gap or export-weighted growth rate of trading partners and the US real interest rate as instruments.

### Data and measurement specifics
- Main data sources:
  - World Economic Outlook database (Fall 2013).
  - US treasury bill rates from the Federal Reserve Board.
  - Trade and commodity price data from the UN Comtrade database.
  - Resource-related fiscal revenues in Chile and Mexico from the IDB Fiscal Resources data set.
- Sample coverage: "Our data set covers the years 1980 to 2012, although actual data availability varies by country and variable."
- Real-time regressions: "for the regressions based on real time data we use the WEO vintages from 1990 through 2012."
- Primary balance construction: "we use our own calculations adding back interest expenditure to overall government net lending." The WEO primary balance deducts interest receipts but reduces sample size; interest receipts are typically small, so the paper uses the approximation for larger sample size. Exception: Brazil uses net interest expenditure.
- Commodity price index:
  - "A newly-calculated price index, based on time-varying weights, lagged by d periods" is used, constructed from log prices in US dollars and export values to ensure index reflects price changes rather than endogenous volume responses.
  - Two alternatives considered: exports of each commodity divided by total commodity exports; commodity terms of trade. "Results were very similar."

### Findings preview (structure of paper)
- The rest of the paper is structured as:
  - Section II: methodological issues, including the definition of cyclical fiscal policy and treatment of commodity and asset price shocks and compositional changes; brief review of existing studies and categorization by definition of procyclical fiscal policy.
  - Section III: data description, including the new commodity price index.
  - Section IV: empirical results, including panel data regressions for advanced and emerging economies and detailed Latin America country-by-country regressions.
  - Section V: brief conclusions.

*Source: _wp1459 - conclusions.*

### appendix Tables A1 to A3). The first regressions were estimated with a simple within-group

### _wp1459 - appendix Tables A1 to A3). The first regressions were estimated with a simple within-group

### Econometric approach
- Initial estimation: within-group (fixed effect) model allowing different intercepts for each country but imposing the same slope within a region; separate regressions run by region to avoid imposing the same slope across regions.
- Three fiscal-balance definitions used in the first set of regressions (equation (7)):
  - Primary balance (traditional)
  - CAPB (cyclically-adjusted primary balance)
  - Adjusted primary balance (as suggested in the text)
- Later specifications address endogeneity and commodity-price effects.
- System-GMM specifications treat the output gap/growth and the lagged adjusted primary balance (or lagged primary spending) as endogenous, using first and second lags as instruments and the collapse option of the xtabond2 command in Stata.
- System-GMM diagnostic expectations:
  - Arellano-Bond AR(1) test: expected to be rejected
  - AR(2) test: should not be rejected
  - Sargan/Hansen test of overidentifying restrictions: should not be rejected

### Panel estimates — key findings (Tables 4 and 5)
- General observation: results are not robust to choice of specification and estimation technique.
- Table 4 (within-group estimates): reported coefficients on the output gap for Δ Primary balance, Δ Adjusted primary balance, and Δ CAPB by region. Selected coefficients and statistics (as presented):
  - Advanced economies (columns 1–3):
    - Output gap: 0.434***, 0.132**, 0.037
    - (standard errors): (0.068), (0.061), (0.132)
    - Lagged fiscal balance: -0.296***, -0.234***, -0.318***
    - (standard errors): (0.028), (0.025), (0.061)
    - Observations: 791, 791, 576
    - R-squared: 0.207, 0.121, 0.180
    - Number of countries: 33, 33, 31
  - Emerging/developing economies (columns 4–6):
    - Output gap: 0.285, -0.102, -0.240**
    - (standard errors): (0.200), (0.116), (0.106)
    - Lagged fiscal balance: -0.587***, -0.572***, -0.217
    - (standard errors): (0.083), (0.107), (0.128)
    - Observations: 2,568, 2,568, 311
    - R-squared: 0.312, 0.302, 0.153
    - Number of countries: 146, 146, 29
  - Latin America (columns 7–9):
    - Output gap: -0.122, -0.335**, -0.346*
    - (standard errors): (0.139), (0.149), (0.183)
    - Lagged fiscal balance: -0.490***, -0.495***, -0.040
    - (standard errors): (0.093), (0.082), (0.111)
    - Observations: 355, 355, 103
    - R-squared: 0.248, 0.299, 0.111
    - Number of countries: 20, 20, 9
- Interpretation from Table 4:
  - Advanced countries: evidence of countercyclical fiscal policy (primary balance tightens in strong activity); CAPB result is insignificant, implying tightening is generally not discretionary.
  - Emerging/developing: mostly insignificant (acyclical cannot be rejected); CAPB shows a negative and significant coefficient on the output gap (procyclical) but on a much smaller sample.
  - Latin America: evidence of procyclical policy; coefficient significant also with CAPB (suggesting discretionary measures have been procyclical among the 9 countries with CAPB data).
  - Adjusted primary balance coefficients always lie between those on CAPB and primary balance.

- Table 5 (various estimations, including commodity-price control and system-GMM):
  - Estimations split by advanced, emerging/developing, and Latin America; top part uses output gap, lower part uses growth rate.
  - For each region three specifications: WG (within-group), WG + commodity-price control, and GMM.
  - Commodity price index: prices weighted by each commodity’s exports as a share of GDP (so countries exporting few commodities have a small weight).
  - System-GMM specification tests: generally passed except regression (3’) where Sargan/Hansen rejects overidentifying restrictions with p-value 4 percent; this rejection does not change overall interpretation for advanced economies.
  - Main patterns:
    - Advanced economies: evidence suggests presence of countercyclical policy (positive and significant coefficients), though results vary by specification; some GMM results indicate countercyclicality.
    - Emerging/developing economies: evidence unclear, mostly insignificant findings.
    - Latin America: results with the output gap, including system-GMM estimates, are significant and indicate procyclicality. Results using the growth rate are all insignificant.
  - Commodity price growth:
    - Across regions the commodity-price coefficient is often significant and always positive when significant.
    - Interpretation: adjusted primary fiscal balance strengthens when commodity price growth is strong unless countries spend more than additional resource revenues.
    - Inclusion of commodity prices does not change other coefficients much.

### Spending-based panel estimates (Table 6)
- Table 6 replaces adjusted primary fiscal balance with the adjusted spending ratio (spending as a share of potential GDP); interpretation reversed (positive coefficient indicates procyclicality).
- Rationale:
  - Avoids many complications of using fiscal balance, which reflects many factors other than fiscal policy.
  - Limitation: cannot fully assess cyclicality if revenue reforms occur; changes in spending accompanied by neutralizing revenue adjustments can mislead.
- Selected coefficients and statistics from Table 6 (output gap specifications):
  - Advanced economies (columns 1–3):
    - Output gap: 0.132**, 0.138**, 0.302***
    - (standard errors): (0.061), (0.057), (0.106)
    - Lagged adjusted prim. spending: -0.234***, -0.228***, 0.042
    - (standard errors): (0.025), (0.025), (0.103)
    - Observations: 791, 760, 760
    - R-squared: 0.121, 0.184, 0.302
    - Number of countries: 33, 32, 32
  - Emerging/developing economies (columns 4–6):
    - Output gap: -0.102, 0.010, 0.066
    - (standard errors): (0.116), (0.149), (0.131)
    - Lagged adjusted prim. spending: -0.572***, -0.585***, -0.741***
    - (standard errors): (0.107), (0.134), (0.145)
    - Observations: 2,568, 2,036, 2,036
    - R-squared: 0.398, 0.299, 0.343
    - Number of countries: 146, 134, 134
  - Latin America (columns 7–9):
    - Output gap: -0.335**, -0.343**, -0.337**
    - (standard errors): (0.149), (0.142), (0.150)
    - Lagged adjusted prim. spending: -0.495***, -0.441***, -0.479***
    - (standard errors): (0.082), (0.082), (0.125)
    - Observations: 355, 333, 333
    - R-squared: 0.248, 0.299, 0.343
    - Number of countries: 20, 19, 19
- Growth-rate specifications in Table 6 yield:
  - Advanced economies: some specifications show significance for growth (e.g., Growth 0.267***, 0.260***, 0.335*** in columns (1)'-(3)'); other entries reported as in table.
  - Emerging/developing and Latin America: mixed and mostly insignificant growth coefficients.
- Overall interpretation from spending-based estimates:
  - Advanced economies: results indicate countercyclical policy, significant mainly in specifications using growth.
  - Emerging/developing: mostly insignificant.
  - Latin America: continuing suggestion of procyclical fiscal policy.
- Robustness: further specifications using alternative commodity-price indicators (commodity prices weighted by total commodity exports; commodity prices applied to net rather than gross commodity exports) produced broadly similar results.

### Country-specific results (Table 7 and related discussion)
- Motivation: panel estimates force same cyclicality coefficient on all countries; country-by-country regressions allow heterogeneity.
- Regressions:
  - Sample: 1990–2012.
  - Include lagged adjusted fiscal balance and commodity price index.
  - Reported: coefficient on the output gap (for readability) with OLS and two instrumental-variable specifications.
  - IV specifications:
    - IV-1: lagged output gap as instrument.
    - IV-2: export-weighted GDP growth of trading partners and the US 1-year treasury rate as instruments (as in Appendix Tables A1–A3).
  - Time-varying regressions: coefficients for period before 2005 and change afterwards (output gap and commodity price index interacted with time indicator). Year 2005 chosen based on panel regressions (Appendix Table A4) showing most significant change in coefficient.
- Empirical patterns:
  - Very small number of statistically significant country-specific results.
  - Two countries identified as having procyclical fiscal policy based on results: Ecuador and Venezuela (but note OLS procyclicality in these two does not hold up in IV estimates for Ecuador and Venezuela, casting doubt).
  - Guyana: positive and significant coefficient in OLS but not supported by IV estimates.
  - No country-specific results provide a strong indication of countercyclical policy based on the presented estimates.
  - Instruments often weak: many first-stage regressions have extremely low R-squared; weak-instrument concerns lead to implausible coefficient levels in some IV regressions (example: IV-1 in Brazil).
- Time-varying findings:
  - Fiscal policy appears to have become more countercyclical since 2005 in Brazil, Chile, Colombia, El Salvador, and Mexico (as noted in the text).
- Additional note from appendices:
  - Country-by-country analysis of adjusted spending ratio (Appendix Table A5) shows even fewer statistically significant results.
  - Analysis of adjusted fiscal balance net of resource revenues (Appendix Table A6) shows all countries employing procyclical fiscal policy based on that measure; interpretation caveat: benchmark is very tight for countercyclical policy because it would require saving the entire increase in commodity revenues and more — politically difficult, and countries appear to let booming commodity revenues replace other revenues.

### Overall conclusions drawn in the text
- Panel-data evidence broadly confirms:
  - Countercyclical fiscal policy in advanced economies (though sensitivity to specification and estimation method exists).
  - Procyclical fiscal policy in Latin America.
  - Inconclusive evidence for the broader sample of emerging and developing countries.
- Commodity-price effects are relevant: commodity price growth coefficients often significant and positive; including them does not substantially change other coefficients.

*Italic: Source: _wp1459 - appendix Tables A1 to A3). The first regressions were estimated with a simple within-group (fixed effect) model… (extracted content as provided).*

### 2005. Only Honduras appears to have become more procyclical.

### _wp1459 - 2005. Only Honduras appears to have become more procyclical.

### Methodology: ex-post vs real-time (historical) output gap
- The results reported use the latest estimates of the output gap, which provide an ex-post assessment of the impact of fiscal policy.
- To assess policymakers’ intentions based on information available at the time, regressions were rerun using a historical output gap series: the estimated output gap for a given year based on data available at the Fall WEO of the previous year (Table 8).
- Instrumental variable approaches considered:
  - IV-1 uses the lagged output gap as an instrument.
  - IV-2 uses the US 1-year treasury bill and the export-weighted growth rate of trading partners as instruments.
- All regressions include a constant, a lagged dependent variable, and the commodity price index; robust standard errors are reported.

### Country-specific findings (high-level)
- On average, fiscal policy in Latin America is confirmed to have been procyclical, compared to counter- or a-cyclical policy in advanced economies.
- Country-specific estimations are mostly insignificant, implying that acyclical fiscal policy cannot be ruled out for most countries.
- For two countries, evidence for procyclicality is significant even at the country level.
- Evidence from real-time (historical) data:
  - Costa Rica: budgets appear to have been aiming to be countercyclical based on data available at the time.
  - Two countries exhibited procyclical fiscal policy even using historical data, indicating procyclicality could have been noticed at the time of implementation.
- Trends in more recent years:
  - Brazil, Chile, Colombia, El Salvador, and Mexico have moved toward more countercyclical policy.
- Summary statement: “Only Honduras appears to have become more procyclical.”

### Key country coefficient estimates (selected, coefficients on the output gap)
- From Table 7 (country-by-country regressions; columns: OLS, IV-1, IV-2, pre-2005 Δ, since 2005):
  - Honduras:
    - OLS: 0.209 (0.142)
    - IV-1: 0.183 (0.371)
    - IV-2: 0.541 (0.574)
    - pre-2005 Δ: 1.510* (0.654)
    - since 2005: -1.422* (0.685)
- From Table 8 (ex ante cyclicality: coefficients on the historical output gap; columns: OLS, IV-1, IV-2):
  - Costa Rica:
    - OLS: 1.512*** (0.329)
    - IV-1: 1.519 (1.962)
    - IV-2: 1.992*** (0.413)
  - Honduras:
    - OLS: 0.900 (0.567)
    - IV-1: 1.498 (0.847)
    - IV-2: 1.115 (0.962)
- Notes on significance notation and standard errors:
  - Robust standard errors are shown in parentheses.
  - Asterisks denote conventional significance levels as used in the tables.

### Robustness and caveats discussed
- Output gap estimates change with updates, so ex-post measures may differ from what policymakers were targeting based on real-time data.
- Whether an IV approach is needed is debatable for historical output gap regressions because the output gap used is a forecast and may only be endogenous to fiscal policy to the extent that fiscal impacts are modeled and actual fiscal policy deviates from projections.
- In Brazil, policy lending (not part of the fiscal balance) has grown in importance. Adding net policy lending to the adjusted fiscal balance increases the coefficient on the output gap interacted with the post-2005 dummy, but the standard error rises even more and the result turns insignificant.
- It remains possible that budget intentions were countercyclical using national definitions of the output gap, or that budget amendments/overruns ultimately produced procyclical outcomes.

### Conclusion (paper’s summary)
- Using a methodology that gives full credit to automatic stabilizers, controls for commodity prices, and allows for endogeneity, the paper confirms that fiscal policy has been procyclical on average in Latin America.
- Country-level evidence is frequently insignificant, so acyclical fiscal policy cannot be ruled out for most countries.
- Two countries show significant evidence of procyclicality at the country level.
- Real-time data indicate Costa Rica intended countercyclical policy (despite insignificant final-data results).
- Recent movements toward more countercyclical policy observed in Brazil, Chile, Colombia, El Salvador, and Mexico.
- The notable outlier toward greater procyclicality is Honduras.

*Source: _wp1459 - 2005. Only Honduras appears to have become more procyclical.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1459.pdf_