## Authorities’ Fiscal Forecasts in Latin America: Are They Optimistic? — Sections 1–4

## Source details

**Canonical URL:** [Authorities’ Fiscal Forecasts in Latin America: Are They Optimistic? — Sections 1–4](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021154-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021154-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021154-print-pdf.pdf.json)

---

### Abstract — Key Findings
- Dataset: authorities’ fiscal forecasts for LA6 (Argentina, Brazil, Chile, Colombia, Mexico, Peru) from annual budget documents, 2000-2018; 117 budget documents reviewed.
- Main findings:
  - (i) for most countries, there is no general optimistic bias in the forecasts for the fiscal balance-to-GDP ratio (though there may be for the components);
  - (ii) fiscal forecasts have improved for some countries over time, albeit they have worsened for others;
  - (iii) forecast errors for the fiscal balance-to-GDP ratio are positively correlated with GDP growth and terms of trade changes and negatively with GDP deflator surprises;
  - (iv) forecast errors for public debt-to-GDP ratios are negatively associated with surprises to GDP growth;
  - (v) budget balance rules seem to help contain the size of the fiscal forecast errors.
- JEL Classification Numbers: E62, H50

### Data and Methodology
- Coverage and construction:
  - Sample period: 2000-2018; countries: Argentina, Brazil, Chile, Colombia, Mexico, Peru.
  - Forecasts for year t taken from documents published in year t-1; outturns for year t from documents published in year t+1.
  - Fiscal series collected (nominal and percent of GDP): overall fiscal balance; expenditure; revenues; public debt.
  - Government levels: general government (Peru), central government (Brazil, Chile, Colombia), public sector (Argentina, Mexico).
  - Brazil annual budget bill documents include only primary balance (used here). For Brazil, Colombia, Mexico total net debt used.
  - Macroeconomic variables collected: nominal GDP; real GDP growth; GDP deflator; CPI inflation; exchange rate.
  - Terms-of-trade (ToT) projections/outturns from corresponding WEO vintages.
- Forecast error definition and empirical strategy:
  - FE = authorities’ forecast minus the actual outturn.
  - Three-stage analysis: visual evolution of FEs; correlation analysis between fiscal FEs and macro variables; formal regression analysis.
- Regression specifications (Section 2):
  - Equation (1): FFE_Y_i,t = β0 + β1 FFE_Δ Real GDP Growth_i,t + β2 FFE_Δ GDP Deflator_i,t + ε_i,t
  - Equation (2): FFE_Y_i,t = β0 + β1 FFE_Δ Real GDP Growth_i,t + β2 FFE_Δ GDP Deflator_i,t + β3 FFE_Δ Terms of Trade_i,t + ε_i,t
  - Estimation: panel regressions with country-specific Fixed Effects (FE) and Random Effects (RE); Hausman tests used to choose FE vs RE.
  - Exchange rate FE tested in alternatives; coefficients were not significant for most fiscal variables.

### Visual and Regression Findings — Fiscal Variables
- Visual patterns (Figures 1–4, 2000–2018):
  - Fiscal balance-to-GDP:
    - Optimistic forecasts (positive FEs) for Argentina and Brazil, especially during 2012–2016.
    - Chile and Colombia: forecasts improved over time, especially past decade.
    - Mexico: forecasts quite accurate most years.
    - Peru: no consistent pattern.
  - Revenues-to-GDP:
    - Argentina: large optimistic revenue FEs in the past decade.
    - Chile and Colombia: overestimation declined over time.
    - Mexico: reversed into underestimation roughly after 2009.
    - Brazil: overestimation recently reversed.
    - Peru: underestimation turned into overestimation in recent years.
  - Expenditure-to-GDP:
    - Patterns similar to revenues; Argentina large overestimation roughly same period.
    - Chile and Colombia: overestimation declined.
    - Mexico: reversed into underestimation.
    - Brazil and Peru: forecasts appear backward-looking (linked to previous-year outturns).
  - Public debt-to-GDP:
    - Forecasts started pessimistic in all countries, turned optimistic in the 2010s, and are reversing in recent years.
    - No consistent official public debt forecasts in Argentina.
- Mincer-Zarnowitz test (Equation (3): Y = α + β Y^F + R):
  - Balance-to-GDP ratio:
    - Pooled: alpha = -0.1880. beta = 0.8750. R^2 = 0.4950. p-value = 0.410
    - Fixed Effects: alpha = -0.1870. beta = 0.8990. R^2 = 0.4060. p-value = 0.503
    - Random Effects: alpha = -0.1900. beta = 0.8820. R^2 = 0.4950. p-value = 0.517
  - Expenditure-to-GDP ratio:
    - Pooled: alpha = 4.5330. beta = 0.7700. R^2 = 0.6210. p-value = 0.000
    - Fixed Effects: alpha = 5.7220. beta = 0.7160. R^2 = 0.5370. p-value = 0.000
    - Random Effects: alpha = 5.4540. beta = 0.7280. R^2 = 0.6210. p-value = 0.001
  - Revenues-to-GDP ratio:
    - Pooled: alpha = 7.2990. beta = 0.6290. R^2 = 0.5640. p-value = 0.000
    - Fixed Effects: alpha = 12.4170. beta = 0.3880. R^2 = 0.2270. p-value = 0.000
    - Random Effects: alpha = 11.2990. beta = 0.4400. R^2 = 0.5640. p-value = 0.000
  - Debt-to-GDP ratio:
    - Pooled: alpha = 0.8470. beta = 0.9470. R^2 = 0.9080. p-value = 0.096
    - Fixed Effects: alpha = 3.6160. beta = 0.8610. R^2 = 0.7090. p-value = 0.000
    - Random Effects: alpha = 1.0590. beta = 0.9410. R^2 = 0.9080. p-value = 0.133
- Formal panel regressions (Tables 4–7) — key coefficients (FE results emphasized):
  - Fiscal balance-to-GDP (Table 4):
    - RGDP growth FE coefficients: 0.366, 0.352, 0.322, 0.306 (standard errors: (0.054), (0.053), (0.055), (0.052); p-values: [0.000]).
    - ToT coefficients (when included): 0.048 and 0.057 (standard errors: (0.019), (0.018); p-values: [0.015], [0.002]).
    - GDP deflator coefficients: -0.021, -0.054, -0.047, -0.079 (p-values vary: [0.470], [0.021], [0.113], [0.001]).
    - Sample/statistics: Observations 78; Countries 66; R-squared up to 0.491.
  - Expenditure-to-GDP (Table 5):
    - RGDP growth coefficients: -0.148, -0.15, -0.123, -0.126 (standard errors: ~0.049-0.052; p-values: [0.004] to [0.020]).
    - GDP deflator coefficients: -0.06, -0.069, -0.045, -0.053 (p-values: [0.024], [0.006], [0.109], [0.042]).
    - ToT coefficients not significant at 10 percent.
    - R-squared up to 0.248.
  - Revenue-to-GDP (Table 6):
    - RGDP growth coefficients: 0.217, 0.213, 0.198, 0.194 (standard errors: ~0.053-0.056; p-values: [0.000] to [0.001]).
    - GDP deflator coefficients: -0.081, -0.095, -0.093, -0.103 (p-values: [0.005], [0.000], [0.003], [0.000]).
    - ToT coefficients not significant at 10 percent.
    - R-squared up to 0.383.
  - Public debt-to-GDP (Table 7):
    - RGDP growth coefficients: -0.813, -0.719, -0.727, -0.672 (standard errors: (0.187), (0.179), (0.201), (0.193); p-values: [0.000], [0.000], [0.001], [0.000]).
    - GDP deflator and ToT coefficients insignificant.
    - Sample/statistics: Observations 64; Countries 55; R-squared up to 0.278.

### Drivers of Forecast Errors — Macroeconomic Links and Exchange Rate
- Bivariate correlations (Figures 9–11) — full-sample relationships:
  - Positive RGDP forecast errors associated with:
    - Overestimated fiscal balance-to-GDP ratios (positive correlation).
    - Overestimated revenues-to-GDP ratios (positive correlation).
    - Underestimated expenditure-to-GDP ratios (negative correlation).
    - Underestimated public debt-to-GDP ratios (negative correlation).
  - GDP deflator forecast errors:
    - Positive deflator forecast errors associated with underestimated fiscal ratios (negative forecast errors) — denominator effect on ratios to GDP can dominate.
  - ToT changes:
    - Positive ToT surprises associated with positive surprises in fiscal balance-to-GDP ratios.
- Exchange rate forecast errors:
  - Insignificant for most fiscal variables.
  - Significant for public debt-to-GDP in some specifications: larger-than-projected depreciation associated with larger-than-projected increase in public debt, but effect not robust to inclusion of ToT (Annex Table A15 shows ExR Dev coefficient 0.114 (0.065) [0.085] in one specification; becomes insignificant in alternatives).

### Fiscal Rules and Forecast Errors
- Visual markers: vertical lines in Figures 1–4 indicate fiscal rule introductions (data from Lledo et al (2017)).
- Regression evidence on fiscal rules (Table A14, FE regressions on absolute FE):
  - D Bal coefficients (impact of budget balance rule on absolute FE):
    - Abs FE Bal/GDP: -0.48 (0.591) [0.420]
    - Abs FE Exp/GDP: -0.78 (0.457) [0.092]
    - Abs FE Rev/GDP: -1.439 (0.547) [0.010]
    - Abs FE PDebt/GDP: -3.144 (1.077) [0.005]
  - D Exp coefficients:
    - Abs FE Bal/GDP: -0.396 (0.638) [0.537]
    - Abs FE Exp/GDP: 0.017 (0.493) [0.972]
    - Abs FE Rev/GDP: -0.334 (0.590) [0.573]
    - Abs FE PDebt/GDP: -2.57 (1.240) [0.043]
  - D Debt coefficients:
    - All coefficients statistically insignificant at 10 percent in the table excerpt.
  - Observations: 78 (Bal/Exp/Rev) and 64 (PDebt); Countries: 66/65.
  - R-squared: up to 0.222.
- Interpretation: budget balance rules appear effective at containing size of fiscal forecast errors for some variables (significant negative coefficients for revenues and public debt), but significance varies across rules and variables; degrees of implementation differ across countries.

### Descriptive and Policy-Relevant Conclusions
- Cross-country descriptive patterns:
  - Optimistic fiscal balance forecasts mainly in Argentina and Brazil (2010s); Argentina’s revenue overestimation larger than expenditure overestimation.
  - Mexico: pre-GFC overestimation of expenditure and revenues, post-GFC underestimation of both by similar magnitudes, yielding accurate overall balance forecasts — interpreted as expenditure adjustments anchoring the balance to revenue outcomes.
  - Chile and Colombia: forecasts improved significantly post-GFC.
  - Peru (and to a lesser extent Brazil): forecasts often backward-looking.
  - Public debt forecasts: initially pessimistic, turned optimistic in 2010s, reversing recently.
- Policy implications:
  - Optimistic forecasts for growth or ToT changes drive optimistic fiscal balance-to-GDP forecasts via overestimated revenue ratios and underestimated expenditure ratios.
  - Persistent fiscal forecast errors and strong association with macro forecasts indicate scope for improved forecasting practices and attention to macro-fiscal linkages.
  - Budget balance rules show preliminary evidence of reducing forecast errors; larger sample required for robust assessment.
  - Suggested operational responses: improve macro-fiscal linkages in forecast models; enhance data quality and forecasting techniques; consider comparing authorities’ forecasts with Consensus Forecasts and other forecasters; analyze nominal-term forecast errors to separate numerator/denominator effects.
- Suggested extensions for future research:
  - Compare authorities’ fiscal forecasts against Consensus Forecasts and other forecasters.
  - Expand analysis using forecast errors in nominal terms rather than ratios to disentangle denominator effects.
  - Explore biased expenditure forecasts around election cycles or asymmetric loss functions by fiscal authorities.

### Dataset Examples and Appendix Highlights
- Chile and Colombia dataset examples described (dataset entries coded by source type: retrieved, calculated, special definitions, visual, or other official source).
- Colombia dataset raw yearly entries (selected values preserved as in source), e.g.:
  - 2000: 8.75; 2,104.72
  - 2001: 7.65; 2,301.74
  - 2002: 6.98; 2,534.22
  - 2003: -12,151.00 -5.40 46,597.00 20.90 34,445.00 15.40 223,668.83 6.50 2,875.05
  - 2004: -13,983.00 -5.50 53,934.00 21.10 39,952.00 15.60 105,514.26 41.20 256,102.56 5.33 8.70 5.50 2,614.80
  - 2005: -13,730.00 -4.80 59,606.00 21.00 45,876.00 16.20 136,197.07 40.30 337,958.00 4.71 26.03 4.84 2,319.26
  - 2006: -13,094.00 -4.10 69,441.00 21.70 56,347.00 17.60 146,154.33 38.30 381,604.00 6.72 5.81 4.48 2,363.75
  - 2007: -11,613.00 -3.20 76,318.00 21.40 64,705.00 18.10 153,405.15 35.80 428,506.00 6.74 5.20 5.69 2,076.57
  - 2008: -11,067.00 -2.30 86,131.00 18.10 75,064.00 15.70 169,176.67 35.50 476,554.00 3.28 7.68 7.68 1,989.35
  - 2009: -20,715.00 -4.20 97,871.00 19.70 77,156.00 15.50 183,576.08 36.60 501,574.00 1.14 4.06 2.01 2,180.19
  - 2010: -20,171.00 -3.70 95,128.00 17.40 74,957.00 13.70 202,390.32 37.20 544,060.00 4.49 3.80 3.16 1,902.50
  - 2011: -17,507.00 -2.80 111,754.00 18.10 94,247.00 15.30 215,420.00 34.80 619,023.00 6.95 6.39 3.73 1,854.02
  - 2012: -15,440.00 -2.30 122,507.00 18.40 107,067.00 16.10 223,946.35 33.60 666,507.00 3.91 3.62 2.44 1,798.73
  - 2013: -16,645.00 -2.36 136,390.00 19.30 119,744.00 16.94 247,076.18 34.60 714,093.00 5.13 1.91 1.93 1,879.53
  - 2014: -18,356.00 -2.40 144,260.00 19.10 125,904.00 16.70 283,037.01 37.10 762,903.00 4.50 2.24 3.66 2,017.85
  - 2015: -24,269.00 -3.00 153,590.00 19.20 129,321.00 16.10 338,775.33 42.10 804,692.00 2.96 2.45 6.77 2,771.55
  - 2016: -34,925.00 -4.00 163,293.00 18.90 128,367.00 14.90 375,745.17 43.50 863,782.00 2.09 5.15 5.75 3,040.09
  - 2017: -33,636.00 -3.60 174,519.00 18.90 144,062.00 15.60 409,609.60 44.50 920,471.00 1.36 5.13 4.09 2,957.52
  - 2018: -30,316.00 -3.10 179,608.00 18.40 149,292.00 15.30 473,246.88 48.00 985,931.00 2.52 4.48 3.18 2,972.04
  - 2019: nana 1,062,342.50 3.32 4.29 3.80 3,299.77
- Appendix descriptive panels present year-by-year Projection - Actual samples and summary statistics for Expenditure/GDP, Fiscal Balance/GDP, Revenue/GDP across country codes ARGBRACHLCOLMEXPER, with averages and extremes preserved in source (e.g., Expenditure/GDP average 0.89 0.65 -0.51 -0.44 0.05 -0.39 0.04; min -1.62 -1.17 -5.01 -2.40 -1.30 -3.70 -5.01; max 4.31 3.04 8.04 0.50 2.20 4.60 8.04).
- Fixed effects panels (Figure A2) and further annex tables (A14, A15) provide robustness checks and sensitivity to inclusion of ToT and exchange rate deviations.

*Source: IMF Working Paper WP/21/154 — Sections 1–4*

### Section 1

### Authorities’ Fiscal Forecasts in Latin America: Are They Optimistic? — Section 1

### Abstract — Key Findings
- Do governments in Latin America tend to be optimistic when preparing budgetary projections? The paper constructs a novel dataset of the authorities’ fiscal forecasts in six Latin American economies using data from annual budget documents over the period 2000-2018.
- Main findings:
  - (i) for most countries, there is no general optimistic bias in the forecasts for the fiscal balance-to-GDP ratio (though there may be for the components);
  - (ii) fiscal forecasts have improved for some countries over time, albeit they have worsened for others;
  - (iii) in terms of drivers, forecast errors for the fiscal balance-to-GDP ratio are positively correlated with GDP growth and terms of trade changes and negatively with GDP deflator surprises;
  - (iv) forecast errors for public debt-to-GDP ratios are negatively associated with surprises to GDP growth;
  - (v) budget balance rules seem to help contain the size of the fiscal forecast errors.
- JEL Classification Numbers: E62, H50

### Introduction — Research Questions and Motivation
- Central issues:
  - The accuracy of official fiscal forecasts has implications for macroeconomic policies, fiscal performance and sustainability, and policy advice.
  - A systematic bias in authorities' fiscal forecasts weakens policy credibility, complicates policymaking, and may persistently worsen public finances.
  - In the context of COVID-19, understanding official fiscal forecast errors informs IMF advice given the central role of fiscal policy in the pandemic response.
- Research questions for the six largest economies in Latin America (LA6: Argentina, Brazil, Chile, Colombia, Mexico, and Peru):
  - Have the authorities’ fiscal forecasts been over-optimistic?
  - What fundamental factors explain official fiscal forecast errors (is there space to improve efficiency)?
  - Do fiscal rules help reduce the size of forecast errors?
- Empirical strategy summary:
  - Construct a dataset of authorities’ fiscal forecasts for LA6 using annual budget documents.
  - Compare forecasts from documents published in year t-1 for budget year t with outturns reported in documents published in t+1 that contain outturns for year t.

### Literature — Context and Prior Findings
- Empirical literature on fiscal forecast errors shows preliminary official data releases tend to be optimistic for fiscal and macroeconomic variables; most studies focus on advanced economies, especially European countries.
- Representative findings from prior work:
  - Bruck and Stephan (2006): evidence of political influence on budget forecasts.
  - Larch and Jonung (2006): systematic optimism in government growth predictions in budget planning.
  - Beetsma et al. (2009, 2011): planned budget balances differ in an over-optimistic manner from ex-post estimates; revision errors mainly caused by over-optimism on revenues at first-release stage.
  - Croushore and Van Norden (2018): Greenbook fiscal forecasts improvement correlated with better forecasting of macroeconomic variables.
  - Kliesen and Thornton (2012) and Auerbach (1994): CBO and OMB forecasts perform worse in recessions and show overoptimism.
  - Merola and Perez (2013): international agencies’ forecasts present optimistic bias the year prior to elections.
  - Frankel (2011, 2013) and Frankel and Schreger (2013): broader sample including emerging markets finds official budget forecasts of advanced economies have an upward bias; significant association between forecast errors of GDP growth and inflation and those of the budget balance.
- This study extends the literature to emerging markets by assembling a new dataset for Latin American countries.

### Data — Dataset Construction and Coverage
- Sample and documents:
  - Sample period: 2000-2018.
  - Countries: LA6 — Argentina, Brazil, Chile, Colombia, Mexico, Peru.
  - Number of annual budget documents reviewed: 117.
  - Forecasts for year t are taken from documents published in year t-1; outturns for year t are taken from documents published in year t+1.
- Government levels and fiscal series:
  - Fiscal variables correspond to:
    - general government (GG) for Peru,
    - central government (CG) for Brazil, Chile, and Colombia,
    - public sector (PS) for Argentina and Mexico.
  - Collected fiscal series (in nominal terms and as percent of GDP): the overall fiscal balance; expenditure; revenues; public debt.
  - For Brazil, the annual budget bill documents only include figures for the primary balance (used in this paper).
  - For Brazil, Colombia, and Mexico, total net debt was used.
- Macroeconomic variables collected:
  - nominal GDP;
  - real GDP growth;
  - GDP deflator;
  - CPI inflation;
  - exchange rate.
- Terms-of-trade (ToT) projections and outturns:
  - Terms-of-trade projections and outturns come from corresponding WEO vintages.
  - Projections and outturns for terms of trade were not consistently reported in the official documents.
- Data collection notes:
  - The 117 budget documents are those sent to Congress for approval each year for each country and are the most updated comprehensive macroeconomic forecasts provided by the Government/Ministry of Finance to legislative bodies.
  - In some cases the approved budget may differ from these forecasts; assembling the finalized legal documents set was not undertaken.
  - Although fiscal forecasts are prepared by fiscal authorities, the underlying macroeconomic assumption sources are not always explicitly mentioned.

### Data Sources — Country Document Types (overview)
- Argentina: Presupuestos de la Administración Pública Nacional (Budget of the National Public Administration)
- Brazil: Projeto de Lei Orçamentária Anual (Annual Budget Bill)
- Chile: Informe de Finanzas Públicas del Proyecto de Ley de Presupuestos del Sector Público (Public Finance Report on the Public Sector Budget Bill)
- Colombia: Marco Fiscal de Mediano Plazo (Medium Term Fiscal Framework)
- Mexico: Criterios Generales de Política Económica (General Criteria of Economic Policy)
- Peru: Marco Macroeconómico Multianual (Multiannual Macroeconomic Framework)

### Dataset Example — Chile (illustrative)
- The paper includes a detailed dataset example for Chile showing authorities’ forecasts and outturns for fiscal variables and macroeconomic indicators (units and values in original budget documents). The dataset entries were coded by source type:
  - retrieved directly from official budget documents,
  - calculated based on data in documents,
  - calculated using special definitions,
  - retrieved from charts/graphs (visual),
  - or collected from another official national source.
- When fiscal balance was not reported explicitly it was calculated from total expenditures and total revenues; alternative definitions (e.g., average vs end-of-period inflation) were used when necessary.

### Methodology — Definition and Empirical Strategy
- Forecast errors (FE) definition:
  - FE = authorities’ forecast minus the actual outturn of the corresponding variable.
- Three-stage analysis:
  - (i) visual analysis of the evolution of fiscal forecast errors across countries and over time;
  - (ii) correlation analysis between fiscal forecast errors and a set of macroeconomic variables;
  - (iii) formal regression analysis to understand drivers of forecast errors.

*Source: IMF Working Paper WP/21/154 — Section 1*

### Section 2

### Section 2

### Regression specifications
- Two panel regression specifications are estimated:
  - Equation (1): FFE_Y_i,t = β0 + β1 FFE_Δ Real GDP Growth_i,t + β2 FFE_Δ GDP Deflator_i,t + ε_i,t
  - Equation (2): FFE_Y_i,t = β0 + β1 FFE_Δ Real GDP Growth_i,t + β2 FFE_Δ GDP Deflator_i,t + β3 FFE_Δ Terms of Trade_i,t + ε_i,t
- Definitions:
  - Y_i,t: a fiscal variable for country i at time t (fiscal balance, total expenditure, total revenue, or public debt), expressed as a share of GDP.
  - FFE_Y_i,t: forecast error for the corresponding fiscal variable.
  - FFE_Δ Real GDP Growth_i,t: forecast error of the growth rate for real GDP.
  - FFE_Δ GDP Deflator_i,t: forecast error of the change in the GDP deflator.
  - FFE_Δ Terms of Trade_i,t: forecast error in the change of the WEO Terms of Trade.
- Estimation approach:
  - Panel data regressions with country-specific Fixed Effects (FE) and Random Effects (RE).
  - Hausman specification tests run to choose FE vs RE in each specification.
  - Alternative specifications adding forecast error for the exchange rate were tested; coefficients were not significant.

### Visualizing the forecast bias (Figures 1–4; period 2000–2018)
- Figures present authorities’ forecasts, actual outturns, and forecast errors (defined as forecasts minus outturns).
- Vertical lines indicate introduction of fiscal rules; fiscal rule data is from Lledo et al (2017).
- Key qualitative findings from Figures 1–4:
  - Fiscal balance-to-GDP ratio (Figure 1):
    - Authorities’ forecasts have been optimistic (positive forecast errors) for Argentina and Brazil, especially during 2012–2016.
    - Chile and Colombia: forecasts for fiscal balance have improved over time, especially in the past decade.
    - Mexico: forecasts have been quite accurate for most years.
    - Peru: no consistent pattern in forecast errors for the fiscal balance.
  - Revenues-to-GDP ratio (Figure 2):
    - Argentina: large optimistic revenue forecast errors in the past decade.
    - Chile and Colombia: overestimation declined over time (in absolute value).
    - Mexico: reversed into an underestimation roughly after 2009.
    - Brazil: overestimation has recently reversed.
    - Peru: underestimation turned into overestimation in recent years.
  - Expenditure-to-GDP ratio (Figure 3):
    - Patterns similar to revenues.
    - Argentina: large overestimation of expenditure roughly over same period as revenue overestimation.
    - Chile and Colombia: overestimation declined.
    - Mexico: reversed into underestimation.
    - Brazil and Peru: no clear signs of bias; forecasts appear linked to previous-year outturns.
  - Public debt-to-GDP ratio (Figure 4):
    - Forecasts started pessimistic (positive forecast errors) in all countries, turned optimistic in the 2010s, and are reversing in recent years.
    - Note: No official forecasts for public debt are provided consistently in Argentina.
- Overall synthesis:
  - No general optimistic bias in authorities’ fiscal forecasts for most countries.
  - Optimistic fiscal balance forecasts mainly in Argentina and Brazil in the 2010s.
  - Argentina: overestimation of revenues larger than overestimation of expenditure, leading to optimistic fiscal balance forecasts.
  - Mexico: expenditure and revenues overestimated pre-GFC and underestimated post-GFC by similar magnitudes, resulting in accurate overall balance forecasts — interpreted as expenditure control anchored to revenue outcomes.
  - Chile and Colombia: forecasts improved significantly in the post-GFC period.
  - Peru (and to lesser extent Brazil): revenue and expenditure forecasts seem backward-looking.
  - Public debt forecasts: initially pessimistic (anticipating higher debt), then more optimistic over time, with recent reversals.

### Fiscal rules and forecast errors
- Fiscal rule introductions are graphed in Figures 1–4 via vertical lines.
- Regression evidence (Annex Table 14 referenced) using fiscal rule data from IMF dataset prepared by Lledo et al (2017) suggests:
  - Budget balance rules seem effective at containing the size of fiscal forecast errors.
  - Potential mechanisms: better information sets and forecasting techniques, improved data quality, and reduced uncertainty.
- Caveat: different degrees of implementation of fiscal rules across countries may play a role.

### Testing for unbiasedness (Mincer-Zarnowitz test)
- Method: regress actual data (Y) on authorities’ forecasts (Y^F) and test joint null α=0 and β=1:
  - Equation (3): Y_i,t = α + β Y^F_i,t + R_i,t
- Summary of results (Table 3):
  - Joint p-values at 10 percent significance level indicate rejection of the null for GDP ratios of expenditure, revenues, and debt in some specifications, suggesting those forecasts are biased.
  - Forecasts for the fiscal balance appear unbiased.
- Extracted coefficient estimates and statistics from Table 3 (by fiscal variable and estimator):
  - Balance-to-GDP ratio
    - Pooled Estimation: alpha = -0.1880. beta = 0.8750. R^2 = 0.4950. p-value = 0.410
      - (standard errors shown in table: (0.222)(0.101)(0.087)(0.129))
    - Fixed Effects: alpha = -0.1870. beta = 0.8990. R^2 = 0.4060. p-value = 0.503
    - Random Effects: alpha = -0.1900. beta = 0.8820. R^2 = 0.4950. p-value = 0.517
  - Expenditure-to-GDP ratio
    - Pooled Estimation: alpha = 4.5330. beta = 0.7700. R^2 = 0.6210. p-value = 0.000
      - (standard errors: (1.531)(0.069)(1.738)(0.079)(1.704)(0.075))
    - Fixed Effects: alpha = 5.7220. beta = 0.7160. R^2 = 0.5370. p-value = 0.000
    - Random Effects: alpha = 5.4540. beta = 0.7280. R^2 = 0.6210. p-value = 0.001
  - Revenues-to-GDP ratio
    - Pooled Estimation: alpha = 7.2990. beta = 0.6290. R^2 = 0.5640. p-value = 0.000
      - (standard errors: (1.372)(0.063)(1.812)(0.085)(1.795)(0.079))
    - Fixed Effects: alpha = 12.4170. beta = 0.3880. R^2 = 0.2270. p-value = 0.000
    - Random Effects: alpha = 11.2990. beta = 0.4400. R^2 = 0.5640. p-value = 0.000
  - Debt-to-GDP ratio
    - Pooled Estimation: alpha = 0.8470. beta = 0.9470. R^2 = 0.9080. p-value = 0.096
      - (standard errors: (1.276)(0.038)(2.270)(0.072)(1.413)(0.042))
    - Fixed Effects: alpha = 3.6160. beta = 0.8610. R^2 = 0.7090. p-value = 0.000
    - Random Effects: alpha = 1.0590. beta = 0.9410. R^2 = 0.9080. p-value = 0.133

### Forecast errors of key macroeconomic variables (Figures 5–8)
- Real GDP growth (Figure 5):
  - Argentina: actual growth much more volatile than forecasts; forecasts generally optimistic over the last decade; growth forecast errors larger than in other LA6 economies.
  - Brazil: tendency to overestimate growth post-GFC.
  - Chile, Colombia, Mexico, Peru: somewhat smaller or less persistent forecast errors.
- Inflation (CPI) (Figure 6):
  - Argentina: actual CPI inflation has consistently exceeded authorities’ forecasts since 2014.
  - Other LA6 economies: inflation forecast errors do not show similar bias.
- GDP deflator (Figure 7):
  - Argentina: GDP deflator consistently underestimated over the past decade and a half; forecast errors have widened over time.
  - Brazil and Mexico: forecast errors very small.
  - Chile, Colombia, Peru: no persistent pattern.
- Exchange rate (Figure 8):
  - Argentina and Brazil: rate of depreciation tended to be underestimated post-GFC; magnitude of forecast errors increased in some recent years in Argentina.
  - Mexico: depreciation rate tended to be underestimated in most recent years to a lesser extent.
  - Chile, Colombia, Peru: substantially smaller forecast errors without visible pattern.

### Understanding the drivers of forecast errors (Section V)
- Two-step approach:
  - Descriptive correlations between forecast errors of fiscal and macroeconomic variables.
  - Panel data regressions to formally assess relationships.
- Bivariate relationships (Figures 9–11):
  - Forecast errors for real GDP growth (Figure 9):
    - Full-sample correlations:
      - Optimistic GDP growth forecasts (positive growth forecast errors) are associated with:
        - Overestimated fiscal balance-to-GDP ratios (positive correlation).
        - Overestimated revenues-to-GDP ratios (positive correlation).
        - Underestimated expenditure-to-GDP ratios (negative correlation).
        - Underestimated public debt-to-GDP ratios (negative correlation).
    - Country-specific relationships mostly align with full-sample correlations, with limited exceptions.
  - Forecast errors for GDP deflator (Figure 10):
    - Full-sample correlations:
      - Optimistic GDP deflator forecasts (positive forecast errors) are associated with underestimated fiscal ratios (negative forecast errors).
    - Explanation:
      - Fiscal variables are expressed as ratios to GDP; a positive surprise in the GDP deflator raises nominal fiscal values (numerator) but raises GDP (denominator) directly, and the denominator effect can dominate, producing the observed negative relationship.
    - Note: positive association between forecast errors for nominal fiscal values and GDP deflator forecast errors is shown in Annex Section C.
  - Forecast errors for terms of trade percentage changes (Figure 11):
    - Full-sample top panel: positive surprises to ToT percentage changes are associated with positive surprises in fiscal balance-to-GDP ratios.

*Source: wpiea2021154-print-pdf - Section 2*

### Section 3

### Section 3

### Regression results: overview
- Formal panel regression analysis presented in Tables 4-7 aims to explain forecast errors of different fiscal variables using FE and RE estimations and Hausman tests.
- Sample and estimation details reported in tables (Observations, Countries, R-squared, Adjusted R-squared, Hausman H-stat and p-values are provided for each specification).

### Fiscal balance-to-GDP ratio (Table 4): key findings
- Forecast errors for the fiscal balance-to-GDP ratio are:
  - Positively associated with positive forecast errors for RGDP growth.
    - FE coefficient(s): 0.366, 0.352, 0.322, 0.306 (standard errors: (0.054), (0.053), (0.055), (0.052); p-values: [0.000] across specifications).
  - Positively associated with positive forecast errors for ToT (change) when included.
    - ToT coefficients: 0.048 and 0.057 (standard errors: (0.019), (0.018); p-values: [0.015], [0.002]).
  - Negatively associated with positive forecast errors for the GDP deflator growth in some specifications.
    - GDP deflator coefficients: -0.021, -0.054, -0.047, -0.079 (standard errors: (0.028), (0.023), (0.029), (0.023); p-values: [0.470], [0.021], [0.113], [0.001]).
- Constants reported: -0.168, -0.224, -0.001, -0.021 with standard errors (0.058), (0.180), (0.088), (0.176) and p-values [0.005], [0.214], [0.994], [0.904].
- Sample/statistics: Observations 78 78 78 78; Countries 66 66 66 66; R-squared 0.414, 0.424, 0.462, 0.491; Adjusted R-squared 0.355, 0.409, 0.40, 0.47.
- Hausman's test H-stat: 4.358 and 3.08 with p-values [0.113] and [0.379].

### Expenditure-to-GDP ratio (Table 5): key findings
- Forecast errors for the expenditure-to-GDP ratio are negatively related to:
  - RGDP growth forecast errors.
    - Coefficients: -0.148, -0.15, -0.123, -0.126 (standard errors: (0.049), (0.049), (0.052), (0.051); p-values: [0.004], [0.002], [0.020], [0.013]).
  - GDP deflator growth forecast errors (insignificant when ToT included in the specification in some cases).
    - GDP deflator coefficients: -0.06, -0.069, -0.045, -0.053 (standard errors: (0.026), (0.025), (0.028), (0.026); p-values: [0.024], [0.006], [0.109], [0.042]).
  - ToT (change) coefficients when included: -0.028 and -0.027 (standard errors: (0.018), (0.018); p-values: [0.131], [0.126]) — not significant at 10 percent.
- Constants: 0.539, 0.520, 0.442, 0.427 (standard errors: (0.054), (0.360), (0.083), (0.385); p-values: [0.000], [0.149], [0.000], [0.268]).
- Sample/statistics: Observations 78 78 78 78; Countries 66 66 66 66; R-squared 0.151, 0.230, 0.179, 0.248; Adjusted R-squared 0.066, 0.210, 0.084, 0.217.
- Hausman's test H-stat: 1.026 and 1.169 with p-values [0.599] and [0.761].

### Revenue-to-GDP ratio (Table 6): key findings
- Forecast errors for the revenue-to-GDP ratio are:
  - Positively related to RGDP growth forecast errors.
    - Coefficients: 0.217, 0.213, 0.198, 0.194 (standard errors: (0.053), (0.053), (0.056), (0.055); p-values: [0.000], [0.000], [0.001], [0.000]).
  - Negatively related to GDP deflator growth forecast errors.
    - GDP deflator coefficients: -0.081, -0.095, -0.093, -0.103 (standard errors: (0.028), (0.027), (0.030), (0.029); p-values: [0.005], [0.000], [0.003], [0.000]).
  - ToT (change) coefficients when included: 0.021 and 0.023 (standard errors: (0.020), (0.019); p-values: [0.278], [0.237]) — not significant at 10 percent.
- Constants: 0.364, 0.338, 0.438, 0.426 (standard errors: (0.057), (0.428), (0.089), (0.516); p-values: [0.000], [0.430], [0.000], [0.409]).
- Sample/statistics: Observations 78 78 78 78; Countries 66 66 66 66; R-squared 0.298, 0.369, 0.310, 0.383; Adjusted R-squared 0.228, 0.353, 0.230, 0.358.
- Hausman's test H-stat: 4.291 and 1.706 with p-values [0.117] and [0.636].

### Public debt-to-GDP ratio (Table 7): key findings
- Negative surprises to GDP growth forecasts (positive RGDP forecast errors) are associated with negative forecast errors for public debt-to-GDP (i.e., overestimating GDP growth implies underestimating public debt-to-GDP).
  - RGDP coefficients: -0.813, -0.719, -0.727, -0.672 (standard errors: (0.187), (0.179), (0.201), (0.193); p-values: [0.000], [0.000], [0.001], [0.000]).
- GDP deflator and ToT forecast errors are insignificant in Table 7:
  - GDP deflator coefficients: -0.056, -0.092, -0.024, -0.052 (standard errors: (0.075), (0.074), (0.080), (0.079); p-values: [0.465], [0.216], [0.770], [0.509]).
  - ToT (change) coefficients: -0.057 and -0.051 (standard errors: (0.050), (0.049); p-values: [0.263], [0.296]).
- Constants: 1.349, 1.283, 1.078, 1.061 (standard errors: (0.141), (0.514), (0.278), (0.652); p-values: [0.000], [0.013], [0.000], [0.104]).
- Sample/statistics: Observations 64 64 64 64; Countries 55 55 55 55; R-squared 0.261, 0.214, 0.278, 0.223; Adjusted R-squared 0.183, 0.189, 0.187, 0.184.
- Hausman's test H-stat: 2.521 and 3.232 with p-values [0.283] and [0.357].

### Exchange rate forecast errors: supplementary finding
- Authorities’ forecast errors for the exchange rate were investigated and found insignif icant for most fiscal variables.
- Exchange rate forecast errors are significant, with the expected sign, only for public debt-to-GDP (larger-than-projected depreciation associated with larger-than-projected increase in public debt), but this effect is not robust to inclusion of ToT in the same specification (see Annex Table 15).

### Interpretation and mechanisms
- The fiscal balance forecast error result reflects:
  - An increase in revenues somewhat stronger than the increase in GDP (positive slope in the revenue chart).
  - An increase in expenditure smaller than the increase in GDP (negative slope in the expenditure chart).
  - Positive ToT surprises are associated with smaller debt-to-GDP ratios (as shown in the fourth chart referenced).
- Mechanisms documented:
  - Positive surprises to GDP growth are associated with lower expenditure and higher revenue forecasts than the outcome, explaining optimistic fiscal balance forecasts.
  - Positive surprises to the GDP deflator are associated with both lower expenditure and revenue forecasts than the outcome, with the revenue effect dominant and driving a negative fiscal balance forecast error.
  - For public debt-to-GDP, the dominant mechanism is a denominator effect (higher GDP) but also numerator improvement (better fiscal balance).

### Descriptive and policy-relevant conclusions
- Descriptive patterns across LA6:
  - Optimistic fiscal balance forecasts mainly in Argentina and Brazil.
  - Argentina: overestimation of revenues larger than of expenditure.
  - Mexico: expenditure and revenues overestimated pre-GFC and underestimated post-GFC by similar magnitudes, yielding accurate overall balance forecasts.
  - Chile and Colombia: forecasts improved significantly post-GFC.
  - Peru (and to a lesser extent Brazil): forecasts appear backward-looking.
- Preliminary evidence suggests that presence of budget balance rules may help contain fiscal forecast errors, but a larger sample is needed for proper assessment.
- A formal test indicates fiscal balance forecasts are generally unbiased for the sample, while revenue and expenditure forecasts appear biased, suggesting component biases offset in the aggregate fiscal balance.
- Policy implications:
  - Optimistic forecasts for growth or ToT changes drive optimistic fiscal balance-to-GDP forecasts via overestimated revenue ratios and underestimated expenditure ratios.
  - Persistent fiscal forecast errors and their strong association with macro forecasts indicate scope for improved forecasting practices and attention to macro-fiscal linkages.
  - Cases like Mexico are consistent with active policy adjustments to meet announced fiscal balance targets (offsetting revenue forecast errors via expenditure adjustments).
- Suggested extensions and future research:
  - Compare authorities’ fiscal forecasts against Consensus Forecasts and other forecasters.
  - Expand analysis using forecast errors in nominal terms (rather than as ratios of GDP) to better disentangle denominator effects.
  - Explore repercussions of biased expenditure forecasts around election cycles or asymmetric loss functions by fiscal authorities; these extensions are left for future research.

*Source: wpiea2021154-print-pdf - Section 3.*

### Section 4

### Section 4

### Dataset Example (Colombia)
- Dataset fields shown: Fiscal balance(% of GDP), Expenditures(% of GDP), Revenues(% of GDP), Public Debt(% of GDP), nominal GDP, real GDP growth, GDP deflator, CPI Inflation, Exchange rate (TRM, average). Units: (billion pesos) and % (change) (dec-dec).
- Yearly entries (selected raw values preserved as in table):
  - 2000: 8.75; 2,104.72
  - 2001: 7.65; 2,301.74
  - 2002: 6.98; 2,534.22
  - 2003: -12,151.00 -5.40 46,597.00 20.90 34,445.00 15.40 223,668.83 6.50 2,875.05
  - 2004: -13,983.00 -5.50 53,934.00 21.10 39,952.00 15.60 105,514.26 41.20 256,102.56 5.33 8.70 5.50 2,614.80
  - 2005: -13,730.00 -4.80 59,606.00 21.00 45,876.00 16.20 136,197.07 40.30 337,958.00 4.71 26.03 4.84 2,319.26
  - 2006: -13,094.00 -4.10 69,441.00 21.70 56,347.00 17.60 146,154.33 38.30 381,604.00 6.72 5.81 4.48 2,363.75
  - 2007: -11,613.00 -3.20 76,318.00 21.40 64,705.00 18.10 153,405.15 35.80 428,506.00 6.74 5.20 5.69 2,076.57
  - 2008: -11,067.00 -2.30 86,131.00 18.10 75,064.00 15.70 169,176.67 35.50 476,554.00 3.28 7.68 7.68 1,989.35
  - 2009: -20,715.00 -4.20 97,871.00 19.70 77,156.00 15.50 183,576.08 36.60 501,574.00 1.14 4.06 2.01 2,180.19
  - 2010: -20,171.00 -3.70 95,128.00 17.40 74,957.00 13.70 202,390.32 37.20 544,060.00 4.49 3.80 3.16 1,902.50
  - 2011: -17,507.00 -2.80 111,754.00 18.10 94,247.00 15.30 215,420.00 34.80 619,023.00 6.95 6.39 3.73 1,854.02
  - 2012: -15,440.00 -2.30 122,507.00 18.40 107,067.00 16.10 223,946.35 33.60 666,507.00 3.91 3.62 2.44 1,798.73
  - 2013: -16,645.00 -2.36 136,390.00 19.30 119,744.00 16.94 247,076.18 34.60 714,093.00 5.13 1.91 1.93 1,879.53
  - 2014: -18,356.00 -2.40 144,260.00 19.10 125,904.00 16.70 283,037.01 37.10 762,903.00 4.50 2.24 3.66 2,017.85
  - 2015: -24,269.00 -3.00 153,590.00 19.20 129,321.00 16.10 338,775.33 42.10 804,692.00 2.96 2.45 6.77 2,771.55
  - 2016: -34,925.00 -4.00 163,293.00 18.90 128,367.00 14.90 375,745.17 43.50 863,782.00 2.09 5.15 5.75 3,040.09
  - 2017: -33,636.00 -3.60 174,519.00 18.90 144,062.00 15.60 409,609.60 44.50 920,471.00 1.36 5.13 4.09 2,957.52
  - 2018: -30,316.00 -3.10 179,608.00 18.40 149,292.00 15.30 473,246.88 48.00 985,931.00 2.52 4.48 3.18 2,972.04
  - 2019: nana 1,062,342.50 3.32 4.29 3.80 3,299.77
- Data provenance and metadata lines in source: Gobierno Nacional Central; Banco Central de Colombia; Official budget report; Calculation / Transformation; Special definition; Taken from a graph (visually); Not available; Official national source.
- Source URL provided in original table: https://www.minhacienda.gov.co/webcenter/portal/EntidadesFinancieras/pages_EntidadesFinancieras/PoliticaFiscal/bgg/balancefiscalgobiernocentral

### Descriptive Findings Overview Tables
- Tables enumerated in source:
  - Table A5. Dataset Example (Mexico)
  - Table A6. Dataset Example (Peru)
  - Table A7. Authorities’ Forecasts for Fiscal Balance-to-GDP Ratio (Source: National authorities and authors’ calculations.)
  - Table A8. Authorities’ Forecasts for Expenditure-to-GDP Ratio (Source: National authorities and authors’ calculations.)
  - Table A9. Authorities’ Forecasts for Revenue-to-GDP Ratio (Source: National authorities and authors’ calculations.)
  - Table A10. Authorities’ Forecasts for Public Debt-to-GDP Ratio (Source: National authorities and authors’ calculations.)
  - Table A11. Authorities’ Forecasts for GDP Growth (Source: National authorities and authors’ calculations.)
  - Table A12. Authorities’ Forecasts for GDP Deflator Change (Source: National authorities and authors’ calculations.)
  - Table A13. Authorities’ Forecasts for CPI Inflation (Source: National authorities and authors’ calculations.)
  - Figure A1. Boxplots: Distribution of the Authorities’ Forecasts (Source: National authorities and authors’ calculations.)
- Selected numerical summaries from panels (preserved exactly as in source) — differences between projection and actual by country code columns ARGBRACHLCOLMEXPER:
  - Expenditure / GDP (Projection - Actual) samples by year (showing values across countries and summary stats):
    - 2000: 0.10 -0.10 0.10 0.10
    - 2001: 0.20 1.30 0.75 0.20 1.30
    - 2002: 0.46 0.15 -0.20 0.04 -0.20 0.46
    - 2003: -0.23 -0.26 -0.10 -0.30 -0.22 -0.30 -0.10
    - 2004: -1.62 -2.86 -0.10 -0.20 -0.30 -1.02 -2.86 -0.10
    - 2005: -0.62 0.21 -3.19 -1.30 -0.10 -0.50 -0.92 -3.19 0.21
    - 2006: -0.24 0.40 -5.01 -0.90 0.00 -2.90 -1.44 -5.01 0.40
    - 2007: -0.01 0.38 -3.37 -0.10 0.00 -3.70 -1.13 -3.70 0.38
    - 2008: -0.34 -0.48 0.91 -0.80 0.10 -1.90 -0.42 -1.90 0.91
    - 2009: 1.56 1.51 8.04 0.50 2.20 4.60 3.07 0.50 8.04
    - 2010: -0.15 -1.17 -0.65 -0.70 0.30 -1.10 -0.58 -1.17 0.30
    - 2011: 0.89 -0.42 -2.08 -1.20 0.20 -2.60 -0.87 -2.60 0.89
    - 2012: 2.31 0.92 -0.96 -0.10 0.30 -0.90 0.26 -0.96 2.31
    - 2013: 1.97 1.20 -0.10 -0.04 0.30 -0.70 0.44 -0.70 1.97
    - 2014: 2.51 1.85 0.72 0.00 -0.40 0.50 0.86 -0.40 2.51
    - 2015: 4.10 3.04 0.24 0.00 -0.10 2.10 1.56 -0.10 4.10
    - 2016: 4.31 1.74 -0.49 0.10 -0.50 -0.70 0.74 -0.70 4.31
    - 2017: 1.00 -0.15 -0.55 0.00 -1.30 0.60 -0.07 -1.30 1.00
    - 2018: -0.49 0.10 -0.25 0.00 --- 0.16 -0.49 0.10
    - 2019: 0.62 -1.13 -2.40 --- 0.22 -2.40 1.13
    - Averages and extremes: average 0.89 0.65 -0.51 -0.44 0.05 -0.39 0.04; min -1.62 -1.17 -5.01 -2.40 -1.30 -3.70 -5.01; max 4.31 3.04 8.04 0.50 2.20 4.60 8.04
  - Fiscal Balance / GDP (Projection - Actual) samples:
    - 2000: 0.27 0.27 0.27 0.27
    - 2001: 1.06 -1.00 -1.03 -1.06 -1.00
    - 2002: -0.14 2.02 2.10 -0.07 -2.10 2.02
    - 2003: -4.80 -2.94 -0.94 -2.70 -0.90 -4.80 2.94
    - 2004: 0.71 -3.21 0.60 2.00 0.10 1.32 0.10 3.21
    - 2005: 3.11 -0.81 1.97 0.90 0.60 0.10 0.98 -0.81 3.11
    - 2006: 1.76 0.61 3.46 0.50 -0.50 1.50 1.22 -0.50 3.46
    - 2007: 0.76 1.62 0.99 0.30 0.80 1.80 1.05 0.30 1.80
    - 2008: 1.97 0.52 -1.84 3.70 -0.60 0.20 0.66 -1.84 3.70
    - 2009: 1.37 0.31 -1.94 0.10 -4.00 -3.80 -1.33 -4.00 1.37
    - 2010: 2.82 1.78 2.61 0.90 -0.50 -0.10 1.25 -0.50 2.82
    - 2011: 4.39 1.38 1.69 0.30 -1.30 1.80 1.38 -1.30 4.39
    - 2012: 3.27 -0.84 1.12 0.20 -0.90 0.10 0.49 -0.90 3.27
    - 2013: 3.50 -1.10 0.76 -0.20 -2.40 -0.10 0.08 -2.40 3.50
    - 2014: 3.86 1.22 -0.92 0.40 -0.60 -1.50 0.41 -1.50 3.86
    - 2015: 1.83 1.31 0.11 0.00 -1.00 0.30 0.42 -1.00 1.83
    - 2016: 0.62 -0.45 0.60 0.10 -2.10 2.20 0.16 -2.10 2.20
    - 2017: 1.56 -0.29 0.58 0.10 0.10 0.60 0.44 -0.29 1.56
    - 2018: 1.21 -1.67 -0.02 -0.30 --- 0.20 -1.67 1.21
    - 2019: 1.32 --1.37 19.00 --6.32 -1.37 19.00
    - Averages and extremes: average 1.62 0.26 0.82 1.66 -0.49 -0.15 0.62; min -4.80 -1.67 -1.94 -0.30 -4.00 -3.80 -4.80; max 4.39 1.78 3.46 19.00 2.02 2.20 19.00
  - Revenue / GDP (Projection - Actual) samples:
    - 2000: 0.18 0.18 0.18 0.18
    - 2001: 0.88 0.20 -0.34 -0.88 0.20
    - 2002: 0.34 1.87 -2.30 -0.03 -2.30 1.87
    - 2003: -5.03 -2.68 -0.84 -2.90 -1.10 -5.03 2.68
    - 2004: -0.92 -0.35 0.50 1.80 -0.30 0.29 -0.92 1.80
    - 2005: 2.50 -0.60 -1.12 -0.50 0.50 -0.40 0.06 -1.12 2.50
    - 2006: 1.51 1.00 -1.55 -0.30 -0.50 -1.40 -0.21 -1.55 1.51
    - 2007: 0.76 2.00 -2.38 0.30 0.80 -2.10 -0.10 -2.38 2.00
    - 2008: 1.63 0.04 -0.92 3.00 -0.50 -1.60 0.27 -1.60 3.00
    - 2009: 2.93 1.82 6.10 0.60 -1.80 0.80 1.74 -1.80 6.10
    - 2010: 2.67 0.60 1.96 0.20 -0.20 -1.10 0.69 -1.10 2.67
    - 2011: 5.27 0.96 -0.39 -0.90 -1.10 -0.90 0.49 -1.10 5.27
    - 2012: 5.57 0.08 0.16 0.10 -0.60 -1.00 0.72 -1.00 5.57
    - 2013: 5.47 0.00 0.65 -0.14 -1.90 -0.50 0.60 -1.90 5.47
    - 2014: 6.35 2.98 -0.20 0.30 -1.00 -1.10 1.22 -1.10 6.35
    - 2015: 5.93 4.35 0.25 0.20 -1.10 2.30 1.99 -1.10 5.93
    - 2016: 4.94 1.28 0.11 0.10 -2.60 1.60 0.90 -2.60 4.94
    - 2017: 2.46 -0.44 0.03 -0.30 -1.40 1.20 0.26 -1.40 2.46
    - 2018: 0.72 -1.47 -0.28 -0.20 --- 0.31 -1.47 0.72
    - 2019: 1.84 --0.23 16.60 --6.07 -0.23 16.60
    - Averages and extremes: average 2.50 0.90 0.31 1.22 -0.44 -0.56 0.65; min -5.03 -1.47 -2.38 -0.90 -2.60 -2.90 -5.03; max 6.35 4.35 6.10 16.60 1.87 2.30 16.60

### Relationship between Fiscal variables and Fiscal Rules
- Table A14. Impact of Fiscal Rules on Authorities’ Fiscal Forecast Errors (note: fiscal rule data from Lledo et al (2017)):
  - Left-hand side: absolute value of forecast error for fiscal variable. Right-hand side: dummies for each fiscal rule (budget balance, expenditure, debt rule).
  - Regression results (Fixed Effects F.E.) with standard errors in parenthesis and p-values in brackets:
    - Dependent variable columns: Abs FE Bal/GDP | Abs FE Exp/GDP | Abs FE Rev/GDP | Abs FE PDebt/GDP
    - D Bal coefficients: -0.48 (0.591) [0.420]; -0.78 (0.457) [0.092]; -1.439 (0.547) [0.010]; -3.144 (1.077) [0.005]
    - D Exp coefficients: -0.396 (0.638) [0.537]; 0.017 (0.493) [0.972]; -0.334 (0.590) [0.573]; -2.57 (1.240) [0.043]
    - D Debt coefficients: -0.96 (0.730) [0.193]; -0.28 (0.564) [0.621]; 0.09 (0.675) [0.894]; -0.58 (1.194) [0.629]
    - Constant terms: 1.941 (0.395) [0.000]; 1.822 (0.305) [0.000]; 2.616 (0.366) [0.000]; 6.886 (1.058) [0.000]
    - Observations: 78 78 78 64
    - Countries: 66 65
    - R-squared: 0.056 0.061 0.160 0.222
    - Adjusted R-squared: -0.053 -0.048 0.063 0.125
  - Note: Figures highlighted in (light) red in source depict lack of significance at the 10 percent level.

### Inspecting the Fixed Effects
- Figure A2. Fixed Effects panels reported:
  - Fixed effects from regressions where:
    - Left charts: regressions include forecast errors of real GDP growth and GDP deflator as regressors.
    - Right charts: regressions also include Forecast Errors of the Terms of Trade.
  - Panels shown for dependent variables:
    - Forecast errors of fiscal balance-to-GDP
    - Forecast errors of public debt-to-GDP
    - Forecast errors of expenditure-to-GDP
    - Forecast errors of revenue-to-GDP
- Source: National authorities and authors’ calculations.

### Impact of the Exchange Rate Forecast Errors
- Table A15. Regression Results: Explaining Forecast Errors for the Public Debt-to-GDP Ratio (FE PDebt/GDP):
  - Two specifications (1) and (2) with Fixed Effects (F.E.), standard errors in parenthesis, p-values in brackets.
  - FE RGDP g coefficients: -0.54 (0.240) [0.029]; -0.524 (0.243) [0.035]
  - FE GDP defl g coefficients: -0.069 (0.075) [0.357]; -0.05 (0.082) [0.545]
  - FE ToT: not included in both (marked "-")
  - FE ToT (change): in column (2) coefficient -0.032 -(0.052) with p-value [0.547] (presented in source as --0.032 -(0.052) -[0.547])
  - ExR Dev (exchange rate deviation) coefficients: 0.114 (0.065) [0.085]; 0.1 (0.069) [0.152]
  - Constant terms: 1.194 (0.165) [0.000]; 1.06 (0.276) [0.000]
  - Observations: 64 64
  - Countries: 55 55
  - R-squared: 0.30 0.304
  - Adjusted R-squared: 0.212 0.203
  - Note in source: Figures highlighted in (light) red depict lack of significance at the 10 percent level. Hausman's test of specification H-stat and p-value referenced (values not provided in excerpt). Standard errors in parenthesis; p-values in brackets.

*Source: National authorities and authors’ calculations as presented in the original PDF section.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021154-print-pdf.pdf_