## ANNEX I. DATA SOURCES AND DESCRIPTIONS

## Source details

**Canonical URL:** [ANNEX I. DATA SOURCES AND DESCRIPTIONS](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020229-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020229-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020229-print-pdf.pdf.json)

---

### I. Introduction and objectives
- Research questions:
  - Are IMF output growth forecasts systematically optimistic?
  - What is the role of planned policy adjustments on forecast optimism?
  - Are program-specific forecasts as optimistic as forecasts for countries under regular surveillance?
- Coverage and data:
  - Real GDP growth forecast errors during 2003–2017.
  - Program-specific forecasts from IMF’s Monitoring of Fund Arrangements (MONA) database for non-concessional, non-precautionary financial arrangements.
  - Non-program observations from vintage-specific IMF World Economic Outlook (WEO) forecasts.
- Identification strategy highlights:
  - Distinguish between planned policy adjustments (baseline macro scenario) and actual policy adjustment during the forecast horizon to attenuate endogeneity.
  - Control for exogenous cyclical and structural factors and examine non-linearities and horizon effects.
- Relationship to prior work:
  - Supplements Blanchard and Leigh (2013, 2014) by broadening scope and investigating non-linearity and other factors across IMF programs and surveillance cases.

### II. Key findings and quantitative results
- Association of planned policy adjustments with optimism:
  - Planned policy adjustments are associated with optimism in both program and non-program observations.
  - Significant non-linearities: larger planned policy adjustments—defined as episodes where planned adjustment exceeded half a standard deviation above the historical average actual adjustment for each country—tend to display stronger association with optimism bias.
- Cross-border and commodity influences:
  - Optimism on trading partners’ growth is positively associated with a country’s own growth optimism.
  - Bullish projections on oil prices are associated with sanguine real GDP growth rate projections in both oil importers and exporters.
- Structural measures:
  - No evidence that planned structural adjustment, proxied by the number of structural benchmarks agreed under an IMF program, plays a statistically significant role.
- Magnitude and horizon effects:
  - IMF forecasts are associated with an optimism bias of around 0.2 percentage points of a country’s real GDP growth rate each year into the projection horizon, resulting in an optimism drift.
  - Over a five-year horizon, the cumulative impact of IMF’s optimism bias adds up to 1 percentage point of a country’s real GDP growth rate in addition to optimism associated from planned policy adjustment.
- Fiscal adjustment and underestimation of multipliers:
  - Strong positive association between planned and actual fiscal adjustment.
  - Instrumental variable analysis indicates actual fiscal adjustment carries a strong positive association with growth optimism, suggesting underestimation of fiscal multipliers may contribute to observed optimism, particularly in programs.
  - Non-linear association may reflect increasing cost to growth from higher adjustment, making high fiscal adjustment programs associated with disproportionately optimistic growth projections.
- Persistence and policy relevance:
  - Tendency for optimism persisted in the period after the Global Financial Crisis (GFC).
  - The strong correlation between magnitude of optimism and expected fiscal consolidation provides a cautionary signal for post-COVID economic projections as countries embark on fiscal adjustment.

### III. Data construction and econometric approach
- Forecast vintages:
  - Use of vintage-specific WEO data to control for information available at forecast time.
- Program sample:
  - Non-concessional, non-precautionary programs approved by the IMF Executive Board during the past two decades (coverage detailed in MONA-derived dataset).
  - Program sample composition: 42 Stand-by Arrangements (SBAs) and 13 Extended Fund Facility (EFF) arrangements.
- Forecast horizon and vintages:
  - Forecast horizon index k ranges from 0 up to 5 (0 = same year; 5 = five years ahead) for WEO vintages; MONA records projections up to four years ahead for program observations.
  - Vintage-specific WEO data: merging 29 databases (one per WEO vintage).
- Dependent variable:
  - Forecast error FE_{i,t,j,k} = E_{i,t,j}(g_{i,t+k}) − g_{i,t+k} (optimism bias labeled positive).
- Main econometric approach:
  - Panel data regressions with observation-specific fixed effects; also estimate random effects and Hausman-Taylor (H-T) instrumental specification.
  - Controls include forecast errors in oil and non-oil commodity prices, interactions with exporter dummies, time-to-forecast (proxy for optimism drift), number of structural benchmarks, and a global financial crisis dummy (2008–2009).
  - Planned policy adjustments matrix P_{i,t,j} decomposed into fiscal and external planned adjustments with indicator variables for “large” adjustments (defined as planned contemporaneous adjustment higher than country i’s average actual adjustment by half of one standard deviation).
- Robustness:
  - Robustness tests available in the Annexes.

### IV. Incidence of large planned adjustments (2003–17)
- IMF-supported financial arrangements:
  - Large planned fiscal adjustment: 258 out of 901 observations (165 out of 263 forecasts; 63 percent).
  - Large planned external adjustment: 107 out of 901 observations (101 out of 263 forecasts; 38 percent).
- Surveillance cases:
  - Large planned fiscal adjustment: 2,453 out of 19,062 observations (1,517 out of 3,989 forecasts; 38 percent).
  - Large planned external adjustment: 1,611 out of 19,062 observations (994 out of 3,989 forecasts; 25 percent).

### V. Key empirical findings (detailed)
- Non-linear association of planned adjustments and optimism:
  - Large planned fiscal and external adjustments are associated with greater optimism in growth forecasts, pointing to potential non-linearities.
  - In program-specific observations, large planned fiscal adjustments are associated with more optimism than smaller planned fiscal adjustments.
  - Large planned external adjustments are associated with higher optimism across the sample but less so in program-specific observations than in surveillance cases.
- Optimism drift over the forecast horizon:
  - IMF forecasts show an optimism bias of around 0.2 percentage points of a country’s real GDP growth rate each year into the projection horizon.
  - Ceteris paribus, optimism accumulates to 1 percentage point over a five-year horizon.
  - Optimism drift increased post-GFC.
- Role of external projections and commodities:
  - Forecast errors in trading partners’ growth and oil price projections have strong impacts on growth forecast errors.
  - Sanguine oil price projections are associated with growth optimism in forecasts of both oil importers and exporters, more so for exporters.
  - Once commodity price optimism is controlled for, being an oil importer or exporter is not significantly associated with growth optimism bias.
- Structural conditionality:
  - The proxy for planned structural adjustments (number of structural benchmarks at program approval) is not significantly associated with growth forecast errors in program-specific observations.
- Idiosyncratic residuals and learning:
  - Residuals from main model (2) are mostly idiosyncratic; controls such as initial public debt, output gap, regional affiliation, exchange rate regime, and past program status are not significantly associated with forecast errors in program subsample.
  - Little evidence that learning from past programs has reduced optimism bias in IMF growth forecasts.

### VI. Selected econometric estimates and statistics
- Main specification (Fixed effects; full sample):
  - partnergrowth_error: 0.441***
  - oil_notexp_error: 0.895***
  - oil_exp_error: 1.836***
  - commodity_notexp_error: 1.480***
  - commodity_exp_error: 0.300
  - time_to_forecast (days*365): 0.225***
  - pb_adjustment: -0.0157
  - prog_pb_adjustment: 0.157*
  - high_pb_adjustment: 0.122***
  - ca_adjustment: 0.0166**
  - prog_ca_adjustment: -0.165***
  - high_ca_adjustment: 0.416***
  - GFC (2008,9=1): 0.663***
  - Observations: 19,963
  - R-squared: 0.2470
  - Number of forecasts: 4,252
- IV (two-stage) fixed-effects results on actual fiscal adjustment (Table 5):
  - pb_adjustment_actual_hat: 0.247*** (standard error 0.0275)
  - prog_pb_adjust_actual_hat: 0.314** (standard error 0.135)
  - oil_g_error: 4.122*** (standard error 0.142)
  - oil_exp_error: 3.711*** (standard error 0.469)
  - Constant: 0.663*** (standard error 0.0284)
  - Observations: 21,028
  - Number of code: 4,493
  - R-squared: 0.059
- Interpretation of IV results:
  - Strong positive association between instrumented actual fiscal adjustment and growth forecast optimism, with higher impact for program cases—supporting hypothesis that underestimation of fiscal multipliers helps explain optimism.

### VII. Mechanisms and interpretation
- Fiscal and external channels:
  - Planned fiscal consolidation can include efficiency gains (reducing multipliers), but large adjustments may exceed efficiency gains and produce disproportionately large contractionary effects.
  - Current account adjustment serves as a close proxy for overall demand-management impact of policy packages, including monetary and structural channels.
- Non-linearity rationale:
  - As easy efficiency gains are exhausted, further tightening may interrupt high-return activities, producing non-linear (larger-than-proportional) negative effects on output.
  - Empirical literature and prior IMF research indicate fiscal policy non-linearities, particularly with large persistent fiscal impulses.
- Contextual drivers noted in prior work:
  - Attribution (2011): underestimation of fiscal multipliers by IMF forecasters possibly explained by the binding zero-lower bound on nominal interest rates, impaired balance sheets of financial institutions, and overall economic slack.

### VIII. Policy implications and cautions
- Forecast construction:
  - Need to account for non-linear effects of planned fiscal adjustment when constructing baseline projections and scenario analyses.
- Fiscal adjustment and multiplier assessment:
  - Reassess assumptions on fiscal multipliers, especially for large consolidation plans, given evidence that underestimation may drive optimism in growth forecasts.
- Surveillance and program design:
  - Increased scrutiny of planned policy adjustment magnitudes and the realism of associated macroeconomic effects in both surveillance and program cases.
- Forward-looking risk management:
  - Given persistence of optimism post-GFC and the link to fiscal consolidation, incorporate stronger downside scenarios when forecasting in periods of large expected fiscal adjustments (policy-relevant caution for post-COVID projections).
- Suggestions for future work:
  - Integrate cross-country information on type of fiscal adjustment (expenditure-based versus revenue-based) to examine whether type of consolidation relates systematically to growth optimism and to test priors on fiscal multipliers by type of adjustment.

### IX. Variable definitions and data sources (selected)
- program_country, PastProgram: IMF programs — Source: MONA database
- gdp_error: Forecast GDP growth error — Source: WEO databases, MONA database, program documents, and IMF staff calculations
- partnergrowth_error: Trading partner’s forecast GDP growth error — Source: Global Economic Environment (GEE) databases
- pb_adjustment_error, prog_pb_adjustment_error, high_pb_adjustment_error: Forecast errors for change in primary fiscal balance — Source: MONA and WEO databases, and IMF staff calculations
- ca_adjustment_error, prog_ca_adjustment_error, high_ca_adjustment_error: Forecast errors for change in current account balance — Source: MONA and WEO databases, and IMF staff calculations
- bm_ngdp_adj_error: Forecast error for change in broad money — Source: MONA and WEO databases, and IMF staff calculations
- oil_g_error, commodity_g_error: Forecast oil and commodity price growth errors — Source: WEO databases
- oilexporter, commexporter: Whether a country is an oil/commodity exporter — Source: WEO database
- oil_exp_error, oil_nonexp_error, commodity_exp_error, commodity_noexp_error: Price-error variables by exporter status — Source: IMF staff calculations
- time_to_forecast: Number of days for forecasts — Source: IMF staff calculations
- time_sb: Number of structural benchmarks at program approval with consideration of number of days for forecasts — Source: MONA database and IMF staff calculations
- exceptional_d: Whether a program has exceptional access — Source: Program documents
- GFC: Global financial crisis — Source: IMF staff calculations
- ggdebt_: Public debt — Source: MONA and WEO databases, and IMF staff calculations
- gap: Difference between forecast GDP growth and potential — Source: MONA and WEO databases, and IMF staff calculations
- fxregime, fxregime2: Exchange rate regime — Source: The Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER)
- isAPD, isAFR, isEUR, isMCD, isWHD: Regional affiliation — Source: WEO database

*Source: ANNEX I. DATA SOURCES AND DESCRIPTIONS (wpiea2020229-print-pdf).*

### ANNEX I. DATA SOURCES AND DESCRIPTIONS _________________________________  21

### ANNEX I. DATA SOURCES AND DESCRIPTIONS

### I. Introduction and objectives
- Research questions:
  - Are IMF output growth forecasts systematically optimistic?
  - What is the role of planned policy adjustments on forecast optimism?
  - Are program-specific forecasts as optimistic as forecasts for countries under regular surveillance?
- Coverage and data:
  - Real GDP growth forecast errors during 2003–2017.
  - Program-specific forecasts from IMF’s Monitoring of Fund Arrangements (MONA) database for non-concessional, non-precautionary financial arrangements.
  - Non-program observations from vintage-specific IMF World Economic Outlook (WEO) forecasts.
- Identification strategy highlights:
  - Distinguish between planned policy adjustments (baseline macro scenario) and actual policy adjustment during the forecast horizon to attenuate endogeneity.
  - Control for exogenous cyclical and structural factors and examine non-linearities and horizon effects.
- Relationship to prior work:
  - Supplements Blanchard and Leigh (2013, 2014) by broadening scope and investigating non-linearity and other factors across IMF programs and surveillance cases.

### II. Key findings and quantitative results
- Association of planned policy adjustments with optimism:
  - Planned policy adjustments are associated with optimism in both program and non-program observations.
  - Significant non-linearities: larger planned policy adjustments—defined as episodes where planned adjustment exceeded half a standard deviation above the historical average actual adjustment for each country—tend to display stronger association with optimism bias.
- Cross-border and commodity influences:
  - Optimism on trading partners’ growth is positively associated with a country’s own growth optimism.
  - Bullish projections on oil prices are associated with sanguine real GDP growth rate projections in both oil importers and exporters.
- Structural measures:
  - No evidence that planned structural adjustment, proxied by the number of structural benchmarks agreed under an IMF program, plays a statistically significant role.
- Magnitude and horizon effects:
  - IMF forecasts are associated with an optimism bias of around 0.2 percentage points of a country’s real GDP growth rate each year into the projection horizon, resulting in an optimism drift.
  - Over a five-year horizon, the cumulative impact of IMF’s optimism bias adds up to 1 percentage point of a country’s real GDP growth rate in addition to optimism associated from planned policy adjustment.
- Fiscal adjustment and underestimation of multipliers:
  - Strong positive association between planned and actual fiscal adjustment.
  - Instrumental variable analysis indicates actual fiscal adjustment carries a strong positive association with growth optimism, suggesting underestimation of fiscal multipliers may contribute to observed optimism, particularly in programs.
  - Non-linear association may reflect increasing cost to growth from higher adjustment, making high fiscal adjustment programs associated with disproportionately optimistic growth projections.
- Persistence and policy relevance:
  - Tendency for optimism persisted in the period after the Global Financial Crisis (GFC).
  - The strong correlation between magnitude of optimism and expected fiscal consolidation provides a cautionary signal for post-COVID economic projections as countries embark on fiscal adjustment.

### III. Data construction and econometric approach
- Forecast vintages:
  - Use of vintage-specific WEO data to control for information available at forecast time.
- Program sample:
  - Non-concessional, non-precautionary programs approved by the IMF Executive Board during the past two decades (coverage detailed in MONA-derived dataset).
- Controls and tests:
  - Controls for exogenous cyclical and structural factors.
  - Tests for non-linearity of planned policy adjustments on forecast errors.
  - Analysis of horizon dependence: optimism bias increases with projection horizon.
- Robustness:
  - Robustness tests available in the Annexes.

### IV. Literature synthesis (selected numeric highlights)
- Early program and surveillance studies:
  - Mussa and Phillips: sample of 69 non-precautionary, non-concessional financial arrangements approved during 1993–1997, involving 47 emerging and transition economies; find optimism in large programs only.
  - IEO (2003): extends sample to 133 programs in 70 countries during 1993–2001; links fiscal revisions, fiscal underperformance, and optimism bias; identifies revenue shortfalls as main culprit in programs with large planned fiscal adjustment (defined as targeting fiscal adjustment of more than 3 percentage points of GDP over a two-year horizon).
- Mixed evidence on consequences of ambitious adjustments:
  - Baqir et al. (2005): sample of 94 IMF-supported programs during 1989–2002; find optimistic growth forecasts but programs with more ambitious fiscal targets led to better growth performance.
- Forecast accuracy relative to markets:
  - Timmerman (2007): WEO forecasts for 178 countries during 1990–2003; WEO overpredicted one-year ahead growth in about 20 percent of advanced economies, 25 percent of Western Hemisphere countries, and 35 percent of African countries; comparable to private consensus.
- Measurement error and time-dependence:
  - Atoyan and Conway (2011): sample of 291 IMF-supported programs during 1993–2009; some optimism in 1990s programs but not later; 194 of the 291 programs were initiated in 1993–2001.
  - Ho and Mauro (2014): baseline forecasts more optimistic than warranted; optimism bias greater the longer the forecast horizon.
  - Eicher et al. (2018): 156 programs in 84 countries during 2002–2015; find forecasts for the year of program approval unbiased but inefficient, driven by low-income countries.
- Fiscal multipliers and forecast errors:
  - Blanchard and Leigh (2013, 2014): negative relationship between growth forecast errors and planned fiscal consolidation during 2010–2011 in a sample of 26 European advanced economies.

### V. Policy implications and cautions
- Forecast construction:
  - Need to account for non-linear effects of planned fiscal adjustment when constructing baseline projections and scenario analyses.
- Fiscal adjustment and multiplier assessment:
  - Reassess assumptions on fiscal multipliers, especially for large consolidation plans, given evidence that underestimation may drive optimism in growth forecasts.
- Surveillance and program design:
  - Increased scrutiny of planned policy adjustment magnitudes and the realism of associated macroeconomic effects in both surveillance and program cases.
- Forward-looking risk management:
  - Given persistence of optimism post-GFC and the link to fiscal consolidation, incorporate stronger downside scenarios when forecasting in periods of large expected fiscal adjustments (policy-relevant caution for post-COVID projections).

*Source: ANNEX I. DATA SOURCES AND DESCRIPTIONS (extracted content).*

### 2011. The authors attribute such finding to an underestimation of fiscal multipliers by IMF

### wpiea2020229-print-pdf - 2011. The authors attribute such finding to an underestimation of fiscal multipliers by IMF forecasters, possibly explained by the binding zero-lower bound on nominal interest rates, impaired balance sheets of financial institutions, and overall economic slack.

### Background and objective
- Study period and coverage:
  - Examines WEO and MONA forecasts of annual real GDP growth recorded during 2003–2017.
  - Dataset covers 34 advanced, 82 emerging, and 54 low-income economies.
  - Includes surveillance (WEO vintage-specific) and program-specific (MONA) observations.
- Primary aim:
  - Measure optimism bias in IMF growth forecasts and assess the non-linear impact of planned policy adjustments on growth forecast errors.
  - Test duration dependence (horizon effects), structural breaks (pre/post GFC), and whether IMF forecasters learn from forecast errors over the lifetime of IMF programs.

### Data, definitions, and methods
- Forecast horizon and vintages:
  - Forecast horizon index k ranges from 0 up to 5 (0 = same year; 5 = five years ahead) for WEO vintages; MONA records projections up to four years ahead for program observations.
  - Vintage-specific WEO data: merging 29 databases (one per WEO vintage).
- Program sample composition:
  - Focus on non-concessional, non-precautionary IMF-supported financial arrangements: 42 Stand-by Arrangements (SBAs) and 13 Extended Fund Facility (EFF) arrangements.
- Dependent variable:
  - Forecast error FE_{i,t,j,k} = E_{i,t,j}(g_{i,t+k}) − g_{i,t+k} (optimism bias labeled positive).
- Main econometric approach:
  - Panel data regressions with observation-specific fixed effects; also estimate random effects and Hausman-Taylor (H-T) instrumental specification.
  - Controls include forecast errors in oil and non-oil commodity prices, interactions with exporter dummies, time-to-forecast (proxy for optimism drift), number of structural benchmarks, and a global financial crisis dummy (2008–2009).
- Planned policy adjustments matrix P_{i,t,j}:
  - Decomposed into fiscal and external planned adjustments with indicator variables for “large” adjustments (defined as planned contemporaneous adjustment higher than country i’s average actual adjustment by half of one standard deviation).

### Incidence of large planned adjustments (2003–17)
- IMF-supported financial arrangements:
  - Large planned fiscal adjustment: 258 out of 901 observations (165 out of 263 forecasts; 63 percent).
  - Large planned external adjustment: 107 out of 901 observations (101 out of 263 forecasts; 38 percent).
- Surveillance cases:
  - Large planned fiscal adjustment: 2,453 out of 19,062 observations (1,517 out of 3,989 forecasts; 38 percent).
  - Large planned external adjustment: 1,611 out of 19,062 observations (994 out of 3,989 forecasts; 25 percent).

### Key empirical findings
- Non-linear association of planned adjustments and optimism:
  - Large planned fiscal and external adjustments are associated with greater optimism in growth forecasts, pointing to potential non-linearities.
  - In program-specific observations, large planned fiscal adjustments are associated with more optimism than smaller planned fiscal adjustments.
  - Large planned external adjustments are associated with higher optimism across the sample but less so in program-specific observations than in surveillance cases.
- Optimism drift over the forecast horizon:
  - IMF forecasts show an optimism bias of around 0.2 percentage points of a country’s real GDP growth rate each year into the projection horizon.
  - Ceteris paribus, optimism accumulates to 1 percentage point over a five-year horizon.
  - Optimism drift increased post-GFC.
- Role of external projections and commodities:
  - Forecast errors in trading partners’ growth and oil price projections have strong impacts on growth forecast errors.
  - Sanguine oil price projections are associated with growth optimism in forecasts of both oil importers and exporters, more so for exporters.
  - Once commodity price optimism is controlled for, being an oil importer or exporter is not significantly associated with growth optimism bias.
- Structural conditionality:
  - The proxy for planned structural adjustments (number of structural benchmarks at program approval) is not significantly associated with growth forecast errors in program-specific observations.
- Idiosyncratic residuals and learning:
  - Residuals from main model (2) are mostly idiosyncratic; controls such as initial public debt, output gap, regional affiliation, exchange rate regime, and past program status are not significantly associated with forecast errors in program subsample.
  - Little evidence that learning from past programs has reduced optimism bias in IMF growth forecasts.

### Selected econometric estimates and statistics (preserving reported values)
- Main specification (Fixed effects; full sample):
  - partnergrowth_error: 0.441***
  - oil_notexp_error: 0.895***
  - oil_exp_error: 1.836***
  - commodity_notexp_error: 1.480***
  - commodity_exp_error: 0.300
  - time_to_forecast (days*365): 0.225***
  - pb_adjustment: -0.0157
  - prog_pb_adjustment: 0.157*
  - high_pb_adjustment: 0.122***
  - ca_adjustment: 0.0166**
  - prog_ca_adjustment: -0.165***
  - high_ca_adjustment: 0.416***
  - GFC (2008,9=1): 0.663***
  - Observations: 19,963
  - R-squared: 0.2470
  - Number of forecasts: 4,252
- IV (two-stage) fixed-effects results on actual fiscal adjustment (Table 5):
  - pb_adjustment_actual_hat: 0.247*** (standard error 0.0275)
  - prog_pb_adjust_actual_hat: 0.314** (standard error 0.135)
  - oil_g_error: 4.122*** (standard error 0.142)
  - oil_exp_error: 3.711*** (standard error 0.469)
  - Constant: 0.663*** (standard error 0.0284)
  - Observations: 21,028
  - Number of code: 4,493
  - R-squared: 0.059
- Interpretation of IV results:
  - Strong positive association between instrumented actual fiscal adjustment and growth forecast optimism, with higher impact for program cases—supporting hypothesis that underestimation of fiscal multipliers helps explain optimism.

### Mechanisms and interpretation
- Fiscal and external channels:
  - Planned fiscal consolidation can include efficiency gains (reducing multipliers), but large adjustments may exceed efficiency gains and produce disproportionately large contractionary effects.
  - Current account adjustment serves as a close proxy for overall demand-management impact of policy packages, including monetary and structural channels.
- Non-linearity rationale:
  - As easy efficiency gains are exhausted, further tightening may interrupt high-return activities, producing non-linear (larger-than-proportional) negative effects on output.
  - Empirical literature and prior IMF research indicate fiscal policy non-linearities, particularly with large persistent fiscal impulses.

### Implications and final remarks
- Persistence and scope of optimism:
  - Optimism in IMF growth forecasts persists post-GFC and rises over projection horizons at about 0.2 percentage points per year (1 percentage point over five years).
  - Strong correlation between optimism magnitude and expected fiscal consolidation signals caution for post-COVID IMF projections as member countries embark on fiscal adjustment paths.
- Policy-relevant conclusions:
  - Large planned fiscal and external adjustments warrant careful calibration of assumed fiscal multipliers and stronger scrutiny of forecast assumptions.
  - Forecasting processes should account for non-linear multiplier effects and increased uncertainty around large planned adjustments.
- Suggestions for future work:
  - Integrate cross-country information on type of fiscal adjustment (expenditure-based versus revenue-based) to examine whether type of consolidation relates systematically to growth optimism and to test priors on fiscal multipliers by type of adjustment.

*Source: wpiea2020229-print-pdf (IMF working paper content provided).*

### Annex I. Data Sources and Descriptions

### Annex I. Data Sources and Descriptions

### Variable definitions and descriptions
- program_country, PastProgram
  - IMF programs
  - Source: Monitoring of Fund Arrangements (MONA) database

- gdp_error
  - Forecast GDP growth error
  - Source: World Economic Outlook (WEO) databases, MONA database, program documents, and IMF staff calculations

- partnergrowth_error
  - Trading partner’s forecast GDP growth error
  - Source: Global Economic Environment (GEE) databases

- pb_adjustment_error
  - Forecast error for change in primary fiscal balance
  - Source: MONA and WEO databases, and IMF staff calculations

- prog_pb_adjustment_error
  - Forecast error for change in primary fiscal balance on programs
  - Source: MONA database and IMF staff calculations

- high_pb_adjustment_error
  - Forecast error for high change in primary fiscal balance
  - Source: MONA and WEO databases, and IMF staff calculations

- ca_adjustment_error
  - Forecast error for change in current account balance
  - Source: MONA and WEO databases, and IMF staff calculations

- prog_ca_adjustment_error
  - Forecast error for change in current account balance on programs
  - Source: MONA database and IMF staff calculations

- high_ca_adjustment_error
  - Forecast error for high change in current account balance
  - Source: MONA and WEO databases, and IMF staff calculations

- bm_ngdp_adj_error
  - Forecast error for change in broad money
  - Source: MONA and WEO databases, and IMF staff calculations

- oil_g_error
  - Forecast oil price growth error
  - Source: WEO databases

- commodity_g_error
  - Forecast commodity price growth error
  - Source: WEO databases

- oilexporter
  - Whether a country is an oil exporter
  - Source: WEO database

- commexporter
  - Whether a country is a commodity exporter
  - Source: WEO database

- oil_exp_error
  - Forecast oil price growth error for oil exporters
  - Source: IMF staff calculations

- oil_nonexp_error
  - Forecast oil price growth error for nonoil exporters
  - Source: IMF staff calculations

- commodity_exp_error
  - Forecast commodity price growth error for commodity exporters
  - Source: IMF staff calculations

- commodity_noexp_error
  - Forecast commodity price growth error for non-commodity exporters
  - Source: IMF staff calculations

- time_to_forecast
  - Number of days for forecasts
  - Source: IMF staff calculations

- time_sb
  - Number of structural benchmarks at program approval with consideration of number of days for forecasts
  - Source: MONA database and IMF staff calculations

- exceptional_d
  - Whether a program has exceptional access
  - Source: Program documents

- GFC
  - Global financial crisis
  - Source: IMF staff calculations

- ggdebt_
  - Public debt
  - Source: MONA and WEO databases, and IMF staff calculations

- gap
  - Difference between forecast GDP growth and potential
  - Source: MONA and WEO databases, and IMF staff calculations

- fxregime, fxregime2
  - Exchange rate regime
  - Source: The Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER)

- isAPD, isAFR, isEUR, isMCD, isWHD
  - Regional affiliation
  - Source: WEO database

### Data construction and calculation notes
- Multiple variables combine MONA database entries, WEO databases, program documents, GEE databases, and IMF staff calculations depending on the variable; sources are specified per variable above.
- Forecast error variables cover several domains: GDP growth (gdp_error, partnergrowth_error), fiscal balances (pb_adjustment_error, prog_pb_adjustment_error, high_pb_adjustment_error), current account balances (ca_adjustment_error, prog_ca_adjustment_error, high_ca_adjustment_error), broad money (bm_ngdp_adj_error), oil and commodity prices (oil_g_error, commodity_g_error, oil_exp_error, oil_nonexp_error, commodity_exp_error, commodity_noexp_error).
- Categorical indicators flag country characteristics relevant for analysis: oilexporter, commexporter, exceptional_d, regional affiliations (isAPD, isAFR, isEUR, isMCD, isWHD), and exchange rate regimes (fxregime, fxregime2).
- Time-related variables: time_to_forecast (number of days for forecasts) and time_sb (number of structural benchmarks at program approval with consideration of number of days for forecasts).

### Key variables for program-related analysis
- program_country, PastProgram — identifies IMF program participation (MONA).
- prog_pb_adjustment_error and prog_ca_adjustment_error — forecast errors specific to programs (MONA and IMF staff calculations).
- time_sb and exceptional_d — program design and conditionality attributes (MONA; program documents).
- ggdebt_ and gap — public debt and forecast vs. potential growth gaps used to assess macrofiscal context (MONA, WEO, IMF staff calculations).
- Oil/commodity exporter indicators and corresponding price-error variables — used to distinguish shock exposures and tailor forecast-error interpretation (WEO database; IMF staff calculations).

*Source: Annex I. Data Sources and Descriptions (from the provided PDF content)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020229-print-pdf.pdf_