Extended IMF WEO Macroeconomic Forecasts and Real Time Outturns Panel Dataset
Description
This dataset extends the IMF World Economic Outlook (WEO) Macroeconomic Forecasts Panel Dataset from three to 17 variables and updates its common multivariable vintage coverage through the October 2026 WEO. It contains 41 Spring and Fall WEO vintages from September 2006 through October 2026 for 200 economies. Extended_WEO_MFPD_Forecasts.xlsx contains 530,014 nonmissing forecast observations for horizons h=0,...,5. Each row identifies an economy, indicator, publication vintage, target year, and forecast horizon. Extended_WEO_MFPD_Outturns.xlsx contains 181,877 nonmissing real-time outturn observations for target years 2006-2025. The outturns use four transparent timing rules: Spring or Fall WEO values observed one or two years after the target year. The panel is defined over 200 economies, of which 198 have at least one available outturn. The 17 indicators cover real GDP growth, period-average and end-of-period CPI inflation, the current-account balance, unemployment, the output gap, export and import volumes, saving, investment, government revenue and expenditure, fiscal balances, and public debt. Historical indicator codes, units, and source formats are harmonized. No interpolation, carry-forward, or cross-vintage substitution is used. An absent row means that the corresponding observation was unavailable in that WEO vintage or under that outturn rule. Each workbook includes Variables and ReadMe worksheets documenting definitions, keys, units, timing rules, sources, and limitations. The author-created organization and documentation are released under CC BY 4.0. IMF-derived statistical content remains subject to the IMF Copyright and Usage Terms.
Files
Steps to reproduce
1. Obtain the official country-level IMF WEO files for the 41 included publications from September 2006 through October 2026. 2. Retain country and economy observations and exclude regional and analytical aggregates. Standardize economy identifiers using ISO3 codes while preserving the country names reported in each vintage. 3. Map the source indicators to the 17 harmonized concepts. Historical mappings include GGB_NGDP to GGXCNL_NGDP, GGD_NGDP to GGXWDG_NGDP, and GGND_NGDP to GGXWDN_NGDP. 4. Reshape each source into country-indicator-target-year observations. Assign the publication vintage, publication date, origin year, and publication round. Calculate the forecast horizon as target_year minus origin_year and retain h=0,...,5. 5. Retain only nonmissing numeric forecast observations. Do not interpolate, carry values forward, or substitute values from another vintage. 6. Construct real-time outturns for target year t using four rules: S_Y1 is the Spring WEO in t+1; F_Y1 is the Fall WEO in t+1; S_Y2 is the Spring WEO in t+2; and F_Y2 is the Fall WEO in t+2. Do not substitute another vintage when the required release is unavailable. 7. For applicable growth rates and ratios, reconstruct a candidate value from published component levels. Select the derived value only when it agrees with the directly published value within 0.051 percentage point. Otherwise retain the direct value and assign the corresponding quality flag. 8. Validate primary-key uniqueness, horizon arithmetic, outturn timing, indicator mappings, units, nonmissing-value rules, worksheet schemas, and final row counts. The resulting workbooks contain 530,014 forecast observations and 181,877 outturn observations.
Institutions
- Centre Hospitalier Universitaire Mohammed VIMarrakesh-Safi, Marrakesh