Replication data and code for: How much fossil generation fills hydro and wind shortfalls? Monthly evidence from 81 countries

This record contains the data and code that reproduce every table and figure in the article “How much fossil generation fills hydro and wind shortfalls? Monthly evidence from 81 countries” (submitted to Applied Energy). The work has two stages. build_data_from_raw.ipynb constructs five analysis files from publicly available sources: electricity data from Ember and the US Energy Information Administration, unit-level capacity from the Global Integrated Power Tracker of Global Energy Monitor, ERA5 reanalysis, HydroBASINS, CRU TS 4.09, Natural Earth boundaries and World Bank income data. replication_final.ipynb reads the five analysis files and produces all tables, tests and figures. The source files are not redistributed; README.md lists each one with its provider and download date. Contents: analysis_monthly.csv: country-month panel, 11,403 rows, 88 countries, January 2000 to August 2026 (generation by source, demand, net imports, instruments, controls) analysis_annual.csv: country-year panel in first differences, 5,285 rows, 209 countries, 2000 to 2025 analysis_country_year.csv: country-year levels of generation, capacity, emissions and trade shares, 2000 to 2025 analysis_weather_year.csv: annual runoff and wind-speed deviations by country analysis_country_shapes.json: simplified country outlines for the maps build_data_from_raw.ipynb and replication_final.ipynb: the two stages replication_outputs.zip: the authors’ run of stage 2 (tables, figures, summary) R_confirmation.R and wildboot_check.ipynb, with their inputs and outputs: independent confirmations in R (fixest) and Julia (WildBootTests.jl) Both notebooks run in Google Colab or any Jupyter environment; stage 2 takes about 30 minutes in Colab. The source data remain subject to the terms of their providers.