Data supporting "Solar and Wind Expansion and Uneven Fossil-Generation Displacement Across 29 Emerging and Developing Economies"

This repository contains the processed data, derived variables, source data for figures and tables, model outputs, and reproducibility materials supporting the study:

"Solar and Wind Expansion and Uneven Fossil-Generation Displacement Across 29 Emerging and Developing Economies”

The study examines whether growth in solar and wind electricity generation was accompanied by absolute reductions in fossil-fuel electricity generation across a balanced panel of 29 emerging and developing economies from 2010 to 2024. The dataset supports descriptive electricity-balance accounting, country and regional comparisons, two-way fixed-effects estimation, small-cluster inference, leave-one-country-out classification, calibration analysis, and robustness checks.

Dataset scope

The harmonized level panel contains:

• 29 countries;
• annual observations from 2010 to 2024;
• 435 country-year observations.

The annual-change panel contains:

• annual first differences from 2011 to 2024;
• 406 country-year observations;
• 317 observations with positive solar-wind generation growth;
• 129 observations with an annual decline in fossil generation.

The dataset includes countries from South Asia, Southeast Asia, the Middle East and North Africa, Sub-Saharan Africa, Latin America, and Central Asia/Eurasia.

Repository contents

The workbook contains:

• the harmonized 2010–2024 country-year panel;
• the 2011–2024 annual-change panel;
• electricity-generation, demand, emissions, trade, and socioeconomic variables;
• gross fossil-displacement ratio and clean-growth-coverage indicators;
• annual transition classifications;
• country-level, regional, and equal-weighted summary results;
• fixed-effects and robustness outputs;
• wild-cluster-bootstrap and multicollinearity diagnostics;
• leave-one-country-out regression estimates;
• out-of-fold classification probabilities and predictions;
• fold-specific logistic-regression coefficients;
• majority-class and sensitivity-model results;
• source-consistency checks against Our World in Data;
• electricity-balance residual checks;
• source data underlying the manuscript figures and tables;
• a complete variable dictionary;
• a workbook README describing individual worksheets.

A separate Markdown README provides additional information on dataset structure, variables, units, sources, missing-value conventions, and the reproduction workflow.

Main indicators

The dataset includes two principal accounting indicators.

Gross fossil-displacement ratio

GDR is calculated when annual solar-wind generation growth is positive. Positive values indicate contemporaneous fossil-generation decline, while negative values indicate that fossil and solar-wind generation increased together. GDR is a descriptive accounting ratio and should not be interpreted as a causal efficiency estimate.

Clean-growth coverage

CGC is calculated when annual electricity-demand growth is positive. Values of at least one indicate that solar-wind growth equalled or exceeded the annual increase in electricity demand.

The dataset also classifies positive solar-wind-growth observations into four mutually exclusive outcomes:

1. fossil decline exceeding solar-wind growth;
2. fossil decline below solar-wind growth;
3. demand-dominant co-expansion;
4. fossil co-expansion despite demand coverage.

Data sources

The processed dataset was constructed from three public sources:

• Ember Yearly Electricity Data for electricity generation by source, electricity demand, net imports, power-sector emissions, and carbon intensity;
• World Bank World Development Indicators for GDP, population, urbanization, industrial structure, manufacturing, and related socioeconomic variables;
• Our World in Data Energy Dataset for source-consistency checks.

The repository contains harmonized and derived research data rather than replacing the original public source datasets. Users should consult and cite the original data providers when reusing the underlying source variables.

Units

The main units are:

• electricity generation, electricity demand, and net imports: TWh;
• power-sector emissions: Mt CO₂;
• carbon intensity: g CO₂ kWh⁻¹;
• generation shares and socioeconomic shares: percent;
• GDP growth and population growth: percent;
• GDP per capita: constant US dollars or log-transformed constant US dollars;
• GDR and CGC: dimensionless ratios;
• fossil-decline outcome: binary indicator.

Analytical outputs

The workbook supports reproduction of:

• aggregate electricity trends;
• country and regional displacement accounting;
• fixed-effects regression estimates;
• country-clustered, Driscoll–Kraay, and wild-cluster-bootstrap inference;
• influence and leave-one-country-out checks;
• prospective structural logistic classification;
• ex-post diagnostic logistic classification;
• full electricity-balance sensitivity analysis;
• ROC, precision-recall, calibration, and Brier-score results;
• manuscript figures, tables, and supplementary tables.

The prospective classifier uses lagged electricity-system and socioeconomic variables. The ex-post diagnostic classifier additionally uses contemporaneous normalized changes in electricity demand, solar-wind generation, and hydropower. These diagnostic models identify annual fossil-decline outcomes after electricity-balance changes are observed and should not be interpreted as causal models or true ex-ante forecasts.

Reproducibility

The associated software repository contains:

• the Google Colab notebook;
• a Python script;
• README.md;
• requirements.txt;
• environment.yml;
• exact package versions;
• fixed random seed and bootstrap settings;
• instructions for obtaining the public input datasets;
• procedures for regenerating the processed panels, model outputs, figures, and tables.

The default analysis uses a random seed of 42, 999 wild-cluster-bootstrap replications, and 2,000 country-cluster bootstrap replications for the principal classification analyses.

Related resource

The complete analysis code and computing environment are deposited in a separate linked Zenodo software record.