Reproducibility Package: No Robust Evidence for a GDP–Methane Relationship in Africa, 1970–2024

Abstract

Reproducibility Package This repository contains the reproducibility materials accompanying the manuscript “No Robust Evidence for a GDP–Methane Relationship in Africa, 1970–2024: A Multi-Specification Test Against the Environmental Kuznets Curve and STIRPAT Frameworks.” The package contains the analysis-ready country-year panel dataset, executable Python analysis code, derived statistical outputs, citation metadata, and an open-source license. The analytical dataset covers 53 African countries over 1970–2024 and was constructed by harmonizing methane-emissions data from the Emissions Database for Global Atmospheric Research (EDGAR) with population and gross domestic product (GDP) indicators from the World Bank World Development Indicators (WDI). The reproducibility workflow implements the principal analyses reported in the manuscript, including: linear STIRPAT-style fixed-effects models; quadratic STIRPAT-style fixed-effects models; the canonical per-capita Environmental Kuznets Curve (EKC) specification; an EKC turning-point fragility assessment; a non-parametric decile-binned assessment of the empirical GDP–methane relationship; and a formal structural-stability test using interactions associated with an independently identified 1995 structural break. The deposited country-year panel is a derived research dataset prepared from the underlying public data sources for reproducible analysis. It should not be regarded as a replacement for the original EDGAR or World Bank datasets. Users who reuse the derived dataset, code, or analytical workflow should cite this Zenodo record. Users who obtain or reuse the underlying source data directly should additionally cite the original EDGAR and World Bank WDI sources as documented in the repository README and citation metadata. The package is intended to support transparent verification, computational reproducibility, methodological inspection, and reuse of the analytical workflow underlying the manuscript. The results reproduced by this package correspond to the study period, country coverage, variable definitions, preprocessing decisions, model specifications, and statistical procedures documented in the associated manuscript and repository materials. Contents socio_panel_full.csv — analysis-ready country-year panel; 01_full_analysis.py — main reproducibility analysis script; table1_stirpat_linear.csv — linear STIRPAT model results; table1_stirpat_quadratic.csv — quadratic STIRPAT model results; table2_countries_above_turning_point.csv — countries identified relative to the estimated turning point; table3_fragility_check.csv — EKC turning-point fragility analysis; table4_canonical_ekc.csv — canonical EKC model results; table5_structural_stability.csv — structural-stability test results; nonparametric_decile_bins.csv — non-parametric decile-binned results; CITATION.cff — machine-readable citation metadata; and LICENSE — software license. The repository should be cited as a specific version. Subsequent revisions, if required, should be released as new versions so that previously cited files remain identifiable and reproducible.