Replication dataset for the study “Climate Vulnerability in a South Asia-Centered Polycrisis: Developing the Climate–ESG–Geopolitical Entanglement Index (CEGEI).”
Description
his repository contains the analytical data and replication materials supporting the study “Climate Vulnerability in a South Asia-Centered Polycrisis: Developing the Climate–ESG–Geopolitical Entanglement Index (CEGEI).” The dataset comprises a balanced panel of 216 country-year observations covering Bangladesh, China, India, Nepal, Pakistan, and Sri Lanka from 1990 to 2025. China is included as an adjacent and systemically embedded upstream and geopolitical actor within a South Asia-centered regional system, rather than being geographically classified as a South Asian country. The analytical framework integrates 33 indicators representing climate vulnerability, governance quality, and geopolitical risk. Principal component analysis was used to derive the domain-level components and construct CEGEI, while partial least squares structural equation modelling was applied to examine the relationships among geopolitical risk, governance quality, climate vulnerability, and the interaction term. The repository includes the final analytical panel, variable definitions, source and harmonization documentation, statistical outputs, and replication instructions. These materials are provided to facilitate transparency, verification, replication, and future extensions of CEGEI in other climate-security and transboundary regional settings. Users should consult the variable dictionary and methodological notes before interpreting the component scores, particularly where the direction of ranks, governance indicators, and interaction terms affects substantive interpretation.
Files
Steps to reproduce
Download the CEGEI data files and open them in software capable of principal component analysis (PCA) and partial least squares structural equation modelling (PLS-SEM). The dataset contains 216 country-year observations for Bangladesh, China, India, Nepal, Pakistan, and Sri Lanka over 1990–2025, with 33 indicators grouped into climate vulnerability, governance quality, and geopolitical risk. Review the variable names, codes, definitions, units, source information, and construct classifications provided in the data files. Confirm that each row represents one country-year observation and each column represents one indicator. Screen for missing values and extreme observations. In the original study, univariate values were checked using |z| ≥ 3.29 and multivariate outliers using Mahalanobis distance at p = 0.001. Check the coding direction of all variables, especially ranks and indicators where higher values represent weaker performance or greater vulnerability. Standardize all indicators using z-scores. Conduct PCA separately for the three domains: climate vulnerability, governance quality, and geopolitical risk. Review eigenvalues, explained variance, and component loadings, then retain the relevant component score for each domain. Construct CEGEI from the standardized domain scores using weights based on the variance explained by the retained components. Estimate the PLS-SEM measurement model and assess Cronbach’s alpha, rho_A, composite reliability, average variance extracted, Fornell-Larcker values, and HTMT ratios. Estimate the structural paths: geopolitical risk → climate vulnerability; governance quality → climate vulnerability; and governance quality × geopolitical risk → climate vulnerability. Apply bootstrapping to obtain path coefficients, standard deviations, t statistics, and p values. Compare the reproduced results with the article. The reported coefficients are: geopolitical risk → climate vulnerability, β = 0.535, p = 0.003; governance quality → climate vulnerability, β = 0.074, p = 0.570; and interaction effect → climate vulnerability, β = 0.205, p = 0.001. Interpret the interaction according to the coding direction of governance and geopolitical-risk scores. The significant interaction confirms moderation, while its substantive direction should be assessed from the coding orientation and, where possible, simple-slope analysis.
Institutions
- Iqra UniversitySindh, Karachi