Mining Externalities and Climate-Smart Agriculture Portfolios: Evidence from Mining-Affected Communities in Ghana

This study examines how mining-related environmental pressure is associated with the breadth and composition of climate-smart agriculture (CSA) portfolios among 485 farming households in Tarkwa Nsuaem and Prestea Huni Valley, Ghana. We develop an adaptation-paradox framework in which environmental stress may strengthen incentives to adopt risk-mitigating practices while severe degradation may simultaneously reduce expected returns and implementation capacity. Because CSA adoption is almost universal in the sample, the analysis focuses on portfolio breadth, complete seven-practice adoption, and practice composition rather than a nearly degenerate adopter/non-adopter outcome. The empirical strategy combines fractional-response estimation, stacked practice-level logit models, nonlinear specifications, alternative exposure measures, small-cluster robustness checks, Double Machine Learning, and repeated nested cross-validation. In the fully adjusted fractional-logit model with community fixed effects, the average marginal association between the core Mining Externality Severity Index (MESI) and CSA portfolio share is close to zero and imprecisely estimated (AME = -0.003; 95% CI: -0.175 to 0.169). However, MESI relationships differ jointly across the seven practices, nonlinear specifications suggest possible non-monotonicity, and results vary with the definition of mining pressure. Gradient Boosting predicts complete portfolios with a mean ROC-AUC of 0.907, while community context contains substantially more predictive information than MESI alone. Overall, the evidence supports a portfolio-based, practice-specific, and place-sensitive interpretation of adaptation rather than a universal positive or negative relationship between mining pressure and CSA adoption.