Climatic and Productive Correlates of Coffee Yield in Latin America: Dynamic Adjustment and Long-Run Heterogeneity

Coffee productivity in Latin America is exposed to climatic variability, input-use pressures, and pronounced cross-country heterogeneity. This study evaluates dynamic associations between coffee yield and climatic, productive, and economic conditions in twelve major producing countries from 1992 to 2023 using an ARDL(1,1) framework with Pooled Mean Group (PMG) and Mean Group (MG) estimators. Both models yield negative and statistically significant adjustment coefficients (−0.866 and −1.087, respectively; p < 0.001). Because coffee yield, the constructed NPK-use/coffee-area proxy, and the temperature anomaly are stationary in levels, these parameters are interpreted as dynamic mean reversion within the ECM parameterization rather than as standalone evidence of cointegration; the Westerlund statistics are therefore treated as supplementary diagnostics. A Swamy-type test rejects common long-run slopes (chi-square = 101.62, df = 55, p < 0.001), while none of the five long-run covariates is significant at the 5% level in either central model. A targeted denominator-free sensitivity check replacing the area-normalized proxy with national NPK agricultural use in tonnes of product also finds no statistically significant fertilizer association. The results support country-sensitive adaptation and productivity strategies and caution against causal farm-level interpretations of national macro-panel associations.