Functional Heterogeneity and Technological Spillovers in Provincial Banana Yields in Ecuador, 2021–2025

Ecuador is the world’s leading banana exporter, yet substantial productivity disparities persist across its provinces. This study applies Functional Data Analysis (FDA) to examine provincial banana yield dynamics during 2021–2025. Using 3034 producer-level observations aggregated into 90 province-year observations across 23 provinces, we estimate smoothed yield trajectories using B-splines and apply Functional Principal Component Analysis (FPCA), functional clustering, function-on-scalar regression (FOSR), and hierarchical functional spatial autoregressive (SAR) models. The first functional principal component explains a substantial proportion of the total variability (ranging from 91.0% to 98.4% depending on the smoothing parameter, with a baseline estimate of 93.3%), indicating that provincial differences primarily reflect yield levels rather than temporal trajectory shapes. Functional clustering identifies three distinct productivity groups: high-yielding (four provinces, 39.6 t/ha), intermediate-yielding (four provinces, 22.8 t/ha), and low-yielding provinces (12 provinces, 5.6 t/ha). Irrigation and producer education are positively associated with yields (both p < 0.001). The hierarchical SAR specification identifies significant dependence associated with Guayas, Ecuador’s principal banana export hub (p = 0.024), whereas geographic contiguity is not statistically significant (p = 0.53). These findings are consistent with the potential importance of export-oriented value chains in technological diffusion, although the underlying transmission mechanisms require further verification. The results highlight irrigation investment, producer education, and stronger productive linkages as priorities for reducing territorial productivity disparities. We hypothesize that: (H1) interprovincial heterogeneity is primarily driven by yield levels rather than temporal shapes; (H2) irrigation and education are the main internal determinants; and (H3) spatial dependence is driven by value-chain hubs rather than geographic proximity.