Drivers of Intention to Expand HDPE-Lined Shrimp Ponds: A Hybrid PLS-SEM and XGBoost-SHAP Approach

Abstract

Decisions to expand an established production technology differ from decisions to adopt it initially because further deployment requires additional capital, land, and management capacity. This study investigates the intention of experienced shrimp-farming households to expand high-density polyethylene-lined pond area in post-merger Can Tho City, Vietnam. Cross-sectional survey data from 356 current users were examined using partial least squares structural equation modeling, followed by extreme gradient boosting and Shapley additive explanations to assess out-of-sample prediction and model-based feature contributions. Perceived value showed the strongest positive association with expansion intention (standardized coefficient = 0.568). Expected economic benefits, social media influence, and perceived climate and disease threats were positively associated with perceived value, whereas perceived investment and operating cost burden was negatively associated with it. Contrary to the proposed hypothesis, technological readiness weakened the association between perceived value and expansion intention (interaction coefficient = −0.191). A theory-informed linear benchmark incorporating this interaction performed better than the tuned boosting model on the holdout sample (predictive explained variance of 0.283 versus 0.163). This finding suggests that model complexity did not improve predictive accuracy in the present dataset. By focusing on households with operational experience, the study identifies a boundary condition in translating favorable technology evaluations into expansion intentions. The analysis distinguishes explanatory associations from predictive patterns, although its cross-sectional, geographically concentrated sample does not support causal or nationwide claims.

GS Cited Analysis