Farmers’ risk attitudes, borrowing behavior, and agricultural production decisions: evidence from southwestern China

Understanding how farmers’ risk attitudes drive their borrowing behavior and agricultural production decisions is of significant importance for promoting rural financial innovation and sustainable agricultural development. Based on experimental and survey data from 447 agricultural operators in China (Sichuan, Chongqing, and Guizhou provinces), this study employs the “Bomb Risk Elicitation Task” to objectively quantify risk attitudes, innovatively utilizing a 3D-tensor-based neural network as a methodological cross-validation tool. To account for unobservable sample selection bias, we employ joint estimation models to systematically evaluate the impact of risk tolerance across the complete credit decision chain. The findings reveal a baseline of profound risk aversion among farmers. However, individuals with higher risk tolerance exhibit significantly stronger borrowing intentions and secure larger loan amounts. Marginal effects analysis demonstrates structural demographic heterogeneity: this risk-driven borrowing is highly pronounced among younger farmers and smaller households, whereas older and larger households exhibit attenuated responsiveness to risk attitudes. Furthermore, conditional on borrowing, higher risk tolerance drives a significant preference for formal credit channels and productive agricultural investments, ultimately leading to improved post-loan income levels. In contrast, farmers with lower risk tolerance show a preference for conservative management strategies and greater reliance on informal lending channels.