The implementation of China’s rural homestead withdrawal policy has been a topic of great interest in recent years. However, little is known about its impact on farmers’ homestead withdrawal behavior from the perspective of psychological resilience. This study aims to evaluate the impact of psychological resilience on farmers’ homestead withdrawal behavior and to explore its mechanism and heterogeneity, as well as its impact on compensation choices. The research data was collected through a survey of 657 rural households in Shaanxi and Shandong provinces. The probit model and mediation effect model were used to analyze the data. The results demonstrate that psychological resilience positively affects farmers’ homestead withdrawal behavior; this finding is robust. Specifically, the positive influence of stability and adaptability dimensions of psychological resilience is more significant. Moreover, the effect of psychological resilience varies across different family life cycles, and it has a stronger promoting effect on the homestead withdrawal behavior of child-raising families. Mechanism analysis indicates that psychological resilience can promote homestead withdrawal behavior by reducing farmers’ land dependence consciousness and improving farmers’ risk preferences. Finally, farmers with strong psychological resilience are more inclined to choose cash compensation. Based on these findings, this paper proposes targeted policy recommendations for rural homestead system reform.

This study examines the determinants of farmers’ climate adaptation strategies across agro-climatic zones, with particular focus on education, household characteristics, and income. A multi-stage random sample of 200 farmers was analyzed using a stochastic frontier model to assess production efficiency and a logit model to evaluate adaptation choices. Results reveal substantial heterogeneity in productivity, with mean 0.619 technical efficiency scores ranging from 0.378 to 0.961. The variance parameter (γ = 0.723) indicates that most performance differences arise from inefficiency rather than random shocks. Among production inputs, tractor use (0.182) and urea application (0.160) significantly increase yield, whereas cultivated area is negatively associated with output (− 0.943), reflecting land fragmentation and managerial constraints. Labor and DAP were not significant contributors. In the logit model (Pseudo-R² = 0.528), farmers’ age and education significantly increase the likelihood of adopting adaptation measures, while larger household size reduces it; agricultural income shows a modest positive effect. These findings suggest that productivity gains are more likely to come from intensification through improved timing, mechanization services, and nutrient management rather than from expanding cultivated area. Policy priorities should focus on three areas: (i) knowledge and extension support, including zone-specific advisories and training; (ii) improved access to mechanization and soil and water management technologies; and (iii) financial tools such as credit and insurance to reduce adoption barriers. Strengthening education and skills, particularly for labor-constrained households, can accelerate the uptake of effective practices, narrow the efficiency gap, and enhance the resilience of maize systems under a warming climate.