Objective/Meaning Taiwan-invested agricultural enterprises serve as a vital vehicle for the cross-strait agricultural exchange and cooperation, and their survival resilience directly affects the sustainable development of Taiwan Farmers’ Entrepreneurship Parks.

Methods/Procedures Based on the survey data from 214 Taiwan-invested agricultural enterprises in the national-level Taiwan Farmers’ Entrepreneurship Parks, an ordered Logit regression model was used to systematically analyze the factors influencing the survival resilience of these enterprises from the four dimensions: market perception, policy resources, industrial characteristics and social embeddedness.

Results/Conclusions The results showed that: (1) The perception of market demand exerted the strongest positive effect on the survival resilience of Taiwan-invested agricultural enterprises (β＝1.577, p＜0.001), verifying the core propositions of the signaling theory and market-based view; (2) The fiscal and financial support had a significantly positive impact on the survival resilience (β＝0.592, p＜0.01), whereas the service-oriented support showed a significantly negative effect (β＝−0.278, p＜0.05), presenting an adverse selection effect (enterprises with operational difficulties tended to seek more service-oriented policy support); (3) The industrial integration (agricultural product processing, leisure agriculture), the relatives’ and friends’ attitudes, and social participation had no significant impact on the survival resilience; (4) The social participation significantly positively moderated the relationship between market demand perception and enterprise survival resilience (β＝0.339, p＜0.05), indicating that the social networks could strengthen the enterprises’ capacity to interpret the market information. This study suggested that the market demand perception and fiscal-financial support were the key pathways to improve the survival resilience of Taiwan-invested agricultural enterprises, while the accuracy and effectiveness of service-oriented policies needed to be further optimized.