Impact of climate change risk on sustainable crop production in China: an EBM-DEA model incorporating an exogenous variable

As agriculture continues its green, low-carbon transformation, evaluating the efficiency of sustainable crop production in China under climate risk is increasingly important. Using 2019–2023 panel data for 31 Chinese provinces, this study incorporates the Climate Physical Risk Index (CPRI) as an exogenous variable into a parallel two-stage Epsilon-Based Measure–Data Envelopment Analysis (EBM-DEA) model to assess sustainable crop production efficiency and examine regional disparities and dynamic evolution. The results show that the national mean overall efficiency declines from 0.7518 to 0.7420 after CPRI is incorporated, indicating that ignoring climate risk may modestly overestimate efficiency. Overall efficiency is highest in central China, followed by eastern and western China; this regional ranking remains broadly stable after CPRI is incorporated, although the efficiency scores and relative positions of some provinces change. At the stage level, Crop Production Efficiency (CPE) is markedly higher than Sustainable Development Efficiency (SDE), while at the substage level, Green, Low-Carbon Efficiency (GLE) is lower than Livelihood Improvement Efficiency (LIE), identifying the sustainable development stage – particularly the green, low-carbon stage – as the principal constraint on further efficiency improvement. Overall efficiency disparities narrow after 2020; inter-regional disparities contribute the most, although the contribution of transvariation density increases in later years. Markov analysis indicates that high- and low-efficiency categories exhibit state dependence, transitions occur mainly between adjacent categories, and stable convergence has not emerged. These findings support incorporating climate risk management into crop production efficiency assessments, shifting policy priorities from production expansion to green, low-carbon transformation, and adopting region-specific measures.