Abstract: As the main carrier of green and low-carbon transformation, improving environmental performance and promoting sustainable development by enterprises is an important manifestation of their active fulfillment of environmental and social responsibilities. Focusing on corporate environmental performance, this paper proposes a novel model that integrates self-evaluation and mutual evaluation. Based on the conventional concept of cross-efficiency, it constructs a mathematical programming model and an interactive iterative algorithm to jointly and continuously optimize environmental performance. By introducing a set of soft constraint variables to continuously promote the approximate Pareto improvement with minimal cost, it obtains the optimal weight combination through the interactive joint decision-making, thereby effectively enhancing the recognition of the evaluation results among all participants. The proposed interactive evaluation model is applied to an empirical study on the environmental performance of listed companies in China’s manufacturing industry from 2019 to 2022. The results show that the overall environmental performance of the sample enterprises is relatively satisfactory, with an average upward trend year by year, indicating that China’s manufacturing sector is beginning to make progress in its green and low-carbon transformation. However, there remain significant differences in environmental performance among enterprises across different years, with a noticeable trend of polarization, which warrants attention. To promote high-quality development across the entire industry, it is essential to strengthen both policy support and regulatory oversight, while establishing and promoting benchmark enterprises that exemplify green production practices.

Key words: corporate environmental performance, interactive evaluation, data envelopment analysis, undesirable output, manufacturing industry

摘要： 作为绿色低碳转型的主要载体，企业提升环境绩效、推动可持续发展，是其积极履行环境责任与社会责任的重要表现。本文聚焦企业环境绩效评价，提出一种融合自我评价与相互评价的新模式，在传统交叉效率模型基础上，构建了促进多方共同持续优化环境绩效的数学规划模型及交互迭代算法。通过引入软约束变量，模型能够以最小代价实现近似的帕累托改进，从而获取交互决策下的最优权重组合，有效提升所有参与成员对评价结果的认可度。将所提出的交互评价模型运用于2019~2022年中国制造业上市企业的环境绩效实证研究，结果显示：样本企业环境绩效整体表现良好，平均水平呈逐年上升趋势，反映出中国制造业绿色低碳转型已初见成效。然而，各年份间不同企业的环境绩效仍存在显著差异，两极分化现象值得关注。为此，建议在政策保障与监管强化的双重推动下，积极树立和推广标杆企业的绿色生产模式，助力全行业实现高质量发展。

关键词: 企业环境绩效, 交互评价, 数据包络分析, 非期望产出, 制造业