科研管理 ›› 2025, Vol. 46 ›› Issue (11): 128-139.DOI: 10.19571/j.cnki.1000-2995.2025.11.013

• 论文 • 上一篇    下一篇

基于文本挖掘的企业社会责任报告质量评价体系构建

骆嘉琪1,2,封垚2,孟斌1,2,贾丹丹1   

  1. 1.大连海事大学综合交通运输协同创新中心,辽宁 大连116026;
    2.大连海事大学航运经济与管理学院,辽宁 大连116026
  • 收稿日期:2024-10-16 修回日期:2025-08-26 接受日期:2025-08-29 出版日期:2025-11-20 发布日期:2025-11-07
  • 通讯作者: 孟斌
  • 基金资助:
    国家自然科学基金项目(72401183);上海市科委启明星计划扬帆专项(23YF1415000);中央高校基本科研业务费专项资金资助(3132024292, 3132025701)。

Construction of a quality evaluation system for corporate social responsibility reports based on text mining

Luo Jiaqi1,2, Feng Yao2, Meng Bin1,2, Jia Dandan1   

  1. 1. Collaborative Innovation Center for Transport Studies, Dalian Maritime University, Dalian 116026, Liaoning, China;
    2. School of Maritime Economics and Management, Dalian Maritime University, Dalian 116026, Liaoning, China
  • Received:2024-10-16 Revised:2025-08-26 Accepted:2025-08-29 Online:2025-11-20 Published:2025-11-07

摘要:    企业社会责任是商业文明的价值跃迁,CSR报告质量是企业铸就可持续竞争力的战略基石。本文以我国A股上市公司2021—2023年的4514份企业社会责任报告为研究对象,基于全球报告倡议组织发布的《可持续发展报告指南》与中国社科院发布的《中国企业社会责任报告指南》等权威标准,构建了包含完整性、回应性、可获取性、可信性、创新性及可比性6个一级指标和18个二级指标的企业社会责任报告质量评价指标体系,通过文本挖掘中的文本情感分析、词频统计等技术对企业社会责任报告质量各维度进行自动评分,采用TOPSIS熵权法确定指标权重并分维度分行业探析。研究发现:(1)2021—2023年企业社会责任报告质量显著提升,行业整体实践呈现积极发展态势;(2)企业在报告“可获取性”与“回应性”上表现突出,但“创新性”与“可比性”仍是普遍短板;(3)行业间社会责任报告质量差异显著,金融业全面领先,娱乐行业亟待提升。本研究有助于高效客观量化企业社会责任报告质量,为政府制定相关政策、企业优化报告编制与提升社会责任实践能力以及行业整体CSR水平的提升提供了有益参考。

关键词: 企业社会责任, 报告质量, 质量评价模型, 文本挖掘

Abstract:    Corporate social responsibility (CSR) represents a transformative evolution in the values of commercial civilization, and the quality of CSR reports serves as the strategic cornerstone for companies to build sustainable competitiveness. By analyzing 4514 CSR reports from China′s A-share listed companies between 2021 and 2023, and based on authoritative frameworks including the Sustainability Reporting Guidelines issued by the Global Reporting Initiative (GRI) and the China CSR Reporting Guidelines developed by the Chinese Academy of Social Sciences (CASS), we constructed a CSR quality evaluation system which comprises six first-level indicators (completeness, responsiveness, accessibility, credibility, innovation, and comparability) and 18 second-level indicators. Moreover, by leveraging text mining techniques—specifically, text sentiment analysis and word frequency statistics, we automatically scored CSR report quality across dimensions. Then, the TOPSIS-entropy weight method was employed to determine indicator weights, followed by dimension- and industry-specific analyses. The key findings include: (1) CSR report quality improved significantly from 2021 to 2023, with industry-wide practices demonstrating positive growth; (2) firms excelled in report accessibility and responsiveness, while innovation and comparability remained widespread weaknesses; and (3) inter-industry disparities in CSR report quality were pronounced, with the financial sector leading comprehensively and the entertainment industry requiring urgent improvement. This research has facilitated efficient, objective quantification of CSR report quality, and it will offer valuable references for government policymaking, corporate optimization of report preparation, enhancement of CSR implementation capabilities, and elevation of industry-wide CSR standards.

Key words: corporate social responsibility, report quality, quality evaluation model, text mining