ISSN 1671-3710
CN 11-4766/R
主办:中国科学院心理研究所
出版:科学出版社

• •    

组织行为元分析研究中的内生性问题:应对方法与应用建议

卫旭华, 龙宇情, 于璟婷   

  1. 兰州大学管理学院, 甘肃 730000 中国
  • 收稿日期:2026-08-15 修回日期:2026-11-05 接受日期:2026-11-12
  • 基金资助:
    国家自然科学基金面上项目(72372063); 甘肃省哲学社会科学规划项目(2023YB013); 甘肃省研究生教育教学改革研究项目(YB030); 中央高校基本科研业务费专项资金战略发展专项项目(2023jbkyzx005)

Endogeneity in organizational behavior meta-analytic research: Remedies and practical recommendations

wei xuhua, long yuqing, yu jingting   

  1. School of Management, Lanzhou University 730000, China
  • Received:2026-08-15 Revised:2026-11-05 Accepted:2026-11-12

摘要: 内生性问题会导致参数估计偏误,并进一步由实证研究传染至基于其开展二次分析的元分析研究,威胁结论的准确性。尽管该问题在实证研究中已受到广泛关注与讨论,但在元分析研究中仍缺乏足够重视。鉴于此,本研究首先回顾了内生性问题的基本内涵,探讨了元分析研究中内生性问题的产生原因及常见来源。随后,系统梳理了检验和修正元分析研究中内生性问题的11种应对方法,介绍了各方法的基本原理、实施步骤及其在组织行为领域的应用现状。最后,总结了元分析研究中内生性问题检验与修正的流程框架及未来展望。本研究有助于为研究者系统检验与修正元分析内生性问题提供方法指导,从而进一步提升元分析研究设计的严谨性与研究结论的准确性。

关键词: 组织行为;元分析;内生性;工具变量;Heckman两阶段模型;交叉滞后模型;元分析结构方程模型

Abstract: Endogeneity biases parameter estimates in empirical studies and transfers to subsequent meta-analyses that rely on them, threatening the accuracy of meta-analytic conclusions. Although endogeneity has been widely discussed in primary empirical research, it remains insufficiently addressed in meta-analytic research. To address this gap, this paper first reviews the fundamental concept of endogeneity and examines its causes and typical sources in meta-analytic research. It then systematically synthesizes eleven approaches to detecting and correcting endogeneity in meta-analytic research, outlining their basic principles, implementation procedures, and current applications in organizational behavior research. Finally, it proposes a procedural framework for detecting and correcting endogeneity in meta-analysis and discusses directions for future research. This paper provides methodological guidance to help researchers identify and address endogeneity in meta-analysis, thereby strengthening the rigor and accuracy of meta-analytic conclusions.

Key words: organizational behavior, meta-analysis, endogeneity, instrumental variables, Heckman modeling (2-stage), cross-lagged panel model, meta-analytic structural equation model