社会学定量分析中的内生性问题测估社会互动的因果效应研究综述_陈云松社会学定量分析中的内生性问题测估社会互动的因果效应研究综述_陈云松.pdf

社会学定量分析中的内生性问题 2010# 4 CJS 30 陈云松 范晓光 : 因果关系是社会学分析的核心议题然而, 基于调查数据的定量分 析都会受到内生性 题的困扰, 从而往往难以作出恰当的因果性推断大量 现有的社会学定量分析虽以解释为己任, 但实质上要么仅仅停留在描述统计 相关阶段, 要么得出错误的因果性结论本文以社会互动研究为例, 详细讨 论了遗漏偏误自选择偏误样本选择偏误和联立性偏误等四种主要的内生 性 题的来源, 并介绍了克服内生性 题的 一系列模型识别方法最后还基 于中国综合调查( CGSS2003) 卷, 提出了通过提高调查数据信息量以利克 服内生性 题的可能性 : 内生性 因果效应 社会互动 Th e Endogen eity P rob lem in Quantitative Analysis: A R eview of Estimating Cau sal Effects of Social In teraction Chen Yunsong FanX iaoguang Abstract: Cause-effect relationsh ips are the core area in sociological analysis. H ow e er, sociological analys is based on sur ey data is confronted by the endogeneity prob lem wh ich p lagues causal inferences. M any existing stud ies ami ing at pro id ing explanations for social phenom ena either merely describes the statistical associations am ong ariables or arri es at prob lematic causal conclusions. Focusing on the social interaction stud ies, this paper addresses the m ajor sources of potential endogeneity b iases, nam ely, the om itted ariab le bias, self-selection b ias, samp le-selection b ias and the smi ultaneity b ias. U seful m odel * : ( Author 1: Chen Yun song, Nuffield College, Oxford Un i ersity) E-m ai:l yunsong. chen@ nuffield. ox. ac. uk; 2: , ( Author 2: Fan X iaoguang, Institute of Sociology, Zhejiang Academy of Social Sciences) ** Tom Sn ijdersPeterH edstrÊmNan D irk deG raaf 赟和郭茂灿等同对本文的批评和建议此外, 还要感谢5社会6匿名 # 9 1# # 2010# 4 iden tification strategies for correcting these prob lem s are re iew ed. Based on CGSS2003, th is paper also d iscusses how to partially correct for the endogeneity