Causal analysis of randomized controlled experiments applied in social science--Taking the selection effect of female reproductive behavior as an example
- 分享到:
摘要：
清楚辨析变量间因果关系是社会科学研究的难点,基于观测数据传统的统计推断方法存在着混淆变量等内生性问题,无法满足“无偏性”和“一致性”估计.提出在研究设计中可利用自然科学随机试验思路控制所有混淆变量,以满足因果分析中的必然性.即在社会现象是多因、随机和未知的前提下,引入随机性自然实验达到“总体相同”的必然性和在社会现象是已知、清晰和可控的前提下,通过倾向值匹配法构造实验组和对照组,控制混淆变量进行因果分析.最后,以女性生育行为的选择效应为案例,详解自然实验法在控制女性多孩生育事件对收入影响分析中混淆变量的应用、倾向值匹配法在控制女性劳动参与对生育行为影响分析中混淆变量的应用.
Clearly identifying the causal relationship between variables is a difficult point in social science research.Traditional statistical inference methods based on observational data have endogenous problems such as missing variables and bidirectional causality,which cannot satisfy'unbiased'and'consistent'estimates.This paper proposes to control all confounding variables by random experiment method to satisfy the inevitability in causal analysis.Specifically,under the premise that social phenomena are multi-causal,random and unknown,the inevitability of introducing random natural experiments to achieve'the same overall'.On the premise that the social phenomenon is known,clear and controllable,the experimental group and the control group are constructed by the propensity value matching method,and the confounding variables are controlled for causal analysis.Finally,taking female reproductive punishment effect and professional female reproductive behavior selection effect as examples,the specific application of the two research designs is explained in detail.
作者：
薛君 魏雷东
Xue Jun;Wei Leidong(Social Work and Social Governance Soft Science Research Base in Henan Provincial,Henan Normal University,Xinxiang 453007,China)
机构地区：
河南师范大学河南省社会工作与社会治理软科学研究基地
出处：
《河南师范大学学报:自然科学版》 CAS 北大核心 2023年第6期65-71,共7页
Journal of Henan Normal University(Natural Science Edition)
基金：
国家社科基金(21BRK038) 河南省高校科技创新人才支持计划(2021-CX-058).
关键词：
因果分析 自然实验 倾向值匹配 惩罚效应 选择效应
causal analysis natural experiment propensity score matching motherhood wage penalty selection effect
分类号：
C923 [社会学—人口学]
