中等收入群体的持续期与退出风险估计——基于EM算法的收入群体划分

Estimation of Duration and Exit Hazard of Middle-income Group——Based on the EM Algorithm for Division of Income Groups

本文在利用EM算法求解有限混合分布模型以识别CFPS数据库中各年度中等收入群体的基础上,使用生存分析方法估计了中等收入群体的持续期与退出风险。两种情境下的估计结果显示,20%以上的中等收入家庭在样本期内表现出了脆弱性,这意味着传统的仅考虑一期家庭收入状态的中等收入群体比例测算可能高估了真实的中等收入群体比例。此外,家庭面临倒U型退出风险,持续期与退出风险存在显著的区域与城乡间差异:东部地区与城市地区的中等收入家庭表现出更强的收入稳固性。本文还进一步分析了家庭退出中等收入群体的决定因素,估计结果表明,家庭收入结构、家庭人口特征、家庭职业特征等因素对家庭退出风险影响显著。

Based on the EM algorithm which can solve the finite mixed model to identify the middleincome group in the CFPS database,this paper uses the survival analysis method to estimate the duration and exit hazard of the middle-income group.In two scenarios,the estimated results indicate that more than 20%of middle-income households exhibit vulnerability in the sample period,which means that the traditional measurement which merely considers a one-year income level may overestimate the proportion of the middleincome group.Besides,families face inverted U-shaped exit hazards,and the duration and exit hazard show significant regional and urban-rural differences:middle-income households in the eastern and urban areas of China show stronger income stability.This paper further analyzes the determinants of households’exit from the middle-income group,and the estimated results show that family income structure,family demographic characteristics,and family occupational characteristics have a significant impact on households’exit hazards.