劳动收入风险测算及其收入分配效应研究——来自机器学习的实证发现

Research on the Measurement of Labor Income Risk and Its Income Distribution Effect:Empirical Findings from Machine Learning

本文以流动人口为研究对象,利用机器学习模型复原劳动者的收入分布,分别使用方差风险刻画收入整体波动性,用偏度风险描述收入增长空间以及峰度风险评估极端收入可能性。研究发现,第一,流动人口承受的劳动收入风险具有显著的时变规律,在人口属性和工作特征维度的分布差距明显。第二,风险是决定工资收入的重要因素,风险补偿在不同时期、不同收入群体和不同风险类别上展现出明显的异质性。第三,风险产生了显著的收入分配效应,总体上拉大了收入差距。方差和偏度风险放大了收入差距,而峰度风险则有助于缩小收入差距。研究结论对于及时化解极端收入风险,稳定收入预期提供了有益的参考。

Taking the migrant population as the research object,this paper uses a machine learning model to restore the income distribution of workers.Variance risk is used to describe the overall income volatility,skewness risk is used to describe the income growth space,and kurtosis risk is used to assess the possibility of extreme income.The research finds that first,the labor income risk borne by the migrant population has a significant time-varying feature,and the distribution gap is obvious in the dimensions of population attributes and work characteristics.Second,risk is an important factor in determining wage income,and risk compensation shows obvious heterogeneity in different periods,income groups and risk categories.Third,risk has a significant income distribution effect,widening the income gap in general.Variance and skewness risks amplify income inequality,while kurtosis risk helps to narrow the income gap.The research findings provide useful references for timely mitigation of extreme income risks and stabilization of income expectations.