基于CFPS数据的反事实分析模型及其应用研究

Research on counterfactual analysis model based on CFPS Data and its applications

目的:对基于Newton-Cotes算法的反事实分析模型的不合理之处进行拓展和改进,以提升模型估算结果的准确性,实现更具现实依据的政策效果模拟和分析。方法:以北京大学中国家庭追踪调查(CFPS)数据为基础,在充分比较不同概率分布函数的拟合效果后,拟合我国家庭收入和医疗支出的概率分布曲线,继而构造由90000个异质性家庭组成的虚拟社群,根据各家庭的属性值估算出相关指标值,并举例说明了模型的应用场景。结果:拓展后的模型对收入和医疗支出期望值的估计结果以及各主要指标的变化趋势与原始模型基本一致,但对两者离散程度的估计更为精准,政策效果模拟和反事实分析结果更具准确度和严谨性。结论:拓展模型可以对原始模型起到一定的补充和借鉴作用,具备较高的应用价值和可操作性。

Objective:Expand and improve the unreasonable aspects of counterfactual analysis model based on the Newton-Cotes algorithm to enhance the accuracy of model estimation results and achieve more realistic simulation and counterfactual analysis.Methods:Based on the data of China Family Panel Studies(CFPS)from Peking University,after widely comparing the fitting effects of different probability distribution functions,the probability distribution curves of household income and medical expenditure in China are fitted.Then,a virtual community consisting of 90000 heterogeneous families was constructed,and relevant indicator values were estimated based on the attribute values of each family.Results:The estimation results of the extended model for household income and medical expenditure expectations,as well as the changing trends of various main indicators,are basically consistent with the original model,but the estimation of the degree of dispersion of the two is more accurate,and the estimation results of policy effect simulation and counterfactual analysis are more correct and rigorous.Conclusion:The extended model can serve as a supplement and reference to the original model,and has high application value and operability.