Unconditional Quantile Regression For Panel Data And Its Application

On the basis of controlling individual heterogeneity,conditional quantile regression for panel data can fully describe the relationship between variables at different quantiles of the explained variable’s conditional distribution.Due to the advantages of both panel data model and conditional quantile regression,conditional quantile regression for panel data is widely used in econometric modeling research.However,the unconditional quantile regression technique is needed to obtain the general marginal effect of the explanatory variable on the explained variable directly,without relying on other covariates.Unconditional quantile regression technique mainly includes the regression based on recentered influence function and the general unconditional quantile regression.Therefore,this paper is devoted to constructing unconditional quantile regression for panel data,and provides a new idea for quantile regression.For recentered influence function regression,unconditional quantile regression model based on first order difference,fixed effect transformation,dummy variable and penalty regression are constructed.The methods of point estimation and interval estimation are given.The simulation results show that conditional and unconditional quantile regression methods are consistent only when no other covariates exist.When other covariates exist,conditional quantile regression is based on all other covariates,while unconditional quantile regression can directly obtain the general marginal impact on a specific quantile of the response variable.The four unconditional quantile regression methods for panel data can obtain accurate and effective estimation,among which the estimation of penalized unconditional quantile regression is the best.Finally,this paper studies the impact of sub-income on consumption for urban residents.The study found that wage income has the strongest stimulatory effect on consumption,and has a greater stimulatory effect on middle and low consumers.For general unconditional quantile regression model of panel data,a general unconditional quantile regression method is constructed based on two moment conditions.The parameters is obtained by using generalized moment estimation.In addition,a method of estimating the confidence interval is given.The simulation results show that general unconditional quantile regression method is optimal under different error distribution,data volume and quantiles.However,its process is a little complicated.Because the optimization algorithm needs to be designed according to different situations.Finally,this paper analyzes the overall income and consumption of urban residents.The study found that the improvement of the income level of the middle and low consumers can stimulate consumption.However,the further increase of the income level of the highconsumption crowd will not obviously promote their consumption.