Two Kinds Of Quantile Regression And Empirical Research

When studying some practical problems,people are usually concerned with the effect of an explanatory variable on the outcome variable,ie,partial effect or marginal effect.OLS regression is a common way to solve this type of problem.However,OLS regression is actually a kind of conditional mean regression.The conditional mean is only an index of the conditional distribution,and it's partial effect is the conditional average partial effect.The quantile regression methods can describe the entire distribution and they can obtain the partial effects on each quantile.The traditional quantile regression is actually the conditional quantile regression,the partial effect it receives is the conditional quantile partial effect.Another kind of quantile regression is unconditional quantile regression,with the natures of influence function and recentered influence function,it can obtain the unconditional quantile partial effect.The quantile regression methods provide more information than OLS regression.This paper tells the differences of partial effects between OLS method and each quantile method through data simulation.This paper uses the OLS as a reference and also uses two kinds of quantile regression to conduct an empirical research on income inequality of gender.It is found that gender has a significant effect on income.There are big differences about various partial effects,and several estimation methods of unconditional quantile partial effect have similar results,it shows the robustness of the estimation methods.At the same time,based on the unconditional quantile regression,this paper uses the counterfactual framework to break down the income differentials of gender,it can be found that gender discrimination plays a dominant role in the income differentials of gender.Due to the significant advantages of the quantile regression methods compared to OLS regression mothod,especially in the case of heteroscedasticity,the quantile regression methods have attracted more and more attention from researchers,in the era of data explosion,they can play a big role.