固定效应面板数据的无条件分位数回归方法及其应用

Unconditional Quantile Regression for Panel Data with Fixed Effects and Its Applicatio

针对含固定效应的面板数据,讨论一般化的无条件分位数回归建模问题。基于两个矩条件,得到面板数据无条件分位数回归的点估计,并通过Bootstrap重抽样技术进一步给出置信区间估计办法。其次,通过计算机蒙特卡洛模拟,详细比较无条件分位数回归估计与条件分位数回归估计的效果。研究结果表明,在数据量、误差项分布、估计参数真实值的不同情况下,UQR的估计偏差和均方根误差都很小,UQR是固定效应的面板数据的有效估计办法。当样本量增加或者估计参数真实值数量级增大的情况下,UQR估计会更有效。UQR估计在0.5分位点的估计效果最佳,低分位点的估计效果优于高分位点的估计效果。最后,根据各省市的消费收入数据进行了实证研究,发现UQR能更好地解释实际的消费情况。

The general unconditional quantile regression model for panel data with fixed effects is discussed.Firstly,based on two moment conditions,the point estimation of unconditional quantile regression is obtained,and the confidence interval estimation method is farther givec by Bootstrap resampling technique.Secondly,through Monte Carlo simulation,the effects of unconditional quantile regression and conditional quantile regression are compared in detail.The results show that UQR has smaller bias and root-mean-square error under different conditions of data size,the distribution of error terms and the real value of estimated parameters.UQR is an effective method for panel data with fixed effects.When the sample size increases or the true value of the estimated parameters increases,UQR estimation is more effective.UQR has the best effect at quantile 0.5,and the estimation effect of low quantile is better than that of high quantile.Finally,according to the consumption and income data of provinces and cities,the empirical study finds that UQR can better explain the actual consumption situation.