Some empirical studies using Panel Quantile Regression have tested stationarity, and some have not. Is it necessary to generate unit root tests? And if the variables are I(1), should quantile regression be performed using the first difference, or the level??