TWO-STAGE QUANTILE REGRESSION FOR DYNAMIC PANEL DATA MODELS WITH FIXED EFFECTS: MONTE CARLO SIMULATION STUDY

The problem of estimators' biasedness arises while estimating dynamic panel data (DPD) models by quantile regression (QR) due to the inclusion of lagged value(s) for dependent variable in DPD models. A new framework for handling the bias is theoretically proposed through the method of two-stage quantile regression for dynamic panel data (TSQRDPD). This proposed version of TSQRDPD relies on estimating both stages by QR at the same rank of quantile. In addition to that, bias performance is evaluated through a Monte Carlo simulation under several quantile ranks for stages, probability distributions and panels' sizes. Monte Carlo simulation results show that the new method is able to minimize estimators' biasedness severely in comparison to other estimation methods of quantile regression.