This paper by Chihwa Kao and Min-Hsien Chiang, published in 1999, focuses on the estimation and inference of cointegrated regression in panel data. The authors introduce a dynamic OLS (DOLS) estimator and compare its finite sample properties with those of the ordinary least squares (OLS) and fully modified OLS (FMOLS) estimators. The key findings are: 1. **OLS Estimator**: The OLS estimator has a non-negligible bias in finite samples. 2. **FMOLS Estimator**: The FMOLS estimator does not generally improve over the OLS estimator. 3. **DOLS Estimator**: The DOLS outperforms both the OLS and FMOLS estimators. The paper provides a detailed asymptotic theory for these estimators and Monte Carlo simulations to illustrate their finite sample properties. The simulations show that the DOLS estimator consistently outperforms the OLS and FMOLS estimators in terms of bias and mean squared error, especially when the panel is heterogeneous. The DOLS estimator is also shown to have a well approximated limiting distribution, making it a reliable choice for cointegrated regression in panel data.This paper by Chihwa Kao and Min-Hsien Chiang, published in 1999, focuses on the estimation and inference of cointegrated regression in panel data. The authors introduce a dynamic OLS (DOLS) estimator and compare its finite sample properties with those of the ordinary least squares (OLS) and fully modified OLS (FMOLS) estimators. The key findings are: 1. **OLS Estimator**: The OLS estimator has a non-negligible bias in finite samples. 2. **FMOLS Estimator**: The FMOLS estimator does not generally improve over the OLS estimator. 3. **DOLS Estimator**: The DOLS outperforms both the OLS and FMOLS estimators. The paper provides a detailed asymptotic theory for these estimators and Monte Carlo simulations to illustrate their finite sample properties. The simulations show that the DOLS estimator consistently outperforms the OLS and FMOLS estimators in terms of bias and mean squared error, especially when the panel is heterogeneous. The DOLS estimator is also shown to have a well approximated limiting distribution, making it a reliable choice for cointegrated regression in panel data.