Bootstrap bias-correction procedure in estimating long-run relationships from dynamic panels, with an application to money demand in the euro area

In dynamic panel data models, which are particularly well-suited to cross-country analysis, the Mean Group estimator (Pesaran and Smith, 1995) is under certain quite strong conditions consistent, but theoretical and empirical evidence indicates that it can be biased when the number of time observations is small. Possible explanations are sample-size bias and omitted variables or measurement errors that are correlated with the regressors. I find support for both hypotheses using a Monte Carlo experiment which analyzes cointegrated systems. A possible solution for the MG estimator bias is a bootstrap bias-correction procedure, but Pesaran and Zhao (1999) show that it performs well only when the true coefficient of the lagged dependent variable is small. In this paper, I test three different bootstrap procedures and obtain an appreciable reduction in the MG estimator bias, especially when the suggestions of Li and Maddala (1997) are applied. Finally, I use bootstrap bias-corrected estimators to investigate the long-run properties of money demand in the euro area.