DOI: 10.36347/sjebm.2015.v02i02.001

Pages: 139-144

This study examines whether the consistency and the efficiency can be improved in estimating panel data models when the sample size is realistically finite. In various cases we compare the quasi- and the first-differencing approach which are designed to control for unit-specific effects. The equations transformed by the two approaches are estimated by instrumental variables. Empirical results from simulated data support that the quasi-differencing approach dominates the first-differencing one in estimating and testing panel data models, particularly for small-sized samples.