Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5058896 | Economics Letters | 2014 | 4 Pages |
â¢In dynamic panel estimation, constants are often omitted from the instrument sets.â¢We examine the effect of this omittance when the instrument means are large.â¢We show that the properties of the estimator depend on the instrument means.â¢We show that existing estimators can be more biased and much less efficient.â¢We provide simple solutions for practitioners.
In the standard generalized method of moments estimation of dynamic panel data models, the constant term is usually omitted from instrument sets. As a result, adding a constant to the dependent variable affects the estimates for models without full period dummies. Omitting the constant term from instrument sets may also result in substantial bias and efficiency loss if the mean of the variable is large in magnitude. In this note, we provide analytical and numerical results and propose convenient solutions for practitioners. We suggest that full period dummies be included as extra exogenous instruments even for models without time effects on the right-hand side.