The Effects of Dynamic Feedbacks on LS and MM Estimator Accuracy in Panel Data Models

Abstract

The Þnite sample behaviour is analysed of particular least squares (LS) and a range of (generalized) method of moments (MM) estimators in panel data models with individual effects and both a lagged dependent variable regressor and another explanatory variable. The latter may be affected by lagged feedbacks from the dependent variable too. Asymptotic expansions indicate how the order of magnitude of bias of MM estimators tends to increase with the number of moment conditions exploited. They also provide analytic evidence on how the bias of the various estimators depends on the feedbacks and on other model characteristics such as prominence of individual effects and correlation between observed and unobserved heterogeneity. Simulation results corroborate the theoretical Þndings and reveal that in small samples of models with dynamic feedbacks none of the techniques examined dominates regarding bias and mean squared error over all parametrizations examined.

Bootstrap-based bias correction for dynamic panels

“…12 For the GMM estimators this conclusion supports previous findings by Blundell and Bond (1998) and Bun and Kiviet (2006). estimator for b is much less biased (see Kiviet, 1995).…”

“…Second, IV and GMM estimators require additional decisions on which instruments to use. For instance, when T is relatively large compared to N more moment conditions are available but the GMM estimators may have substantial small sample biases when too many of these conditions are used (see Ziliak, 1997;Bun and Kiviet, 2006).…”

On the effect of mean-nonstationarity in dynamic panel data models

“…We show that the same applies to the dynamic panel data models considered here. To see this, note that using an asymptotic expansion, we have Bun and Kiviet, 2006). With regard to the numerator, Alvarez and Arellano (2003) show that …”

Indirect inference for dynamic panel models

“…In such cases, the GMM/IV estimator of the autoregressive parameter can suffer from substantial bias and large variation. In other circumstances, when the number of moment conditions becomes large, the GMM/IV estimator also suffers from large finite sample bias (Bun and Kiviet, 2006;Ziliak, 1997). Finally, the GMM/IV estimator is designed for linear dynamic systems and is not readily applicable to nonlinear models.…”