The ability to correct the bias in the stable AD(1,1) model with a feedback effect

The behavior of a bias-corrected (BC) estimator assuming strongly exogenous regressors is compared to the behavior of a BC estimator assuming weakly exogenous regressors, when in fact the marginal model contains a feedback mechanism. To this end, the effect of a feedback mechanism on the first-order least-squares coefficient estimation bias is examined through large-sample asymptotics in the stable AD(1,1) model. In the simulations, it appears that the valid BC estimator based on the whole system is less biased than the invalid BC estimator based on the conditional model only. Substantial efficiency gains are possible for parts of the parameter space that determine the feedback effect, although the picture is less clear when results are averaged over the whole parameter space.