We use the quasi-likelihood ratio statistic to test the hypothesis that a power transformation is ineffective. The null is not standard because a parameter is unidentified, and we can also let the power coefficient be zero for the null.

This note derives the correct limit distributions of the Anderson–Hsiao (1981) levels and differences instrumental variable estimators, provides comparisons showing that the levels IV estimator has uniformly smaller variance asymptotically as the cross section (n) and time series (T) sample...