Diff-GMM vce (cluster clustervariable)

[Raphael George]

In Diff GMM if I use cluster-robust standard errors, then why would the instruments yi,t−2, yi,t−3, . . . would become invalid and the GMM estimator inconsistent. Does this mean if I use cluster-robust standard errors, then I should just leave the differenced lag dependent variable as it is?

[Sebastian Kripfganz]

Consistency and inference using cluster-robust standard errors are two different pairs of shoes. Whether you use cluster-robust standard errors has no implication for the validity of the instruments.