Simple and trustworthy cluster-robust GMM inference

This paper develops a new asymptotic theory for GMM estimation and inference in the presence of clustered dependence. The key feature our alternative asymptotics is that number clusters G regarded as fixed sample size increases. Under fixed-G asymptotics, we show Wald t tests two-step are asymptotically pivotal only if recenter estimated moment process covariance estimator (CCE). Also, J statistic, trinity statistics (QLR, LM, Wald), statistic can be modified to have an standard F distribution or distribution. We suggest finite-sample variance correction further improve accuracy approximations. proposed very appealing practitioners because test simple modifications conventional statistics, critical values readily available from tables. No simulations resampling methods needed. A Monte Carlo study shows more accurate than large-G inferences.