Specification tests for non‐Gaussian maximum likelihood estimators

We propose generalized DWH specification tests which simultaneously compare three or more likelihood‐based estimators in multivariate conditionally heteroskedastic dynamic regression models. Our are useful for Garch models and many empirically relevant macro finance applications involving Var s regressions. determine the rank of differences between estimators' asymptotic covariance matrices under correct specification, take into account that some parameters remain consistently estimated distributional misspecification. provide finite sample results through Monte Carlo simulations. Finally, we analyze a structural proposed to capture relationship macroeconomic financial uncertainty business cycle.