Dynamic Specification Tests for Dynamic Factor Models

摘要: We derive computationally simple and intuitive expressions for score tests of neglected serial correlation in common idiosyncratic factors dynamic factor models using frequency domain techniques. The implied time orthogonality conditions are analogous to the obtained by treating smoothed estimators innovations latent as if they were observed, but account their final estimation errors. Monte Carlo exercises confirm finite sample reliability power our proposed tests. Finally, we illustrate empirical usefulness an application that constructs a monthly coincident indicator US from four macro series.