Can We Estimate Macroforecasters’ Mis-Behavior?

We answer positively to this question by using Maximum Lq-Likelihood (orDeformed Likelihood) estimator. This is based on a parameter which measuresthe degree of data contamination. We apply this estimator, for the first timein econometric literature, to dynamic system and derive a robust version ofthe Kalman Filter—the Deformed Kalman Filter (DKF). The evidence fromU.S. data suggests that the contamination dynamics exists and is correlated(but not coincident) with the phases of the Business Cycle. Furthermore itsknowledge improves in-sample as well as out-of-sample estimation