Three EDS researchers publish an article in the Journal of Econometrics where they introduce new theoretical results on how modern statistical filters behave when used to track economic and financial indicators. Their main contribution is to show, in simple terms, that these filters still work well even when the underlying model is wrong.
The paper by Eric Beutner, Yicong Lin and André Lucas studies a popular method used to track changing economic conditions, called score driven filtering. The authors show that these filters reliably follow the true underlying patterns in data—even if the model is imperfect—and describe how the remaining forecast errors behave. This matters for economic policy because policymakers often rely on imperfect models. It is also important for econometrics, as it further strengthens the theoretical foundations of widely used score-driven dynamic models.
The article Consistency, distributional convergence, and optimality of time-varying parameters in score-driven models in the Journal of Econometrics is available online.