Implicit score-driven filters for time-varying parameter models

We propose an observation-driven modeling framework that permits time variation in the model parameters using an implicit score-driven (ISD) update. The ISD update maximizes the logarithmic observation density with respect to the parameter vector, while penalizing the weighted L2 norm relative to a one-step-ahead predicted parameter. This yields an implicit stochastic-gradient update. We show that the popular class of explicit score-driven (ESD) models arises if the observation log density is linearly approximated around the prediction. By preserving the full density, the ISD update globalizes favorable local properties of the ESD update. Namely, for log-concave observation densities, whether correctly specified or not, the ISD filter is stable for all learning rates, while its updates are contractive in mean squared error toward the (pseudo-)true parameter at every time step. We demonstrate the usefulness of ISD filters in simulations and empirical illustrations in finance and macroeconomics.