Title: A score-driven smoother for general state-space models
Authors: Giuseppe Buccheri - University of Rome Tor Vergata (Italy) [presenting]
Giacomo Bormetti - University of Bologna (Italy)
Fulvio Corsi - University of Pisa and City University London (Italy)
Fabrizio Lillo - Scuola Normale Superiore (Italy)
Abstract: A simple approximate smoother is introduced which extends the score-driven estimation approach to also include present and future observations. The newly proposed Score-Driven Smoother (SDS) can be used to improve the estimation of time-varying parameters in nonlinear non-Gaussian state-space models. In contrast to complex and computationally demanding simulation-based methods, the SDS has the same simple structure of the Kalman backward smoothing recursion but uses the score of the true observation density. Through an extensive Monte Carlo study, we provide evidence that the performance of the approximation is very close (with average differences of 2\% in mean square errors) to that of simulation-based techniques and is superior to other approximate methods. Empirically, the effectiveness of the SDS is shown in recovering accurate estimates of time-varying volatilities and correlations of inflation rates in the euro area.