Finance & Financial Econometrics: 
Time: 10.30 am
Date: 11th of May 2023
Room 3001

Dick VAN DIJK (Erasmus University Rotterdam) “Robust Observation-Driven Models Using Proximal-Parameter Updates”

Abstract : We propose an observation-driven modelling framework that permits time variation in the model’s parameters using a proximal-parameter (ProPar) update. ProPar maximizes the observation log-density with respect to the parameter vector, while penalizing the weighted L2-norm relative to the one-step-ahead prediction. This yields an implicit stochastic-gradient update; taking instead the explicit version would produce the popular class of score-driven models. For log-concave observation densities (even when misspecified), ProPar’s robustness is evident from its muted response to outliers, stability under poorly specified learning rates, and global contractivity towards a pseudo-truth. We illustrate ProPar’s usefulness for estimating time-varying regressions, volatility, and quantiles.

Jean-Michel ZAKOIAN (CREST)

Sponsors:
CREST