I will be presenting my recent work on score-driven high-dimensional approximate dynamic factor models in an upcoming seminar hosted by CREST – Center for Research in Economics and Statistics.
Score-Driven High-Dimensional Approximate Dynamic Factor Models: Estimation and Inference
🗓️ January 8, 2026
🕙 10:00–11:00 am
📍 Room 3049, Finance–Insurance
ENSAE Paris building, Palaiseau campus
Institut Polytechnique de Paris
🔗 More info: https://crest.science/event/enzo-dinnocenzo-bologna-university-italy-t-b-a/
We propose a dynamic factor model for high-dimensional time series in which multivariate score-driven dynamics generate the latent factors, allowing the model to capture non-linearities and heavy tails. We estimate the model in two steps. First, we extract the factors using Principal Components or alternative robust methods. Then we estimate the parameters of the score-driven model for the extracted factors via Quasi Maximum Likelihood. We consider models for both the conditional mean and the conditional variance. The parameter estimates are consistent and asymptotically normal as both the number of time series and the sample size diverge to infinity. We also construct valid asymptotic prediction intervals. Numerical results confirm the good performance of the estimator.
This is joint work with Matteo Barigozzi (Economics and Econometrics, University of Bologna – Alma Mater Studiorum).
Many thanks to Jean-Michel Zakoian and Christian Francq for the kind invitation and the opportunity to present this work.
For more information you can contact me here.