Statistical Methods for High-Dimensional Volatility
-
Series
-
SpeakerMark Podolskij (University of Luxembourg)
-
FieldEconometrics, Data Science and Econometrics
-
LocationErasmus University Rotterdam, Campus Woudestein, ET-14
Rotterdam -
Date and time
April 16, 2026
12:00 - 13:00
Abstract
In recent years, there has been growing interest in statistical methods for high-dimensional volatility processes in continuous-time models. In such settings, classical estimators, such as realized (co-)variance, often exhibit poor performance. To address this, existing approaches typically impose sparsity assumptions on the integrated volatility matrix and rely on shrinkage-based techniques, such as LASSO.
In contrast, this talk focuses on the estimation of the spectral distribution of the integrated volatility matrix without imposing sparsity constraints. We propose a consistent estimator for the spectral distribution based on an inversion of the celebrated Marčenko–Pastur theorem from random matrix theory. The results are based on joint work with Grégoire Szymanski.