Modeling and predicting financial risk measures with factor augmented models

This project aims at developing a novel econometric framework for forecasting financial risk measures bycombining matrix-valued time series models, multi-level factor structures and high-frequency realized risk indicators. The motivation arises from the limitations of existing univariate and multivariate approaches when applied to large cross-sections of assets, where the number of realized predictors (realized variances, quantiles, tail expectations, skewness, kurtosis) grows rapidly and standard regression methods suffer from dimensionality and collinearity issues.