News-Optimized Risk Management

This doctoral dissertation is concerned with the improvement of the market risk measurement. Since the crucial element in many usual models or methods used to compute market risk measures (especially the Value at Risk and the Expected Shortfall) is the volatility, our main focus in this thesis is the improvement of the volatility refinement. Our optimization consists of including additional input factors in existing models. These input factors are firm-specific news, which can have an influence on a firm and especially on its stock value. First, we extract from the news, through an ordinal stochastic volatility model, the information that will serve later for the volatility refinement. The improvement is then performed using a GJR-GARCHX model (exogenous extension of the GJR-Generalized Autoregressive Conditional Heteroskedastic model) of Glosten et al.