Skip to content

Tous les séminaires

Jeudi 26 Mars 2015

16H15 : Bernd SCHWAAB (European Central Bank) "Modelling Financial Sector Joint Tail risk in the Euro Area"

17H30 : Diaa NOURELDIN (American University in Cairo) "Volatility Prediction Using a High-Frequency-Based Component Model"

 

  1. Bernd SCHWAAB (European Central Bank)

MODELING FINANCIAL SECTOR JOINT TAIL RISK IN THE EURO AREA

 

We develop a novel high-dimensional non-Gaussian modeling framework to infer conditional and joint risk measures for many financial sector _rms. The model is based on a dynamic Generalized Hyperbolic Skewed-t block-equicorrelation copula with time-varying volatility and dependence parameters that naturally accommodates asymmetries, heavy tails, as well as non-linear and time-varying default dependence. We demonstrate how to apply a conditional law of large numbers in this setting to define risk measures that can be evaluated quickly and reliably. We apply the modeling framework to assess the joint risk from multiple financial firm defaults in the euro area during the 2008-2012 financial and sovereign debt crisis. We document unprecedented tail risks during 2011-12, as well as their steep decline after subsequent policy actions.


 

17H30 : Diaa NOURELDIN (American University in Cairo)

VOLATILITY PREDICTION USING A HIGH-FREQUENCY-BASED COMPONENT MODEL

This paper presents a high-frequency-based component model suitable for short- and medium-term volatility prediction. In a joint model for the daily return and a realized measure (e.g. realized variance), the conditional variance of the daily return has a multiplicative component structure: a long-run (secular) component, and a short-run (transitory) component driven by a high-frequency-based signal. Despite being a fixed-parameter model, its component structure allows for a time-varying intercept as well as time-varying dynamic parameters, that evolve with the long-run components of both the return and the realized measure. We discuss the model properties and estimation by quasi-maximum likelihood. The empirical analysis reveals strong out-of-sample performance at both short- and medium-term forecast horizons compared to benchmark models.

Prochaine séance : 28 Mai 2015

Dennis KRISTENSEN (Univ. College London), Sébastien LAURENT (AMSE – Aix Marseille Université)