Tid: 22 november 2004 kl 1515-1600
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Marcus Josefsson.
Titel: copula-EVT based approach for measuring tail related risk: Applied on the swedish market. (Examensarbete)
Sammanfattning: The aim of this master thesis is to present a A copula-EVT based approach to estimate portfolio Value-at-Risk. More specific, Monte Carlo scenarios for asset log-returns are generated from a conditional multivariate distribution with gaussian or student's t-copula and marginal distributions normal in the centre and in the profit tail, and generalized pareto distributed in the loss tails.
Three different conditional Copula-EVT models are applied on a portfolio consisting of 28 Swedish assets and the VaR at high quantiles (95% or more) over a one-day holding period are estimated. For comparision, the same estimates are done for an unconditional normal, a conditional normal, a conditional t4 model and a conditional normal model with t10 copula.
Furthermore, backtesting the seven models over a time window of six years at 99%, 99.5% and 99.9% VaR enables their effectiveness to be tested.