Tid: 26 januari 2004 kl 1615-1700
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Jens Svensson.
Titel: Realized Volatility and Multivariate GARCH: Theory and Application to Foreign Exchange Portfolios. (Examensarbete)
Sammanfattning: Many financial applications rely on accurate forecasts of both volatility and correlation. An understanding of volatility is therefore essential for making rational financial decisions. This thesis presents some recent advances in volatility measurement and forecasting. Problems of volatility measurement are summarized and realized volatility is examined as an alternative measure. Using autoregressive processes, realized volatility is successfully forecasted and compared to standard methods such as GARCH. The analysis is then extended to the multivariate case and the most popular covariance matrix estimation methods are compared to some more advanced methods. Our comparison includes multivariate GARCH models as well as autoregressive processes. Different methods are evaluated using portfolio optimization and significant differences in performance between the models are found.