Tid: 20 juni 2011 kl 15.15-16.00.Seminarierummet 3721, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Fang Li
Titel: Predicting future returns with investor views (Examensarbete - Master thesis)
Abstract The mean-variance framework developed by Markowitz (1959) for portfolio analysis is probably one of the most elegant and widely accepted concepts in financial theory. A major problem in portfolio optimization is the estimation of the input parameters in the distributions of the underlying assets. Jobson and Korkie (1980) conclude that classical estimation techniques, where unknown parameters as means and covariances are replaced by their sample values perform very poorly. Small variations in sample means often lead to extreme portfolio allocation readjustments. To compliment historical data with new information that helps better predict future market outcome Black and Litterman (1992) found a way in which the investors could combine their market views with historical information. Meucci (2008) extended the idea into Copula-opinion pooling which is the numerical equivalence to the Black-Litterman framework. Also an effort has been made by Stein (1956) on the frontier of statistics to construct a mean estimator better than the observed average of the data sample, the shrinkage estimator. In this thesis a thorough comparison between Black-Litterman, Copula-opinion pooling, and Black-Litterman with shrinkage estimators will be made to find out their respective strengths and shortcomings. Certain simplifications and assumptions are made to make comparisons more convenient. New methods are suggested to increase stability of the models and cope with some of the problems encountered. The result indicates that Black Litterman combined with the shrinkage estimator does give consistently better results compared to the other methods.
|Sidansvarig: Filip Lindskog