Multivariate GARCH models and Black-Litterman approach for tracking error constrained portfolios: an empirical analysis

In a typical tactical asset allocation set up managers generally make their investment decisions by inserting private information in an optimisation mechanism used to beat a benchmark portfolio; in this context the sole approach a' la Markowitz (1959) does not use all the available information about expected excess return and especially it does not take two main factors into account: first, asset returns often show changes in volatility, and second, the manager's private information plays no role in the optimisation process. This paper provides an empirical work for large scale tactical asset allocation strategy in which a multivariate GARCH estimation is used in portfolio optimisation, given a tracking error constraint (Jorion, 2003). Moreover, the use of Black and Litterman (1991, 1992) approach allows for the possibility to tactically manage the selected portfolio through a very short time, combining informations taken from the time varying volatility model with some personal "view" about asset returns.