Portfolio management using multivariate time series forecasts

We use a multivariate VAR-GARCH model to predict the monthly returns and the variance matrix of the MSCI North America, MSCI Europe and MSCI Pacific indices from February 1990 until September 1999. We are interested in the following questions: First, can time series forecasts be successfully transformed into portfolio weights, and second, what kind of forecasts improves the portfolio performance? We compare two minimum-variance portfolios: the first portfolio is based only on the forecasted variance matrix, while the second portfolio uses the predicted returns and the predicted variance matrix. As a benchmark we choose the MSCI World index.