Inquire Europe Research Proposal

After more than half a century mean variance analysis still remains the primary conceptual tool for understanding the trade-off between risk and return and for the definition of optimal investment portfolios. Despite its theoretical appeal practical applications have always experienced major difficulties. Optimal portfolio weights appear extremely sensitive to minor changes in expected returns or the covariance matrix. Typical solutions in practice are to put restrictions on the portfolio weights (usually minimum and maximum holdings) or direct the weights to a priori likely values in a Bayesian framework. In the proposed research project we consider a different approach to obtain more sensible optimal portfolios. Robust optimisation is an approach that is well known in the operations research literature, and has recently received much attention in finance. With robust optimisation an optimal portfolio is chosen such that it performs well under a worst case scenario, where the worst case is based on the uncertainty about the input parameters. Our aim is to apply robust optimisation to the mean variance problem. Robust optimisation should make the optimal portfolios less sensitive to estimation error in the parameters. If this works, the technique would be a useful practical tool for portfolio management.