Dynamic Portfolio Optimization: Beyond MPT

Optimization based solely on the REIT returns in a historical time window is severely restricted by that set of realized historical returns, leaving the portfolio vulnerable to downturns unseen in the historical data. Dynamic portfolio optimization, which determines portfolio composition using a massive ensemble of return predictions that are statistically consistent with historical returns but include extreme events safeguard against this vulnerability. Dynamic optimization, based upon ARMA-GARCH models with heavy-tailed innovations and non-Gaussian copulas, is developed in this Chapter for mean variance and conditional value-at-risk measures as well as for the Black–Litterman model. Dynamically optimized portfolios comprised of domestic REITs are computed and their performance compared to corresponding portfolios optimized under the classical historical return approach. Fairly dramatic performance improvement is seen under dynamic optimization