Using Economic Theory to Build Optimal Portfolios

Given expected returns and return covariances, portfolio weights are known in closed form in a mean-variance framework. The real difficulty is in estimating these parameters. Using recent advances in Bayesian techniques, we show how investors can incorporate any prior information for optimal portfolio selection. We apply our method to 27 domestic and international data sets. We find that our tangency portfolios have three essential and attractive features. i) They perform better in terms of out-of-sample Sharpe ratio. ii) Their weights are guaranteed to be economically "reasonable": positive, stable, and without extravagant position in any asset. iii) Turnover is very low.