William N. Goetzmann
Franklin R. Edwards

Efficient frontiers based upon long-horizon inputs differ from those based upon short-horizon inputs. Inputs to mean-variance optimization change significantly as investor horizons change because of autocorrelation between certain asset classes and the use of short-horizon inputs to estimate long-term correlations. The authors employ simulation techniques to estimate long- horizon inputs to the mean-variance model and to construct efficient frontiers. These portfolios are then compared to efficient portfolios generated with short-horizon input. The authors find that some significant differences occur, especially with respect to the minimum-variance portfolio.