The Windham Portfolio Advisor 111
Appendix —WPA features
Return estimation
The WPA offers several methods for estimating expected returns, including:
● Historical full sample
● Equilibrium based on each asset’s beta with respect to a reference portfolio
● Implied returns that assume the current portfolio is optimal
● User-specific views
● Historical blend of historical returns and user views, blended according to user con-
fidence in views
● Equilibrium blend of equilibrium returns and user views about idiosyncratic com-
ponent of returns, blended according to user confidence in views
● Black – Litterman blend of equilibrium returns and user views about systematic com-
ponent of returns, blended according to user confidence in views
The user may choose to estimate returns after taxes.
Risk estimation
The WPA offers several options for estimating standard deviations and correla-
tions, including:
● Historical full sample
● Exponential decay factor placing higher weight on recent performance
● Turbulent regime based on periods representing turbulent market conditions
● Quiet regime based on periods representing quiet market conditions
● User-defined weighting of turbulent and quiet periods
● User-specified return or risk thresholds, such as downside risk
● User-defined or imported from other risk model estimation module
The user may choose to estimate risk after taxes.
The WPA checks to ensure that the correlations are positive semidefinite.Parametric optimization
The WPA enables the user to perform a variety of parametric optimizations,
including mean – variance optimization, mean – tracking error optimization, or
mean – variance – tracking error optimization.
The user sets the minimum and maximum allowable asset weights.
The user may add other linear or group constraints.
Full-scale optimization
The WPA also allows the user to construct portfolios on the basis of full-scale
optimization, which identifies the optimal portfolio by examining a sufficiently
large set of asset combinations. Full-scale optimization accommodates a wider
range of investor preferences and return distributions than mean – variance