Tactical Asset Allocation with Pairwise Strategies

Abstract

An alternative approach to quantitative tactical asset allocation (TAA) is based on time series forecasting models and mean-variance optimization. The central concept is pairwise TAA, and the correct metric for assessing forecast quality is the pairwise information coefficient. TAA using mean-variance optimization is generally equivalent to a linear combination of pairwise TAA, with the relative weights of the pairs in the combination directly connected to the covariance matrix used in the optimization. TAA managers should avoid using a covariance matrix without knowing its implied pairwise weights, but instead should use pairwise information to influence the choice of the pairwise weights. The expected long-term performance of TAA strategies for a given set of pairwise weights as well as optimal sets of pairwise combinations that attain the maximum information ratio is also derived.