Unconstrained Strategies and the Variance-Kurtosis Trade-Off

In this paper, we study unconstrained strategies through a re-specification of classic meanvariance utility and, as a reference implementation, a long-only strategy based on Canadian and U.S. bond markets. First, we capture the underlying economic forces that drive benchmark indices in the two economies as orthogonal components of yields. We find that bond indices in the two markets are sensitive to components that account for lesser total yield variability. Next, we develop a new polynomial utility function that captures the kurtosis effects found in the sensitivities to lower-eigenvector components. In our unconstrained strategy, excess kurtosis triggers portfolio adjustments and the resulting returns outperform those of traditional meanvariance optimization. The re-specified utility function introduces iso-risk contour lines that account for abrupt adjustments of portfolios to eigenvectors of hidden influence