An alternative quantitative approach to tactical asset allocation using the Kalman filter

Abstract

Tactical asset allocation (TAA) is a dynamic investment strategy which seeks to actively adjust fund allocation to a variety of asset classes by systematically exploiting inefficiencies and temporary imbalances in equilibrium values. TAA adds value by underweighting fund allocation to those assets whose returns have been forecasted to underperform on a relative basis and overweighting those whose returns were forecasted to indicate outperformance. This approach contrasts with strategic asset allocation (SAA) in which a long-term investment view target allocation is established using a combination of target return and risk tolerance. Portfolio managers who employ TAA as an investment strategy aim to benefit from market timing, a non-trivial exercise involving the entry and exit of selected asset classes based on future performance.

TAA decisions are governed by three major considerations: valuation-based approaches, macroeconomic scenarios and technical/quantitative analyses. This work explores a quantitative analytical approach for TAA which adjusts portfolio weights based on forecasted returns of constituent asset classes. Asset returns are forecasted using the Capital Asset Pricing Model (CAPM), complemented with results obtained from the Kalman filter, a Bayesian forecasting tool whose recent application to time-dependent variable estimation has shown promising results. Using a decade of recent monthly return data, the performance of the TAA and SAA approaches are compared using a range of diagnostic metrics. The TAA approach outperforms its SAA counterpart for most of these metrics, even when the most recent returns (i.e. those affected by the coronavirus pandemic) are excluded.