Optimal Allocation to Time-Series and Cross-Sectional Momentum

The authors examine the optimal combination of time-series (absolute) and cross-sectional (relative) momentum in a pure trend-following strategy. They show that the solution depends on two main inputs: (1) the signal (trend) strengths of and (2) the covariances between the different instruments in the investment universe. Time-series momentum leads to superior results when all instruments have similar trends and correlations are low. Conversely, cross-sectional momentum (relative momentum) is preferable in periods in which the correlations are high and the dispersion in the signals is large. The authors construct an (ex ante) optimal momentum portfolio and provide empirical evidence that an (ex ante) optimal dynamic allocation to absolute and relative trends outperforms (ex post) the pure time-series and cross-sectional strategies. Based on data for 59 futures covering all major asset classes between 2000 and 2018, they find that the average optimal allocation to time-series (cross-sectional) momentum is around 80% (20%). TOPICS:Statistical methods, portfolio construction, performance measurement Key Findings ▪ A trend follower should dynamically amend the allocation to time-series and cross-sectional trends to adapt to the changing market conditions. ▪ The ex ante optimal allocation to time-series and cross-sectional trends depends on the trend strengths of and the covariances between the instruments traded. ▪ The higher the trend dispersion and the correlation between the instruments the more a trend follower should weight cross-sectional trend bets ceteris paribus.