Smart Beta and Statistical Significance

Abstract

The author proposes an alternative weighting mechanism for equity mandates based on the statistical significance of the factor loading on the benchmark. Specifically, the weight of each security entering the active portfolio is directly proportional to the t-statistic of the factor loading (β) with the benchmark. The author show that this amounts to overweighing securities with high correlations with the benchmark and vice versa. The t-statistic of market beta within a single-factor model is a monotonic transformation of its correlation with the benchmark. He tests the out-of-sample performance of this alternative weighing scheme with industry portfolios as well as individual US stocks and find that this strategy has higher correlations with the benchmark than other popular alternatives, has lower tracking error, unitary exposure to the benchmark, very good Sharpe ratios, and substantial ex-post realized returns relative to the underlying benchmark.

TOPICS: Wealth management, statistical methods, simulations