Does Factor Investing Improve Investor Welfare? A Multi-Asset Perspective

TLDR: In this paper, the authors show that a set of benchmark portfolios made of individual securities offers the same investment opportunities as these securities if, and only if, their returns are asset pricing factors, meaning that they collectively explain differences in expected returns.

Abstract: Investment decisions in delegated money management are typically made in two stages. In the first stage, a centralized decision maker allocates capital to the different asset classes. In the second stage, asset managers decide how to allocate available wealth to the individual securities within the corresponding asset class. Although commonly used in investment practice, this approach potentially involves a loss of efficiency compared with a single-step optimization procedure. The authors show that this welfare loss is zero if, and only if, the benchmarks completely explain the cross section of expected returns of the securities. This theoretical proposition justifies factor investing as an investment method that constructs investable proxies for pricing factors. Mean-variance spanning tests confirm that investment opportunities offered by standard equity, bond, and commodity factor benchmarks are not spanned by corresponding broad asset class benchmarks. The authors then conduct out-of-sample tests with these factors and find that they improve the risk-return profile of multiclass portfolios with respect to the case in which these portfolios are invested in broad asset class benchmarks. TOPICS:Wealth management, factor-based models, portfolio theory, portfolio construction Key Findings • A set of benchmark portfolios made of individual securities offers the same investment opportunities as these securities if, and only if, their returns are asset pricing factors, meaning that they collectively explain differences in expected returns. • Mean–variance spanning tests lead the authors to reject the hypothesis that traditional broad equity, bond, and commodity asset class benchmarks generate the same efficient frontier as factor benchmarks drawn from these three classes. • Out of sample, introducing factor benchmarks in multi-asset portfolios that contain traditional asset class benchmarks improves their Sharpe ratio.