@article {Benderjpm.2019.1.117,
author = {Bender, Jennifer and Sun, Xiaole},
title = {A Closer Look at the Factor-to-Specific Risk Ratio in Factor Portfolios},
elocation-id = {jpm.2019.1.117},
year = {2019},
doi = {10.3905/jpm.2019.1.117},
publisher = {Institutional Investor Journals Umbrella},
abstract = {One of the metrics that portfolio managers often use to gauge the success of factor investing strategies is the factor-to-specific risk ratio. There is much confusion, however, regarding its use and interpretation. Specifically, what is a reasonable level that investors might expect for long{\textendash}short versus long-only factor portfolios, and how much deviation is normal? Are there limitations with this metric, and is a higher ratio necessarily always better? The authors find that even pure long{\textendash}short factor mimicking portfolios do not have 100\% of risk coming from factors. Using the value factor as an example, a long{\textendash}short factor mimicking portfolio has an average ex ante factor-to-total-variance ratio of 80\%. The long-only version has an even lower ratio of 61\%. More importantly, factor-to-total-variance ratios are misleading in that they do not reflect how much risk and return come from the targeted factors versus the nontargeted factors. Counterintuitively, a simple quintile portfolio can have a very high ratio because of exposure to nontargeted factors. Other metrics such as exposure per unit of tracking error, percent of variance contribution, and the factor efficiency ratio are needed to understand how much exposure the portfolio has to the targeted factor and how much risk comes from the latter.TOPICS: Analysis of individual factors/risk premia, factor-based models, style investingKey Findings{\textbullet} Even long{\textendash}short factor mimicking portfolios that exactly replicate the pure factor returns do not have 100\% of risk coming from factors. Using the value factor, a long{\textendash}short factor-mimicking portfolio has an average ex ante factor-to-total-variance ratio of 80\%.{\textbullet} Different aspects of portfolio construction can affect all of these metrics. Portfolio tracking error, misalignment between the estimation and investment universes, and factor definition misalignment can significantly affect the factor-to-total-variance ratio.{\textbullet} Factor-to-total-variance ratios are misleading in that they do not tell us how much risk and return comes from targeted factors. Other metrics such as exposure per unit of tracking error and percent of variance contribution are needed.},
issn = {0095-4918},
URL = {https://jpm.pm-research.com/content/early/2019/11/01/jpm.2019.1.117},
eprint = {https://jpm.pm-research.com/content/early/2019/11/01/jpm.2019.1.117.full.pdf},
journal = {The Journal of Portfolio Management}
}