Will Your Factor Deliver? An Examination of Factor Robustness and Implementation Costs

publication image

Will Your Factor Deliver? An Examination of Factor Robustness and Implementation Costs

September 2016
Read Time: 30 min

Executive Summary

The academic literature on factors supposedly driving investment returns is vast and growing. Which ones, however, are likely to prove profitable in practice? In order to help investors discriminate among the hundreds of published factors, the authors apply heuristic guidelines to narrow the range of choices and simulate the performance of selected factors using various definitions in diverse geographical markets both before and after estimated trading costs. They find that value, momentum, illiquidity, and low beta are meaningfully more robust than are size and quality, and that illiquidity and momentum are associated with significantly higher trading costs than are other factors. The extensive empirical evidence presented in this article stands to inform investors’ decisions about which factor-based strategies to use and whether to implement them through indexation or active management.

Building on prior research, the authors employ the Hsu-Kalesnik-Viswanathan framework to identify and evaluate promising factor-driven strategies. This heuristic approach applies four rules. Factors should be 1) grounded in a long and deep academic literature, 2) robust across definitions, 3) robust across geographies, and 4) simulated results should be adjusted for trading costs. These guidelines shape the study. The authors construct standard long–short portfolios by definition and geography on the basis of the strategies cited most often in the Social Science Research Network (SSRN) database. Using US data over the period 1967–2014 and international data over the period 1987–2014, Beck et al. evaluate the strategies’ robustness in terms of the long–short portfolios’ excess return t-statistics, which reflect benchmark risk, and the long and short portfolios’ Sharpe ratios, which reflect total risk. In addition, the selected strategies’ downside risk and upside/downside capture characteristics are compared. Using the Aked–Moroz trading-cost model, the impact of turnover on investment results is also estimated.

Following standard practice, the authors first divide the universe into large and small stocks, and then partition the large- and small-stock subsets by factor strategy—value, momentum, low beta, quality, and illiquidity—to construct high-characteristic and low-characteristic portfolios weighted by market capitalization. They then construct combined portfolios invested equally in the resulting large and small portfolios. For comparisons across geographical regions, stocks are grouped into the United States, Japan, United Kingdom, Europe ex UK, and Global. Portfolios are rebalanced annually in January, with the exception of momentum portfolios, which are rebalanced monthly.

Tests reveal that low-beta and low-volatility stock portfolios consistently have more attractive Sharpe ratios than high-beta portfolios. Moreover, the differences in Sharpe ratios are economically and statistically significant. The authors observe, however, that due to large tracking errors the t-stats of the differences in excess return are not statistically significant. They conclude that although the low-beta strategy is potentially inappropriate for investors who are averse to benchmark risk, it has proven to be a robust source of risk-adjusted performance for investors who can accept the requisite tracking error.

In addition to the customary book-value-to-price ratio, the authors choose three other ratios to define the value strategy: earnings to price, cash flow to price, and dividends to price. For all definitions, they observe economically significant differences in returns between value and growth stocks. The high-dividend-yield portfolios do not generate statistically significant differences in excess returns. Given that stocks with high-dividend-to-price ratios have low risk, the authors suggest this outcome is a result of the volatility of the difference portfolio. In this interpretation, dividend yield, like low beta, is a doubtful strategy for investors who evaluate results in terms of the information ratio. On the other hand, led by high dividend yield, all definitions of value provide statistically better Sharpe ratios. Outside the UK large-stock universe, the international data also reveal a robust pattern of value outperforming growth.

The authors’ examination of the size factor discloses that, on average, small-cap stocks provide higher returns than large-cap stocks. The differences in Sharpe ratios reveal, however, that no definition of small-stock portfolios delivers statistically significant risk-adjusted benefits. The authors find the size premium lacks robustness, but further data provided in an online appendix help investors reach their own conclusions.

Studies of the momentum factor typically look at the past year of returns, skipping the most recent month to adjust for short-term mean reversion. Using this definition along with others, and varying the formation and holding periods, Beck et al. find that the momentum strategy is more reliable in the small-cap than in the large-cap universe. Across regions, the standard definition of momentum produces statistically significant benefits everywhere except in Japan, where value investing predominates.

Whether defined with reference to 1-month, 6-month, or 12-month average daily volume, the illiquidity factor produces economically and statistically significant benefits in the United States. Internationally, however, illiquidity does not appear to offer a premium. Citing another published study, the authors state they are not ready to rule out the existence of an illiquidity premium, but the evidence presented is mixed.

Many definitions of quality exist in the literature. In their study, the authors use return on equity, gross margins, and leverage in addition to the most popular measure, gross profitability. They find very few signs of a persistent premium across definitions or geographical regions. Here, too, an online appendix with further data enables readers to examine quality investing in greater depth.

An examination of the factor portfolios’ downside risk and upside/downside capture percentages brings interesting information to light. For example, momentum investing is subject to abrupt crashes. Low-beta strategies tend to outperform in bear markets, yet have the potential for long periods of underperformance. Value strategies may also prove disappointing for extended periods, but can recoup losses so quickly that investors who wait for a recovery may miss out on the gain. These and other insights may help practitioners weigh factor allocations.

Finally, recognizing that different factor-based strategies have different turnover rates, Beck et al. explore the impact of estimated trade costs on investment returns. Among the factors previously adjudged robust, the large-cap and small-cap value strategies and the low-cap low-beta strategy predominately preserve their return advantages after adjustment for trading costs. The liquidity-taking strategies, however, do not fare as well on a trade-cost-adjusted basis: both the large-cap and small-cap momentum strategies and the large-cap illiquidity strategy lose their attractiveness. The authors suggest that active management, rather than indexation, may be a more effective approach to executing these strategies. They state, “Interestingly, the value that active managers can provide arises not necessarily from their stock-picking skills, but rather from their ability to actively manage transaction costs in liquidity-taking strategies.”  For factor-based strategies that can be implemented efficiently—notably the value and low-volatility strategies—lower-fee indexing seems more advantageous. 

Summarized by Philip Lawton, CFA