Since the development of the capital asset pricing model (CAPM) about 50 years ago, academic researchers have documented several hundred “anomalies” that generate a significant positive alpha. There are now so many that professor John Cochrane referred to them as the “factor zoo.”
There are certainly large incentives to find these anomalies, both for academics and investment firms. And that raises the concern that many of the discovered anomalies could be nothing more than the result of data mining exercises. These concerns can be mitigated by out-of-sample tests, as well as strong risk or behavioral explanations.
An additional concern is that many academic research papers identify theoretical anomalies, meaning they might exist only on paper. However, it’s really an anomaly only if you can exploit it after considering real world transactions costs.
While trading costs—in the form of commissions as well as bid/offer spreads—have indeed fallen dramatically over the past 50 years, costs still matter, especially for high-turnover strategies and smaller-cap stocks.
A Study Of Anomalies
Robert Novy-Marx and Mihail Velikov—authors of the December 2014 paper, “A Taxonomy of Anomalies and Their Trading Costs”—studied the performance of 23 anomalies after accounting for estimated transaction costs. They also examined the effectiveness of three strategies designed to mitigate transaction costs.
The first strategy is to limit trading to low expected transaction cost stocks. The second strategy is to reduce rebalancing frequencies (at the expense of some staleness in the signals on which the strategies are based).
The study’s authors note that this technique is popular among large institutional money managers. For example, the AQR momentum indexes, which are designed to track a momentum strategy with limited trading costs, are rebalanced quarterly instead of monthly.
The third strategy is to realize lower turnover by introducing a buy/hold range and holding stocks that would be no longer be bought. This strategy has long been used by Dimensional Fund Advisors (among others) in its small-cap funds and is also now used in MSCI indexes.
Novy-Marx and Velikov grouped their 23 anomalies by the level of turnover in each strategy. The three groups they generated were low, mid and high turnover. These groups corresponded roughly to strategies where the long- and short-side turnover was, on average, less than once per year; between one and five times per year; and more than five times per year, respectively.
Among the low-turnover (and thus low-cost) anomalies are size; gross profitability; value; combining value and profitability; and investment. Among the mid-turnover strategies are momentum; combining value and momentum; combining value, momentum and profitability; and idiosyncratic volatility. Among the high-turnover strategies are industry momentum and industry-relative reversals.
The following is a summary of the authors’ findings:
- The cost of trading the low-turnover strategies is generally quite low, often less than 10 bps per month.
- Most of the anomalies considered that had one-sided monthly turnover lower than 50 percent continued to remain significantly profitable, at least when the strategies were designed to mitigate transaction costs.
- The cost of trading the high-turnover strategies, at least when they were designed with complete disregard for trading costs, always exceeded 1 percent per month. Transaction costs significantly exceeded the gross spread for all but two of the anomalies examined, with only the high-frequency combo strategy and the low-volatility industry-relative reversal strategy achieving positive net excess returns. Transaction costs in all cases reduced the strategies’ profitability and associated statistical significance, raising the barriers to arbitrage and increasing concerns related to data snooping.
- The buy/hold range was the most effective cost-mitigation technique for most of the anomalies considered, though for very-high-turnover strategies (for which transaction-cost mitigation is most important), a combination of all three techniques yielded greater performance enhancements.
In summary, Novy-Marx and Velikov provide strong evidence that at least some “anomalies” (including value, momentum, profitability and investment) survive transactions costs.
What’s more, their paper provides further evidence supporting the findings of a 2012 study—“Trading Costs of Asset Pricing Anomalies” by Andrea Frazzini, Ronen Israel and Tobias Moskowitz—that also concluded value and momentum survive trading costs. Novy-Marx and Velikov also provide evidence that “intelligent design” (some would use the term “smart beta”) can improve the performance of simple indexing strategies by reducing transactions costs.
Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.