Abstract
This article compares the efficacy of three common transaction-cost-mitigation techniques: limiting a strategy to cheap-to-trade securities, rebalancing a strategy less frequently, and “banding,” which imposes a higher hurdle for actively trading into a position than for maintaining an established position. All three strategies significantly reduce transaction costs, but the techniques that reduce turnover have a less negative impact on strategy gross performance than limiting trade to low-cost securities has. Banding is more effective than simply reducing rebalancing frequencies, because banding yields similar trading-cost reductions while maintaining a better exposure to the underlying signal used to select stocks.
Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of Richmond or the Federal Reserve System. We thank Barbara Petitt, CFA, Stephen Brown, and Milena Novy-Marx for discussions and comments. Robert Novy-Marx provides consulting services to Dimensional Fund Advisors, an investment firm headquartered in Austin, Texas, with strong ties to the academic community. The thoughts and opinions expressed in this article are those of the authors alone, and no other person or institution has any control over its content.
Editor’s Note
Submitted 28 June 2018
Accepted 27 September 2018 by Stephen J. Brown
Notes
1 In untabulated results, we examined an alternative definition according to which large caps are in the top 80% of total market capitalization, small caps are those in the next 14%, and micro caps are those in the bottom 6% excluding the bottom 0.1%. Using this alternative universe definition, our conclusions are largely unchanged.
2 Specifically, costs are estimated using a Bayesian–Gibbs sampler on a generalized Roll (1984) model of stock price dynamics, which accounts for marketwide moves in prices. For a detailed description of the estimation procedure, see Hasbrouck (2009) or Novy-Marx and Velikov (2016). The SAS code to estimate the effective bid–ask spreads is available on Joel Hasbrouck’s website: http://pages.stern.nyu.edu/~jhasbrou/.
3 Appendix A provides detailed definitions of the variables used for the construction of the strategies.
4 Results are provided in Tables S1 and S2 in the online supplemental material (available at https://www.tandfonline.com/doi/suppl/10.1080/0015198X.2018.1547057).
5 This staggering, whereby we rebalanced a third of the portfolio each month instead of 100% of the portfolio each quarter, is common in the literature but inconsequential for our results. The staggering did yield slightly better strategy diversification, but the underlying portfolios, rebalanced on the basis of the same observed signal but on the different quarterly rebalancing cycles, always held similar stocks, were highly correlated, and had similar Sharpe ratios.
6 Tables S3 and S4 in the online supplemental material (available at https://www.tandfonline.com/doi/suppl/10.1080/0015198X.2018.1547057) show that our results are robust to using either a narrower (15%/25%) or broader (10%/40%) no-trade band.
7 Our implementation follows Novy-Marx and Velikov (2016), but similar trading rules have been proposed by both academics (Davis and Norman 1990) and practitioners (Donohue and Yip 2003; Sun, Fan, Chen, Schouwenaars, and Albota 2006). Banding is also widely used in index reconstitution; see, for example, CRSP (2018, 8); FTSE Russell (2018, 29); Research Affiliates (2018, 35).
8 The unmitigated strategies are designed so as not to incur unnecessary transaction costs. In particular, they do not trade out of positions that they would have maintained on the basis of the stock selection signal just because a stock leaves a strategy’s size universe.
9 None of the versions of the large-cap defensive strategy considered produced positive performance, but here the strategy with quarterly rebalancing realized less negative returns than the one that uses banding.
10 These two strategies are particularly complementary because their trading is negatively correlated. Stocks become small, entering the small-cap universe, after realizing poor relative performance. An investor following a small-cap strategy thus tends to buy momentum losers. The momentum screen consequently captures more momentum when used on a small-cap strategy than when used on other strategies for which the stock selection signal is not correlated with past stock price performance.