AgPa #29: Cost-Mitigation Techniques

Comparing Cost-Mitigation Techniques (2019)
Robert Novy-Marx, Mihail Velikov
Financial Analysts Journal 75(1), 85-102, URL/SSRN

After recognizing that frequent trading of larger portfolios is quite costly, I was looking for tools to improve that. As a consequence, this week’s AGNOSTIC Paper examines three techniques to mitigate trading costs of systematic equity strategies and compares them by means of after-cost performance.

Everything that follows is only my summary of the original paper. So unless indicated otherwise, all tables and charts belong to the authors of the paper and I am just quoting them. The authors deserve full credit for creating this material, so please always cite the original source.

Setup and Idea

The paper is build around a fundamental trade-off for high-turnover strategies. As markets tend to react quickly on new information, most systematic strategies perform better with fast and frequent rebalancing.1This is called information- or factor decay and means that trading on a timely signal is better than trading on a stale one. On the other hand, trading is not free and trading costs reduce performance significantly. In practice, it is therefore very important to somehow manage them. In many cases, even giving up some gross-performance through non-optimal trading is more than compensated by lower costs. I know this sounds unsatisfying in the first place, but if the “lower cost” mechanism dominates, you can actually have higher net-performance from less gross-performance.

This is exactly the point where this week’s paper comes in. The authors present three specific techniques to manage turnover and trading costs for systematic equity strategies. In general, there are only two mechanisms to approach the issue. Either reduce turnover (trade less) or reduce the costs of turnover (trade cheaper).2Of course, a combination of both also works. The techniques in the paper use both mechanisms and the results are surprisingly clear.

Data and Methodology

To evaluate their cost-mitigation techniques, the authors apply them to seven well-known systematic equity strategies from the literature and industry. Specifically, they group them into accounting-, defensive-, and momentum strategies. I will not go into too much detail about the construction of those portfolios. But essentially, the authors just rank stocks by the corresponding characteristics and go long (short) the top (bottom) quintile. The critical variable to evaluate the cost-mitigation techniques is the realized net-return of the respective strategy.

The sample covers on average about 4,500 US stocks within the period between 1975 and 2016. Market- and fundamental data is from CRSP and Compustat, respectively. For trading costs, the authors estimate bid-ask-spreads according to existing methodology from the literature. They also compare their estimates with other studies and conclude that they are fairly consistent.

Unsurprisingly, trading costs strongly depend on companies’ size. The authors therefore split their sample into large-, small-, and micro caps. They define large caps as stocks within the top 90% of market capitalization in each month, small caps as the next 9%, and micro caps as the final 0.9%.3The remaining 0.1% are probably penny stocks that are not relevant for almost all investors. As of December 2016, these breakpoints were $3.95B for large caps, $468M for small caps, and $78M for micro caps. According to the authors, this is quite similar to the methodology of the well-known Russell 1000/2000 indices.

Important Results and Takeaways

Trading costs decreased but are still important

The following chart shows average trading costs (in basis points) by market capitalization over time. Just to give you a feeling for the numbers: for the most recent 5 years between 2011 and 2016, the average trading costs for large, small, and micro caps are 18, 27, and 48 basis points, respectively.

Figure 1 of Novy-Marx and Velikov (2019).

The chart also shows that trading costs strongly decreased over the last decades. This trend particularly accelerated around the mid 1990s which falls together with the rise of electronic trading and high-frequency market makers. It is also observable that trading costs spike during crises, for example around 2000 and 2008. My personal summary of this chart: trading costs (fortunately!) strongly decreased over the last years but they still matter. So it makes sense to look at mitigation-techniques.

Technique #1: focus on “cheap-to-trade” securities

The first cost-mitigation technique is to simply avoid stocks that are expensive to trade. The authors show that stock-level trading costs are fairly persistent over time, so stocks that were cheap to trade in the past tend to remain cheap in the future. Note that with this technique, the turnover of the strategies remains unchanged. So we are not trading less but we try to trade cheaper. The authors implement this technique by restricting the strategies to the cheaper half of the respective size universe at each point in time. The following table summarizes the results.

Table 3 of Novy-Marx and Velikov (2019).

The most important part are the “Net Gain” lines which present the additional annual return from applying the cost-mitigation technique (in percentage points). Except for micro-caps, the results are not very promising. For most strategies, the net gain is actually negative which means that the improved trading costs are mostly offset by lower gross-performance from restricting the universe. In my opinion, this is not too surprising. Systematic strategies live from broad diversification and a large opportunity set. So just throwing out expensive-to-trade stocks doesn’t seem to be a good technique. So let’s go the next one.

Technique #2: rebalance less frequently

The next idea is to rebalance portfolios at a lower frequency. This is now the other mechanism. We try to trade less and don’t care about how much it costs. The important advantage of this approach is that the full universe remains available. The disadvantage is that some strategies (for example, momentum) live from fast rebalancing. At a lower frequency, such strategies trade on stale signals which probably reduces gross-performance. To implement this technique, the authors simply shift from monthly- to quarterly rebalancing which reduces turnover by 2/3. The following table again summarizes the results.

Table 4 of Novy-Marx and Velikov (2019).

Overall, the numbers look somewhat better than for the first technique. Except for the two momentum strategies, the net-performance of all strategies increases. Unsurprisingly, the gains are again highest for the micro-cap universe. But even for large caps, some net gains almost reach up to 1%-point per year. While this is promising, the authors still discard the technique. They argue that the strong reduction of turnover (by 2/3 or 66.67%) doesn’t reduce trading costs by the same factor. So although the net gains are positive in absolute terms, you still give up a lot of gross-performance for relatively little cost-savings.

Technique #3: create better trading rules

From the style of the previous paragraphs, you probably expected that I saved the best for the end. The final technique again aims to reduce turnover while leaving the cost of trading unchanged. Specifically, the authors introduce a buy/hold spread to establish higher thresholds for executing trades. What does this mean?

Suppose you implement a quintile portfolio with monthly rebalancing. Each month you sell all stocks that left the quintile and buy those that entered. So you maintain timely exposure to your signal with probably quite a bit of turnover. Now suppose you do the same thing with a 20%/40% buy/hold spread. You still buy all stocks that move into the quintile (top 20%), but you only sell stocks when they leave the top 40%. This is a compromise of exposure to the signal and turnover. You still buy all stocks with the strongest and most timely exposure, but you also hold those with deteriorating exposure somewhat longer to reduce turnover.

The authors implement this technique with a 10%/30% buy/hold spread. That means they keep stocks in their decile-portfolios as long as they remain in the top (for short portfolios, bottom) 30% of the universe. Once again, the table summarizes the results.

Table 5 of Novy-Marx and Velikov (2019).

This technique yields positive net gains for all strategies in all size-categories. Especially for small- and micro-caps, the results are impressive as net-performance sometimes increases by up to 6.6%-points per year. The authors also mention that applying this cost-mitigation technique accounts for roughly half of the strategies’ net-performance. Once again, this highlights that having a good stock selection model is just half of the story. The other half is efficient implementation and cost-control.

Value- versus equal-weighted portfolios

In the last part of the paper, the authors also comment on the important issue of equal- versus value-weighted portfolios. While equal-weighted strategies often produce better gross-performance, they are much harder to implement in practice. The key problem is that equal-weighting naturally overweights small and micro caps which are, as we have seen in the very first chart, much more expensive to trade.

The authors re-estimate their strategies with equal-weighted returns and find that trading costs are about 5x larger than for value-weights. Apart from the more expensive small and micro caps, this difference also comes from higher turnover. Equal-weighted portfolios must not only rebalance to maintain exposure to the signal, they must also rebalance to maintain equal-weights.4As long as there no changes because of the signal, value-weighted portfolios remain value-weighted… The authors even find that the high trading-costs of equal-weighted strategies typically offset their higher gross-performance. So basically you can also use value-weights right away.

Conclusions and Further Ideas

In my experience, it is quite rare to find papers with such clear and unambiguous practical takeaways. Of the three cost-mitigation techniques, the authors clearly recommend the somewhat more sophisticated buy/hold spreads. Of course, you can argue that this is just one empirical backtest and only a set of seven strategies. It is absolutely fair to demand further robustness of such results.5I can reassure you, there is more evidence on such rules. Robeco, a quantitative asset manager, recently published a paper in which they use buy/hold spreads to increase net-performance of high-turnover strategies for global equities (MSCI World constituents). On the other hand, I do believe that the underlying idea of the buy/hold spreads is very plausible. It is a balance between maintaining timely exposure to a signal, keeping the opportunity set as large as possible, and still reducing turnover.

In the end, there are no “solutions” to the fundamental trade off between fast rebalancing and trading costs. At least I don’t see any reasons why bid-ask-spreads should disappear, so we have to work with them and the buy/hold spreads appear to be one of the best tools we have.



This content is for educational and informational purposes only and no substitute for professional or financial advice. The use of any information on this website is solely on your own risk and I do not take responsibility or liability for any damages that may occur. The views expressed on this website are solely my own and do not necessarily reflect the views of any organisation I am associated with. Income- or benefit-generating links are marked with a star (*). All content that is not my intellectual property is marked as such. If you own the intellectual property displayed on this website and do not agree with my use of it, please send me an e-mail and I will remedy the situation immediately. Please also read the Disclaimer.

Endnotes

Endnotes
1 This is called information- or factor decay and means that trading on a timely signal is better than trading on a stale one.
2 Of course, a combination of both also works.
3 The remaining 0.1% are probably penny stocks that are not relevant for almost all investors.
4 As long as there no changes because of the signal, value-weighted portfolios remain value-weighted…
5 I can reassure you, there is more evidence on such rules. Robeco, a quantitative asset manager, recently published a paper in which they use buy/hold spreads to increase net-performance of high-turnover strategies for global equities (MSCI World constituents).