Q&A on Short-Run Reversals with Mamdouh Medhat and Robert Novy-Marx

Dimensional is known for its deep ties to the academic community, allowing the firm to stay at the forefront of financial science. A recent example is the research paper “Reversals and the Returns to Liquidity Provision,” coauthored by Professor Robert Novy-Marx and Dimensional researchers Savina Rizova, Wei Dai, and Mamdouh Medhat. Robert Novy-Marx is a world-renowned academic in the field of asset pricing, with publications in top journals on topics such as profitability, momentum, and transaction costs. In this Q&A, Mamdouh asks the professor for his thoughts on the recent paper, its implications, and the collaboration with Dimensional.

Mamdouh: Robert, thanks for taking the time to do this Q&A on our paper! I’ve put together some questions that I hope will help shed light on the paper’s findings and implications. So, let’s start with the basics. Short-run reversal is well-known in academic circles but perhaps less so more broadly. Could you define it for us?

Robert: Sure. Short-run reversal is the tendency for recent winners to underperform recent losers over short horizons. If you rank stocks on their past performance over, say, the most recent month, you tend to see that ranking reverse over the next month or so. Gene Fama was the first to document the effect back in 1965.1 Short-run reversal is, in a sense, the opposite of the momentum effect, which is the tendency for winners over the last six to 12 months to continue to outperform losers. It’s the reason people skip the most recent month when constructing momentum signals for individual stocks.

Mamdouh: Given we’ve known about reversals for nearly 60 years, some might be wondering why we would want to study them again. In the paper, we argue it’s because understanding reversals can help us understand liquidity. Could you elaborate on that connection?

Robert: In theory, the returns to reversal strategies approximate the returns earned by liquidity providers, such as market makers, for the services they offer.

When investors want to quickly sell a large amount of a stock, they will generally be forced to sell to market makers at a discount because of the demand for immediate liquidity. The price will then tend to revert to a higher level once the liquidity demand is alleviated. This reversal compensates market makers for providing liquidity since, in doing so, they are exposed to the risk of the price falling before they can offload their inventory. You can imagine the opposite happening when investors want to quickly buy a large amount of a stock.

Our paper contributes to the literature that uses reversals to study liquidity. An influential paper from 2012 uses reversals to quantify how much more costly it is to demand liquidity during highly volatile periods, such as the financial crisis of 2007–2009.2 That analysis focuses on the co-movement over time between reversal-strategy returns and measures of aggregate market liquidity. Our paper complements those findings by using reversals to shed light on how the returns to liquidity provision vary across different groups of stocks.

Mamdouh: Given that reversals are in theory linked to liquidity, there are surprisingly few results in the literature on how reversals vary across stocks with different liquidity. That’s the main contribution of our paper, and I think the results are particularly compelling because they’re so intuitive. Could you unpack the paper’s main results for us?

Robert: Yes, in theory, the less liquid the stock, the higher the expected return to providing liquidity for it. Reversals should therefore be stronger among less liquid stocks. Liquidity, however, is complex and multidimensional. No single variable is a comprehensive measure of it. So, in the paper, we study how reversals vary across groups of stocks that differ on three characteristics associated with different aspects of liquidity: size, volatility, and turnover.

Intuitively, smaller and more volatile stocks are riskier to hold in a market maker’s inventory. Similarly, a stock that trades less often is likely to stay in the inventory for longer. Reversals should therefore be stronger among smaller and more volatile stocks, and more persistent among stocks with lower turnover. We confirm this empirically and find that the effects of volatility and turnover are particularly dramatic. More volatile stocks see faster, initially stronger reversals, while stocks with lower turnover see more persistent, ultimately stronger reversals. The paper’s Figure 1 shows those effects for US stocks. We find broadly consistent results in non-US developed markets and in emerging markets.

Mamdouh: A more subtle point is how one constructs reversal strategies. In the paper, we argue for using past performance adjusted for news-related effects. We also argue for using shorter look-back periods and measuring strategy returns more frequently than in the related literature. Why are those considerations important?

Robert: The way you construct reversal strategies matters a lot when you’re using them to study liquidity. In theory, when a stock’s price changes due to liquidity trades, we should see a reversal because such price changes are temporary. In contrast, when a stock’s price changes due to information, such as news about the stock, we should not see a reversal. So, if we remove the conflating effects of news from a stock’s past returns, we are more likely to isolate liquidity-driven reversals. This should also improve the performance of reversal strategies because, empirically, news-related effects are associated with momentum, not reversal.

In the paper, we show that standard reversals, based on unadjusted past returns, appear weak because the signal you’re sorting on is polluted by at least two important news-related effects: post-earnings-announcement drift and short-run industry momentum. Reversal strategies based on past returns adjusted for earnings announcements and industry momentum correspond better to the liquidity-driven reversals suggested by theory. They also perform much better.

Further, the standard construction of reversal strategies in the literature is to sort stocks on past one-month returns and evaluate the strategy’s performance a month later. The monthly horizon, however, is simply too long among the stocks where reversals are most short-lived. For high-volatility and high-turnover stocks, only the most recent past returns are informative about reversals, and the reversal itself lasts a few weeks at most. Among such stocks, it’s more sensible to use a shorter look-back period and evaluate strategy performance more frequently.

Mamdouh: One way the paper helps us understand liquidity is that it explains seemingly surprising findings from the earlier literature. Could you give us an example of that?

Robert: Several recent papers highlight the remarkably strong performance of reversal strategies among low-volatility stocks. The machine-learning literature, for instance, absolutely loves this strategy because it appears so strong.³ This has left the impression that reversals are stronger among less volatile stocks. That’s not the case. Reversals are in fact stronger among more volatile stocks, which is more intuitive, but are also more short-lived there. High-volatility reversals only look weak because the literature has constructed them using a one-month look-back window, which is too long, and evaluated the strategy’s performance monthly, which is too late.

There are other examples of surprising findings in the literature that can be explained through the paper’s results. More generally, the paper’s results highlight the importance of studying and implementing market phenomena at their natural frequency.

Mamdouh: The paper’s main results use US data going back to the 1970s. A question we often get is whether liquidity issues are still as relevant today. How would you respond to that? 

Robert: It’s true we’ve seen improved market liquidity and generally lower transaction costs over time, and especially over the last two decades. Nonetheless, a better understanding of liquidity is still important and relevant. Liquidity varies over time and can quickly dry up during volatile periods, implying dramatic increases in the costs of trading. It also varies by region and across stocks in nontrivial ways. This is why liquidity is still an active research topic for academics. Moreover, understanding it better is equally relevant for practitioners because this can improve execution.

More specifically, the average returns to reversal strategies have indeed come down over the last 20 years, consistent with improved market liquidity. Still, we show in the paper that they remain reliable over this latter period. In addition, the cross-sectional effects we document are qualitatively consistent before and during the last two decades of our sample.

Mamdouh: The average returns to reversal strategies are reliable, even in the last 20 years. Still, we argue this does not imply one can directly benefit from a reversal strategy because the effect is short-lived. How else might we benefit from the paper’s findings?

Robert: I think the paper has clear real-world implications. A better understanding of liquidity can improve execution. Trading in a way that accounts for reversals can reduce the cost of demanding liquidity and increase the compensation for providing it.

Importantly, this is not limited to market makers but applies more generally to anyone who trades. A standalone reversal strategy is impractical because it will have extremely high turnover and trading costs. Nonetheless, you can increase the expected return of a strategy based on more persistent return drivers by incorporating a reversal screen into the strategy’s natural rebalancing process. The screen would delay trades that demand costly liquidity but delay them no longer than necessary. For a broadly diversified strategy, such a screen can have measurable benefits with minimal impact on the strategy’s long-term focus.

Mamdouh: We at Dimensional are certainly excited about implementing reversal screens based on the paper’s findings. It’s nice when academic research not only helps us understand an important topic but also has direct practical implications.

Robert: Definitely. I see finance as an applied academic field. For a paper to have real-world impact, I think it’s necessary to be explicit about why people outside of academia should care about it and to pay attention to practical issues like implementation and trading costs. That’s been one of the nice things about collaborating with Dimensional on this paper. I think it’s in capable hands to be used to help improve real-world strategies.