Momentum Turning Points Can Be Costly. Here’s How Investors Can Prepare.

Momentum, the tendency of past winner stocks to outperform past loser stocks over the next several months, is one of the most well-documented and well-researched asset pricing anomalies. In our book Your Complete Guide to Factor-Based Investing, Andrew Berkin and I presented the evidence of a momentum premium that has been persistent across long periods of time, pervasive around the globe and across asset classes, and robust to various definitions and that survives transactions costs.

In the asset pricing literature, momentum is generally defined over the short, medium, and long term in the following manner:

• Short-term reversals, defined as the prior month’s (t - 1) return.
• Medium-term momentum, defined as the returns from month t - 2 to t - 12.
• Long-term reversals, defined as the returns from month t - 13 to t - 60.

The most recent month is excluded from the medium term because it tends to show a reversal. However, in their study “Short-Term Momentum,” published in the March 2022 issue of The Review of Financial Studies (an older version can be found here), Mamdouh Medhat and Maik Schmeling demonstrated that the short-term reversal effect existed among thinly traded, low-turnover stocks, whereas high-turnover stocks exhibited short-term momentum.

The fact that the largest, most-liquid stocks actually exhibited short-term momentum with economically and statistically significant premiums was an important new finding, allowing long-only investors to exploit the short-term momentum of stocks with very low transaction costs. Medhat and Schmeling also found that short-term reversal’s high transaction costs subsumed the high gross return. However, that finding does not mean the information is without value, as a fund considering buying a micro-cap stock could delay buying based on the signal provided by the short-term reversal effect.

Shorter-Term Momentum Signals

In our study “Short-Term Trend: A Jewel Hidden in Daily Returns,” published in the November 2020 issue of The Journal of Portfolio Management, Marat Molyboga, Junkai Qian, and I examined the performance of time-series momentum using daily rather than monthly returns with standard lookback periods of one, three, and 12 months and a rebalancing period of one month. We introduced a shorter duration momentum strategy with weekly rebalancing frequency and found that the short-term momentum strategy was a strong diversifier to the longer-term strategies—the short-term strategy reduced the risks around turning points.

Momentum’s Achilles’ heel

While average time-series momentum returns have been high, momentum investors have also experienced huge drawdowns at turning points (which mark reversals in trend from uptrend to downtrend or vice versa) because momentum strategies are prone to place bad bets. As a result, time-series momentum exhibits both high kurtosis and negative skewness. As Maik Dierkes and Jan Krupski, authors of the 2022 study “Isolating Momentum Crashes,” published in the March 2022 issue of the Journal of Empirical Finance, noted:

“Since 1926, there have been several momentum crashes that feature short but persistent periods of highly negative returns. From June to August 1932 the momentum portfolio lost about 91%, followed by a second draw-down in April to July 1933. Another prominent crash took place in 2009 when momentum lost more than 73% within a period of three months. In addition to that, there have been several smaller crashes in 1938/1939, 1974/1975, and 2001/2002. While being smaller in size, each of those comprised at least one monthly loss of more than 19%. Therefore, high monthly returns of 1.15% come with a large kurtosis of 16.6 and a highly negative skewness of -2.3.”

They added: “Remarkably, crashes are driven by large gains of previous losers while winners still exhibit modestly positive returns.”

The good news is that researchers have uncovered strategies that have reduced crash risk—crashes are at least partly forecastable, tending to occur in “panic” states following market declines and when market volatility is high, and are contemporaneous with market rebounds. The authors of the 2015 study “Momentum Has Its Moments,” the 2016 study “Momentum Crashes,” the 2017 study “A Century of Evidence on Trend-Following Investing,” the 2018 study “The Impact of Volatility Targeting,” and the 2019 study “Portfolio Management of Commodity Trading Advisors with Volatility Targeting” have found:

• Long-only momentum strategies are not subject to deep crashes.
• Scaling momentum based on momentum’s mean and variance dramatically reduces the risk of crashes and greatly improves the Sharpe ratio of momentum strategies.

Extreme Volatility States

Dion Bongaerts, Xiaowei Kang, and Mathijs van Dijk, authors of the 2020 study “Conditional Volatility Targeting,” found that scaling strategies could be improved upon by adjusting risk exposures conditional on (extreme) volatility states—their strategy reduces risk exposures during high volatility states, increases risk exposures during low volatility states, and maintains an unscaled exposure otherwise. They found that not only did the conditional strategy significantly reduce drawdowns and tail risks across all major equity markets and momentum factors but also significantly reduced turnover. Their findings were consistent with those of Georg Cejnek and Florian Mair, authors of the 2020 study “Understanding Volatility-Managed Portfolios,” who found that once volatility exceeded a certain threshold, the volatility-managed portfolio always outperformed the unmanaged strategy.

Latest Momentum Research

Christian Goulding, Campbell Harvey, and Michele Mazzoleni contributed to the literature with their May 2023 study “Momentum Turning Points” published in the September 2023 issue of the Journal of Financial Economics. They began by saying:

“The speed (or sensitivity to recent data) of the momentum signal balances the tension between reducing the impact of noise and reacting quickly to turning points. This tension plays out differently for different speeds. Either the momentum signal attempts to reduce the influence of noise by having a relatively long look-back window (e.g., 12 months) but thus is slow to react to a turning point (Type II error of missed detection), or the momentum signal attempts to be fast to react to a turning point by having a relatively short look-back window (e.g., one month) and, therefore, is more influenced by noise (Type I error of false alarm).”

While noting that the literature has documented that time-series-momentum strategies based on slow momentum tend to perform better than strategies based on fast momentum, to address that issue they developed a model of expected returns to examine connections between the performance of different speeds and the following unobservable variables: trend (persistence in expected returns), turning points, and noise levels in realized returns.

They found that their model indicated that when the slow strategy, or SLOW, outperformed the fast strategy, or FAST, in the long run, it was because expected returns were relatively persistent and realized returns were relatively noisy. Yet, in the short run after turning points, higher persistence translated into short-run average gains for FAST, while both higher persistence and higher noise translated into higher losses for SLOW. Thus, the same conditions that make SLOW attractive overall relative to FAST can also make it prone to suffer more around turning points, and at these times, FAST can be more effective.

Thus, they hypothesized: “These connections suggest that the union of information embedded in SLOW and FAST positions can be useful in detecting market turning points. When bets indicated by SLOW and FAST disagree, the intuition is that the market is more likely to be at a turning point. The agreement of SLOW and FAST to go long (short) is more likely to indicate the market is in the midst of an uptrend (downtrend).”

Bull and bear markets were defined as states where SLOW and FAST agreed. When SLOW and FAST disagreed, they called it a “correction” state if SLOW indicated a long position and a “rebound” if SLOW indicated a short position.

Finally, they also studied dynamic strategies with speeds that might vary each month based on observed market cycles. They derived the state-dependent speed rule, which yielded the maximum Sharpe ratio. They estimated this strategy using historical returns. Their data sample spanned the 50-year period of 1969-2018. Following is a summary of their key findings:

• Bull months were the most common, with a relative monthly frequency of 48.3%, followed by correction months (24.5%), bear months (16.7%), and rebounds (10.5%). Thus, about 35% of the time, the slow and fast signals were in disagreement.
• When bets indicated by SLOW and FAST disagreed, the market was more likely to be at a turning point.
• The agreement of SLOW and FAST to go long (short) was more likely to indicate the market was in the midst of an uptrend (downtrend).
• Bull market states were followed by relatively high average returns with low volatility, while bear market states were followed by negative average returns with the highest relative volatility.
• Correction states were followed by deteriorating average returns, increased volatility, and severe downside outcomes—possibly a lead-up to a bear state.
• Rebound states were followed by average returns and skewness similar to bull phases but with higher volatility—possibly a lead-up to a bull state.
• Corrections and rebounds were significant in frequency, combining for more than one third of the time.
• Market cycles had close connections to the macroeconomy and the business cycle; in particular, bear states had negative expected returns and were closely linked with high macroeconomic risk states.
• Monetary policy shocks were linked to market states—surprise cuts were associated with rebound phases and hikes with correction phases.
• Bear states were the highest relative frequency state in the early months of recessions—occurring at twice the relative frequency of the other three states combined.
• Bull and rebound states, which predicted positive subsequent returns, increased to dominant relative frequencies late in periods of recession—over 90% combined relative frequency.
• Intermediate-speed momentum strategies had higher Sharpe ratios than the average Sharpe ratios of SLOW or FAST. They also further reduced exposure to extreme downside events by scaling down after corrections and rebounds. They linked this behavior to the volatility of returns following turning-point states.
• Market-timing drove about two thirds of the alpha of time-series momentum strategies on the U.S. stock market, with volatility timing driving the remaining one third.
• The Sharpe ratios of intermediate-speed strategies were higher than the average of the Sharpe ratios of SLOW and FAST uniformly across all 20 international markets (data began for most countries in 1980).
• The Sharpe ratio of the dynamic-speed strategy was higher than the highest static-speed strategy for most countries.
• State-dependent speed analysis elected slower-speed momentum after correction months and faster-speed momentum after rebound months.

Their findings led Goulding, Harvey, and Mazzoleni to conclude:

“We find relatively high persistence and high noise, properties that make the union of information in slow and fast momentum signals relevant for detecting trends and momentum turning points.”

They added: “Intermediate-speed strategies, formed by blending slow and fast [time-series] momentum, vary the exposure to the good bets associated with uptrend (bull) or downtrend (bear) phases and the bad bets associated with turning points (correction or rebound). For the U.S. stock market, we find that intermediate-speed strategies empirically exhibit many advantages over slow and fast strategies, including higher Sharpe ratios, less severe drawdowns, more positive skewness, and higher significance of alphas.”

They also found that despite being simple in construction, the four market states (bull, bear, correction, and rebound) possessed predictive information for stock market returns and had close connections to the macroeconomy and the business cycle.

Investor Takeaways

The empirical research demonstrates that, on average, investing in previous winners and short-selling previous losers has offered significant returns that cannot be explained by other common risk factors. But momentum also displayed huge tail risk, as there were short but persistent periods of highly negative returns. Crashes occurred particularly in reversals from bear markets when the momentum portfolio displayed a negative market beta and momentum volatility was high.

Fortunately, the empirical evidence has found that crashes were at least partially predictable and that the predictability and, thus, the performance of momentum strategies can be improved by employing systematic strategies.

The views expressed here are the author’s. Larry Swedroe is head of financial and economic research for Buckingham Wealth Partners, collectively Buckingham Strategic Wealth, LLC and Buckingham Strategic Partners, LLC.

For informational and educational purposes only and should not be construed as specific investment, accounting, legal, or tax advice. Certain information is based on third party data and may become outdated or otherwise superseded without notice. Third party information is deemed to be reliable, but its accuracy and completeness cannot be guaranteed. The securities mentioned should not be construed as a recommendation. By clicking on any of the links above, you acknowledge that they are solely for your convenience, and do not necessarily imply any affiliations, sponsorships, endorsements or representations whatsoever by us regarding third-party websites. We are not responsible for the content, availability or privacy policies of these sites, and shall not be responsible or liable for any information, opinions, advice, products or services available on or through them. The opinions expressed here are their own and may not accurately reflect those of Buckingham Strategic Wealth or its affiliates. R-23-564