AgPa #79: The Momentum OGs – 30 Years Later

Momentum: Evidence and insights 30 years later (2023)
Narasimhan Jegadeesh, Sheridan Titman
Pacific-Basin Finance Journal, URL/SSRN

Momentum is one of the strongest phenomena in financial markets. Narasimhan Jegadeesh and Sheridan Titman were among the first who documented the factor in the academic literature back in 1993. Now, 30 years later, they wrote a little overview about what happened since then. In this week’s AGNOSTIC Paper, they particularly focus on Asian stock markets and the potential explanations for sustained momentum profits. Having a good idea why someone takes the other side of a winning trade is crucial to really understand a strategy and I think this paper is quite interesting in this respect.

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

When one of the earliest papers on momentum (URL) celebrates its 30th birthday, you don’t need much motivation and ideas to do a follow-up. The authors show that momentum profits persisted over the 30 years since the original publication and provide an overview of potential explanations. Such out-of-sample tests, in this case a different time period and different regions, are the most effective tools against data mining concerns of factor strategies (URL). Having an idea why the factor worked instead of just knowing that it worked increases both evidence and conviction that the pattern is actually real. I therefore regard this week’s paper as additional robustness for the momentum factor. Compared to some pieces of the now huge momentum literature, however, the empirical analyses in the paper are not overly sophisticated.

Data and Methodology

The authors examine momentum in countries of the MSCI Asia-Pacific index (including Japan) and the G7 ex Japan. Since other papers found relatively weak evidence for momentum in Asian countries, the authors use those two groups to examine the momentum differences across regions and countries. They source US stock returns from the Center for Research in Security Prices (CRSP) database and all other returns from Datastream. Both are high-quality data providers and the authors follow the literature for best-practices of data cleaning.

They form 10 monthly momentum-decile portfolios within each stock market over the sample period from January 2000 to December 2020. The momentum signal is as classic as it can be – the return from month t-12 to t-2. As common in the literature, they skip the last two monthly returns to reduce the impact of short-term reversal. Portfolios are equal-weighted and must contain at least 10 stocks to be considered in the analysis. If a country doesn’t reach this threshold, the authors exclude it for the respective month.

This momentum signal and the decile-portfolio-sort is the common, but unrealistic, academic approach to momentum. Few real-world investors implement momentum like this and there are now countless ways to measure it besides simple 12-month returns. On the other hand, this simple and transparent methodology probably still captures a lot of the underlying phenomenon. So for the sake of this intellectual exercise, I think it is sufficient.

Important Results and Takeaways

Momentum worked internationally and out-of-sample

The authors’ simple analysis reveals that momentum, buying recent winners and selling recent losers, still worked from January 2000 to December 2020 on average. That is pretty impressive out-of-sample evidence for the late 1990s publications on momentum. It is also fully in-line with what I and many others have already written about the pervasive evidence for momentum (URL).

Surprisingly, the authors mention that “All countries in the G7 ex-Japan region, except the US, experience significant momentum” in their sample period. This was somewhat surprising to me and probably has to do with the specific time period. At least to the best of my knowledge, US momentum was actually a quite good strategy over the last years.

The authors also find the same regional patterns as the literature and explain that momentum profits tend to be smaller in Asia than in the G7 ex-Japan countries. They also show that the correlation among momentum returns across countries increased over time. They attribute this result to the fact that potentially more asset managers are using momentum strategies nowadays.

Momentum is most likely not data mining

Any anomaly in financial markets has to defend itself against data mining and momentum is not different. Rightly so! Our human psychology is prone to see creative patterns where no real ones are. In addition to that, incentives of academia or the financial industry can easily lead researchers into data mining (URL).

Just to get everyone on the same page, data mining refers to the collection of biases that cause researchers to mis-interpret random patterns as real mechanisms. For example, you can easily create a profitable backtest by testing 273 different strategies and picking the best one. Needless to say, this is hindsight bias and the chance that this one particular strategy will also be the best in the future is very low.

Jegadeesh and Titman therefore provide and explain a whole bunch of robustness tests. First, they mention that they tested 36 combinations of rankings and holding periods in their original 1993-paper for joint significance. Momentum remained pervasive. If the same strategy works across 36 different specifications, chances are quite good that it is not just a random pattern.

Second, there are countless momentum studies that test the strategy across regions, sample periods, signals, and asset classes. The results are mostly similar. While we never can rule out a small chance of seeing patterns that only exist by chance, the evidence for momentum is pretty strong.

Having said that, the authors do mention that momentum profits could have decreased as more and more people started to do it. This is a fundamental feature of markets, however. Greedy investors go after profitable patterns and arbitrage them away by doing so.

The evidence speaks against risk-based explanations

There are generally two ways to explain why an investment produces or should produce excess returns. Either it carries some risk for which we receive a compensation or it is indeed an inefficiency where some investors make mistakes that we can exploit. Depending on how strong you believe in market efficiency, you either prefer the first or second.

There are some strong believers in the efficient-market-hypothesis who argue that people do not make the same errors over and over again without adjusting their behavior. In contrast, proponents of behavioral finance (URL) argue that humans suffer from psychological biases that are very hard to eliminate and provide countless examples of persistently stupid behavior (URL).

It is neither my nor the authors’ intention to convince you of something. Everyone must find their own position on that spectrum and the authors just provide an overview about potential explanations from both categories. They do provide evidence, however, why they believe that behavioral explanations are more appealing (see below). But let‘s first go to the risk-based arguments.

When momentum stocks are more risky than non-momentum stocks, it is “okay” that they deliver higher returns. Market are efficient when they offer adequate compensation for risks and when no one can outperform without taking more risk. So we need a measure of risk. This is a question worth many posts itself, but the authors just stick to what academic finance defines as risk – exposures to factors. They first show that long-short momentum portfolios tend to have negative exposures to the market (URL), the size factor (URL), and the value factor (URL). So by this definition, momentum is actually less risky and still produced profits. That doesn’t make sense and therefore speaks against the risk-based explanation.

So they also look at another type of risk. Momentum is a form of trend-following and when you follow a trend, things get really ugly when the trend suddenly changes. Daniel and Moskowitz (2016) therefore brought up the notion of momentum crashes as such strategies suffered painful losses during periods of sudden market recoveries, like for example in January 2023 (URL). So you could again argue that momentum profits compensate for exposing yourself to the risk of such crashes. The authors measure this risk by skewness and the magnitude of past crashes.1Momentum returns exhibit negative skewness. The strategy makes small positive returns most of the time, but goes through a few crashes with deeply negative returns.

Neither skewness nor past crash magnitudes are significantly related to momentum profits across countries, however. The authors argue that if momentum compensates for such risk, we should find higher profits in countries with more negative skewness or deeper crashes. Since this is not the case, they again question the risk-based explanation for momentum. I like the idea of this argument and the way the authors test it, but I personally need some more empirical evidence to finally discard it.

Underreaction and noise traders seem a plausible explanation

Coming to behavioral explanations, the authors provide the following ideas. Suppose there are two types of investors. Active investors (like institutions) who are somewhat overconfident and tend to underreact on new relevant information, and noise traders (like some retail investors) who react on irrelevant information. According to the authors, such a market composition can explain both the empirical evidence for momentum and short-term reversal. Short-term reversal is the idea that recent losers tend to outperform over short periods (about one month) before the longer-term (one to twelve months) momentum effect kicks in. Noise traders generate short-term reversal by bidding up beaten-down stocks on the basis of actually irrelevant information. Active investors generate momentum by gradually correcting their initial underreaction.

This is of course elegant theory, but what says the data? The authors provide some interesting examples, but unfortunately do not test their theory more thoroughly. They mention, for example, that a lot of momentum profits accrue around and after earnings announcements. In addition to that, past winners and losers exhibit more analyst earnings forecast revisions. According to them, both data points suggest that momentum indeed comes from an initial underreaction to important information that gradually finds its way into the price.

Another aspect that supports the authors’ theory is the composition of investors in different markets. Asian markets, for example, tend to have relatively more retail investors which is consistent with their lower momentum profits. The authors also mention Chinese A and B shares as natural experiment. A-shares are mostly owned by local retail investors and tend to exhibit no momentum but short-term reversal. In contrast, there are quite some foreign institutional investors active in B-shares and those exhibit significant momentum without too much short-term reversal. Finally, they mention that there is significant momentum among Singaporean large caps but none among small caps. Once again, the former tend to have more foreign institutional shareholders whereas the latter are mostly held by local retail investors or smaller institutions.

Overall, the authors therefore speak in favor of the behavioral explanation with active investors and noise traders. While there is certainly some truth in it, I believe there are two potential challenges. First, a collection of empirical observations is no rigor test of a theory. It would be very interesting to see how the (good and interesting!) model of the authors holds up in more rigorous tests. Second, while I do believe in many insights from behavioral finance, behavioral explanations still require an endless supply of people who make errors without adjusting their behavior. I see a lot of evidence for that in reality, but it remains a stronger assumption than the risk-based explanations. The idiots may stop trading when they recognize they are losing, but if you expose yourself to a risk, you should be compensated as long as the risk doesn’t go away.

Conclusions and Further Ideas

In my view, the paper is a really nice overview and summary over the academic history of momentum and potential explanations why it exists. It is not a sophisticated and deep empirical analyses, however. The authors replicate the evidence for momentum beyond their original sample period and around the world.

While more empirical analyses are definitely required, I like the behavioral explanation with noise traders and believe there are some quite interesting implications for investors. If momentum profits are indeed related to the composition of a market’s investor base, maybe we can take advantage of that. If the theory holds, we should focus our momentum strategies to markets with a relatively high share of smart investors and avoid momentum in countries with more retail or other unsophisticated investors. This would definitely help for portfolio construction and maybe even unlock some alpha. Even more extreme, we could also combine sophisticated-region momentum with unsophisticated-region short-term reversal. Those are just two examples that illustrates why it is helpful to not only understand that a strategy works, but also why it works.

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.


1 Momentum returns exhibit negative skewness. The strategy makes small positive returns most of the time, but goes through a few crashes with deeply negative returns.