Momentum Investing

Should You Be a Momentum Investor?

At some point, every argument about whether or not financial markets are reasonably efficient comes to the subject of momentum. Technically, this refers to two separate phenomena. First, in a given period, a portfolio that is long assets with the highest returns in the previous period (i.e., "winners") and short those with the lowest performance (i.e., "losers") will usually generate positive returns that cannot be explained by varying exposure to different risk factors (e.g., the overall market, value, or small size). Second, from period to period, there is a tendency for the past price of an asset to predict its future price. In other words, these returns are not independent of each other (technically, they have a positive correlation). In both cases, however, the bottom line is the same: investors want to know if they can "beat the market" by using momentum-based trading strategies.

At first, the answer appears to be "yes". Wherever you seem to look, you find studies that show that momentum strategies yield market-beating returns. These include investments in different countries (see, for example, "The Speculative Dynamics of World Equity Markets" by De Bondt, Fung, and Lam, or "International Momentum" by Geert Rouwenhorst); across industries, and style categories ("Momentum and Autocorrelation in Stock Returns" by Jonathan Lewellen, or "Cross-Industry Momentum" by Menzly and Ozbas), and even across asset classes (see "Passive Momentum Asset Allocation" by King, Silver, and Guo). Not that market-beating returns are guaranteed, mind you. As Griffin, Ji, and Martin note in "Global Momentum Strategies: A Portfolio Perspective", the momentum approach has generated negative returns for extended periods of time. Still, you have to wonder: should I be doing this with my portfolio?

To answer this question, let's begin with two important questions. First, what factors cause momentum? Is there a reason to expect these factors to continue to operate in the future? And second, how much evidence is there that momentum is a profitable strategy?

As befits an issue that has caused so much consternation to so many professors, there is no shortage of explanations for what may be causing the momentum effects everyone sees in the historical returns data. Broadly speaking, these fall into three classes: the cognitive limitations of some investors; information and liquidity factors, and rational, risk-based factors.

Perhaps the most common school of thought about momentum locates the cause of this phenomenon in the tendency of some investors to under and over-react to information. For example, in "Investor Psychology and Security Market Under and Over-Reactions", Daniel, Hirshliefer and Subrahmanyam describe a model in which investors use information they believe is not available to all investors ("private information") to form their initial view about the value of a stock. They will subsequently pay more attention, and give more weight to information they later receive that supports their initial view, and pay less attention (and give less weight) to information which contradicts it. For example, if a firm has previously reported 11 quarters of rising earnings, investors will probably under-react if it reports flat earnings this quarter. This is formally known as the "confirmation bias", and it affects many areas of human thinking. For example, it was the underlying problem identified by the investigations into the 911 attack, and the missing WMDs in Iraq.

Back in the world of investing, the confirmation bias can cause investors to become overconfident about the accuracy of their views, and to keep buying a stock and driving its price higher. In "Short Term Momentum and Long-Term Reversal: An Experimental Investigation", Bloomfield Taylor and Zhou show how some investors' reluctance to realize their losses (the so-called "disposition effect") can further contribute to the under-reaction of stock prices to new information. Together, these factors sometimes cause a stock price to rise above its fundamental value. However, it will not fall until sufficient "disconfirming" evidence accumulates in the minds of enough investors (there is an old saying that it takes twice as much information to change an opinion than it does to form it in the first place).

In "The Role of Delisted Firms in the Momentum Puzzle", Assaf Eisdorfer looks for evidence of under and over-reaction. He shows how roughly ten percent of the stocks in the "winner" and "loser" portfolios used in many U.S. momentum studies are delisted during the holding period, but account for 40 percent of reported momentum profits. De-listed winners are firms that are acquired; de-listed losers are those that go bankrupt. Eisdorfer concludes that under-reaction to bad news seems more important to momentum profits than over-reaction to good news, with the shares of merged firms being more accurately priced than those that eventually go bankrupt. However, in "The 52 Week High and Momentum Investing", George and Hwang find evidence for over-reaction, and show how nearness of a stock's price to its 52 week high is a better predictor of future momentum profits than even its past returns.

In "Simple Forecasts and Paradigm Shifts", Hong and Stein note that while the actual process generating stock returns is a very complex one (involving both firm level variables and the actions of other investors), our cognitive limitations force us to think about it using relatively simple models, which we only slowly update. Hong and Stein show how these "learning effects" generate momentum effects very similar to those produced by the confirmation bias. In another paper ("A Unified Theory of Underreaction, Momentum Trading and Overreaction in Asset Markets"), the same two authors assume a market populated by two different types of investors. "Newswatchers" essentially trade on the fundamentals, based on the information they receive. "Momentum traders" trade on the basis of changes they observe in the price of a stock. If information diffuses to the newswatchers with some delay (i.e., some get it before others), this will initially cause returns (or price changes) in successive periods to be the same. In turn, this will attract the attention of the momentum traders, who will push the price of the stock still higher (or lower, as the case may be). This process can easily send a stock price to levels above or below its fundamental value. The key question is the percentage of newswatchers relative to momentum traders in the market (or, more accurately, the amount of capital each group controls). If there is a relatively high percentage of newswatchers (or "fundamental traders" as they are called in similar studies), any overpricing caused by the momentum traders is quickly corrected (and momentum profits are small). On the other hand, if there is a relatively high percentage of momentum traders, the process can go on for quite some time before it is reversed.

A number of papers focus in on the issue of when there will be relatively more momentum traders active in a market. In "Market States and Momentum", Cooper, Gutierrez and Hameed show how the profitability of a momentum strategy critically depends on the state of the market. They define a positive state as one where the past three years' returns have been positive, and a negative state as one when they have not. In the United States, they find that between 1929 and 1995 all momentum profits have come during positive market states. In "Market States and International Momentum Strategies," Dayong Huang finds qualified support for this conclusion in markets outside the United States.

Even if the percentage of fundamental traders is reasonably high, prices can still become substantially overvalued if there are barriers that prevent them from completely arbitraging away the excesses created by the momentum traders. For example, Bloomfield, Taylor and Zhou also found in their experimental market that the existence of potential margin calls reduced fundamental traders' willingness to take risk to earn arbitrage profits by shorting the shares momentum trading had overvalued. In their market, you could sell short a stock that you believed overvalued; however, to keep your position open you had to keep putting up more money if the stock price kept going higher. If you didn't have enough cash to meet a margin call, your position could be closed out at a big loss. This limited the amount of arbitrage that took place, resulting in more volatile stock prices. Here's another example of this phenomenon: how many people do you know who believed U.S. equities were overvalued in 1998 and 1999, and spent a lot of money buying put options that always expired out of the money as the market kept rising? By 2000, a lot of them had stopped buying the puts, only to find out (painfully) that they had been right in their fundamental view, just off on their estimate of how long it would take a critical mass of investors to share it.

These examples raise another critical point. Momentum profits may also be due in part to the actions of rational fundamental investors who choose to "join the bandwagon" in the belief that they will be able to get off before it goes over the cliff. Indeed, in "Hedge Funds and the Technology Bubble", Nagel and Brunnermeier show that a number of hedge funds took exactly this view between 1998 and 2000.

An altogether different cognitive explanation of momentum is put forth by Jonathan Lewellen in his paper "Momentum and Autocorrelation in Stock Returns." He finds that rather than under and overreaction, momentum is a result of the excessive correlation between returns caused by investors incorrect belief that information they receive about one company provides valuable information about another one. This is reflected in levels of correlation between stock returns that are higher than the correlations between their dividends and cash flows. In their paper "Limited Attention and Asset Prices", Peng and Xiong provide a more elaborate explanation of this phenomenon, showing how limited investor attention causes people to group companies into aggregate categories (e.g., growth vs. value, sectors, small cap, etc.) that cause equity returns to covary more than is justified by their underlying fundamentals. Two other papers, "Style Investing" by Barberis and Shleifer and "Style Effects" by Teo and Woo find evidence for this in the historical data, in the form of style-level momentum effects.

The second major explanation for momentum profits looks to information and liquidity factors rather than limited investor attention and cognitive resources. In "Information Uncertainty and Stock Returns", Frank Zhang defines "information uncertainty" as ambiguity about the valuation implications of information received by an investor. He employes different proxies to measure this (e.g., the dispersion in analyst earnings estimates for a stock), and finds that under-reaction is linked to information uncertainty. In other words, even when all investors receive a piece of information at the same time, to the extent that its meaning is uncertain, its impact on stock prices will only take place gradually, giving rise to momentum.

Another take on the information issue is found in "Predicting Stock Price Movements from Past Returns: The Role of Consistency and Tax Loss Selling" by Grinblatt and Moskowitz. They find that returns are strongly negative in December for losing firms in the U.S., which implies that "tax loss trading accounts for a good portion of the profitability of momentum strategies." Information is also closely related to the extent to which investors and market makers will provide liquidity to the market as a whole, and to the individual markets for different stocks. In "Liquidity Risk and Asset Pricing", Ronnie Sadka shows how liquidity varies over time, which causes investors to demand a return premium for bearing liquidity risk. Moreover, this risk is associated with both changes in information, as well as the extent of momentum traders' activity in the market. The paper finds that momentum profits are in good measure compensation for bearing higher levels of liquidity risk. In "Liquidity, Market Sentiment and Momentum", Akiko Fujimoto ties a number of strands of research together and finds that momentum profits are highest when the market is in a positive state (following successive positive returns) and liquidity is high (implying a high level of activity by momentum traders).

In addition to liquidity, there are two other possible risk-based explanations for momentum profits. The first is that they are related to varying levels of macroeconomic risks. However, in "Momentum Investing and Business Cycle Risk", Griffin, Ji, and Martin present substantial evidence from around the world that this is not the case. The second risk-based explanation for momentum profits shows much more promise. Beginning with a critical paper by Berk, Green, and Naik ("Optimal Investment, Growth Options, and Security Returns"), a number of writers have explored the (common sense, but long overlooked) idea that what goes on inside companies has something to do with the equity returns we observe. The key to their argument is that firms change over time, in terms of their riskiness and the percentage of their market value that reflects cash flows from current assets versus as yet unrealized future growth options.

For example, managers seek new investments with low risk and high expected returns; when they make them, they reduce the overall riskiness of their firm's cash flows. This causes investors to perceive the firm as less risky, and require a lower rate of return to hold its stock. This lower rate of return increases the net present value of the firm's future cash flows, and hence its market value. And the increase in its market value lowers the expected future return on its stock. This process also works the other way, when low risk investments either wear out or become obsolete (perhaps due to changes in technology, customer needs, or the competitive environment) and must be replaced with higher risk investments. Berk, Green and Naik show how these normal processes in the life of a firm give rise to many of the effects we observe in equity prices, including value (book to market) and small cap return premiums, and momentum profits.

In a closely related paper ("Rational Momentum Effects"), Timothy Johnson shows how momentum profits are a logical result of variation in firms' expected rates of growth. Liu, Warner and Zhang confirm Johnson's findings in their paper, "Economic Fundamentals, Risk and Momentum Profits", as do Yeh and Vos from the Reserve Bank of New Zealand in their paper "A Proposal to Reach a Middle Ground." All of these papers find that a shift to higher growth is risky (something any corporate manager could confirm!), and hence investors initially demand a higher rate of return for holding these stocks. As an aside, in their paper "The Level and Persistence of Growth Rates", Chen, Karceski and Lakonishok show that investor uncertainty about future growth rates is not misplaced; in fact, future growth is almost always overestimated. (For other papers that further build on Berk, Green, and Naik, see "Anomalies" by Lu Zhang, "When Good News Means Higher Risk" by Sagi and Seasholes, and "Dynamic Beta, Time-Varying Risk Premium, and Momentum" by Hong Zhang).

Last but not least, a different set of research papers show how investment managers, also acting rationally in their own self-interest, could contribute to the momentum phenomenon. In "Momentum, Reversal, and the Trading Behavior of Money Managers," Gutierrez and Pirinsky show how professional investment managers have a strong incentive invest in stocks with high recent returns if it will help them outperform the index to which their performance is compared. Russ Wermers reached a similar conclusion in his paper "Is Money Really Smart?" where he showed how mutual fund managers tend to buy past winners and sell past losers. Of course, these active managers are also making an assumption that, due to some combination of superior information or a superior model, they will be able to sell these high momentum shares before their price eventually drops. However, as we have repeatedly noted, as with any other active management approach, this one is also exceedingly hard to apply with consistent success.

So what causes momentum? The most likely conclusion is that all of these factors, cognitive, information and liquidity related, and rational, contribute to the momentum effect we observe in the historical data. However, this begs the question, "but can I make money using this strategy?"

Lots of research papers give us plenty of reason to suspect that in practice, momentum strategies are much less profitable than they first appear. In "The Illusory Nature of Momentum Profits", Lesmond, Schill and Zhou find that "those stocks which generate large momentum returns are precisely those with high trading costs." As a result, he infers that the profitability of many theoretically attractive momentum strategies is probably overstated. In "The Cost of Trend Chasing and the Illusion of Momentum Profits", Donald Keim documents the actual costs for an institutional investor implementing a momentum strategy. He finds that "the actual costs [e.g., commission, spread, and price impact] of momentum based trades indicates that the returns reported in previous studies of simulated momentum strategies are not sufficient to cover the costs of implementing those strategies."

On the other hand, two other papers are somewhat more optimistic. In "Are Momentum Profits Robust to Trading Costs?", Korajczyk and Sadka show that while profits decline with portfolio size, momentum strategies may still be viable below a certain maximum level of deployed capital. Similarly, in "Feasible Momentum Strategies", Rey and Schmid describe a profitable strategy that was implemented using only the shares of large, liquid Swiss firms.

Last but not least, we also took a look ourselves at the potential returns and risks from possible momentum strategies. The results were instructive about the practical limits of a momentum strategy for an index-oriented investor.

Our first analysis was based on the real annual returns for eight U.S. dollar asset classes between 1989 and 2004. They included real return bonds (using estimated returns for 1989 to 1997), investment grade bonds, foreign currency bonds, commercial property, commodities, domestic equity, foreign developed market equity and emerging markets equity. Using different strategies, we formed our first momentum portfolio in 1990 (using the 1989 returns) and then updated it annually. We tested alternate portfolio decision rules, including increasing the portfolio weight (by different amounts) of the best performing asset class from the previous year; increasing the best performer's weight and decreasing the worst performer's weight by different amounts, and simply allocating equal amounts of the portfolio to a limited number of asset classes (e.g., the three or four best performing from the previous year). We did not include the possibility of taking short positions in any asset class, and we did not include transaction costs. After calculating the annual returns for the different strategies, we compared them to the returns from simply allocating a constant 12.5% of the portfolio to each asset class every year. Specifically we calculated the average excess return (alpha) of the momentum strategy, its incremental risk (i.e., its "tracking error", which equals the standard deviation of the annual alphas), and its Information Ratio (alpha divided by tracking error, a measure of incremental return per unit of incremental risk). Finally, we ran a statistical test to see if we could be 95% confident that the true Information Ratio for the strategy was different from zero.

We could not identify a momentum strategy that delivered an Information Ratio that was statistically different from zero. For example, our highest alpha strategy was to allocate 33.3% of the portfolio to the three best performing asset classes from the previous year. This generated an alpha of 1.50%, but with a tracking error of 7.38%. In other words, we generated higher returns than our equally allocated benchmark, but took on a lot of additional risk to get them. The Information Ratio for this strategy was only .203, and it was not statistically different from zero. And don't forget, in the real world transaction costs would have eaten up a good chunk of that 1.50% alpha.

We then repeated the same analysis using historical returns for 1989 to 2004 for the same eight asset classes in the U.K. (i.e., we used UK real return bonds, property, equity, etc.). Again, our highest alpha momentum strategy was to allocate 33.3% of the portfolio each year to the previous year's top three performing asset classes. In this case, the alpha was .76%, with a tracking error of 6.93% and a statistically insignificant Information Ratio of .109.

Finally, we looked at a momentum strategy applied to ten different sectors within the U.S. equity asset class. Once again, the 33.3% strategy produced the highest alpha, of 3.04% (versus the Dow Jones Total Market Index, not including transaction costs) over the 1992 to 2004 period. This strategy had a 6.46% tracking error, and an IR of .47, which was not statistically significant.

So where does this leave us? Most people will look at the above results and focus on that 3.04% alpha from the industry sector momentum strategy. Tempting, isn't it? That's the way human nature works. However, before anybody over-reacts to it, we emphasize the following points. First, it is not statistically different from zero. Second, transaction costs will undoubtedly reduce it. Third, as markets for sector based exchange traded funds become more liquid, there is every reason to believe that this alpha will be much smaller in the future, and closer to the alphas in the two broad asset class momentum strategies. Think of it this way: if we can identify this strategy, so can a lot of hedge fund managers. Finally, let's not forget that it took a significant amount of additional risk to produce this alpha. In sum, we can't help but thinking that big, long-term momentum profits are like anything else in life that seems too good to be true -- they're probably not.