Momentum Investing

Markets have no mass. Velocity in a market is not a vector quantity; it does not conserve; it is not equal to the time derivative of position because there is no position. And yet the dominant English word for what happens when an asset’s price keeps rising is momentum, a Latin contraction of movimentum that Isaac Newton defined precisely, in 1687, as the quantity of motion arising from the velocity and quantity of matter conjointly — mass times velocity, the thing that conserves in a closed system, the property whose change requires a force.

The metaphor should have failed. Markets are not closed systems and prices have no inertia in any physical sense. But the metaphor did not fail, and that is the first interesting thing about momentum investing: a strategy named after a quantity that does not exist in the substrate it describes turns out to predict that substrate better than the leading theory said it should.

In April of 1720 Isaac Newton was Master of the Royal Mint, age 77, and a director of the East India Company. He bought shares in the South Sea Company at the start of its run, sold them in April at a profit of about £7,000, and watched, from the sidelines, as the price kept rising. By summer he could not bear it. He bought back in near the top. The bubble collapsed in September. Newton’s documented loss is reconstructed from his niece’s later account at around £20,000 — perhaps £4 million in present-day terms — most of his accumulated wealth.

The remark attributed to him afterward, I can calculate the motions of heavenly bodies, but not the madness of people, is almost certainly apocryphal in its exact wording. The pattern of behavior is not. Newton had executed, in 1720, the canonical momentum mistake: bought into a rising trend, panicked into a profitable exit, and then chased the trend back at higher prices. He died seven years later, having never written about the loss in any surviving manuscript. The richest man in England, the author of Principia Mathematica, the formalizer of the word momentum itself — undone, in part, by his own coined quantity in a domain where his physics did not apply.

In 1933, Alfred Cowles published an article in Econometrica with one of the most famous titles in the history of finance: Can Stock Market Forecasters Forecast? He had analyzed the forecasting record of 45 professional agencies — insurance companies, financial publications, market services — over the period 1928 to 1932. The conclusion, set down in three words at the end of the paper, was It is doubtful. For sixty years afterward, the dominant academic position on price prediction was a version of Cowles’s: that publicly available information, including past prices, was already reflected in current prices, and that no strategy based on past returns should beat the market on a risk-adjusted basis. Eugene Fama formalized this into the efficient markets hypothesis in 1970.

Sixty years after Cowles, in 1993, Narasimhan Jegadeesh and Sheridan Titman published a single paper in The Journal of Finance that the efficient markets framework could not absorb. The title was Returns to Buying Winners and Selling Losers. They had ranked U.S. stocks by their returns over the past three to twelve months, formed a portfolio that bought the top decile and shorted the bottom decile, and held it for three to twelve months. The strategy returned, on average, almost 1.5% per month — roughly 18% annualized — and the return was not explained by exposure to standard risk factors. The paper has been cited over twenty thousand times. The anomaly it documented, momentum, is still working three decades later, in datasets the authors never saw, in asset classes they did not test, and in countries where they had no data.

The behavioral case for why momentum works is, at heart, a single asymmetric mistake. In 1985, Hersh Shefrin and Meir Statman published a paper in The Journal of Finance with a self-describing title: The Disposition to Sell Winners Too Early and Ride Losers Too Long. Drawing on Daniel Kahneman and Amos Tversky’s prospect theory, they argued that investors hate realizing losses more than they enjoy realizing gains — at roughly a 2:1 ratio — and that this asymmetry produces a systematic error in selling decisions. The trader who is up 15% looks for an excuse to sell. The trader who is down 15% looks for a reason to hold. The aggregate effect, across millions of accounts, is that winners get sold prematurely (which slows their continued rise toward fair value) and losers get held past the point of capitulation (which slows their continued decline).

Momentum, in this reading, is not a paradox. It is the visible byproduct of a behavior pattern that one might call almost universal. The strategy that works — buy winners, short losers — is the exact opposite of what the disposition effect leads the average investor to do. Momentum is the trade that exists because most market participants cannot stomach making it themselves.

The standard academic specification of cross-sectional momentum is not simply buy stocks with the highest twelve-month return. It is buy stocks with the highest return from twelve months ago to one month ago, skipping the most recent month. The one-month skip is not cosmetic. Returns at the one-month horizon mean-revert: stocks that ran hard in the last twenty trading days tend to give some of it back in the next twenty. Returns at the one-month-to-twelve-month horizon trend. Returns at the three-to-five-year horizon mean-revert again, this time over a longer arc.

The same asset can therefore be a buy, a sell, and a buy on three different timescales simultaneously. A strategy that ignores the time horizon collapses into noise. The published momentum effect is not a claim about recency; it is a claim about a specific window, with a specific gap, on a specific scale.

In 2004, Thomas George and Chuan-Yang Hwang published The 52-Week High and Momentum Investing in The Journal of Finance. Their finding was uncomfortable. The proximity of a stock’s current price to its 52-week high — not its past return, not its trend, not any directly economic quantity — predicted future returns better than the Jegadeesh–Titman signal did. The closer a stock was to printing a new 52-week high, the more likely it was to keep rising over the next several months.

The 52-week high is not a fundamental quantity. It is a number on a financial-data page, refreshed daily, computed mechanically over a rolling window. It exists nowhere outside the page. And yet, in eighteen of twenty international markets, the page itself appears to be doing causal work — investors anchor on the printed number, hesitate to bid above it, then commit once it is decisively breached, and the breach pulls the next wave of capital. The number on the page becomes the cause of the next number on the page. Whether this is a behavioral artifact, a coordination point, or a self-fulfilling prophecy is contested. That it works, in the data, is not.

Eugene Fama won the Nobel Memorial Prize in Economic Sciences in 2013 for the empirical work that produced the efficient markets framework. In 1989, four years before the Jegadeesh–Titman paper, a 23-year-old PhD student named Cliff Asness entered the University of Chicago’s finance program and became Fama’s research assistant. By 1994 Asness had completed his doctoral dissertation. The dissertation argued — drawing on years of empirical work done partly at Goldman Sachs, where Asness had taken a job during the program — that momentum was a real, robust, profitable phenomenon in U.S. equities. It argued, in other words, against the framework his Nobel-laureate advisor had built.

Fama signed it. Asness later said in interviews that Fama’s actual instruction was, if the data says momentum exists, write what the data says. Asness founded AQR Capital Management four years later, in 1998, on the explicit premise that both value (Fama’s signal) and momentum (the anomaly) were real, persistent, and could be combined in ways the textbook denied. The firm grew, at its peak, to over $200 billion in assets under management. Asness has spent thirty years being one of momentum’s most public defenders. Fama, asked about momentum, has historically said it is the biggest embarrassment to the theory. They remain, by all reports, on warm terms.

In 2012, Tobias Moskowitz, Yao Hua Ooi, and Lasse Pedersen published Time Series Momentum in the Journal of Financial Economics. They had tested a single, mechanical rule on 58 different futures and forward contracts — equity indices, currencies, commodities, government bonds — across twenty-five years of data. For every one of those 58 instruments, returns from the past month to the past twelve months predicted the direction of returns over the next month. Sixty-three percent of monthly observations confirmed the sign of the past year’s return. A diversified portfolio applying the rule across all 58 produced a Sharpe ratio of roughly 1.4 — twice the Sharpe of the equity market itself.

A year later, Cliff Asness, Moskowitz, and Pedersen published Value and Momentum Everywhere in The Journal of Finance. This paper widened the lens further: it tested value and momentum in U.S., U.K., continental European, and Japanese stocks; in country equity indices; in government bonds; in currencies; in commodities. Both effects were present in every category. And the two strategies — value, which buys what is cheap by fundamentals, and momentum, which buys what is rising regardless — were negatively correlated with each other, at about −0.60 globally. A 50/50 portfolio of value and momentum had a markedly higher Sharpe than either strategy alone. The two main known anomalies in finance, both contradicting efficient markets in different ways, hedge each other.

The point at which momentum most reliably fails is not the point most investors expect. In March and April of 2009, in the weeks following the absolute bottom of the global financial crisis, the standard academic momentum portfolio lost 45.6% of its value. It was not the worst month of the bear market for the broad equity market — the broad market was rallying. It was the worst month for momentum because the broad market was rallying. The short side of the portfolio consisted of stocks that had been the deepest losers of 2008 — financials, deeply cyclical names, the carcass of Lehman-era panic. When the market turned, those names did not just recover. They ripped, often doubling or tripling in weeks. The short side suffered catastrophic losses while the long side, full of defensive winners, gained modestly. The spread collapsed.

Kent Daniel and Tobias Moskowitz documented this pattern in Momentum Crashes, published in the Journal of Financial Economics in 2016. The result, simplified: momentum’s tail risk is not symmetric. It crashes hardest not at the top, where the strategy has been long the high-flyers, but at the bottom, where the strategy has been short the most-hated names and those names are exactly the candidates for the most violent reversion. The 1932 momentum crash, after the Great Depression bottom, was even more severe — over −90% in a year. The pattern that ruined Newton in 1720 has a mirror in the data that ruined the strategy designed to exploit Newton’s mistake.

A strategy named after a quantity that does not apply in the medium it describes; documented as an anomaly forty years after the dominant theory predicted it could not exist; profitable, in the original specification, at about a third the rate Newton lost in 1720; written into a dissertation by a 26-year-old whose Nobel-laureate advisor signed it without changing the framework that contradicted it; demonstrated to work in 58 of 58 major futures contracts across two and a half decades of data; whose explanatory mechanism nobody can fully agree on; whose worst losses come not at the top of the market, where it is most exposed, but at the bottom, where the most-hated assets stop being hated; whose negative correlation with the only other proven market anomaly produces a free diversification that the textbook denies should exist; and which, three decades after the 1993 paper, continues to work in datasets the original authors never imagined and asset classes nobody had a name for when the paper was written. The metaphor of momentum, in finance, was wrong on its physics and right on its shape. The strategy that exploits it is the same — wrong on the framework that says it cannot exist, right on the data that says it does.