The Momentum Factor: Why Winners Keep Winning

The Most Dangerous Free Lunch in Finance

Momentum has delivered a 7 to 8 percent annual premium in U.S. equities since 1927, making it one of the most robust anomalies in finance. But it is not a free lunch. Momentum strategies are vulnerable to sharp crashes during regime changes, and the factor behaves very differently across geographies. Korean investors should use industry-adjusted momentum. Japanese momentum has strengthened since 2010. In emerging Asian markets, momentum works best among large-cap liquid stocks.

What Is Momentum?

At its core, momentum is a simple observation: stocks that have gone up tend to keep going up, and stocks that have gone down tend to keep going down. This runs counter to the efficient market hypothesis, which predicts that past returns contain no information about future returns. Yet the pattern has persisted across decades, asset classes, and geographies.

The academic foundation was laid by Jegadeesh and Titman in their landmark 1993 paper. They showed that buying stocks with the highest returns over the prior 3 to 12 months and selling stocks with the lowest returns earned significant positive returns over the subsequent 3 to 12 months. The most common implementation (buying the top decile of 12-month performers while skipping the most recent month) became the canonical momentum factor.

The one-month skip matters. The most recent month tends to exhibit short-term reversal rather than continuation, likely due to microstructure effects like bid-ask bounce. Skipping it materially improves performance.

How Large Is the Momentum Premium?

According to AQR Capital Management data through their 2025 update, a long-short momentum portfolio in U.S. equities has delivered an average annual return of approximately 7 to 8 percent since 1927. This makes momentum comparable to the value premium, and substantially larger than the size premium over the same period.

The Sharpe ratio of standalone momentum has historically been around 0.5 to 0.6 in U.S. equities, roughly on par with the market portfolio. However, momentum's low correlation with the market means it contributes significant diversification benefits.

Perhaps the most striking finding is the breadth of evidence. Asness, Moskowitz, and Pedersen (2013) established that momentum is pervasive, appearing in stock markets spanning the US, UK, continental Europe, and Japan, and extending beyond equities into fixed income, foreign exchange, and commodities. This is not a statistical artifact confined to one market.

Why Does Momentum Exist?

The persistence of momentum is a genuine puzzle. If markets are efficient, a pattern this well-documented should have been arbitraged away. Several competing explanations exist.

The honest answer is that momentum persists because of a combination: behavioral biases create the initial trend, institutional frictions slow the correction, and inherent crash risk limits arbitrage capital.

How Momentum Differs Across Markets

Recent research from AQR and KCMI reveals that momentum is not monolithic; it behaves very differently by market.

Momentum Crashes: The Risk You Must Understand

Daniel and Moskowitz (2016) documented that momentum strategies suffer infrequent but devastating drawdowns, particularly during bear-to-bull transitions.

During prolonged downturns, momentum portfolios accumulate large short positions in beaten-down stocks and long positions in defensive winners. When the market abruptly reverses, losers snap back violently while defensive winners lag. The portfolio is caught wrong on both sides.

The 2009 crash is the canonical example. From March to May 2009, U.S. momentum lost roughly 40 percent; wiping out the prior five years of accumulated premium in a single quarter.

Several approaches help manage crash risk. Dynamic strategies that reduce exposure when momentum portfolio volatility spikes have shown promise. Combining momentum with value also helps, since the two factors are negatively correlated; momentum crashes tend to coincide with value rallies.

Practical Implementation

Signal construction: The standard signal is 12-month cumulative return excluding the most recent month. Many practitioners use 6-month or 9-month lookbacks, or a blend. For Korean equities, industry-adjusting the signal materially improves performance per KCMI findings.

Portfolio formation: Sort the universe into quintiles by momentum score, go long the top group. Rebalance monthly, or weekly with partial turnover to smooth transitions.

Risk management: At minimum, implement volatility-scaling that reduces exposure when trailing momentum volatility rises above its historical norm. More sophisticated approaches include the value-momentum barbell, sector-neutralization, and tail-risk hedging.

Transaction costs: Momentum is higher-turnover than value or quality. Use patient execution, limit turnover through partial rebalancing, and focus on liquid names. In emerging markets, the gap between paper and live returns can be substantial.

Independent Backtest: Momentum Factor by Decade

The following table presents the decade-by-decade performance of the Fama-French UMD (Up Minus Down) momentum factor, revealing both the premium's magnitude and its extreme regime dependence.

Methodology: Using monthly returns from the Fama-French UMD factor, long top-decile 12-month winners (skipping the most recent month) minus short bottom-decile losers, January 1927 through December 2025. Returns are gross of transaction costs.

The 2000s decade illustrates momentum's defining risk: the March-May 2009 crash that wiped out approximately 40% of the factor's value in under three months, dragging the decade's Sharpe ratio to 0.15. Daniel and Moskowitz (2016) documented that this crash pattern is not random; momentum crashes cluster at bear-to-bull market transitions when former losers snap back violently. The 1990s represent the golden decade, with an 11.2% premium driven by the technology trend that momentum captured early.

Reported returns are gross of all implementation costs and therefore overstate what any real-world strategy would capture. Frazzini, Israel, and Moskowitz (2018) showed that turnover-related costs alone absorb 40-50% of the raw momentum premium at institutional scale, implying a net achievable premium in the range of 3.5-4.5% per year.

Cross-Market Evidence

Momentum's universality is one of its strongest claims to factor legitimacy; and one of its most nuanced realities. The evidence is strong but geographically uneven.

In their landmark study "Value and Momentum Everywhere," Asness, Moskowitz, and Pedersen (2013) assembled the most expansive cross-market evidence to date, identifying momentum effects in eight distinct asset classes spanning several geographies. The breadth of these findings matters because it undercuts data-mining objections: the same behavioral mechanism (investor underreaction to new information) can plausibly account for the pattern wherever liquid financial markets exist.

Japan's historical weakness in momentum is the most notable exception. Chaves (2012) and Asness et al. (2013) both noted that Japanese momentum was statistically insignificant through approximately 2010. The post-2010 strengthening coincides with Japan's corporate governance reforms (the Stewardship Code (2014) and Corporate Governance Code (2015)) which increased foreign institutional participation and may have introduced the information-processing dynamics that generate momentum elsewhere. This structural shift offers a natural experiment supporting behavioral explanations for the factor.

Fama and French (2012) confirmed momentum's presence across North America, Europe, Japan, and Asia Pacific in "Size, value, and momentum in international stock returns," though they noted that the size and statistical significance of momentum varied meaningfully across regions.

The Ongoing Debate

The momentum factor occupies a unique position in financial economics: it is simultaneously one of the most robust empirical findings and one of the most theoretically uncomfortable. Several fundamental questions remain unresolved.

The persistence puzzle is the deepest. If momentum reflects behavioral underreaction, as Daniel, Hirshleifer, and Subrahmanyam (1998) proposed, why has it not been arbitraged away? McLean and Pontiff (2016) documented that a typical academic anomaly surrenders nearly a third of its excess return in new out-of-sample windows and sheds about a quarter more after journal publication. Momentum has shown some post-publication decay but remains highly significant, suggesting that implementation barriers (transaction costs, crash risk, capacity constraints) are sufficiently large to prevent full arbitrage. Frazzini, Israel, and Moskowitz (2018) showed that transaction costs alone consume 40-50% of gross momentum returns for institutional portfolios, creating a natural floor below which arbitrage becomes unprofitable.

The crash risk question is equally fundamental. Daniel and Moskowitz (2016) documented that momentum crashes are predictable in the sense that they cluster after market downturns, when former losers snap back. This predictability has spawned dynamic momentum strategies that reduce exposure during high-crash-risk periods, but the improvement is modest and comes at the cost of reduced average returns. Barroso and Santa-Clara (2015) proposed volatility-managed momentum, which scales exposure inversely with recent momentum portfolio volatility, showing meaningful Sharpe ratio improvement. However, the practical implementation of such strategies requires real-time volatility estimation and frequent rebalancing.

The relationship between time-series momentum (trend following) and cross-sectional momentum (relative strength) adds another dimension. Moskowitz, Ooi, and Pedersen (2012) documented strong time-series momentum across dozens of futures markets in "Time Series Momentum," showing that assets with positive recent returns tend to continue rising, and vice versa. While conceptually related to cross-sectional momentum, time-series momentum captures a distinct pattern and has different risk characteristics.

For practitioners, the combined evidence supports momentum as a genuine and persistent anomaly that earns a meaningful premium even after accounting for transaction costs and crash risk. The optimal implementation combines cross-sectional stock selection with dynamic risk management, uses patient execution to minimize market impact, and pairs momentum with value (their negative correlation, documented by Asness, Moskowitz, and Pedersen 2013, produces substantial diversification benefits). The premium is real, but so is the risk; and the investors best positioned to capture the momentum premium are those who understand and can tolerate its distinctive crash profile.