Momentum Crashes: Why Winners Become Losers Overnight

Forty Percent Gone in Three Months

In March 2009, the S&P 500 hit its crisis low of 676. What followed should have been a moment of triumph for momentum investors -- after all, the strategy had been selecting defensive winners and shorting battered financials and cyclicals for months. Instead, it became the worst quarter in momentum's recorded history. Between March and May 2009, the canonical long-short momentum portfolio lost roughly 40 percent of its value. Five years of accumulated premium vanished in a single quarter. The stocks that momentum was short -- beaten-down financials like Citigroup and Bank of America -- surged over 100 percent. The defensive winners on the long side barely moved.

This was not a freak event. It was the predictable consequence of a mechanism that Daniel and Moskowitz (2016) would rigorously document in their landmark paper "Momentum Crashes," published in the Journal of Financial Economics. Their contribution went far beyond cataloguing historical disasters. They identified why crashes happen, showed that they are predictable, and proposed a dynamic strategy that substantially improves momentum's risk-adjusted performance. This article examines their findings in detail.

The Anatomy of a Momentum Crash

Daniel and Moskowitz began by establishing a comprehensive empirical record. Analyzing U.S. equity data from 1927 through 2013, they found that momentum's long-run performance is exceptional -- an average monthly return of 1.47 percent with a t-statistic above 4. But the distribution of returns is deeply asymmetric. Momentum exhibits strong negative skewness and excess kurtosis: the left tail is fat and dangerous.

The paper identified several major momentum crashes across the sample period:

The pattern is unmistakable. Every major crash occurred at a bear-to-bull market transition -- the precise moment when a prolonged market decline abruptly reversed. This is not coincidence; it is the direct consequence of how momentum portfolios are constructed.

The Optionality Mechanism: Why Losers Become Explosive

The core theoretical insight of Daniel and Moskowitz (2016) concerns the optionality embedded in past loser stocks. During a sustained bear market, the momentum strategy accumulates short positions in stocks that have been declining for months. Many of these are firms in severe financial distress -- their equity has been battered, and some are approaching insolvency.

Here is the critical observation: as a firm's equity value declines toward zero, its equity begins to behave like an out-of-the-money call option on the firm's assets. This follows directly from Merton's (1974) structural model of corporate debt, where equity is a call option on firm value with the strike price equal to the face value of debt. When the firm is near default, small improvements in asset value produce large percentage gains in equity.

This convexity creates an asymmetric payoff profile. If the market continues to decline, the loser stocks cannot fall much further -- many are already near zero. But if the market reverses, these distressed equities can appreciate by hundreds of percent in weeks. The momentum portfolio is short precisely these option-like securities, creating an exposure that resembles a written straddle: limited upside with catastrophic downside.

Daniel and Moskowitz formalized this by showing that the beta of the momentum portfolio is time-varying and becomes sharply negative during market recoveries. In normal periods, the momentum portfolio's market beta is close to zero or mildly positive. But conditional on the market being in a recovery from a bear market, the portfolio beta can plunge to -2 or below, meaning that every 1 percent market gain translates into a 2 percent or greater momentum loss.

Predictability: The Bear Market Signal

Perhaps the most consequential finding in the paper is that momentum crashes are not random. They are predictable -- not with certainty, but with enough reliability to build a practical risk management framework.

Daniel and Moskowitz showed that two observable variables forecast momentum crashes with significant statistical power:

1. Prior market state: Crashes cluster after prolonged bear markets. The longer and deeper the downturn, the more distressed firms accumulate on the short side and the more dangerous the optionality effect becomes.

2. Realized volatility of the momentum portfolio: When the trailing volatility of the long-short momentum spread spikes, it signals that the portfolio's constituent stocks have entered a high-dispersion regime. This elevated volatility is itself a predictor of large negative returns.

The authors demonstrated that the conditional expected return of momentum is strongly related to these state variables. When the market has recently experienced large declines and momentum portfolio volatility is elevated, the expected return of momentum drops sharply and can turn negative. Conversely, in calm markets with no recent crash, momentum's expected return is highly positive.

This predictability distinguishes momentum crashes from true black swan events. A black swan is, by definition, unpredictable. Momentum crashes are better described as gray swans -- rare, severe, and preceded by identifiable warning signs that most investors choose to ignore.

The Dynamic Momentum Strategy

Armed with the predictability result, Daniel and Moskowitz constructed a dynamic momentum strategy that adjusts its exposure based on the forecasted variance of momentum returns. The approach is conceptually straightforward: take a full-sized momentum bet when conditions are favorable, and reduce exposure when crash risk is elevated.

Specifically, the strategy scales the portfolio weight inversely with the conditional variance of momentum returns. The conditional variance is estimated using a simple model that incorporates recent momentum volatility and the bear market indicator. When volatility is low and the market has not recently crashed, the strategy takes maximum exposure. When the bear-market signal is triggered and volatility spikes, it dramatically reduces position size.

The results are striking:

The dynamic strategy nearly doubles the Sharpe ratio while cutting the maximum drawdown from a catastrophic 96.5 percent to a painful but survivable 28.4 percent. The annualized return actually increases slightly because the strategy avoids the deep losses that erode compounded wealth. This is a rare case in finance where reducing risk also improves returns -- a direct consequence of the convexity of compounding.

The improvement is not confined to the 2009 crash. The dynamic strategy also avoids or reduces losses during the 1932 crash, the 2001 dot-com reversal, and other historical episodes. The strategy's benefit is robust across sub-periods and not driven by a single outlier event.

How This Relates to Volatility-Managed Portfolios

Daniel and Moskowitz's dynamic approach is closely related to the contemporaneous work by Barroso and Santa-Clara (2015), who proposed a volatility-managed momentum strategy that scales exposure inversely with realized momentum volatility. Both papers share the insight that momentum's time-varying risk is predictable and can be managed.

The key difference lies in the conditioning information. Barroso and Santa-Clara use only the trailing six-month realized volatility of momentum returns. Daniel and Moskowitz incorporate additional information about the market state -- specifically whether the market is recovering from a bear market. This richer conditioning set allows the Daniel-Moskowitz strategy to capture predictability that pure volatility scaling misses.

The two approaches are complementary rather than competing. In practice, many quantitative investors now implement some variant of volatility-scaled momentum, often combining elements from both papers. The core principle is the same: when momentum's recent volatility rises, reduce exposure, because the elevated volatility signals heightened crash risk.

Moreira and Muir (2017) later generalized the volatility-management approach beyond momentum, showing that scaling by inverse realized variance improves the Sharpe ratio of most equity factors. However, they found that the improvement is largest for momentum -- precisely because momentum's crashes are the most severe and most predictable among standard factors.

Why This Matters for Factor Investors

The Daniel-Moskowitz findings have profound implications for anyone running a multi-factor portfolio. The main lessons are:

Momentum risk is not captured by standard models. The CAPM, Fama-French three-factor model, and even the five-factor model fail to capture the tail risk of momentum. The time-varying beta and option-like behavior of loser stocks create a non-linear risk profile that linear factor models cannot represent. This means that investors who size their momentum allocation based on normal-times volatility will be dramatically underprepared for drawdowns.

Factor timing is feasible for momentum. While factor timing is notoriously difficult for most factors -- the value premium, for instance, has shown little exploitable predictability -- momentum is an exception. The bear-market signal and volatility signal provide genuine out-of-sample predictive power. This makes momentum one of the few factors where dynamic allocation demonstrably improves risk-adjusted returns.

The alpha of momentum partially reflects crash risk compensation. One interpretation of momentum's high average return is that it compensates investors for bearing infrequent but severe losses -- similar to how selling insurance or writing put options generates steady income punctuated by catastrophic payouts. Daniel and Moskowitz show that roughly half of momentum's alpha (measured against standard factor models) survives even after accounting for crash risk, suggesting that behavioral explanations (underreaction, herding) remain important. But the other half may indeed reflect rational compensation for bearing tail risk.

Limitations and Open Questions

The Daniel-Moskowitz dynamic strategy, while powerful, is not a complete solution. Several caveats deserve attention.

The bear-market signal requires judgment about market regimes. In real time, it is not always clear whether a market decline constitutes the kind of prolonged bear market that precedes momentum crashes. The 2018 and 2020 market selloffs, for instance, were sharp but brief -- momentum suffered temporary losses but recovered without a full crash. Implementing a binary bear-market indicator in real time involves look-ahead bias risks that the paper's in-sample analysis cannot fully address.

Transaction costs are amplified by dynamic scaling. When the strategy rapidly reduces and then restores momentum exposure, it generates substantial turnover. Frazzini, Israel, and Moskowitz (2018) estimated that transaction costs consume 40 to 50 percent of gross momentum returns under static implementation; dynamic strategies likely face even higher costs due to the urgency of crash-period trading.

The sample of crash events is small. While the paper covers nearly a century of data, there are only a handful of true momentum crash episodes. This makes it difficult to assess the statistical reliability of the predictive model with high confidence, despite the large t-statistics. The out-of-sample track record since publication includes the COVID-19 crash of 2020, which produced a modest momentum drawdown consistent with the dynamic strategy's predictions, but a sample of one additional event does not settle the question.

Conclusion: A More Honest View of Momentum

Daniel and Moskowitz (2016) changed how sophisticated investors think about the momentum factor. Before their paper, momentum crashes were treated as unfortunate but unpredictable events -- acts of God that one simply had to endure. After it, the field understood that crashes follow a specific and identifiable pattern: bear markets create distressed firms whose equity behaves like call options; when the market recovers, these options explode upward, devastating momentum shorts.

The practical takeaway is clear. Static momentum is a deeply flawed strategy -- its exceptional Sharpe ratio masks a drawdown profile that can destroy portfolios. Dynamic momentum, scaled by conditional variance, preserves the factor's alpha while making the tail risk manageable. For any serious quantitative investor, the Daniel-Moskowitz framework is not optional reading. It is the operating manual for surviving momentum's most dangerous moments.