Momentum & factors

Momentum crashes: why winners-minus-losers can lose so much so fast

Momentum & factors · Published 2026-10-21

Momentum strategies earn steady returns most of the time and then, occasionally, lose a large fraction of their value in a few weeks. These episodes, called momentum crashes, are not random accidents. They cluster in a specific market situation: a sharp rebound after a long decline. Understanding why is essential for anyone studying or running a momentum portfolio.

Educational analysis based on published research; not a forecast or trading advice.

The shape of momentum returns

A long-short momentum portfolio buys recent winners and sells recent losers, as described in our overview of momentum strategies. Its historical average return has been high relative to its volatility, but the distribution of monthly returns is far from normal. It shows negative skewness, meaning the large moves are mostly losses, and excess kurtosis, meaning extreme months happen more often than a bell curve would suggest. A strategy can look attractive by Sharpe ratio and still carry a risk of sudden, deep losses.

What happened in 1932 and 2009

Kent Daniel and Tobias Moskowitz (2016), in "Momentum crashes" in the Journal of Financial Economics, studied U.S. momentum from 1927 onward. The worst episodes were concentrated in two periods. In the summer of 1932, as the market rebounded from the depths of the Depression, the winners-minus-losers portfolio lost the vast majority of its value within a couple of months. In the spring of 2009, after the financial crisis, momentum suffered another severe drawdown as the market turned up and the most beaten-down stocks rallied hardest.

The pattern was the same both times: a prolonged bear market, high volatility, then a fast recovery. The losers, the stocks that had fallen most, were often highly leveraged or distressed companies. When the market recovered, they rose much more than the market, and the momentum portfolio was short exactly those stocks.

Why rebounds hurt momentum

Time-varying beta

After a long decline, the winners are typically defensive, low-beta stocks that fell least, and the losers are high-beta stocks that fell most. The long-short portfolio therefore ends up with a strongly negative market beta at exactly the moment the market is most likely to bounce. In normal times, momentum's beta is near zero, so this exposure is easy to overlook. Grundy and Martin (2001) had already shown that momentum's factor exposures swing with past market returns.

An option-like payoff

Daniel and Moskowitz describe the loser portfolio in bear markets as behaving like a call option on the market. Distressed firms have equity that resembles an option on the value of their assets: if the economy recovers, their shares can multiply. Shorting the losers is then like selling call options. Selling options earns small premiums most of the time and suffers large losses when the underlying moves sharply, which matches momentum's crash pattern.

Volatility

Crashes come in high-volatility periods. Momentum's own volatility is somewhat predictable: periods of high past volatility in the strategy tend to be followed by more high volatility. This predictability is the starting point for most of the defensive responses below.

Responses proposed in the literature

Volatility scaling

Pedro Barroso and Pedro Santa-Clara (2015) scaled the momentum portfolio each month to a constant target volatility, using the realised volatility of daily momentum returns over the previous six months. When recent volatility was high, exposure was cut. In their sample, this nearly doubled the Sharpe ratio and greatly reduced the worst losses, largely because high-volatility periods were the ones in which crashes happened.

Scaling uses only past information, so it can be applied in real time. It does require leverage in calm periods to hit the target, which brings financing costs and its own risks. Moreira and Muir (2017) found that volatility management improved several factors, not only momentum, although later work has debated how robust those gains are once trading costs and estimation are taken into account.

Dynamic weighting

Daniel and Moskowitz proposed a dynamic strategy that forecasts momentum's mean and variance from market conditions and scales exposure accordingly, cutting it sharply in bear markets with high volatility. In their tests, this improved risk-adjusted returns over the static strategy. The approach relies on a forecasting model and is therefore exposed to model error out of sample.

Market state filters

Cooper, Gutierrez and Hameed (2004) found that momentum profits depended on the state of the market: they were positive following market gains over the prior three years and close to zero following losses. A simple rule that reduces momentum exposure after a long market decline uses this finding. Our article on market timing models for a momentum strategy discusses such filters, and hidden Markov models offer a statistical way of estimating market states.

Hedging the beta

Because the crash mechanism runs through negative beta, another response is to estimate the portfolio's current beta and offset it with index futures. This neutralises the market bet but needs an accurate beta estimate during fast-moving markets, when estimates are least reliable.

Long-only and combination

Much of the crash comes from the short leg. Long-only momentum portfolios avoid the explosive loss on shorted losers, though they keep market risk. Combining momentum with value, which tends to do well when losers rebound, also smooths the path, as Asness, Moskowitz and Pedersen (2013) documented.

What this means for research

Sources