Momentum Crashes

Picture a rubber band stretched further and further as a market falls: the more beaten-down and high-beta a stock gets, the more tension builds. Momentum's short leg is made of exactly these stocks — the ones that fell hardest. That's fine while the market keeps falling. But the moment the market snaps back, the rubber band doesn't ease gently — it fires back hardest on the stocks that were stretched furthest, and those are precisely the stocks momentum is short. A momentum crash is that snap-back: a short, violent loss concentrated in the weeks right after a market bottom, not during the selloff itself.

Why the short leg is a coiled spring

After a sustained downturn, the momentum strategy's short book fills with stocks that (a) fell the most and (b) tend to be high-beta, high-volatility, distressed names — think small, leveraged, cyclical companies. Call the short leg's average market beta $\beta_S$. If the market then rallies by $r_M$, the short leg's expected loss from beta exposure alone is approximately

$\text{loss} \approx \beta_S \times r_M$

In words: how much the short-leg drags against you is roughly its average sensitivity to the market times how big the market's bounce is. During a normal bull run $\beta_S$ and $r_M$ are unremarkable. Right after a crash, $\beta_S$ is unusually high (the short leg is full of exactly the wrong kind of stock) and $r_M$ tends to be unusually large (bounces off a bottom are sharp) — the two multiply together at the worst possible moment.

Worked example 1. In the two months after the March 2009 market bottom, the S&P 500 rallied roughly 25%. A standard 12-1 cross-sectional momentum short book at that point was loaded with distressed financials and cyclicals carrying an average beta near 3 relative to the market (some studies estimated momentum's short-leg beta spiking above 2.5–3.0 in exactly this window, versus roughly 1.0 in calm periods). Plugging in: $\text{loss} \approx 3.0 \times 25\% = 75\%$ against the short position. Momentum's long-short portfolio, which had looked like a steady, low-volatility strategy for a decade, lost over 40% in a matter of weeks — the single worst drawdown in the momentum factor's live and backtested history, dwarfing anything seen in 1929–2008.

Run the mean-reverting path a few times — momentum crashes are best understood as this kind of sharp reversion, not a random walk further downward: after being stretched far from its recent trend, the price snaps back hard, and it's the size and speed of that snap that momentum's short leg is exposed to.

Timing, not magnitude, is the real problem

Barroso and Santa-Clara (2015) showed that momentum's own realized volatility spikes sharply right before a crash — because the strategy's beta exposure and dispersion both widen as the short leg fills with distressed names. That gives a practical hook: scale the momentum position down when its recent volatility is elevated, rather than trying to predict market direction.

$w_t = \frac{\sigma_{target}}{\hat\sigma_t}$

In words: today's momentum position size is the target volatility divided by momentum's own estimated recent volatility — when momentum has been unusually turbulent, cut the size, and scale back up when it calms down.

Worked example 2. Suppose a fund targets 10% annualized volatility on its momentum book ($\sigma_{target} = 10\%$). In a calm period the strategy's trailing realized volatility is $\hat\sigma_t = 10\%$, so $w_t = 10\%/10\% = 1.0$ — full size. Heading into early 2009, momentum's trailing realized volatility had risen to roughly 40% annualized as dispersion widened, so $w_t = 10\%/40\% = 0.25$ — a vol-managed version would already be running at a quarter size before the crash hit, cutting the subsequent 40%+ drawdown to something closer to 10%. Barroso and Santa-Clara found this single rule roughly halved momentum's crash risk over the full sample without materially hurting its average return.

What this means in practice

A momentum sleeve run without any volatility management will, at some point in a multi-decade track record, hand back a large chunk of its cumulative gains in a few weeks — that's not a bug in the implementation, it's the structural signature of the strategy. Desks that run momentum alongside value or quality often lean on the fact that these crashes tend to coincide with periods when value is doing well, since both are reacting to the same sharp reversal, which is part of why multi-factor blends are more common than standalone momentum books.