Momentum: why winners keep winning, and how it crashes

The evidence

Jegadeesh and Titman (1993) showed that US stocks with the best returns over the previous 3–12 months kept beating the worst ones by about one percentage point a month over the next 3–12 months. The pattern has since been found across countries, sectors, bonds, currencies and commodities, which makes it one of the best-documented effects in finance.

Researchers usually skip the most recent month (the 12-1 return), because very short-term returns tend to reverse. Our Trend component uses the 12-1 and 6-1 month total returns, a risk-adjusted version and the distance from the 52-week high.

Why it might work

Explanations include slow reaction to news, when investors underreact and then catch up, and herding later on. Whatever the cause, the effect fades and then reverses: the winners of the past two or three years tend to lag afterwards.

The crashes

Momentum loses badly in panic states: after a long decline, with high volatility, when the market suddenly rebounds. Daniel and Moskowitz recorded that between March and May 2009 past losers gained about 163 % while past winners gained about 8 %. Barroso and Santa-Clara found that scaling momentum down when its own volatility is high virtually eliminates the crashes and nearly doubles its risk-adjusted return.

Our scoring uses a simpler guard: in a region whose index has fallen over two years and is unusually volatile, the momentum weight is halved.

Momentum next to value

Asness, Moskowitz and Pedersen found value and momentum strongly negatively correlated, about −0.5 on average, and an equal blend had a much higher risk-adjusted return than either. In plain words, cheap stocks that have started to recover are a different bet from cheap stocks that are still falling. That is why the Smart Score combines both.

Using it sensibly

• Treat momentum as one input among several, not as a trading signal.
• Expect turnover: the list of winners changes, and trading costs add up.
• Be careful about adding to positions in the weakest trend group unless the fundamentals are strong.
• Remember the haircut: published effects lost more than half of their returns after publication (about 58 % on average).