How long can a factor lose?
Every factor premium in the record was earned by someone who held through a stretch in which it looked finished. Here are two of those stretches, measured, because "factors can underperform for years" is a sentence people nod at and then sell after eighteen months.
Value, 2017 to 2020
Measured on 2026-09-04 from adjusted closes. From 3 January 2017 to 31 August 2020 the large-cap value fund IWD.US returned about 18% and its growth twin IWF.US about 124% — a gap of over a hundred points in three years and eight months, on two funds cut from the same thousand names. The value fund was profitable, so nothing on a statement said loss; what it said was that the cheap half of the market had been left a hundred points behind by the expensive half. The valuation spread was at a record. Every argument for value was intact. Money left anyway, because a three-year comparison is the horizon on which most people are judged.
From 31 August 2020 to 30 December 2022 the same two funds returned about +31% and −4%. Thirty-five points the other way in two years and four months. The premium arrived — after the horizon on which most holders had stopped waiting for it.
Momentum, spring 2009
The momentum crash is the other shape: not slow, fast. In March 2009 the market turned, and the shares that rebounded hardest were the ones that had fallen hardest — banks and cyclicals that had lost most of their value. A 12-1 momentum sort was short those and long the defensive names that had held up. Daniel and Moskowitz's 2016 paper is the account: in their long-short series the winners-minus-losers portfolio lost roughly three quarters of its value over three months, most of it in a few weeks. A long-only momentum fund lagged rather than collapsed, because it was not short the rebound; it simply did not own it.
The mechanism is a feature of the factor. Momentum's losers are, at a market bottom, the highest-beta shares in the universe, and a portfolio short them is a leveraged short on the recovery. Momentum crashes when markets turn up sharply from a crash, which is to say at the moment the rest of a portfolio is doing best.
Why the drawdown is the price
The premium-or-mispricing lesson said a factor that never lost would not be a risk premium. This is what that means in years: value paid because it can lose for four; momentum paid because it can lose a decade's gain in a quarter; small paid because the smallest companies fail. A factor allocation is sized by the risk capacity rule from the allocation course, and the drawdowns above are the capacity test — not "would I be fine", but "would I still be holding it in August 2020".
In the data
The value window rests on six prices: the value fund and its growth twin, adjusted closes on 3 January 2017, 31 August 2020 and 30 December 2022:
The momentum crash has no such table. The momentum series is long-short and not a fund; the nearest holdable shadow is a momentum fund, whose history begins in 2013 and does not reach 2009 — which is why the lesson quotes the paper rather than a chart.
Try it now
Reproduce the value window from the first two rows of each table above and write the two returns down. Then compute the same from the second and third rows, 31 August 2020 to 30 December 2022. Two numbers each side; that is the whole lesson.
The market through the same window, so the value fund's "18%" has a context:
Measure 3 January 2017 to 31 August 2020, then 9 March 2009 to 30 June 2009. The second is the momentum crash from the index's side — the sharpest rise in the record, which a momentum portfolio was positioned against. 3. In writing, and truthfully: at what month of the 2017–2020 window would you have sold?