Can you time a factor?
If value has lost for three years, is it cheap? The question sounds like it should have a yes, and the answer is that the evidence for timing factors is weak, the one signal with some evidence is slow, and the cost of trying is certain. This lesson is that answer with its reasons.
The valuation spread
There is one measurement that says something. Take the cheap group and the expensive group of a value sort and compare their price-to-book ratios. When the expensive group trades at three times the cheap group's multiple, value is "normally" priced; when it trades at six or eight times, the spread is wide and value is historically cheap relative to growth. In early 2000 and again in 2020 the spread reached the widest readings on record, and the years that followed were among value's best.
That is the case for timing. It has two problems. The spread was wide for years before it paid, and a portfolio that leaned into value in 2017 on the strength of a wide spread lost for three more years — from 3 January 2017 to 31 August 2020 IWD.US returned about 18% against 124% for IWF.US, measured on 2026-09-04. And the signal has fired perhaps four times in a century: a rule with four observations is a story with a chart.
Why momentum-in-factors fails too
The other candidate is factor momentum: buy the factor that has been working. It has a literature, and the literature's own finding is that it works mostly through the momentum factor itself and adds turnover to everything else. Asness and colleagues' 2017 title says the rest — contrarian factor timing is deceptively difficult.
What the spread is good for
Not timing. Sizing and expectation. A wide spread says the premium on offer is larger than usual, which is a reason to hold a value allocation at its full intended weight rather than cut it after losses; a narrow spread says the premium is thin, which is a reason not to add. Neither is a trade. And the spread is the best available answer to "how much of this year's loss was the factor being repriced rather than the factor failing?" — a repricing is temporary by definition, a failure is not, and the two feel identical in a monthly statement.
The cost that is certain
Timing a factor means trading between factor funds, and each switch is turnover with spreads and, in a taxable account, realised gains. A switch that is right half the time and costs half a percent each way needs the factor to move two full percentage points in the direction guessed to break even — the right call has to pay for itself and for the wrong one that follows. The implementing unit's turnover lesson is that arithmetic at scale.
In the data
The spread is a ratio of two aggregates: the capitalisation-weighted price-to-book of the expensive group over that of the cheap group, each built from every name's published figures. Its raw material looks like this, two large consumer-facing companies in the same universe:
On 2026-09-04 Verizon traded at about 2.0 times book and Coca-Cola at about 10.5, a five-to-one spread inside two names that sell things people buy every day. The group-level spread averages hundreds of pairs like this one.
Try it now
- Compute a two-name spread: the price-to-book of NVIDIA divided by that of Exxon Mobil, both below. Write the ratio down with the date on the tables. It is one point on a chart that needs twenty years to mean anything.
- The growth end of the market over five years, so the repricing has a shape:
Measure the sharpest fall on it. A wide spread narrowing looks like that fall on the growth side and a rise on the value side, and the spread tells you nothing about which month it starts. 3. Your call: what would you have needed to know in January 2017 to avoid the value fund's next three years — and could any spread reading have told you?