Factor Valuation Audit
Separate genuine factor returns from gains caused by rising valuations
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 99%
The Factor Valuation Audit treats an investment factor as an asset that can itself become cheap or expensive. Begin by decomposing its historical return into income yield, fundamental growth, and changes in valuation multiples. A factor may appear powerful because investors increasingly paid more for the securities it favors, not because its underlying mechanism generated durable alpha. Next, compare today's valuation with the factor's history and test whether a plausible behavioral or economic reason should sustain the premium. Finally, model what happens if valuation multiples stop expanding or revert toward normal. This process prevents investors from extrapolating inflated backtests, selecting recently loved factors, and mistaking a temporary revaluation windfall for repeatable excess return.
Origin
Extracted from Masters in Business, where Rob Arnott explained research behind the paper How Can Smart Beta Go Horribly Wrong? and Research Affiliates' failed attempts to enhance fundamental indexing.
Core principles
- 01Past performance can be inflated by rising valuation multiples
- 02Factors and strategies can become expensive just like stocks
- 03A sound mechanism matters more than a mined historical relationship
- 04Current valuation changes the return implied by a backtest
- 05Mean reversion can reverse a strategy's apparent alpha
How to run it
- 1
Establish the claimed premium
Document the factor's reported historical return, benchmark, testing period, and implementation assumptions.
Pro tip Prefer evidence spanning multiple markets and economic regimes.
Watch out A large factor catalog makes false discoveries likely.
- 2
Identify the mechanism
State why the factor should affect future prices or why investor behavior should repeatedly create the opportunity.
Pro tip Write the mechanism before looking at the most favorable backtest.
Watch out Historical correlation without a forward-looking reason may be data mining.
- 3
Decompose past returns
Separate returns into yield, fundamental income growth, and valuation-multiple change. Quantify how much apparent alpha came from investors repricing the factor.
Pro tip Use the same decomposition for the factor and its benchmark.
Watch out Do not label multiple expansion as permanent alpha.
- 4
Measure current valuation
Compare the factor's present valuation spread with its own long-term distribution. Determine whether recent success has made the strategy unusually expensive.
Pro tip Track relative valuation rather than relying only on absolute market levels.
Watch out A strategy can own cheap stocks while the strategy itself is expensive relative to history.
- 5
Stress-test mean reversion
Estimate prospective returns if valuation remains unchanged, normalizes gradually, or reverses sharply.
Pro tip Show results under several normalization horizons.
Watch out Mean reversion has uncertain timing and cannot be treated as a precise catalyst.
- 6
Set a decision rule
Adopt, reduce, or reject the factor based on mechanism quality, implementation viability, and valuation-adjusted expected return.
Pro tip Require a margin of safety after fees and trading costs.
Watch out Do not restore exposure merely because a formerly expensive factor has fallen slightly.
In the wild
A factor shows excellent five-year returns, but most of the excess return came from its valuation multiple expanding. The audit removes that tailwind, compares the factor with its historical valuation, and models a return to normal multiples.
→ The investor avoids buying a newly expensive strategy whose backtest overstates its prospective return.
Research Affiliates tried to improve fundamental indexing by dynamically favoring the measures with the best results. Those measures had performed well because the market increasingly loved them, leaving them expensive and poised to disappoint.
→ The failed enhancement revealed that valuation must be assessed for strategies and factors, not only individual stocks.
Common mistakes
Equating backtest alpha with expected alpha
A backtest can include a one-time valuation increase that cannot recur indefinitely.
Ignoring the factor's current price
A factor with a valid mechanism can still produce poor forward returns when its valuation is stretched.
Selecting the recent winner
Favoring the metric with the best recent record can systematically buy what has just become most expensive.
Is it for you?
Best for
It is best for investors comparing smart-beta factors, quantitative strategies, or recently successful investment styles.
Not ideal for
It is not ideal for decisions based solely on short-term tactical price forecasts.
From the transcript
“If you know those three numbers, you know your rate of return.”
“when a strategy becomes massively more expensive, it'll have wonderful past returns and terrible future returns.”
“Today, I have a better understanding that valuation matters for factors and strategies just as much as it does for stocks.”
From the episode
Robert Arnott Discusses the Process Behind Research Affiliates
Robert Arnott