TL;DR
Out-of-sample data since the original 1992 Fama-French paper shows value, size and momentum premiums roughly half their pre-publication magnitudes — but still positive and statistically meaningful over long horizons.
The criticism that won't go away
Every few years, someone publishes a "value is dead" article. Sometimes it's "size is dead", sometimes "momentum is dead." The thesis is always the same: the academic finding got popular, ETFs were launched, capital flowed in, the anomaly got arbitraged away.
The argument has serious empirical support. R. David McLean and Jeffrey Pontiff published "Does Academic Research Destroy Stock Return Predictability?" in the Journal of Finance in 2016. They studied 97 documented return predictors from academic papers. The headline finding: post-publication returns fell by an average of 58% versus pre-publication magnitudes. That's strong evidence that publication itself partly destroys the anomalies.
But it doesn't say the anomalies disappear. It says they shrink. And shrinkage doesn't mean death.
What's actually shrunk
A factor-by-factor look at out-of-sample performance since the original Fama-French publication in 1992:
- Size (SMB): pre-1992 premium ~4%, post-1992 premium ~1%. Significant decay, possibly to zero. The pure size effect has been the biggest victim.
- Value (HML): pre-1992 premium ~4%, post-1992 premium ~2%. Roughly halved. The 2010s were brutal; 2020–2024 was a partial recovery.
- Momentum (UMD): pre-1992 premium ~10%, post-1992 premium ~6%. Modest decay, still the largest factor.
- Profitability (RMW): published 2013, post-publication premium ~2.5% (similar to pre-publication estimates). Least decayed.
- Investment (CMA): published 2015, post-publication premium ~2.5%. Similar resilience.
The pattern: the older the publication, the more decay. The cleaner the academic case (profitability, investment), the less decay. Pure size — the weakest pre-publication case — essentially disappeared.
The Asness vs Arnott debate
Cliff Asness (AQR) and Rob Arnott (Research Affiliates) have argued the question publicly for over a decade. Their respective positions, simplified:
Asness: Yes, premiums have shrunk, but they're still positive and statistically meaningful. Plan as if future factor returns will be 50–70% of historical levels. That's enough to matter.
Arnott: Some factors (especially value) may be permanently impaired. Investors should not extrapolate even shrunken premiums forward without scepticism.
The honest middle is closer to Asness's view. Even at 50% of historical magnitudes, the factor edge over decades is significant. But Arnott's caution about treating decayed-but-positive as "permanent" deserves weight — the premiums could shrink further or disappear in specific factors.
Why some shrinkage is structural
Three reasons future premiums are likely lower than historical:
- Arbitrage by smart money. As soon as a factor is documented, hedge funds and asset managers start trading it. The original gross-of-cost return gets eroded by their activity.
- Lower trading costs. The factors were originally documented in an era of 1%+ brokerage commissions. Today's near-zero trading costs let institutions hold tighter factor exposures, which compresses the premium.
- Capacity limits. As factor ETFs absorb hundreds of billions of dollars, the strategies become capacity-constrained. The marginal investor in the factor earns less than the academic strategy did at the start.
All three effects argue for forward expectations of factor premiums at roughly half their historical magnitudes. That's the conservative planning baseline.
The case the critics get wrong
The "factors are dead" argument oversells. Three reasons:
- Cycles, not extinction. Value spent the 2010s underperforming, then strongly outperformed 2020–2024. If the factor were truly arbitraged away, recovery wouldn't happen. The cycles continue because the underlying behavioural and risk-based mechanisms continue.
- International evidence. Even if US factor premiums have shrunk, the same factors continue to work in European, Asian and emerging markets — often with less decay.
- Risk-based factors should persist. If value reflects compensation for distress risk (rather than just mispricing), the premium should persist as long as investors remain risk-averse.
What to plan around
Honest forward planning for FIRE investors:
- Assume factor premiums are 50–70% of historical levels. Plan for 1.5–2% per year of multi-factor uplift over the market, not the historical 3%+.
- Diversify across factors. Pure value or pure size could disappoint. A multi-factor mix is more robust against single-factor decay.
- Don't extrapolate any single decade. 2010–2020 was brutal for value; 2020–2024 was excellent. Both are within historical variation.
See our factor investing vs index funds article for how the shrunken-but-positive premium compares to the cost-and-behavioural advantage of plain index funds.
Test what a 1.5–2% per year factor uplift does to your FIRE date in our factor comparison tool. Even at the lower end of forward expectations, the difference is usually 1–2 years off the median FI date.
Frequently asked questions
- Should I just ignore historical premium sizes?
- No — but discount them. A useful rule of thumb is to plan as if future premiums will be 30–60% of their historical level.
- Has value really stopped working?
- It performed terribly from 2010–2020 and very well from late 2020 onwards. The premium isn't dead, but it's noisier than the long-run averages suggest.
- Which factor has held up best out-of-sample?
- Profitability (RMW) and investment (CMA) have decayed least since their publication. They're newer factors with cleaner risk-based foundations, which probably helps.
Stress-test your own FIRE plan
FIRE Wealth OS runs your savings rate and expenses against every historical market starting point since 1871. Free to use, no card required.