Skip to content

Alice's Adventures in Factorland: Three Blunders That Plague Factor Investing: summary and key takeaways

The paper argues that investors misjudge factor investing in three ways: they expect too much return, they underestimate tail losses, and they overstate diversification across factors. Using US long-short factors, it shows returns fell after publication and that factor portfolios suffered a deep, clustered drawdown that normal-distribution models miss.

Arnott, R.; Harvey, C.R.; Kalesnik, V. et al. · 2019 · The Journal of Portfolio ManagementSummary updated 20 November 2026
Published2019Journal article
Asset focusEquitiesFactor investing
TypeResearchQuant Crises & Methods
Reading time5 minthis summary

Alice's Adventures in Factorland: Three Blunders That Plague Factor Investing at a glance

The question
Why do factor portfolios disappoint relative to backtests, and how large are the tail and diversification risks that standard tools understate?
Dataset & period
US equity long-short factors built from CRSP/Compustat, July 1963 through June 2018, monthly. The paper covers the 15 factors most followed by investors plus a wider set of 46 factors in total, each scaled to 10% annualized volatility. A recent window from July 2003 is treated as largely out of sample.
Method
The authors build size-and-characteristic sorted long-short portfolios in the Fama-French style, scale each to equal risk, and combine them into equal-weight factor portfolios. They compare actual drawdowns and recent returns with simulated ones. The simulations assume normality, or resample history independently, in one-month blocks, or in 12-month blocks, which preserve fat tails, cross-factor correlation and serial correlation in turn.

Key findings of Alice's Adventures in Factorland: Three Blunders That Plague Factor Investing

  • Over 400 factors have been published in top academic journals through year-end 2018, which the authors read as a sign that many are spurious.
  • Across the 46 factors, average return in the 10 years after the original study sample ends is less than half the average of the 10 years before it.
  • Apart from the market, no popular factor delivered a statistically significant excess return since 2003. The portfolio of the six most popular factors averaged 7.38% over the full sample but 3.10% in the recent window.
  • Momentum lost 24% in April 2009 alone, a worst-ever month, and fell 44% from its peak. The authors call it an 8-sigma event under normality.
  • For 11 of the 14 main factors, the worst month would be expected less than once in 2,000 years if returns were normal.
  • The six-factor portfolio had a worst drawdown of 33.4%. Only 2.7% of normal simulations were that bad, while with 12-month block resampling more than a quarter of paths exceeded it.

Robustness, as reported in the paper

  • Resampling in one-month and 12-month blocks shows that cross-factor and serial correlation make large drawdowns far more probable than a normal model suggests.
  • The investment factor earned 5.4% a year before mid-2003 and 0.3% after; simulations based on the earlier period rated that recent result as unlikely.
  • For the six-factor portfolio, recent return of 1.9% had a probability of just 0.3% under 12-month block resampling of earlier history.
  • The paper says results hold for worst one-, three- and five-year periods, shown in its online appendix, which is not in the supplied text.

Caveats and limits

  • The recent 15-year window began just after the 2002 bear market low and ended in a long bull market, so market-relative comparisons are sensitive to the sample choice.
  • Reported factor returns are before trading costs and fees, which the paper notes would lower them further.
  • Operating profitability was published in 2013, so much of its recent period is not truly out of sample.
  • Factor betas to the market are time-varying, so the average betas in the exhibits are rough guides (the paper's own warning).
  • All evidence is US equities; the paper does not test other regions (inferred).

Items marked (inferred) are our reading of the design, not statements by the authors.

Why it matters

Backtested factor averages can overstate what an allocator should expect, and normal-distribution risk models understate drawdowns. Correlations and market betas across factors move with conditions, so diversification measured in calm periods may not hold in stress. Sizing and risk budgets built on those assumptions deserve a check.

Questions about this paper

Do factor returns weaken after publication?

Yes. Average return in the 10 years after the original sample ends is less than half the average of the prior 10 years across the 46 factors studied.

Does combining factors remove tail risk?

Not fully. The six-factor portfolio had a worst drawdown of 33.4%, and large losses across factors tend to occur at the same time.

How bad was the momentum crash?

Momentum fell 24% in April 2009 and 44% from its peak, which the authors describe as an 8-sigma event under normality.

Where can I read the original paper?

The original is at its publisher: https://doi.org/10.3905/jpm.2019.45.4.018. We link to the paper and do not host the PDF.

Follow new quant research

Download MacroFinance market data as CSV or Excel, or pull it by API, and test the ideas in these papers yourself. New summaries are added as we review them.

Create a free account
Summaries are written by MacroFinance from the published papers, in our own words, for information only. They are not investment advice and may omit detail. Papers are not hosted here: each page links to the original at its publisher.