Microcap Contamination of Anomaly Returns
Many published factor returns look strong mostly because of a small sliver of tiny, illiquid microcap stocks that a real portfolio could barely trade.
Prerequisites: The Factor Zoo and the Replication Crisis
Academic factor tests typically sort every stock on an exchange — including companies worth $30 million with almost no daily trading volume — into portfolios by some characteristic, then look at the average return across deciles. Microcap stocks are a tiny fraction of total market value but often a large fraction of the count of stocks, especially at the extreme ends of a sort like deep value or small size. If those tiny names happen to have wild, hard-to-trade returns, they can dominate an equal-weighted portfolio's average and make a factor look far more profitable than any real fund could ever capture.
An equal-weighted academic portfolio gives a $30 million company the same vote as a $300 billion one. Because factor effects are often strongest, and returns most volatile, among the smallest names, a lot of "anomaly" return sits in stocks too illiquid to actually trade at scale.
Why the weighting scheme matters this much
Consider a value anomaly built from ten deciles of the whole market. The bottom size decile alone can contain over half the number of stocks in the sample even though it holds only a percent or two of total market capitalization. If that decile also happens to show the strongest value effect — common in the literature — then an equal-weighted portfolio's headline return is quietly being driven by names that would move 5% on a $50,000 order.
Worked example
A value strategy tested with equal weighting shows a 12% annualized return spread between cheap and expensive deciles. Rerun with value weighting (each stock's return contributes in proportion to its market cap, so a mega-cap barely moves the number while a microcap barely registers), the spread falls to 4%. A researcher then removes stocks below $100 million market cap or below $1 share price entirely, and the value-weighted spread among the remaining names is 3.2% — close to what a real institutional value fund would actually realize, once realistic position sizing forces it to hold mostly larger, liquid names.
What this means in practice
Before trusting a backtest, check how it weights and screens its universe. A study that reports equal-weighted decile returns across all listed stocks, including sub-$1 names, is measuring something closer to a lottery-ticket effect than a tradeable strategy. Reputable factor research now routinely reports both equal- and value-weighted results, and applies price and size screens (commonly excluding stocks under $1 or below the NYSE 20th percentile in market cap) specifically to keep microcap contamination from inflating the headline number.
Removing microcaps entirely doesn't just shrink the return, it also shrinks the statistical significance of the finding, because it removes most of the observations. A factor that looks strong and significant across the full universe can become weak and statistically shaky once realistically investable — and that gap is exactly the information a practitioner needs before allocating capital.
Related concepts
Practice in interviews
Further reading
- Fama, French, 'A Five-Factor Asset Pricing Model' (Journal of Financial Economics)
- Hou, Xue, Zhang, 'Replicating Anomalies' (Review of Financial Studies)