The Capacity Limits of Published Anomalies

A backtest reports a factor's return assuming every trade fills at the observed historical price, no matter how large the position. In reality, buying $500 million of a small-cap value stock will push the price up as you buy, and selling it later will push the price down as you sell — the act of trading a strategy at scale erodes the very mispricing the strategy is trying to capture. Capacity is the size at which a strategy's trading costs, driven by its own market impact, grow large enough to cancel out its edge.

Why capacity is concentrated in the smallest names

Anomalies that are strongest among microcap and small-cap stocks — which is common, since mispricing is more likely to persist where fewer large investors are looking — are also the ones with the least capacity, because those are exactly the stocks with the thinnest trading volume and widest spreads. A factor that looks fantastic on paper because its returns are dominated by $50 million companies might only be able to absorb a few tens of millions of dollars of real capital before its own trading moves those same stocks enough to erase the edge.

Worked example

An academic paper finds a small-cap value anomaly earning 10% annualized on a backtest using the full universe of small caps, equal-weighted. A fund tries to trade it with $2 billion. Suppose the anomaly's investable universe (stocks the fund is willing to hold in reasonable size) has $40 billion in combined market cap; a $2 billion book means owning roughly 5% of the float across many of these names, and trading in and out of positions that size against thin daily volume generates estimated market impact costs of 3-5% annualized, on top of ordinary spread costs. The realistic net return for this fund is closer to 4-5%, not 10% — and doubling the fund's size again would likely push net return toward zero or negative, since the market impact grows roughly with the square of position size relative to daily volume in standard impact models.

What this means in practice

Capacity estimation is now a standard step before any institutional allocation to a factor strategy: run the same signal at realistic size using an impact model calibrated to daily volume and spread data, not the frictionless backtest. This is also why many well-known anomalies remain "real" in an academic sense while being nearly irrelevant to large asset managers — the mispricing exists, but not at a scale that matters for a multi-billion-dollar fund.