Factor Alpha After Trading Costs
A factor's academic backtest ignores trading costs entirely; once bid-ask spreads, market impact, and turnover are subtracted, high-turnover factors can lose most or all of their apparent edge.
Prerequisites: The Factor Zoo and the Replication Crisis
An academic factor backtest usually assumes a stock can be bought or sold at its closing price with zero cost. Real trading involves a bid-ask spread, market impact from moving the order book, and commissions — and a factor that rebalances its entire portfolio every month racks up those costs on every single trade, every single month. A factor with an impressive 8% annual gross return can turn into a 2% net return, or less, once realistic costs are applied — and the deciding variable is almost always turnover.
Turnover is the hidden multiplier on every factor's cost. A slow-moving factor like value, which reshuffles maybe 30% of its portfolio a year, survives realistic costs easily. A fast-moving factor like short-term reversal, which can turn over its entire portfolio every month, can see most or all of its gross alpha eaten by trading costs.
Why turnover, specifically
Every rebalance means selling names that fell out of favor and buying names that newly qualify, and each trade pays the spread plus whatever market impact the order size creates. A factor's annual cost is roughly proportional to (round-trip trading cost per dollar traded) × (annual turnover). A factor earning a large gross premium but trading its entire book monthly can easily have annualized turnover above 500%, meaning the position is essentially replaced five times a year — five times the cost drag of a factor that trades its book once.
Worked example
A short-term reversal factor reports a gross annualized return of 15% in an academic backtest, with monthly rebalancing and roughly 100% turnover each month (about 1,200% annualized, since positions are effectively replaced monthly). Applying a realistic round-trip cost of 40 basis points per dollar traded on liquid large caps, the annual cost drag is approximately against turnover value, but because reversal trades concentrate in higher-cost, more volatile names, realistic studies find effective costs closer to 60-100bp round trip, pushing total drag to 8-12% a year — largely or fully consuming the 15% gross number.
Compare a value factor with 12% gross annual return and 30% annual turnover: cost drag is roughly , leaving net return essentially unchanged at about 11.9%.
What this means in practice
Gross backtested returns should never be compared across factors without adjusting for turnover, because two factors with identical gross Sharpe ratios can have completely different net Sharpe ratios once costs are applied. This is one reason capacity and net-of-cost analysis has become a standard part of any serious factor evaluation, and why some of the most "significant" academic anomalies — often high-turnover ones like short-term reversal — are also among the least useful in practice.
Cost estimates from academic papers are usually based on average historical spreads for the whole market, which understate the true cost of trading the specific stocks a high-turnover factor concentrates in — smaller, more volatile names with wider spreads than the market average.
Related concepts
Practice in interviews
Further reading
- Novy-Marx, Velikov, 'A Taxonomy of Anomalies and Their Trading Costs' (Review of Financial Studies)
- Frazzini, Israel, Moskowitz, 'Trading Costs' (working paper)