Definition
Factor turnover is the rate at which a portfolio's holdings rotate as the underlying factor scores change over time. It is usually measured as the one-sided dollar value of trades executed per period divided by portfolio gross market value, or as the fraction of names that enter or exit the holdings on each rebalance. A factor whose ranking is stable across months — quality, for instance — generates low turnover. A factor whose ranking flips on noisy signals — short-horizon reversal — generates high turnover.
The reason it matters is mechanical. Every rotation costs money: half the bid-ask spread, the commission, the market impact of moving the price against yourself, plus any taxes triggered by the sale. Those costs compound. An in-sample backtest with 200% annual turnover and a 4% gross factor premium can deliver 0% net after honest cost assumptions.
Why it matters
How it works
A common pattern is to refresh factor scores monthly and rebalance the portfolio to a target weight on each refresh. Without smoothing, even tiny score changes near the edge of the ranking trigger trades — a name ranked 100th drops to 101st, the top-100 portfolio sells it and buys the new 100th. Multiplied across hundreds of names and a year of rebalances, the cumulative trade volume can exceed the portfolio's gross value several times over.
The defences are familiar to every practitioner. Score smoothing averages the factor score over a window so single-period noise cannot flip a holding. Holding bands keep a name in the portfolio until its rank drops past a wider threshold, creating hysteresis. Trade-cost penalisation in the optimiser charges each candidate trade an explicit cost so the optimiser only rotates when expected alpha exceeds expected impact. Each of these techniques reduces in-sample raw alpha by a few basis points but typically increases net alpha by far more — the difference between a paper-profitable and a live-profitable factor.