Costs, capacity and crowding: turnover, implementation shortfall and factor decay after publication

What you will learn

• How to convert turnover and per-trade costs into an annual drag in percentage points and euros.
• What implementation shortfall measures and why it is larger than the quoted spread.
• How to judge capacity with a simple days-to-trade calculation.
• What the evidence says about returns shrinking after a factor is published and widely used.

Explanation

From paper premium to net premium

Academic factor returns use closing prices with no trading costs. Between that paper return and your account sit the fund's own trading costs (driven by turnover), the fund's fee, your dealing costs (course "ETF Investing") and taxes. This lesson covers the first layer and two reasons the paper premium itself may shrink: capacity and crowding.

Turnover and implementation shortfall

One-way turnover is the smaller of purchases and sales in a year divided by average fund assets; 120 % means the fund replaced 1.2 times its assets and traded 2.4 times its assets in total. The all-in cost per euro traded is implementation shortfall: the difference between the price when the decision was made (for an index fund, the close used by the index) and the price actually achieved, including spread, market impact and delay. Rules of thumb for liquid developed-market large caps: half-spread 0.02–0.10 %, impact a few more hundredths to tenths of a percent depending on order size relative to volume; small caps and emerging markets several times more. Approximate annual drag:

Drag ≈ 2 × one-way turnover × cost per euro traded (one side)

Typical one-way turnover: value and quality roughly 20–40 % a year, low volatility 20–30 %, momentum often 100 % or more, which is why implementation decides everything for momentum (lesson 3).

Capacity

Capacity is the amount of money a strategy can run before its own trading removes the premium. A practical proxy is days to trade: position size divided by (participation rate × average daily volume, ADV). Managers rarely trade more than 10–20 % of daily volume without visible impact; a strategy needing 50 days per position cannot follow a monthly signal. Small-cap and momentum strategies have the lowest capacity, large-cap value and quality the highest. Published capacity estimates differ by a factor of ten, so use the concept, not a single number.

Crowding and decay after publication

Once a pattern is published, investors trade on it. The best-known study (McLean and Pontiff, Journal of Finance, 2016) examined 97 published predictors of US stock returns and found long-short returns about 26 % lower out of sample (after the sample ended, before publication) and about 58 % lower after publication. The first part is data-mining shrinkage; the extra post-publication part is investors learning. Two cautions: the study covers mostly narrow anomalies, and a long-only tilt carries only a fraction of a long-short factor, so the euro effect on an ETF investor is smaller than the headline. Crowding indicators (short interest, valuation spreads, fund flows) have weak forecasting power; the decay numbers are the more reliable planning input.

Worked example

Hypothetical example: a momentum ETF with one-way turnover of 120 % a year in developed-market large caps. Assumed cost per euro traded, one side: 0.10 % half-spread plus 0.15 % market impact = 0.25 %. Assumed gross long-only momentum premium 3 % a year. Fund fee, your own dealing costs and taxes are excluded.

Step 1 – trading drag: 2 × 1.20 × 0.25 % = 0.60 % a year. On a €100,000 position that is €600 a year, every year, inside the fund's performance.

Step 2 – planning haircut for decay: 3 % × (1 − 0.58) = 1.26 % if you apply the full post-publication figure; 3 % × (1 − 0.26) = 2.22 % if you apply only the out-of-sample shrinkage.

Step 3 – net planning range: pessimistic 1.26 % − 0.60 % = 0.66 %; moderate 2.22 % − 0.60 % = 1.62 %, before the fund fee and your own spread.

Step 4 – capacity check: the fund has €2 billion and a 1 % position (€20 million) in a stock with ADV of €4 million. At 10 % participation it can trade €400,000 a day, so the position takes 20 million / 400,000 = 50 trading days: just workable semi-annually, impossible for a monthly signal.

Reading: 3 % gross may be worth well under 1 % net in the pessimistic case; lesson 12 uses this range.

Common mistake

Practice activity

On paper: a small-cap value ETF reports one-way turnover of 35 %. You estimate one-side costs at 0.20 % half-spread and 0.30 % impact. The fund holds €600 million; its average position is 0.4 % of assets in stocks with ADV of €1.5 million. (a) Compute the annual trading drag in percent and in euros on a €20,000 holding. (b) Compute days to trade an average position at 15 % participation. (c) Is the fund near its capacity limit? Give one reason.

Key takeaways

• Annual drag ≈ 2 × one-way turnover × one-side cost; momentum's high turnover makes it the most cost-sensitive factor.
• Implementation shortfall, not the quoted spread, is the right cost measure; it includes impact and delay.
• Days to trade = position / (participation × ADV) is a quick capacity test.
• Predictor returns fell roughly a quarter out of sample and more than half after publication; plan with a haircut.

Next step

Lesson 11 asks why any premium should survive at all, and quantifies the long underperformance that drives investors out before it does.