How Liquidity, Turnover and Trading Costs Affect Factor Strategies

What are rule-based smart beta funds?

Some funds do not employ someone to choose shares. Instead they follow a written rule — buy the cheapest-looking companies, or the ones whose prices have been rising, or the ones with the steadiest profits. In the industry these rules are called factors, and a fund built on one is often marketed with a chart showing how well the rule would have done over the past twenty years.

That chart is produced by running the rule over old price data and recording what it would have bought and sold. This is called a backtest, and there is nothing dishonest about it. But a backtest and a real fund are not doing the same thing, and this page is about the gap between them.

What does a smart beta backtest assume about trading?

When a computer tests a rule on historical data, it buys and sells at whatever price is written in the file. It gets however many shares it asks for. Nothing moves against it, and placing the order costs nothing.

A real fund manager placing those same orders lives in a different world:

• There are two prices, not one. At any moment you can buy at one price and sell at a slightly lower one. That gap is called the spread, and a full round trip — buying then later selling — loses the whole of it once: half on the way in, relative to the mid-price, and half on the way out. It does not appear as a line on any statement.
• Buying pushes the price up. If you want a lot of shares, you use up the people willing to sell cheaply and have to pay more for the rest. The bigger your order, the worse the price you end up with.
• Sometimes the shares are not there at all. In a company that few people trade, you may not be able to buy the amount you wanted at any sensible price.

None of that shows up in a backtest, for the simple reason that a backtest never has to find anybody to trade with.

Why do trading costs hit some factor funds harder?

The cost above is charged per trade. So the total bill depends on how much buying and selling a fund does — and funds differ enormously in that.

The industry word for this is turnover. If a fund replaces half its holdings during a year, that is turnover of 50%. If it replaces everything, that is 100%. Some strategies go well beyond that.

Here is what the trading itself costs, before anyone asks whether the rule is any good. The cost per trade is quoted in basis points, where one basis point is a hundredth of a percent — so 30 basis points is 0.3%.

Portfolio replaced each year	Cost a year	Cost over ten years
25%	0.07%	0.75%
50%	0.15%	1.50%
100%	0.30%	3.00%
200%	0.60%	6.00%
400%	1.20%	12.00%

Read the bottom row. A fund that replaces its holdings four times a year hands over around 1.20% of your money annually just in trading. The "over a decade" column is that same annual figure simply multiplied by ten — roughly 12.00% — which is a useful approximation rather than an exact compounded loss; the true ten-year cost of a steady annual drag compounds to a slightly smaller number than the straight multiple does. Either way, it is paid whether the rule works or not.

That is the number to hold in mind when a strategy claims it can beat the market by two percentage points a year. If it needs heavy trading to do so, a large part of that advantage is spent before you ever see it.

How much does a factor fund actually trade?

Turnover is easy to wave through as an abstraction, so here is an actual rule with its trading counted.

The rule is a simple one: hold the index while its price is above its own average of the last 100 days, and hold nothing while it is below. An average like this is just the typical price over recent months, and the idea is to stay invested during rises and step aside during falls. It has one instruction and nothing to adjust. It sounds like something that would sit still for long stretches.

It does not. Across the whole record it moved in and out about 4.5 times a year. Counting both the selling and the buying, that comes to around 9.0 times the value of the portfolio traded every year. At 30 basis points a round trip, the trading costs 1.35% a year on average, which subtracted from the rule's own annualised return takes it from +11.53% before costs down to +10.18% after them. That subtraction is an approximation — the actual cost was paid on each of the rule's own trade dates rather than smoothly across the year — not an exact re-run of the backtest with every cost charged on the day it would have been paid.

Why does such a simple rule trade so much? Because a price that is drifting near its own average crosses back and forth over it repeatedly, and every crossing is a trade — including the many that turn out to have meant nothing. The amount of trading is decided by the rule, not by anyone's intention. Nobody set out to trade nine times a year. The rule does that on its own, and a rule based on a jumpier signal does it more.

Which trading costs does turnover miss?

Even when a fund reports its turnover honestly, several costs sit outside that figure.

The spread is paid once on every round trip, but it is always there. Half on the way in and half on the way out, relative to the mid-price — the full quoted spread, once per round trip, not twice.

Large orders move prices against themselves. A backtest assumes the price stays put. A fund big enough to matter is part of what sets that price. This is worst in smaller companies that few people trade, which is awkward, because that is exactly where several of these rules claim their strongest advantage.

Rebalancing dates are public. Indexes announce in advance which companies are joining and leaving, and on what day. Every fund tracking that index has to trade at a moment the whole market already knows about, and other traders can get there first. The extra cost lands on the fund's own investors.

Success makes it worse. A strategy that worked with a small amount of money becomes harder to run as more money follows it, because the same trades now push prices further. This is why a fund's early results are a poor guide to its later ones.

None of this is misconduct. It is the ordinary cost of turning a rule into actual shareholdings, and it is missing from the chart because the chart never had to buy anything from anyone.

What is worth asking about a factor fund?

• How much does it trade in a year? Ask for the fund's published turnover figure.
• What does it hold? A fund holding large, heavily traded companies faces much lower costs than one holding small, rarely traded ones.
• How big is the fund compared with what it owns? A large fund holding small companies has a problem it cannot trade its way out of.
• What are the returns after all costs? Not the backtest — the fund's own record since it launched, which is the only figure that had to survive real trading.

Why are precise trading rules costly to follow?

A rule that has to be followed precisely, on particular days, is more fragile than a patient one, because every cost above gets worse when a trade cannot wait.

Funds that deliberately trade less — by allowing their holdings to drift further before correcting, by capping how much money they accept, or by avoiding rarely traded companies — give up part of the theoretical advantage and hand more of what remains to the people who own the fund. The strategies that have survived contact with real money tend to be the ones that never needed precision in the first place.

How is a factor fund best judged?

Judge one of these funds on what it actually returned to investors after costs, not on the chart of what the rule would have done. A small theoretical advantage that requires heavy trading is not an advantage waiting to be collected — most of it is handed to the people on the other side of the fund's trades.