3. Tracking difference: index and ETF do not match

Costs 4: Funds, ETFs and products

The TER is the one cost figure almost everybody looks at, and it is the one that carries the least information about the final outcome. This module is about the gap between what a fund declares it charges and what actually separates you from the index, which is sometimes larger, sometimes smaller and sometimes carries the opposite sign to the one you expect.

The TER is a tariff, not a result

The TER (ongoing charges) rolls together the management fee, the depositary fee, the audit and other recurring expenses. It is deducted daily from the net asset value, so you never see it as a charge. What it leaves out is everything that is not a fixed tariff: the dealing costs when the fund rebalances, the taxes the fund bears on the dividends it receives, the effect of cash waiting to be invested, the income from securities lending and, in synthetic replication, the cost of the swap.

None of those items is in the TER, and added together they can come to more than it does.

Tracking difference and tracking error

Two ideas that get muddled constantly need keeping apart:

• The actual deviation from the index (*tracking difference*) is the plain subtraction: the fund's return minus the index's return over a period. It is what it has genuinely cost you. It can be negative (the fund lags) or positive (the fund beats the index).
• The variability of that deviation (*tracking error*) measures whether the fund strays erratically, not how much it strays in total. A fund can have a minuscule tracking error and still finish half a point behind every year with exasperating regularity.

For the long-term investor, the figure that matters is the first, measured and accumulated over several years.

Why an ETF can beat its index

It sounds like an arithmetic mistake, but it happens often and for structural reasons.

• Securities lending income. What the fund earns from lending its shares goes into the fund's assets and is not in the index, which lends nothing.
• Withholding on dividends according to the fund's domicile. This is where the big effect lies, and it is worth taking apart slowly.

When an American company pays a dividend to a non-resident, the general rate of withholding at source is 30%. Total return indices come in two versions: the gross one, which assumes the whole dividend is received, and the net one, which deducts the maximum withholding applicable to a foreigner, that 30%. Almost every European ETF uses the net version as its benchmark.

Now, a fund domiciled in Ireland comes under the double taxation treaty between Ireland and the United States and bears 15% rather than 30%. That is, it collects more dividend than the net index assumes it collects. And a synthetic ETF on an American index can, through the swap, receive the return equivalent to the gross index, with no withholding on those dividends, thanks to an exception written into American tax rules for certain broad indices.

Result: the tax advantage can outweigh the declared cost and leave the deviation in positive territory.

The calculation, worked through

Take a broad American index with a dividend yield of the order of 1.5% a year, a figure close to that of the S&P 500 in recent years. We compare three things against the net version of the index, which has already deducted 30% of withholding, that is, 0.45 percentage points a year (30% of 1.5%).

• Irish physical ETF, TER 0.07%. It bears 15% instead of 30%, that is 0.225 points rather than 0.45. It recovers 0.225 points against the benchmark. Subtracting the TER: 0.225 minus 0.07 comes to plus 0.155 points. If it also earns 0.02 points net from securities lending and loses 0.03 in dealing costs and cash drag, it ends up around plus 0.145 points a year above the net index.
• Synthetic ETF, TER 0.05% and swap cost 0.10%. It receives the equivalent of the gross index, so it recovers the full 0.45 points. Subtracting 0.05 and 0.10: plus 0.30 points. On paper the tightest of the three, at the price of taking on counterparty risk.
• Physical ETF domiciled outside a favorable treaty, TER 0.07%. It bears 30% just as the net index does, so it recovers nothing. Its deviation is roughly minus the TER plus the costs: around minus 0.12 points.

Note the order of magnitude: between the first and the third there is some 0.27 points a year of difference with a practically identical TER. The choice of domicile and structure weighs more here than the declared fee.

What to compare in earnest

The operating rule that follows from all this: between two ETFs on the same index, do not compare TER against TER. Compare the cumulative actual deviation against the same benchmark index and over the same period, at three and five years. Two cautions: the same version of the index, net against net, because setting one measured against the gross version alongside another measured against the net gives you a false winner; and the same currency, because measuring a dollar-quoted fund in euros mixes the exchange rate in with the tracking.

A real case: two funds, the same index, different results

The cleanest case is the one that pits domiciles against each other. With the S&P 500 yielding around 1.3%-1.5% in dividends in recent years, the difference between bearing the Irish 15% and the general 30% is worth of the order of 0.20 percentage points a year, and bearing none at all is worth of the order of 0.40. That comes to three times and six times respectively the typical TER of an index ETF on that index, which is around 0.07 points.

That is why many Irish physical-replication ETFs on the S&P 500 have shown cumulative deviations close to zero or even slightly favorable against the net version of the index, despite charging a fee. It is not the manager's skill: it is international taxation, securities lending and a benchmark that assumes a worse withholding than the real one. It works exactly the same in the other direction: a product with a low TER and an unsuitable domicile carries a permanent disadvantage that no sales sheet ever highlights.