Two funds report against the same index. One returned 40 basis points below it in each of the last three years. The other returned 90 basis points above, 150 below, and 30 above. Which tracked better?
The first fund has a tracking difference of 40 basis points a year and almost no tracking error. The second has an average tracking difference near zero and a large tracking error. Tracking difference is the level of the shortfall: the fund's return minus the index's return over a stated period. Tracking error is the volatility of that shortfall: the standard deviation of the periodic relative returns, annualised. The first number tells you what holding the fund costs. The second tells you how confidently you can predict what it will cost next year.
For a passive mandate, the difference is the cost and the error is the uncertainty about the cost, and a client who cares about compounding should care mostly about the first. For an active mandate the roles invert entirely: tracking error is the risk budget being spent, and a manager with none of it is not managing anything.
Advertisement
What causes tracking difference?
Tracking difference decomposes, and the decomposition is the single most useful piece of diligence in the passive world because every component is separately observable.
The expense ratio is the largest line and the only one printed on the front page. It is charged daily on net assets, which produces a subtlety worth holding on to. In a year when the index rises 21.54%, a 60 basis point expense ratio costs more than 60 basis points of return, because the fee is levied on an asset base that grew while the return is measured against the base at the start. Worked example 3 computes that drag at 72.62 basis points.
Cash drag is the fund's uninvested balance multiplied by the gap between the index's return and what the cash earned. It costs the fund in a rising market and helps it in a falling one, which means a tracker's tracking difference improves in bad years for a reason that has nothing to do with skill.
Dividend withholding tax applies wherever the fund's income is taxed at source and the index assumes it is not. A total-return index built gross of withholding tax is a benchmark no fund subject to that tax can ever match, and the gap is the dividend yield multiplied by the withholding rate.
Securities-lending revenue works the other way, and it is the one component that adds to a tracker's return whatever the market does. Cash turns into a credit as well, but only in a year when the index returns less than the cash rate, so lending is the only line a fund can count on in a rising market. The fund lends its holdings against collateral, splits the fee with a lending agent, and credits its share to the fund. It is disclosed, it is recomputable from the annual report, and it carries real risks that the ETF risks page takes seriously.
Rebalancing and reconstitution costs are the trading the rulebook forces, priced at whatever the fund actually pays. This is the line that separates a large tracker from a small one, because the large one can work its orders and the small one takes the close.
Do the components add up, and which convention applies?
Add them with care, because they do not simply sum. Each friction is levied on an asset base that is itself compounding, so the additive presentation understates the total in a rising year and overstates it in a falling one. Worked example 3 shows the same five components reconciling exactly through a multiplicative chain, and quantifies the error the linear version makes at 16.61 basis points on a 118 basis point total. Say which convention you used.
One further ambiguity is worth naming before it costs you an argument. The arithmetic tracking difference, R_fund - R_index, and the geometric relative return, (1 + R_fund) / (1 + R_index) - 1, are both defensible and both in use. On the fund in Worked example 3 they are 118.42 and 97.43 basis points respectively, a gap of 21 basis points that is entirely convention. A factsheet that quotes one and a consultant who computes the other will disagree, and neither of them is wrong.