Why a tracker may not own every holding in its index

A tracker is often assumed to hold everything in its index in the correct proportions. Many do not, and the deviation is a deliberate cost decision rather than a shortcut.

Full replication is not always practical

An index containing thousands of constituents includes many that are very small and thinly traded. Buying and maintaining tiny positions in each carries transaction costs disproportionate to their weight.

Those costs are borne by the fund and therefore by its investors. A fund that replicates perfectly may lag its index by more than one that approximates it cheaply.

The trade-off is between precision and cost. Beyond a certain point additional precision costs more in trading than the tracking improvement is worth.

How sampling is constructed

Sampling selects a subset chosen so the portfolio's characteristics resemble the index. Sector weights, size distribution and other measurable attributes are matched even though individual names are omitted.

Large constituents are almost always held directly, since they dominate the index and are cheap to trade. The omissions concentrate at the small end where weights are negligible.

Some approaches use optimisation to select the subset that minimises expected deviation. Others apply simpler rules such as holding everything above a size threshold.

The consequence is tracking error

A sampled portfolio will not move exactly with the index. The difference is usually small and can run in either direction on any given day.

Tracking error measures the variability of that difference, and it is distinct from tracking difference, which measures whether the fund lagged or led over a period.

A fund can have higher tracking error and still deliver a smaller shortfall, because the noise around the index and the average gap to it are separate things.

Where it matters more

Sampling has little effect on funds tracking concentrated indices of large companies, where a small number of holdings account for most of the weight.

It matters more for broad indices covering many small companies and for those covering markets where trading is expensive or access is restricted.

Bond indices are the extreme case. Many contain issues that trade rarely or not at all, so full replication is generally impossible rather than merely expensive.

What disclosure to look at

Fund documents state the replication approach and the number of holdings, and comparing that number to the index constituent count reveals immediately whether sampling is in use.

Historic tracking difference is more informative than the method. It reports what the approach actually produced rather than what it was designed to produce.

Neither figure supports a judgement about the future. They describe how a fund has been run, which is a different question from how a market will behave.