To achieve this, a number of replication techniques have emerged, each of which has different implications in terms of tracking performance and underlying counterparty or operational risks. The two main types typically described are direct, physical replication and indirect, swap-based, often also referred to as synthetic – replication.
This unofficial classification system is not fully informative, however. To better understand how ETFs work it is necessary to subdivide these categories further:
• Full direct physical replication;
• Partial physical replication with stratified sampling;
• Partial physical replication with optimisation;
• Funded swap based ETFs; and
• Unfunded swap based ETFs.
Full, direct physical replication involves buying and managing all the underlying constituent securities of the index being tracked.
Managers of these ETFs have to dynamically trade their portfolio to keep up with changes to the index, so transaction costs can impact tracking performance. Partly to mitigate this, providers often engage in securities lending, which involves lending out the physical securities held by the ETF.
This exposes the ETF to counterparty risk that the borrower of the securities will default and the ETF will suffer a loss. To limit this risk, the ETF lender demands that the borrower deposits collateral with a third party custodian bank.
The idea behind this is that, if the counterparty fails, the collateral is liquidated in compensation to the fund.
Full physical replication can become difficult and/or expensive to implement when tracking broad indices that reference a high number of individual securities or when all or some of the underlying securities being tracked are relatively illiquid.
To overcome these problems, physical replication providers have developed tracking methods that involve only holding a portion of the underlying securities of the index. One method is called stratified sampling, and it involves the ETF provider holding a selection of representative securities only.
This could entail splitting the index into, for example, sector based subgroups, and then purchasing a sample of securities from each group. The choice of which securities will form part of the sample may be taken by the ETF manager or by a computer driven, quantitative model.
The main advantage of this approach is reduced costs to the fund. But it can also increase the possibility of the fund experiencing relatively significant tracking difference vis à vis the underlying index.