ETF Structures
ETF providers can either use either physical or synthetic replication techniques in order to achieve their investment objectives.
Physical Replication
As the name suggests, ETFs using physical replication own most or all of their reference index’s constituents in order to replicate the index’s performance. The key selling points of this method are transparency and simplicity. However physical replication funds tend to levy higher total expense ratios (TERs) and exhibit larger tracking error relative to synthetic replication funds.
As physical replication involves buying most or all component stocks of the reference index, this strategy is inherently more labour intensive and costly than synthetic replication. These additional costs will ultimately be passed along to the investor in the form of higher TERs--which tend to be the largest and most explicit source of tracking error.
Another potential source of tracking error within physical replication funds is sampling. Depending on the size of the reference index and/or the liquidity of the index constituents, an ETF manager may choose to hold just a sample of the benchmark’s constituents. While a full physical replication of the DAX 30, for instance, is fairly simple as the index consists of just 30 highly liquid securities, other indices like the MSCI World Index are much broader (The MSCI World is comprised of about 1600 stocks), containt some fairly illiquid members, and are reviewed and/or re-balance periodically. In these instances full replication would be less efficient. In cases such as this, ETF providers make use of an optimised sampling approach. Applying this method, the ETF invests in a select subset of the index constituents which are deemed to have very similar risk and return characteristics to the overall index. This model works well during normal market conditions. However, in an adverse market environment historical statistical attributes are thrown out of the window, and sampling can result in higher than expected tracking error.
Other potential sources of tracking error within physical replication funds include the tax treatment and the timing of dividend payments.
When ETFs receive dividends from multiple tax-jurisdictions, the associated dividend withholding taxes cannot always be fully reclaimed. Moreover, clawing them back usually involves substantial time and effort. Swap based ETFs employ methods to receive more favourable tax treatment on dividend income and can therefore reduce the amount of tracking error caused by withholding taxes.
As for the timing of dividend payments and dividend reinvestment assumptions, index constituents are listed ex-dividend the day after a dividend is announced. The actual dividend payment happens at a later date. This time-lag between the announcement and the payment could affect the funds’ tracking performance. This is because many indices assume that dividends are reinvested on the ex-date. This results in a discrepancy in dividend reinvestment assumptions between the benchmark and the fund is a potential source of tracking error.
Physical ETFs can also engage in securities lending to generate extra income for the fund. While this additional income can partially offset higher TERs and mitigate some of the aforementioned sources of tracking error, the practice also exposes the investor to counterparty risk. Specifically, there is always a risk that the borrower will fail to return the security at the end of the term. To mitigate this risk, ETF providers require borrowers to post high quality liquid collateral in an amount in excess of the value of the securities on loan.
It is important to note that there are a number of additional potential sources of tracking error, but we have limited our discussion to the largest and most common sources for purposes of this article.
Synthetic Replication
Most ETF providers in Europe use synthetic replication methods to achieve their investment objective. Synthetically replicated funds will enter into swap agreements with a single or multiple counterparties to deliver the return of the fund’s benchmark index. These swaps can either be fully funded or unfunded.
In a fully funded swap, the ETF provider transfers the cash received from investors to the swap counterparty(ies). The swap counterparty(ies) then buy(s) a basket of securities which is pledged as collateral into a segregated account with an independent custodian.
With unfunded swaps, the money received from investors is used by the ETF provider to purchase securities which are often completely unrelated to the index’s constituents. This ‘substitute basket‘ usually consists of highly liquid blue-chip equities or investment grade government bonds. Most providers disclose the composition of their ETFs’ collateral/substitute baskets, either on their website or upon request. The industry trend is towards publishing the composition of these baskets online on a daily basis.
Funded and unfunded swaps have both pros and cons in terms of efficiency, costs, counterparty risk, etc. and therefore each provider prefers a different approach.
Irrespective of the swap structure, the swap-counterparty agrees to exchange the index return for the return of the collateral/substitute basket. This clearly introduces counterparty risk for the investor as the swap-provider could fail to fulfil its obligation in the case of default. The use of multiple counterparties can serve to reduce such risk but not eliminate it. Moreover, under UCITS III, the maximum exposure to any swap counterparty is limited to 10% of the fund’s net asset value. However, most providers limit their funds’ exposure to a level well below the UCITS mandated maximum and many even over-collateralise their synthetic ETFs. In practice, once the counterparty exposure reaches a provider’s threshold limit, swaps will be reset, bringing the counterparty exposure back to zero. Swaps can also be reset periodically or whenever ETF shares are redeemed or new ones are created.
As the swap provider is responsible for delivering the index return, the ETF provider does not have to worry about some of the aforementioned sources of tracking error inherent to physical replication funds. In many cases, especially when it comes to following less liquid, more exotic indices, this helps to reduce tracking error and ensure closer tracking of the index than physical replication. However, it is important to note that swap-based funds are not without their own potential sources of tracking error. For instance, swap spreads (the cost of having the relevant investment bank provide the index’s return) are not included in the TER and will therefore show up in the tracking error of synthetic replication funds.
Physical vs Synthetic Replication
The main advantage of physical replication is simplicity. However, this often comes at the expense of greater tracking error and a higher TER. The main advantages of synthetic replication tend to be lower tracking error and lower fees. But much like those physical replication funds that engage in securities lending, synthetic ETFs carry counterparty risk, though limited to less than 10% of the fund’s net asset value in most cases.
For some indices physical replication is more difficult to implement given the breadth of the underlying index or a lack of liquidity in its constituent securities. In those cases, synthetic replication is often more efficient. A good example is those ETFs tracking single country emerging market equity indices where many shares are either fairly illiquid or are otherwise not easily accessible to investors.
Part III of this guide will look into ETCs and ETNs in greater detail, focusing on how they are structured and what investors should watch out for in terms of counterparty risk and collateral.
