Each Method of Handling Illiquid Index Components has its Own Pros and Cons
Unlike traditional actively-managed open-end funds which aim to best their benchmark, the goal of index-linked passive investment vehicles like ETFs is to track their benchmark index as closely as possible. Aside from the inevitable management fees, one of the biggest obstacles to perfectly tracking the performance of an index is the costs involved in buying and selling the underlying holdings. This applies to both physical- and synthetic-replication products alike, albeit in different ways.
In the case of physical-replication ETFs, the fund itself does not actually go into the market to buy or sell securities. Instead, on the primary market, authorised participants (APs) purchase baskets of securities which they then swap for new shares in the ETF. Fierce competition for the trading profits from buying or selling ETF shares versus their portfolio of securities keeps ETF market prices close to their net asset value. But these arbitrage opportunities can sometimes be insufficient to entice APs in cases where the index contains illiquid securities that would cost far too much to buy or sell. The cost of this market impact for APs is ultimately be borne by ETF investors in the form of a wider spreads to buy or sell in the secondary market.
The main advantage boasted by synthetic or swap-based ETFs is that they track their benchmark index with higher fidelity than physical replication funds. Within synthetic-replication funds, the fund enters into a total return swap, typically with a large investment bank. The fund will receive the performance of its reference index, in exchange for a swap fee and the performance of the collateral or substitute basket. The counterparty cares about the liquidity of the constituents of the ETF's underlying index because it will often buy or sell those securities in an effort to hedge away the risk it assumes by agreeing to provide the index return. Much as is the case within physical replication vehicles, the costs of tracking indices containing less liquid or difficult to locate securities will ultimately be borne by investors in the form of higher expense ratios, big spreads versus the NAV on secondary markets, higher swap fees, or--most likely--some combination thereof.
While the aforementioned issues concerning illiquid securities rarely impact ETFs which track indices composed of large, developed-market stocks and bonds, they do emerge in less liquid categories like emerging market equities and high-yield corporate bonds. The liquidity of the underlying securities can also be a concern within ETFs tracking more mundane fixed income indices, which may also have components that trade infrequently. To keep their funds trading fluidly in even the rockiest of markets, there are several methods ETF providers use to reduce the effects of illiquid index constituents, each with its own benefits and drawbacks.
Optimisation
Optimisation is where the index provider takes investability into account in structuring the selection and weighting of individual securities. However, this introduces a trade-off between greater investability and the extent to which the index adequately represents its chosen market. An example of this technique is found in Source's STOXX European sector ETFs. The ETF provider contracted with the index provider STOXX to create a series of European sector indices optimised for liquidity, so that the ETFs would be able to more faithfully track the performance of the index, thus making them a favourite among traders and institutional investors who are concerned about minimising trading costs.
Synthetic or Swap-based ETFs
As we previously mentioned, synthetic ETFs reduce tracking error as the fund does not hold the underlying securities but instead enters into a swap agreement to gain the returns on the benchmark index. One specific example of where a synthetic fund exhibited superior tracking to a physically-replicated fund can be seen with two funds tracking the MSCI Emerging Markets index. The physically-replicated Credit Suisse ETF (Lux) on MSCI Emerging Markets (XMHB) underperformed the benchmark by nearly 3% in 2009, while the swap-based db x-trackers MSCI Emerg Mkts TRN (XMEM) Index ETF only underperformed by 1%. The nearly two percentage point difference between these funds 2009 returns obviously can't be fully accounted for by the 0.05% difference in the total expense ratio of the two funds, or by the difference caused by the accumulation or distribution of dividends. So with this admittedly narrow example, we're left with the superiority of using swaps to track the performance of an index composed of relatively illiquid stocks compared to holding the underlying securities to explain the discrepancy.
Representative Sampling
For physically-replicated ETFs tracking indices where it is difficult or prohibitively expensive to fully replicate the benchmark index, ETF issuers have used representative sampling as a solution. Representative sampling employs mathematical models to create a portfolio that excludes the least liquid, most difficult to locate securities from the full index portfolio whilst simultaneously seeking to ensure the fund tracks its benchmark closely. These models examine correlations between the securities in the index so that the securities which are expensive or otherwise difficult to acquire can be replaced by other securities while maintaining the portfolio's risk and return characteristics.
Of course, correlations can change over time, so optimisation introduces a potential source of tracking error that a fully replicated physical replication ETF wouldn't have. Essentially, if the market is stable, and the models used to underpin the sampling are accurate, the savings from lower transaction costs passed on to the ETF investors would outweigh any potential concerns related to tracking error. However, the opposite can be true if the correlations change, making the potential for a large tracking difference ever present. Investors in these ETFs are trading the known benefits of lower transaction costs for the potential for higher tracking error. A good example of this phenomenon is the iShares MSCI Emerging Markets ETF (IEMM). During 2007, the ETF only underperformed the index by 4 basis points, but in 2008 (a much more volatile year for emerging market equities), the tracking difference increased to 163 basis points.
In some cases, there can be multiple ETFs tracking the same index, where one uses sampling whilst another opts for full replication. Defenders of sampling claim that because it is more cost-effective, over the long run ETFs using sampling will outperform fully-replicated ETFs. A good example of this takes place with the aforementioned MSCI Emerging Markets index. Credit Suisse and iShares both have physically-replicated ETFs which track the performance of the index, with Credit Suisse opting to use full replication and iShares' ETF using sampling--holding less than half of the more than 750 securities in the index. The iShares' ETF tracked the benchmark very well in 2009 as its return bested the index by 15 basis points while the Credit Suisse ETF underperformed that year by more than 3 percentage points, although differences in share lending income may have contributed to that disparity.
Index sampling is neither exclusively positive nor negative, but it is an important concept for investors in physically-backed ETFs to understand. As many investors are not even aware of the existence of this practice, it is critical that ETF providers make their clients aware when this technique is being used, and for investors themselves to balance the potential plusses and minuses of sampling when choosing an ETF investment.
