The article was created for www.etfmarketpro.com.
Investable indexes are designed to be realistic representations of actual market spaces (the S&P 500 index represents the securities in the S&P500); however, in some asset classes, indexes are too idealistic to be managed as ETFs in a cost effective and efficient manner. Many Exchange-Traded Fund companies employ index sampling techniques called index optimization to help track the index more effectively. This is where the index will not attempt to purchase every stock in the index, but purchase a sampling of securities. There are drawbacks and benefits to optimization techniques, which vary among asset classes and ETF issuers.
The main reasons that an index would need to be sampled by an ETF are liquidity, trading costs and large numbers of holdings. There are many issues surrounding this topic and investors should be aware and include these risks in their analysis.
A fundamental feature of ETFs is that they trade over exchanges with intraday market prices. These market prices are ideally supposed to be in line with the intraday movements of the underlying holdings of the fund, but some variance is normal. A basket of underlying securities or assets that the ETF represents is called a creation unit. The intraday price of a creation unit is tracked by an ETF’s intraday indicative value (IIV), which is the fund’s net asset value (NAV). The ETF’s NAV is managed and has one objective-to track the index benchmark.
A creation unit is one basket of securities, which together make up the ETF’s NAV and underlying investments. Each day, the creation unit is set by the ETF’s management so that market makers can create liquidity for the ETF by trading in ETF creation units for ETF shares or delivering in ETF shares for creation units.
Since it is market makers that create this liquidity and fair pricing for ETFs, it is important to have a liquid and tradeable creation unit so that this process can be smoothly done. This is why splitting ETF shares makes a lot of sense and is extremely practical. This is unlike stock splits, which add no actual value. Since creation units are traded in units of 50,000, splitting an ETF from $100 to $50 (with no change in value) allows a creation unit to be handled with only $2.5 million instead of $5 million. This is a useful ability when cash is harder to obtain or tie up.
In the same way, when a creation unit is made up of a small number of very liquid assets, the creation units are easier to obtain and trade and therefore make the ETF more efficient. For example, the stocks within the S&P 500 are so easy to obtain and trade that on any day, that S&P 500 ETFs are some of the world’s most efficient ETFs.
Below is a chart showing the difference between the number of holdings in two ETFs that track the same index compared to the index which they both track. The differences between the two ETFs and the index are vast and each has different benefits and risks.
Vanguard, in general, does not optimize or sample indexes, but includes all or most of the underlying holdings of the index. They do this to minimize the risk of tracking error and desire to represent the index without any management decision. The ETF included below actually has more holdings than the index; this is because of the international nature of the emerging market investments and the need to have underlying liquidity during US trading hours while international markets are closed. Vanguard merely “doubles up” on holdings when the exact same international stock is available as an ADR investment.
IShares, in general, practices sampling of the indexes which they track. As seen in the chart below, the iShares emerging market ETF has half the holdings of the index. The reason for this, as addressed above, is that the creation units of the ETF have 386 stocks instead of the 751 stocks in the index. This smaller basket makes creating and redeeming creation units much easier, especially since liquidity may be an issue with some of the stocks from emerging market countries. The risk of this kind of sampling is that tracking error may happen. A report on the tracking error of these two funds is here.
Data: Morningstar Office 9/30/2009
Emerging market ETFs are a good example of the differences between sampling and full replication ETF products. The tracking errors and cost have been extremely different in the past. The example of emerging markets is not the only place with these differences between funds. However, it is a good example to show how to break down and explore these differences in any asset class.
Liquidity in many asset classes is an issue, especially when dealing with ETFs. It is widely thought that ETF liquidity only matters when concerning bid/ask spreads, but the most important liquidity factor is in the underlying holdings of a creation unit of an ETF.
Like the example of emerging markets above, liquidity is something that ETF managers optimize since it is so important. For example, when a stock within an index represents .001% of the overall index, it has a low marginal effect on the overall returns of the index. In this case, sampling the index to exclude stocks with this little amount of overall effect allows the creation unit to be free from obtaining this stock (which may be costly) with little effect on tracking error. When an index contains a majority of very small holdings, the issuing becomes harder. The creation unit is more costly to put together and trade for market makers, and when the small holdings are extremely illiquid, as they will often be since most indexes are market cap weighted, the pricing on these stocks, which are the majority of the fund, can be inefficient.
Investors must choose between index sampling ETFs and full replication products. Each asset class has a unique set of issues and risks surrounding sampling and should be dealt with on an individual asset class basis. An investor should be looking for the least amount of structural risk within an ETF product that will allow for the best representation of the assigned category.