Wall Street’s creative genius never runs out of new investment strategies for another exchange-traded fund based on one index or another. But in many cases, the funds barely reflect the actual securities in the indexes to which they are tied. In fact, many ETFs hold less than 10 percent of the components of the index they are marketed as representing. And critics say the returns of ETFs with what they call major “tracking error” can vary so much from those of the index as to be misleading.
Last March the technical committee of the International Organization of Securities Commissions, an international body of securities regulators, asked its members if regulators should make ETFs explain to investors how a new fund replicates its index and the degree to which its returns may vary. The comment period runs until late June.
“How 50 to 60 issues in an ETF will track 1,000 issues in a bond index is a good question,” says James Squyres, president of Buyside Research, a Darien, Connecticut–based research firm that analyzes ETFs based on the fundamentals of the securities they own. “Why not just create a fund with 50 to 60 issues and not track any index? What does tracking an index bring to the party?”
A well-known index can give a new fund instant gravitas and prospective investors a performance record on which to base their decisions. And skillful managers can use modeling techniques to produce index-like results in a fund without holding all the index components, thus saving a bundle in inventory transaction costs.
But tracking error can create big deviations between the fund’s annual returns and those of its benchmark. Firms applying to the SEC to create an ETF must state the maximum deviation they expect tracking error to produce, and that’s usually 5 percent. From 2008 to 2011, most ETFs stayed well below that ceiling, according to Lipper, with the exception of commodities funds, whose returns diverged by an average of 5 percent to 9.5 percent.
But analysts say other types of funds are subject to such divergence, particularly in fixed income.
As a rule, “a small universe of portfolio holdings relative to a highly diversified index will result in high tracking error“,” says Greg Davis, who heads up the bond indexing group for the Wayne, Pennsylvania–based Vanguard Group, a provider of mutual funds and ETFs. “Folks need to be careful since the fund returns could be much different than the index returns.”
For example, Barclays Capital U.S. High Yield Bond index holds about 1,900 components. Keeping in mind that a 30-year bond will have a very different risk profile from a 2-year bond, tracking error for a fund holding only 50 of the Barclays index components would produce 240 basis points of deviation in return, according to Davis’s models. If you could increase the fund’s holding to 1,000 securities, the tracking error would decline to about 19 basis points, which is considered more than acceptable for a highly diversified portfolio.
How a fund issuer prices its portfolio can also yield deviations in returns. “What gets published by the fund is the number the fund uses from its pricing service, which is not necessarily the same pricing service the index would use,” notes Davis. “There may be a 90 basis point difference in return that needs to be taken into consideration.”
The difference between the index price and the fund price can vary even among the same bond issues held in the exact same weights. Davis’s group compares the index and the fund numbers from its pricing service with the benchmark’s prices. The difference is called the pricing adjustment, which allows Davis to see how much of a difference in return is pricing-related. “It can be a relatively small amount,” he says. Its impact depends on given market conditions and should disappear in the long term. But by stripping it out, Davis says, they “can see how effective we’ve been in matching the benchmark pricing.”
Tracking error has been associated with ETFs in the past. In August of 2010 Credit Suisse referenced the fact in a press release announcing its plans to use swaps to ensure that its new line of emerging-markets ETFs would follow their benchmark. Until then, the bank had physically matched components to replicate its indexes, at considerable expense. But by last fall, Credit Suisse reversed course, saying it would return to replication and begin levying a swap-related charge on in its remaining synthetic funds. In November the bank switched four funds back to physical replication — all emerging markets: the MSCI Australia, MSCI Brazil, MSCI Mexico Capped and MSCI South Africa funds.
The challenge portfolio managers face is how closely one can track an index without having transaction and inventory costs eat all of the gains. “The trade off is how close you can come to the index while living with the liquidity issues you have,” says Dan DiBartolomeo, president of Boston-based Northfield Information Systems, which makes the software for testing and monitoring tracking error. Indexes dominated by a few big securities can help minimize the problem. DiBartolomeo notes that the 80 biggest issues on the 300-stock Australia Stock Exchange, for instance, probably represent 90 percent of its total capitalization. Conversely, tracking error can be magnified with a small-cap index, which usually has a smaller number of stocks, even if it is cap weighted. In that case, he observes, a stock that suddenly triples or a company that gets indicted can quickly skew the whole index.
When using ETFs to hedge exposures within emerging-markets positions, tracking error becomes even more important. If a hedge fund buys the ETF to cover a short-term exposure to a specific African country, but the fund hasn’t been able to buy bonds today for that sovereignty, the position may be inadvertently left naked. “Indexes of obscure securities, such as equities in Africa, can be a problem,” says DiBartolomeo. “Replicating volatile securities by sampling may not follow the index.”
Technology has helped managers create better models for replicating or optimizing an index, which requires active management and specific expertise even in a passive fund. Buy-side clients can also use models to monitor risk exposures in a fund and its index, such as MSCI Barra and Axioma models for equities and DiBartolomeo’s Northfield models and analytical tools. Some advisers will run the models for their clients, says John Tucker, managing director and head of U.S. Index ETFs for State Street Global Advisors in Boston. “It could be useful information to know especially in very illiquid markets.”
Tucker points out another rule investors need to remember: The index never includes transaction costs, which matters a lot in the case of emerging markets. Transaction taxes can also be significant there. And funds need not include these costs in their expense ratios. But as for tracking error, Tucker contends it is not necessarily a negative. “Constantly trading (for a large fund) to keep up with an index racks up costs and will cause the fund to underperform,” he says. The key lies in the expertise behind the fund. “To do it well requires lots of infrastructure, monitoring index details with precision and accuracy and staying on top of lots of different data, through a strong operating staff with lots of systems.”
