GIVEN that investors’ prime reason for buying exchange-traded funds (ETF) is to gain exposure to indices, in an ideal world the performance of an ETF ought to reflect movements in the benchmark index as closely as possible. Unfortunately, ETFs do suffer from tracking error, which is defined as a divergence between the performance of the underlying index and the ETF that tracks it. Normally – although not always – this mistracking causes an ETF to underperform its benchmark index. Research by Deutsche Bank shows that European ETFs had an average tracking error of 0.3 per cent (30 basis points) in the second quarter of 2010.
The main reason for the disparity between the performance of the underlying index and the ETF is cost. Even if the ETF tracks perfectly, it will always underperform the benchmark by the fees that are deducted from the fund’s returns. While important, an ETF’s fees are readily available and can easily be accounted for. The other causes of tracking error are less easily identifiable but can amplify the underperformance. The Deutsche Bank research showed that ETFs that track more volatile benchmarks have experienced a higher tracking error – for example, tracking errors were higher for all types of ETFs in 2008 than in 2009 and 2010. (See chart.)
There are three factors that can contribute to elevated volatility in physically replicated ETFs, which translates into higher tracking error.
First, if a benchmark contains illiquid constituents – this is often the case with emerging market indices – then attempted replication can generate mistracking by selling securities to redeem ETF units. Second, Christos Costandinides, ETF strategist at Deutsche Bank, says that ETFs are typically denominated in a single currency. Hence, optimising a physical portfolio of a benchmark whose constituents are denominated in different currencies also requires managing a currency basket. This currency hedge needs to be continuously optimised and this exercise may often give rise to tracking error and costs, he warns.
Third, it can be very difficult for ETF providers to physically replicate indices that have hundreds of constituents. Therefore, statistical portfolio sampling techniques attempt to optimise replication by holding a fraction of an ETF benchmark’s total constituents. The magnitude of mistracking will depend on the number of constituents held and market events, which are often more pronounced when an ETF has multi-country constituents.
Investors need to be careful if their ETFs’ benchmarks display any or all of these characteristics. If so, then the ETF might not mirror the underlying index’s performance precisely and you are liable to suffer from greater tracking error before you even take the fees into account.