International index funds and ETFs showed large tracking errors in 2009. In my first post on this topic, I explained how the time differences between North American and overseas markets can make it appear as if international funds are not tracking their indexes closely. Now let’s look at another reason that international ETFs can diverge from their index: representative sampling.
Many ETFs track their underlying indexes almost perfectly. For example, S&P 500 funds generally hold all 500 stocks in the index, in almost the exact same proportions. The iShares S&P/TSX 60 Index Fund (XIU), too, holds all 60 companies in this Canadian large-cap index. But few international indexes are as easy to track as these popular benchmarks. Sometimes it’s because the indexes are much larger, but mostly it’s because overseas stocks are often less liquid and more expensive to trade.
For example, the MSCI EAFE Index — the most popular yardstick for developed markets in Europe, Australasia and the Far East — includes 952 companies in 21 countries. The MSCI Emerging Markets Index, meanwhile, includes 765 stocks in 24 nations. As you can imagine, managing a portfolio of that many stocks, which trade in a host of different currencies, can be difficult and expensive.
When an index is hard to replicate fully, ETF providers use representative sampling. Rather than buying every company in the index, they buy a smaller sample that the fund manager expects will closely mirror the index’s performance. When chosen properly, the sample will have characteristics very similar to the overall market in terms of industry weighting, market cap (the mix of large and small companies), dividend yield, and the like. This can be a smart trade-off: buying fewer stocks than the index can cause tracking error, but it also reduces transaction costs.
Let’s look at some examples. The Vanguard Europe Pacific ETF (VEA) holds 934 of the stocks in the MSCI EAFE index, while the iShares MSCI EAFE Index Fund (EFA) and its Canadian-listed equivalent, XIN, hold 859. iShares also uses representative sampling in its MSCI Emerging Markets ETF (EEM) and its Canadian version, XEM: the fund selects 611 of the index’s 765 holdings.
Sampling need not always cause negative tracking error: about half the time it should achieve returns slightly higher than the index. iShares’ EEM, for example, had an ugly –6.72% tracking error in 2009, but the year before it outperformed by 3.33%. Over the last five years, its annualized tracking error is 0.86%, which is almost entirely explained by its management fee. In other words, the tracking error from representative sampling (if it’s done properly) is random and should even out, so it isn’t a major concern for long-term investors.
However, this assumes that the sample is large: Vanguard and iShares international ETFs typically replicate 90% of the index or more. In the case of Claymore’s international ETFs, however, the representative samples are much smaller, making up as few as one-quarter of the stocks in the index. That was a major reason why these funds showed huge tracking errors in 2009.
Claymore’s Japan Fundamental ETF (CJP) tracks an index of 254 stocks, but holds just 135 of them. The ETF had a tracking error of –8.54% in 2009. Claymore’s International Fundamental (CIE), which is benchmarked to an index of 1,000 stocks, lagged by 6.73%. One reason was that the fund’s largest holding in 2009 was CJP. In March 2010, the fund stopped using CJP and now holds its Japanese stocks directly, but its representative sample is still just 364 of the 1,000 companies in the index.
Meanwhile, Claymore’s US Fundamental ETF (CLU) had a tracking error of –9.67% in 2009. Last year the fund held 250 of the 1,000 stocks in the FTSE RAFI Fundamental Index. The sample size has since been increased to 350 securities, which should reduce the tracking error going forward, but is still quite small.
To be fair, the tracking errors on CLU and CIE were tiny in 2008, and CIE beat its benchmark in 2007. (CJP, on the other hand, has lagged for three years running.) Perhaps the errors will even out over the long term. But Claymore’s job is tougher, because it lacks the economies of scale that benefit larger players. The Vanguard and iShares EAFE funds have $10 billion to $40 billion in assets. By comparison, CIE holds about $110 million, while CJP has a mere $15.3 million. For funds that small, fully replicating a large international index is probably prohibitively expensive.
Of course, that’s not your problem. As a Couch Potato investor, you have the right to expect an ETF’s performance to closely track its index. Advocates of fundamental weighting (of which Claymore is the most vocal) argue that their indexes are superior to traditional cap-weighting. They might be right. But a superior index doesn’t do you any good if the ETF tracking it underperforms that benchmark by 7% to 10%.
Claymore has made some improvements to their sampling in 2010, and as their funds grow larger, they should be able to reduce these tracking errors. We’ll be watching.
[My thanks to Claymore for providing me with the data I needed to prepare this post. I appreciate their openness and transparency.]
Good on Claymore for trying to make improvements on their tracking errors.
Thanks for another very informative post CCP.
In the last section you mentioned fundamental indexing. Do you have any future plans on posting about that in more detail? There doesn’t see to be too much material on PF blogs about this topic in particular.
Brian: I have been thinking about a post on fundamental weighting for a while, but something else always seems to come up! The Claymore link I include at the end of the post is a good start. The most complete resource you’ll find on this topic is Rob Arnott’s book, The Fundamental Index. It may be more detail than you want, but the first couple of chapters lay out the case very convincingly. I found it quite readable. There’s a link on my Library page
Of the PF blogs, your best bet is WhereDoesAllMyMoneyGo.com. Check out the series he starts here: