Exchange-traded funds (ETFs) track indexes, measuring the price and yield performance of various asset classes ; however, they don’t promise to follow every twist and turn. As the ETF fact sheets say, they only seek to replicate, to the extent possible.

When these funds fall short, the result is called a tracking error. Let’s take a look at this divergence in the returns between ETFs and their underlying indexes by looking at the size or the errors, the at-risk ETFs and the causes.

Size of Tracking Errors

An annual study by investment dealer Morgan Stanley found that the average tracking error for U.S.-listed ETFs (excluding inverse and leveraged versions) was 0.52% in 2008. If this average is weighted by assets under administration, the average comes in lower, at 0.39%.

It could be argued that tracking error is even lower than reported because the performance of some ETFs in 2008 exceeded their benchmarks. and the absolute values of these tracking errors were included in the calculations. As this kind of tracking error would be welcomed by investors, one wonders if it should be included in the calculations. Leaving it out would, of course, reduce the average. (To learn more, see Benchmark Your Returns With Indexes .)

There is potential for a misleading impression on another count. A discrepancy of 0.39% may seem small if the ETF returns were 12%, but not so if the return was 1%. So it may be useful to also express the error as a proportion of the return earned, especially when comparing ETFs across categories and over time.

ETFs at Risk

The averages mask a high degree of variability. To illustrate how extreme the divergences can get, check out the three ETFs registering the worst errors during the volatile year of 2008 (using Morgan Stanley’s absolute tracking errors): iShares FTSE NAREIT Mortgage REIT (11.8 percentage points under), Vanguard Telecommunication Services (5.7 percentage points under) and iShares MSCI Emerging Markets Income ETF (4.1 percentage points over).

Sector. international and dividend ETFs tend to have higher absolute tracking errors; broad-based equity and bond ETFs tended to have lower ones. Management expense ratios (MER) are the most prominent cause of tracking error and there tends to be a direct correlation between the size of the MER and tracking error. But other factors can intercede and be more significant at times, as described next.

Sources of Tracking Errors

Premiums and Discounts to Net Asset Value

Premiums or discounts to net asset value (NAV) may occur when investors bid the market price of an ETF above or below the NAV of its basket of securities .

Such divergences are usually rare. In the case of a premium, the authorized participant typically arbitrages it away by purchasing securities in the ETF basket, exchanging them for ETF units and selling the units on the stock market to earn a profit (until the premium is gone). And vice versa if a discount exists.

However, authorized participants (such as specialists on the exchange or institutional broker or dealers) are sometimes prevented from doing these arbitraging operations. For example, the United States Natural Gas ETF traded to a premium as high as 20% in August of 2009 because the creation of new units was suspended until there was clarity on whether or not regulators would allow the ETF to take larger futures positions on the New York Mercantile Exchange. (For more information, see A Brief History Of Exchange-Traded Funds .)

Moreover, the arbitraging mechanism may not always be fully functional during the first and last few minutes of daily trading on stock exchanges, particularly on volatile days. Premiums and discounts as high as 5% have been known to occur, particularly for thinly-traded ETFs.

Optimization

When there are thinly-traded stocks in the benchmark index, the ETF provider can’t buy them without pushing their prices up substantially, so it uses a sample containing the more liquid stocks to proxy the index. This is called portfolio optimization.

A prime example is the iShares MSCI Emerging Markets ETF. In 2007, it trailed the MSCI Emerging Market Index by more than four percentage points; in 2008, it beat the index by a similar amount, as noted above. This variability is attributed to the use of a sample less than half the size of the index basket. Another example of an optimized ETF is the iShares Russell Micro Cap Index. (To learn more, see An Evaluation Of Emerging Markets .)

Diversification Constraints

ETFs are registered with regulators as mutual funds and need to abide by the applicable regulations. Of note are two diversification requirements: no more than 25% of assets are permitted in any one security and securities with more than a 5% share are restricted to 50% of the fund.

This can create problems for ETFs tracking the performance of a sector where there are a lot of dominant companies. The Vanguard Telecommunication Services ETF, mentioned above, is a case in point: AT&T Corp. makes up about half of the index, but the ETF can only give it a 25% weight. Other ETFs in the same boat include: iShares Dow Jones US Energy Sector, iShares MSCI Brazil and iShares MSCI Mexico.

The iShares FTSE NAREIT Mortgage REIT, mentioned above, was off its benchmark in large part because one REIT zoomed to over 50% of the index (offsetting the declines in most other mortgage REITs during 2008). The ETF could not participate fully in this gain because it had to keep trimming its position back to 25%. (Learn about the differences between mutual funds and ETFs in Mutual Fund Or ETF: Which Is Right For You? )

Cash Drag

Indexes don’t have cash holdings, but ETFs do. Cash can accumulate at intervals due to dividend payments, overnight balances and trading activity. The lag between receiving and reinvesting the cash can lead to a variance. Dividend funds with high payout yields are most susceptible. An example of this would be the iShares DJ Utilities ETF. The contribution of cash drag to tracking error so far has been rather small.

Index Changes

ETFs track indexes and when the indexes are updated, the ETFs have to follow suit. Updating the ETF portfolio incurs transactions costs. And it may not always be possible to do it the same way as the index. For example, a stock added to the ETF may be at a different price than what the index maker selected.

Several dividend ETFs had a hard time in 2008. Turnover in their indexes was high because many high-yielding companies significantly reduced, or eliminated dividends, requiring their removal from the index.

Capital-Gains Distributions

ETFs are more tax efficient than mutual funds, but have nevertheless been known to distribute capital gains that are taxable in the hands of unitholders. Although it may not be immediately apparent, these distributions create a different performance than the index on an after-tax basis.

Indexes with a high level of turnover in companies (e.g. mergers. acquisitions and spin-offs ) are one source of capital-gains distributions. The higher the turnover rate, the higher the likelihood the ETF will be compelled to sell securities at a profit.

One of the most dramatic cases was the 2007 distribution by the Rydex Inverse 2x Select Sector Energy ETF amounting to about 86% of the unit price. Inverse and leveraged ETFs use swaps. futures and options. and when an authorized participant wishes to redeem a block of units, they usually have to be sold to generate cash for the redemption, which can trigger capital gains. Regular ETFs can do a non-taxable, in-kind transfer. (Read How To Reduce Taxes On ETF Gains to learn more about this issue.)

Securities Lending

Some ETF companies may offset tracking errors through security lending. which is the practice of lending out holdings in the ETF portfolio to hedge funds for short selling. The lending fees collected from this practice can be used to lower tracking error if so desired.

Currency Hedging

International ETFs with currency hedging may not follow a benchmark index due to the costs of currency hedging, which are not always embodied in the MER. Factors affecting hedging costs are market volatility and interest-rate differentials, which impact the pricing and performance of forward contracts. (For more information, see Hedging With Currency Swaps .)

Futures Roll

Commodity ETFs in many cases, track the price of a commodity through the futures markets. buying the contract closest to expiry. As the weeks pass and the contract nears expiration, the ETF provider will sell it (to avoid taking delivery) and buy the next month’s contract. This operation, known as the roll, is repeated every month.

If contracts further from expiration have higher prices (contango ), the roll into the next month will be at a higher price, which incurs a loss. Thus, even if the spot price of the commodity stays the same or rises slightly, the ETF could still show a decline. Vice versa, if futures further away from expiration have lower prices (backwardation ), the ETF will have an upward bias. (Learn more about the futures curve in Contango Vs. Normal Backwardation .)

The divergence can be large. For example, a year or so after its inception in 2006, the United States Oil Fund trailed its benchmark, West Texas Intermediate crude oil, by 13%.

Maintaining Constant Leverage

The tracking error is minuscule for the daily change in indexes but some investors may not understand how leveraged and inverse ETFs work and will be under the impression that they are supposed to return twice or three times the direct or inverse index change for periods longer than a day. However, these ETFs won’t accomplish this most of the time because daily rebalancing of the derivatives changes the amount of principal compounded each period.

The Bottom Line

As you can see, ETFs do not always follow the sector or exchange that they’re supposed to due to a number of rules, regulations and shortcomings. It’s important to understand these tracking errors before investing in ETFs. ( To learn more about ETFs, see Dissecting Leveraged ETF Returns and Five ETF Flaws You Shouldn’t Overlook . )