Guide to tracking error

As ETFs continue to rise in popularity among both retail and institutional investors, there is a growing need for better understanding of some of the structural elements of how these investment vehicles operate.

ETFs continue to be successful because of their efficiency, low cost and liquid means to access various indexes, but it’s important for investors to understand how closely the fund has tracked the index to which it is benchmarked.

This difference is commonly called tracking error. It can be an important indicator of the quality of a funds structure, the potential hidden costs, and even the competence of the ETF manager.

What is tracking error?

Tracking error should be thought of as the difference between the performance of a fund and the performance of the index to which it is benchmarked. There are a couple of ways to come up with this measurement.

Traditionally, institutional investors have used a statistical measure — this approach calculates the variance (as opposed to the level) of the performance difference between a portfolio and its benchmark index using specific intervals over a period of time.

This measurement is focused on portfolio management skill, and will ignore the effects of fees and structure. For that reason, it’s less useful for most investors when comparing ETFs, since it doesn’t paint a complete picture.

For most investment advisors and their clients, it makes sense to measure tracking error as the realized difference in the performance of a fund and its benchmark index. The calculation involves a simple fund performance minus index performance calculation, and focuses on actual outcomes.

This approach is also handy because it will work for both short and long periods of time (performance for periods longer than a year should be annualized). Like any performance number, this tracking error calculation tells you more when it covers a long period of time (e.g. five or more years) than a short one.

Many advisors prefer to look at performance based on market price when measuring tracking error, which is understandable, because investors trade on price, not NAV.

But for a few reasons, we think it’s preferable for investors to look at performance based on NAV for a more meaningful comparison to the underlying index. It is true the market price of an ETF should closely follow its NAV. Some ETFs can come as close as one cent from the NAV.

However, since market price is affected by all market participants, there can sometimes be a wider difference. This is especially true for ETFs that provide access to less liquid investments, like small- or micro-cap stocks, emerging market stocks, and even some corporate bonds.

NAV, on the other hand, is calculated in a manner that’s similar to the index, and so NAV-based performance is more directly linked to how well the ETF is structured and managed to track its benchmark.

That’s why calculating tracking error using NAV-based performance will give you a more accurate view of an ETF’s quality; it won’t be affected by the short-term noise you might get when using performance based on market price.

Why does tracking error occur? For a “perfect” (that is, theoretical) ETF, tracking error wouldn’t exist — fund performance would be identical to index performance.

But in the real world, where investors pay fees and transactions have costs, a variety of factors affect how well an ETF tracks the underlying index.

Here are three such factors.

01 MERs

Everything else being equal, the higher an ETF’s MER, the higher the tracking error will be.

Given that all ETFs have management fees, the next best thing to a perfect ETF is one whose tracking error is caused only by its MER. Large, well-structured funds with plenty of liquidity come within one-to-two basis points of achieving this goal.

But smaller funds, or funds that must invest in less liquid securities, can have tracking errors that are substantially higher than their MERs.

02 Liquidity and Transaction Costs

Why can tracking error be greater than MER at all? There are several reasons. All ETFs, for one, incur costs when they rebalance to reflect securities that enter or leave the benchmark index. Many funds rebalance on a quarterly or semi-annual basis.

But perhaps the most significant (non-MER) factor in explaining why tracking errors vary is liquidity — or lack of it. ETFs that track large, liquid indexes enjoy more efficient pricing of securities. Bid/ask spreads are slim, and so transaction costs are kept to a minimum.

On the other hand, funds that invest in relatively illiquid markets or securities face higher transaction costs because of wider bid/ask spreads. Those spreads make for less efficient trades, higher costs, and therefore higher tracking error.

03 Sampling or Optimized Strategy

To get around the illiquidity inherent to certain markets, ETF managers can employ a sampling or optimized strategy, where the ETF invests in a reduced portfolio of securities (relative to the index). It’s selected so it should still closely replicate the return of the benchmark index.

This lowers transaction costs. But the practice, which is common in fixed-income ETFs, requires the manager to exercise more discretion than when replicating an index (as with most large equity funds), and therefore requires greater skill.

This approach introduces a gap between the ETF’s composition and the benchmark index, creating inevitable sampling noise, which can be reflected in higher tracking error. Good fund managers will limit the impact of that noise to a large extent, but it’s rarely completely eliminated.

Take the iShares DEX Universe Bond ETF (XBB) as an example of a fund that holds only 720 of the 1,242 securities in the index it is benchmarked to. The 5-year annualized performance of XBB is 6%, while the index is 6.35%, as of year-end 2012. The fund’s MER is 0.33%, so investors are left with a difference of two basis points.

Some fund providers also seek to deliver lower tracking error (and sometimes lower costs) through alternatives like synthetic ETFs, which gain exposure to underlying markets through swaps. While it may reduce tracking error, it also introduces counterparty risks that investors directly invested in ETFs are not exposed to.

When should investors be concerned about tracking error?

Tracking error is an important metric for investors to consider. It may be an indicator of poor fund construction or management, but there are other reasons for it as well. The level of tracking error investors find acceptable should also depend on the asset class they’re trying to access.

For instance, consider an ETF that tracks the emerging-market bond markets. These are small and illiquid markets. An ETF is an excellent way to get access to such hard-to-reach markets.

However, because of their illiquidity, the underlying markets will be more expensive to replicate (i.e., transaction costs will be higher), which will increase tracking error.

But in comparison to the other risks in the underlying markets, and the potential portfolio benefits, an investor might find a relatively high tracking error an acceptable cost of entry.

As such markets mature, we can expect that tracking error in corresponding ETFs will decrease — a trend we have seen over the past few years for emerging-market equity funds.

At the same time, while volatile or illiquid underlying asset classes can naturally drive tracking error up, unusually high error can also occur simply as a result of poor fund structure. The fact is for large, well-established ETFs replicating large, liquid indexes, there is very little excuse for persistent, significant tracking error.

In general, recognizing tracking error’s relationship with liquidity can be key to evaluating its significance in any ETF. In some cases, where the underlying market is relatively illiquid, a high tracking error might be acceptable; in others, where the market is liquid, a high tracking error might be cause for greater concern.

Mary Anne Wiley, CFA, is managing director and head of iShares, at BlackRock Asset Management Canada Limited.