The number of exchange traded funds (ETFs) listed on markets around the world is growing rapidly. However, the launch of so many new ETFs has also made this part of the investment landscape a much more confusing place. With that in mind, we have prepared this short guide to understanding exchange traded funds.

We will start with a short review of investment theory, and then use it to classify the many ETFs that are now offered in markets around the world. Let’s begin with the definition of an asset class. In our view, asset classes should be broadly defined, as they are distinguished by significantly different underlying economic return generating processes. Different statistical techniques can be used to perform this analysis, including correlation (true asset classes should have returns that have low correlations with each other) and principal components analysis (true asset classes should have different loadings on different return generating factors). For example, consider the difference between domestic investment grade bonds and emerging market equity. The underlying economic processes that generate the returns on these two investments are quite different from each other.

In contrast, the processes generating returns on large cap emerging market equities and small cap emerging market equities are quite similar, as evidenced by the high correlation between their respective returns. Hence we regard these two categories not as distinct asset classes, but rather as an example of tilts or sub-segments within the emerging market equity asset class.

An asset class’s return generating process can be broken down into two parts. The first is common to all the securities that make up the asset class. It is often called the systematic or non-diversifiable return on the asset class. The second return source is either unique to a specific company, or common only to a subset of companies within the overall asset class (e.g. companies in the energy sector). Here is a simple example. Consider an asset class made up of only two securities, which have equal weightings on all possible measures (e.g. their market capitalization, their book value, their sales, etc.). The return on security A is 7%; the return on security B is 3%. The average return is 5%, which represents the systematic return on the asset class, which would be received by an investor who owned both A and B. The unique return on Security A is 2%, and on B it is (2%).

This simple example illustrates a number of critical points. First, at the asset class level, the unique returns (also called alpha returns) cancel each other out, leaving only the systematic return. Traditionally, this has been referred to as the market or beta return. Investing with the objective of earning only this broad asset class return should, in our view, be called asset class investing, market investing, beta investing or passive investing. As you can see, the distinguishing characteristic of the market return for an asset class is that it requires no ability or attempt to forecast A and B’s future returns. It simply seeks the return that comes from owning all the securities in the asset class.

Second, return is compensation for bearing risk. At the asset class level, you receive only systematic market return, which compensates you for bearing systematic market risk. This systematic return is composed of two parts: the risk free rate (which compensates you for deferring consumption) and an asset class risk premium. Most important, earning this asset class risk premium does not depend on skill.

A third insight from our simple example is that, when you take on additional unique risks that can be avoided through diversification, you may receive additional compensation in the form of positive alpha. In the short term, this can be due to either luck or skill. However, as you keep taking on unique risks over longer and longer periods, the probability increases that your average return will be either zero or (after your higher costs) negative — unless you have better than average forecasting skill, and can better (than the average active investor) distinguish between those unique risk exposures that will earn positive and negative alphas. This last point is an important one, and is often overlooked. Successful active investment doesn’t depend just on having some forecasting skill. Rather, it requires that your forecasting skill be superior to the average level possessed by the other active managers against whom you are competing. And that is a far higher bar than most people like to admit to themselves.

A fourth point is that forecasting skill must be based on either access to superior information and/or use of a superior model to make sense of publicly available information. It also must be based on the existence of a financial market that is not perfectly efficient. In an efficient market, skilled forecasting is impossible because market prices already incorporate all the public and private information available about a security, and the pricing insights generated by many different models. As we have noted many times in our writing, we believe that financial markets are not always efficient. Rather, we see them as a complex adaptive system that, while strongly attracted to efficiency, seldom achieves it. Hence, we believe that skilled forecasting and successful active management that generate alpha are possible, though quite rare in practice (especially after the additional costs incurred are taken into account).

The fifth point from our simple example is that asset class or passive investing, as we define it is not quite the same as indexing which is often used as a synonym for it. In our example, we measured the performance of our two-security asset class by constructing an index, which put equal weights on A and B. The index return was equal to 5%. But suppose that A and B differed from each other in ways besides their return. For example, suppose A had higher revenues and book assets, but lower market capitalization than B. What weights would we then use to construct our index to measure the systematic return on our asset class? Reasonable people can and do disagree on the right answer to the question of how best to measure market returns. This is a critical point, because the use of two different benchmark indexes to measure market return will lead to two different estimates of the size of an active manager’s alpha. If that manager’s compensation is in any way tied to the amount of alpha he or she generates (as is usually the case), this creates a potential conflict of interest. An active manager has an incentive to choose a market benchmark that maximizes reported alpha, while the investor who is hiring the active manager has an incentive to use the most accurate market benchmark possible. As a practical matter, this benchmark decision is often made by a third party. In the case of institutional investors (e.g. pension plans), it is often made by a consultant hired by the plan sponsor. In the case of individual investors, it is often made by either a financial adviser or a rating service (e.g. Morningstar or Lipper). However, we note that the potential conflict of interest will still be present if the party making the benchmark decision (i.e. the consultant or rating service) derives any economic benefit from the generation of more alpha rather than less by active managers. In our view, this is often the case, with too little recognition of the conflicts of interest that are present.

With that introduction, let us move on to the market for managed investment products (e.g. hedge, mutual and exchange traded funds). Broadly speaking, these fall into three categories. Some products offer only systematic (beta) returns. Their objective is to provide a return equal to the average return on a broad asset class, such as domestic equity (however it is measured by the index provider). Because capturing these returns requires no forecasting skill and only minimal trading, these asset class index funds charge very low expenses.

At the other end of the spectrum, some funds offer only unique (alpha) returns. An example of this type of fund is an equity market neutral hedge fund. The manager of such a fund attempts to do two things: (a) utilize his or her superior forecasting skill to identify securities and transactions that will produce positive alpha; and (b) use other transactions to eliminate the fund’s exposure to systematic (beta) returns. Because of the additional operations involved in comparison with a beta only fund, this pure alpha fund must charge higher expenses. Also, based on the assumption that it is easier to run a beta only fund than one that earns pure alpha, the manager of the alpha fund also expects to receive a larger portion of the fund’s returns, as compensation for the use of his or her relatively scarcer skill. We also note that you can construct an index to measure the average performance of all equity market neutral fund managers, even though there is no systematic (beta) return involved. As we said, beta investing is not the same thing as indexing.

Other products seek to provide investors with a bundled mix of beta and alpha returns. Most actively managed mutual funds are in this category. They buy and sell securities, but don’t eliminate their exposure to beta returns. To maximize their forecasting advantage, many active managers restrict their investing activities to a sub-segment of the broad asset class (e.g. small companies’ shares, or shares of healthcare companies), which provides a convenient basis for classifying these funds into sub-segments of the broad asset class.

Finally, in between pure beta and pure alpha funds lie products that use a low-cost indexed approach to track the performance of different sub-segments of the broad asset class, that are identified using a clear, publicly disclosed set of rules. For example, sector and style (e.g. small cap value) exchange traded funds are examples of these products, as are bond ETFs that track different maturity indexes. Clearly, because they hold portfolios of securities that differ from the composition of the overall asset class, the returns they produce are a type of alpha. Yet, because these approaches to earning alpha have become well-known and embodied in a rules-based index, they are, confusingly, often called beta or factor beta. (Alpha in beta clothing, if you will). As a result, the meaning of alpha has shrunk, and is now sometimes taken to include only active manager returns net of the return not only on the broad market but also on one or more sub-segment indexes (i.e. factor betas) that an investor can buy for a relatively low price. To put it differently, alpha is now often taken to mean only an active manager’s gross return, less the return on the market and the return on relevant factor betas to which the active manager has decided to be exposed via the investments his or her fund makes.

This alpha arguably comes four sources. The first is market timing, or the skill to profitably switch between asset class or factor exposures. The second is security selection, which reflects superior skill in forecasting the returns of individual assets. The third is skill in profitably providing insurance to other investors, for example by selling them put options that limit their downside risk (but increasing the manager’s). This is theoretically an attractive source of alpha, as it is not an inherently zero sum game. The fourth source of alpha involves earning a fee for providing liquidity to other investors. Again, this has the advantage of not being an inherently zero sum game. However, liquidity within some asset classes (with equities in the lead) is rapidly migrating to factor-beta status.

In our view, the creation of factor beta index products has been both a blessing and a curse. On the one hand, they have made it possible to implement a wider range of forecasts at lower cost. On the other hand, they have probably created a dangerous amount of confusion in many investors’ minds. Too many people appear to be under the illusion that they can earn alpha over a long-term holding period simply by using these factor beta index funds to permanently tilt their portfolios one way or another. In a reasonably efficient market, this should be impossible. Rather than alpha, a long-term one way factor beta tilt (e.g. toward small cap value stocks) should produce either lower returns but with lower risk than the overall market, or higher returns with higher risk. To believe that it will produce positive alpha requires acceptance of two additional premises.

The first is that some investors will systematically, over long periods of time, and for one or many reasons, make valuation mistakes. There is some evidence that this may happen. For example, immediate liquidity needs will always force some investors to sell securities they know are undervalued. And some investors will, because of overconfidence or their use of a momentum strategy, tend to buy securities that are overvalued. However, the second premise is that there are permanent barriers that prevent other investors from arbitraging away most of the alpha that these mistakes are expected to produce, by buying (and bidding up the price of) the undervalued securities, and selling short the overvalued securities. The evidence suggests that this premise is much weaker than the first one (see, for example, The Limits of the Limits to Arbitrage by Brav and Heaton). Moreover, if both these premises are true, historical data should show significant positive risk adjusted returns from permanently tilting one’s portfolio towards a sub-segment. But this is not what we find.

One way to measure the effectiveness of an active management strategy is by using something called the Information Ratio. To calculate this, you start with the return on the sub-segment tilt (e.g. the return on a small cap value index) and subtract from it the return on the broad asset class index. Over many periods, the average of this result is the active return on this strategy or its gross alpha. If you subtract the expenses you pay to the active manager from this, it is the net alpha. The next logical step is to relate this to the amount of risk that was taken to earn the alpha. This is measured by the standard deviation of the alphas, which is known as active risk or tracking error. The Information Ratio therefore measures the risk adjusted return of the active strategy, by dividing the active return (alpha) by the active risk (tracking error) that was taken on to earn it. Information ratios of .50 or more are generally considered excellent performance by an active manager (although this varies by asset class, with higher IRs generally needed for top quartile performance in asset classes where returns are more volatile).

The following table shows the annualized net alphas, tracking errors, and information ratios for four common sub-segment tilts over two different ten year periods, covering 1979 to 1988, and 1989 to 1998. All the data are in nominal terms, and are based on the Wilshire Indexes.