Asset Class Allocation and Portfolios: Critique and Complication

Advisor Perspectives welcomes guest contributions. The views presented here do not necessarily represent those of Advisor Perspectives.

This article is intended for the educated layman. It was written as part of a continuing series of articles on a variety of investment topics. To view all the articles in this series, click on “More by the same author” in the left margin.

In Part 1 of this essay, I explained that for asset class allocation to become an investment practice, it required a foundation of theory. And Modern Portfolio Theory was that foundation. But today, most financial journalists and investment advisors who proffer advice centered on asset class allocation are—if I may judge from their writings—oblivious of this. And why shouldn’t they be? Theory is abstract and difficult to apprehend.

So a foundational theory alone was insufficient for building such a prominent edifice of practice. There also had to be a superstructure of evidence, concrete facts that clearly attested to the efficacy of asset class allocation. Plausible evidence, which was easily apprehended by financial journalists and investment advisors, appeared in a research paper that was published in 1986.1 This is one of the most widely cited and misunderstood papers in the field of investing. Even its own authors seem to have misunderstood it, though not as egregiously as the many careless investment writers who imagined that they were echoing its conclusions to the public. But all the same, with this paper, asset class allocation at last became identified with good investment practice.

The paper’s authors looked at pension plans, not mutual funds or private portfolios, and only to the extent that those pension plans held stocks, bonds, and cash. (Back in the early 1980s, few pension plans invested to any great degree in other asset classes or in alternative investments, like hedge funds.) The authors’ conclusions, from their analysis of a database of pension plans’ changing values over a span of time, were that asset class allocation, rather than the selection of investments within the asset classes, was far and away the major source of investment returns, and that, on average, asset class allocation explained more than 90% of the variability of a pension plan’s returns.2

Let’s scrutinize that last statement. It has led to countless mistaken repetitions of the claim that asset class allocation is the source of more than 90% of a portfolio’s returns. Please note, unlike so many investment professionals, the unsubtle difference in wording: The paper did not say 90% of the returns, but 90% of the variability of returns. Yet, to cite but one example, in the week that I am writing this paragraph, I read in Barron’s that “Most investors have heard the axiom that 90% of portfolio performance can be explained by their allocation decisions.”3

Even the paper’s stated finding, concerning variability (or volatility), is unremarkable; it could have been foretold without crunching numbers. The returns of stocks, as an asset class, are much more volatile than the returns of bonds, let alone cash. If one portfolio is invested 90% in stocks and 10% in cash, and a second portfolio is 10% in stocks and 90% in cash, which one will have the more volatile returns? The first, of course. This will be true even if stocks end up at the same price from which they started during the period of our analysis, or if they go down. It will also very likely be true whether stocks, as an asset class, are represented by only a handful of names or by an index fund, or whether the managers make judicious shifts over time in response to changing economic forecasts or keep their allocations constant.

Consider, by analogy, that you’ve been informed that the ability of a group of people to walk directly from point A to point B, as measured by their cumulative wanderings off the straight course, is determined by the allocation of ages within the group: what percentage are under five years old, what percentage are between five and ten years old, what percentage between ten and fifteen, and so on. Of course it is! It’s almost entirely dependent upon the percentage consisting of peripatetic children under the age of five. Everyone else, of whatever age, will follow the instructions to walk directly from A to B. The relative proportions of the other age groups are beside the point. So, similarly, the conclusion that asset class allocation explains most of the variability of returns was both obvious and trivial.4

To appreciate why the 90% claim, if made with respect to the returns to asset class allocations, rather than the variability of returns, doesn’t amount even to a rule of thumb, you must first put out of your mind a clarification that I made in Part 1 of this essay. There, I said that when allocating among asset classes, you should define the asset classes in a way that reflects how you intend to use them in the portfolio, not as they might be defined by some standard, ready-at-hand comprehensive benchmark. For example, if you intend to fulfill the stock allocation of your portfolio with—just to be extreme and ridiculous—only oil company stocks, then you shouldn’t define “stocks” as the S&P 500 and the statistics appertaining to that benchmark; the returns and risks of oil company stocks will almost certainly be very different from the returns and risks of the S&P 500 index. You should, instead, define “stocks” by an index of oil company stocks.5 Most practitioners of asset class allocation, however, fail to heed this distinction, as did the authors of the 1986 paper. To them, all stocks are stocks, and that means that the S&P 500 index represents them, and all bonds are bonds, and some comprehensive bond index, perhaps the Barclays Capital Aggregate Bond Index, represents them.

From the point of view of these practitioners and financial journalists, the critical distinction, then, is between asset class allocation and security selection, which is the choice of individual investments from within asset classes. And their misunderstanding, because they’re under the misapprehension that the 1986 paper proved something that it didn’t, is that 90% of the returns to a portfolio are determined by asset class allocation and 10% by security selection. This, if true, would be a “Fancy that!” factoid, because—as I, too, have been at pains to explain, but for entirely different reasons—it runs counter to the traditional and longstanding view that the only thing that matters is stock selection.

Let’s undermine their mistake from a different direction: The one thing that everyone correctly knows about diversification is that it files down the rough edges of returns, leaving fewer and smaller deviations, over time, from the average return. Let’s consider three portfolios, starting on January 1, 2001: The first portfolio is 100% bonds, the second is 100% stocks, and the third is 50% bonds and 50% stocks. The one that is 100% stocks is invested entirely in an S&P 500 index fund, and so is very diversified. The one that is 100% bonds is invested entirely in an index fund that mimics the Barclays Capital Aggregate Bond index. And the one that is 50/50 invests its bond portion in the Barclays Capital Aggregate Bond index fund, but for stocks, it relies, not on an index fund, but on just three select stocks of large companies, two “hot” ones and a “blue chip”: Enron, WorldCom, and General Motors—which means that the stock portion is barely diversified. After thirteen years, as of December 31, 2012, the three portfolios’ annualized rates of return (before subtracting fees and taxes) were:

100% stock index portfolio: 2.6% rate of return

100% bond index portfolio: 5.8% rate of return

50% select stocks and 50% bond index portfolio: 0.0% rate of return.6

Obviously, asset class allocation cannot account for the differences between the third portfolio’s return and the returns of the other two portfolios. If it could, the third portfolio’s return would have fallen between the other two.

Now, you may object that I set this up as an absurd case by choosing only the stocks of companies that went bankrupt, and that the third portfolio is therefore unrealistic. My reply is that the argument holds regardless. In less extreme cases, yes, asset class allocation will matter more than it did here. But until we give up the idea that all investments in stocks, as an asset class, are at all times and in all portfolios defined by the S&P 500, and that other asset classes are also defined by the least-considered choice of index, or unless the holdings of stocks and bonds are very diversified, then security selection, not asset class allocation, will likely account for a large proportion of many portfolios’ returns.7 And not in a good way: If the markets are as efficient as I have argued that they are, security selection will almost always detract from the investment performance that could have been achieved by investing in index funds that represent the asset classes.8

I am basically reworking my argument in Part 1 of this essay, that the definition of an asset class for the purpose of constructing a portfolio should reflect what is actually being stuffed into the portfolio.

1. Gary P. Brinson, L. Randolph Hood, Gilbert L. Beebower, “Determinants of Portfolio Performance,” Financial Analysts Journal, July-August 1986, pp. 39-44.

2. “But the investment policy return [that is, long-term asset class allocation] … explained on average fully 93.6 per cent of the total variation in actual plan return; in particular plans it explained no less than 75.5 per cent and up to 98.6 per cent of total return variation.”

3. Barron’s, March 11, 2013, vol. XCIII, no. 10, p 33.

4. A thorough critique of the paper is Roger G. Ibbotson and Paul D. Kaplan, “Does Asset Allocation Policy Explain 40, 90, or 100 Percent of Performance?” Financial Analysts Journal, January-February 2000, pp. 26-33.

5. This is a gross simplification. Benchmarking is complicated. But if, in this example, we were being pragmatic, then the technology stock benchmark would probably be good enough.

6. I am assuming that the third portfolio begins with 50% stocks and 50% bonds and that there is no subsequent rebalancing. If there were, the annualized return of our 50/50 portfolio would have been less than 0%.

7. Mark Kritzman and Sébastien Page, “Asset Allocation versus Security Selection: Evidence from Global Markets,” Journal of Asset Management,Vol. 3 No. 3 (2002), pp. 202–212.

8. With the standard proviso that it is returns in relation to risk, not returns alone, that are the best measure of performance.