IN THE ANNUALS OF FINANCIAL RESEARCH, IT IS THE RARE STUDY THAT survives the test of time. "Determinations of Portfolio Performance" is widely considered one of the precious few. Making its debut in a 1986 issue of the Financial Analysts Journal, the article has remained a widely cited staple in the investing canon, courtesy of its message: asset allocation matters. To quote the paper more precisely, it matters more than securities selection or market timing.
Prodding investors to watch their asset allocations is, by current standards, something of a yawn, considering how far and wide the counsel's been dispensed, praised, and otherwise debated over the years. It's hard to imagine an informed investor who's still unfamiliar with the paper, or at least with its central conclusion. In 1986, by contrast, asset allocation was somewhat less celebrated--a shortcoming that inspired Gary Brinson, L. Randolph Hood and Gilbert L. Beebower to pen their spare but potent six pages of analysis.
That's not to say that everyone now accepts the study in full, or that the paper has escaped controversy or misinterpretation. Twenty years in the limelight has given critics ample time to raise questions. To name but a few of the objections: The paper analyzes portfolio volatility rather than returns; it doesn't factor in trading costs; and although asset allocation is important, it's not nearly as crucial as the study finds.
Yet it's also telling that even the critics usually stop short of dismissing the paper's principal argument: Asset allocation decisions warrant a primary role in crafting investment strategy. One of the study's sharpest critics, William Jahnke, nonetheless conceded in a 2004 essay in The Journal of Financial Planning that there's "little doubt that asset allocation is an important determinant of portfolio performance," and so the authors of the 1986 research "are to be commended for their work, upon which others will build."
As the paper marks its 20th anniversary, Wealth Manager called on Brinson for some reflection. Might the paper's conclusion require adjustment for the 21st century? Did he get some things wrong? We recently posed these and other questions to Brinson, who at the apex of his career in the 1990s was overseeing some $1 trillion of assets at Brinson Partners. These days, he runs a smaller private investment shop, GP Brinson, in Chicago and manages the $100 million endowment of The Brinson Foundation, the non-profit organization he set up in 2000. All of which allows him to keep a hand in putting his asset allocation notions to the test in the real world.
There's been debate about your 1986 paper over the years. Let's start by asking you to summarize the research.
The basic conclusion is that when thinking about an investment portfolio, the most important decisions are what asset classes to utilize, and what percentage, or allocation, to use for those asset classes. Those two decisions have the biggest impact on what the ultimate performance and rate of return on the portfolio will be. That's not to diminish the importance of security selection, or what specific stocks and bonds, etc. to own. Sometimes this point is misunderstood. It's just to say that in the order or magnitude of importance for influencing a portfolio's return, the first two decisions--which asset classes do I play with, and what weights do I assign to them?--will have the biggest determination on how the portfolio will perform.
Do the paper's original conclusions still stand in 2006?
I think they do, and even more so. The world has become a very interesting place in terms of different asset class categories that one can and should consider and utilize in the portfolio. We could spend a lot of time discussing what exactly is an asset class, but as I work with portfolios there are eight major asset classes:
3. Money market instruments
4. Emerging markets stocks
5. Emerging markets debt
6. Private equity/venture capital
7. Real estate
8. High yield/distressed debt
Those are fairly simple in the sense that each of the categorizations or asset classes is distinctly different from the others. You might wonder, for example, why break out emerging markets stocks from equities in general? They're just stocks. But as it turns out, stocks in emerging markets aren't highly correlated with the developed countries' stock markets. When something's highly uncorrelated with something else, it makes sense to look at it as a separate asset class.
Are there any qualifications to your study's findings?
The asset allocation and asset weighting decisions are of overwhelming importance. That said, this assumes that the underlying asset classes are reasonably diversified. Let's say you have a portfolio whose equity asset class is represented with only two stocks--Google and Microsoft, for instance. In that case, asset allocation isn't going to show up as being very important if the portfolio is made up of just a few securities.
Because Google and Microsoft have high degrees of volatility unassociated with the stock market in general. Those two stocks are apt to perform radically different than an index of stocks, and so that narrowly defined, highly undiversified portfolio will be dominated by the fact that the investor chose to have a concentrated exposure in just a few securities. In that case, the importance of asset allocation greatly diminishes. But in a portfolio that's reasonably diversified, the asset allocation tenets discussed in the paper hold with great force. As I said, the only exception would be if the portfolio is highly undiversified.
Your paper's conclusions regarding asset allocation's value came about by studying the variation of portfolio returns. Why did you take that approach?
It's a test that examines, in a rigorous statistical way, how much of the variation in portfolio returns is explained by asset allocation. It's relevant in the world of statistical analysis. Analysis of variance is a commonly accepted testing methodology for trying to explain how much something's influenced by something else. It's not a casual observation; it's not a clumsy observation. The paper used a rigorously tested statistical methodology that's been around for more than half a century in academic and scientific work.
Does the paper offer what you'd consider to be timeless conclusions?
I believe that to be true. Any investor, whether an individual or institution, is best served by thinking long and hard on those two questions: What asset classes do I utilize, and what weight do I assign to those asset classes?
What inspired the original research?
What was bothering us at the time, in the mid-1980s, was that even at a sophisticated level--the institutional level--a lot of people were focused too myopically on security selection: What stocks or bonds should I own? Meanwhile, the asset mix was almost an afterthought. In some cases, asset allocation was a residual of making decisions about individual securities. As a result, the asset mix was a byproduct of focusing on the individual securities rather than, as we're suggesting, focusing on the asset class choices first and then on the securities.
Has the investment management businesses heeded the advice in the 20 years since the paper was published?
Yes, institutionally, that has occurred. That's why institutional portfolios are much more diversified across the types of assets I mentioned. On the individual level, some of that has occurred, although individuals have been somewhat slower and less responsive to the paper's message. Individual portfolios still tend to be too country-centric. That is, most U.S. investors still think primarily about investing in U.S. markets; most Japanese think primarily in their markets, and so on. In terms of the global aspect, [asset allocation] hasn't happened with as much force at the individual level as it has with institutions.
Do you think your paper has been misinterpreted or its findings abused? For example, L. Randolph Hood, one of your co-authors, recently wrote that the widely quoted 93.6% number that the study attributes to the policy, or asset allocation decisions is, "often misquoted or taken out of context."
I think that's right. People will take anything and sometimes misinterpret or misuse it for their own purposes. And that's too bad. To some extent, individuals make claims about the article that aren't factually correct, and I suppose that's undesirable. On the other hand, anything that gets people focused on the issue of asset allocation is desirable.
Is there any particular misinterpretation or misuse that you find especially annoying?
There's one quote that I see periodically. Let's say a portfolio had a return of, say, 10 percent, and someone will say, "93.6% of that return is due to asset allocation." That's not true. That's not what the paper said; that's not the application of the paper. First of all, it didn't apply to an individual portfolio; it applied on average to the portfolios we were examining. The number is an average number, not a number that applies to every single portfolio.
So, maybe Warren Buffett's portfolio in Berkshire Hathaway isn't asset allocation driven, for example?
Does the finding in your paper naturally lead to indexing?
No, not naturally so. But it does point out where the preponderance of return is likely to come from. This was not the intent of the paper, nor did we get into it in the paper, but it clearly shows that it's difficult for active managers--people trying to outperform an index--to outperform in the aggregate, net of fees and costs associated with active management.
If you could go back to 1986, would you change anything in the paper before it was published?
I think we'd probably leave it as is. Maybe we could have inserted an additional paragraph discussing the proper way to interpret the data if we'd known what some of these misinterpretations might have been.
One could argue that the paper's observations are quite clear, and so the issue of prioritizing asset allocation should be beyond debate. Yet Wall Street seems conspicuously dismissive or uninformed of the 1986 paper and its intellectual offspring.
I'll answer by pointing out the annualized returns for the eight asset classes I mentioned, for the five years through November 2005, are as follows:
Money market: 2.1%
Emerging market stocks: 18.6%
Emerging market debt: 12.8%
Private equity: -0.6%
Real estate: 20.5%
High yield debt: 9.0%
What jumps out at me from empirical observation is that when you look across these eight basic asset classes, the distribution, or dispersion of returns, is quite large. Even to a lay person, the results are all over the place. Clearly, it would have made a big difference in how you allocated your assets across the asset class categories.
Point taken. And yet, there doesn't seem to be much focus on asset allocation when it comes to Wall Street. The number of strategists on the Street is noticeably low compared with the army of analysts studying individual companies and issuing "buy," "hold," or "sell" recommendations on stocks. To that extent, the lessons of your paper aren't being heeded.
But it's in the interest of Wall Street to have the focus it has because that's where the commissions are. Unfortunately, that's the driver, the bias, if you will.
Arguably, the so-called buy side--pension funds, for instance--are slightly more enlightened than Wall Street when it comes to asset allocation.
Yes. If you look at David Swenson's asset allocation, it's radically different from most endowment portfolios. (Swenson managed Yale University's pension fund and generated spectacular returns, in large part by deviating from the conventional asset mix favored by the institutional crowd. Brinson's point is that asset allocation matters, and so exploiting it can produce impressive results, at least in the hands of a talented manager.)
Some critics say that your original study mistakenly favors fixed asset allocation weights as opposed to a more dynamic asset allocation. Your reaction?
That's simply untrue. The study is silent on that. That said, there are so-called policy asset allocation weights. Whether for individuals or institutions, these are the weights that reflect at a moment in time the investor's circumstances and risk preferences. Over time, it's likely that circumstances and risk preferences will change, as they do with aging individuals. But the policy weights ought to be an accurate reflection of those circumstances and risk preferences. In the study, we examined the institutions' average asset allocation weights across time and took that as a proxy for what we would believe the policy weights to be. Maybe this is getting into too much detail, but there's something else called strategic weights that may or may not fit an investor's activity or desires. Strategy weights ask, given the policy you've come up with, are the markets today in a set of configurations that you believe are normal? If the markets today are in a normal state, you ought to be at the policy weights. However, if you believe some of the asset classes are in an abnormal state--such as the U.S. equity market in the late 1990s--in terms of future return or risk characteristics, then it might be sensible for someone to have an allocation to stocks that's less than normal in that moment in time.
So, the complaint that the paper advocates a fixed asset allocation is wrong?
There's also criticism that your paper's results are flawed because the study doesn't factor in trading costs.
But it does. I don't know where that comes from. The data we collected came from the institutions that we used in the study, and it was actual returns data. That pension plan data, by definition, is net of the transaction costs and trading costs. It's spelled out in the paper that we got the data from the institutions, and any knowledgeable person would know that the plan data is in effect net of any trading costs.
Switching gears, can you give us a quick overview of the Brinson Foundation?
I set the foundation up when I retired so that my family and I could do something in a societal sense, using philanthropy. We funded the foundation with what's now $100 million for grant-making activity, primarily in areas of education, health and science.
Are you managing that endowment?
Yes, although the one area that we use outside managers for is real estate, which is to say actual properties. But for the bulk of the portfolio, we manage it internally. Overall, we allocate across those eight asset classes that I mentioned. The way we think about it, given the risk profile that's suitable for the foundation, is developing a "normal" asset mix. At any instant in time, we ask, is there a good reason to deviate positively or negatively away from that normal mix because markets might be offering us risk and return opportunities at a moment in time.
Do you often deviate from the normal mix?
I wouldn't say often, but we do deviate at times. In 2000, when we started the foundation, we didn't want a large exposure to common stocks because we felt that there was an element of excessive exuberance in stock markets. As a result, we had a very meager weight toward stocks at that time.
James Picerno (email@example.com) is senior writer at Wealth Manager.