In a series of three articles published in Investment Advisor (September ’05, January ’06, February ’06), Craig Callahan and C. Thomas Howard levy several criticisms of the so-called investment style box. According to the authors, use of the style-box concept hurts investment performance, without yielding appreciable benefits to advisors. Thus, state the authors, style boxes are “unnecessary.”
As you might expect, Morningstar disagrees.
The problem, from our view, lies with the authors’ two key assumptions–advisors are reliably foolish and money managers are reliably correct. Each of these assumptions is required to sustain the authors’ arguments. However, it is unlikely that either supposition is correct.
Let’s begin with the first assumption: That advisors are reliably foolish. The authors portray a world where advisors are intensely focused on investment-style purity, such that they will hound portfolio managers who break the rules.
Who are these advisors? Certainly, they are not the ones who pay any attention to Morningstar, which has stated loudly and publicly for 15 years that style boxes are descriptions, not prescriptions. Neither are they the advisors who have poured billions of dollars into American Funds’ U.S. equity funds, Dodge & Cox Stock, Legg Mason Value, and Davis New York Venture–all style-inconsistent funds by Morningstar’s measures. Surely, if the penalties to portfolio managers are so great for violating the style box precepts, those funds should not have been among the biggest successes in the advisor marketplace over the past decade.
We deny that the typical use of style boxes by advisors is anything like what the authors portray. Rather, we assert that the large majority of readers use the boxes as they were intended: to sort funds into categories; as a tool to understand how a fund’s approach may change over time; and as a gauge for understanding how an investor’s portfolio is spread among the different types of stocks.
The second assumption, that money managers are reliably correct, follows from the authors’ insistence on managerial freedom. It is in money managers’ interest to argue for greater freedom, because that enables them to argue for greater management fees. But in aggregate, there is no logic for accepting the argument, and since advisors are not foolish, they don’t. Instead, they pick their spots, giving certain managers certain amounts of flexibility, depending upon the role that the manager plays in the client portfolio and the credibility of the manager’s claim for adding future alpha.
Let’s turn to the authors’ specific arguments. According to them, style boxes:
- Damage portfolio manager performances
- Do not qualify as an asset class
- Fail as a risk management tool
The evidence for the first claim comes from a laboratory experiment. The authors apply four famed investment formulae to the S&P 1500, tracking for each formula: 1) the formula’s ranking of each stock, 2) into which of the 9 style-box grids each stock falls, and 3) the subsequent performance of each stock.
The authors conclude that because the higher-ranked stocks by each of the four formulae outgain the lower-ranked stocks, and because the higher-ranked stocks do not fall into a single style-box grid, that using style boxes to constrain funds hampers returns. They then run a regression that determines the difference between the strategies’ theoretical returns and those obtained from creating portfolios that fall into a single style box, to be more than 300 basis points per year.
This is no doubt true, however, the applicability of this study in the living, breathing world of financial management is very much open to question. Among the unstated assumptions are:
- That the results from these four strategies can be generalized;
- That managers who feel constrained by investment style boxes will place 100% of their assets into a single style grid;
- That managers who do not feel constrained by investment style boxes will invest strictly in order of their stock rankings.
In reality, none of the three assumptions is likely to be true.
Nothing so drastic as investing 100% in a single style grid is required for maintaining “style purity.” Even if a portfolio manager were morbidly concerned about adhering to Morningstar’s style definitions, the manager would have much more flexibility than is implied by this study. That is because the Morningstar style designations are averages of a portfolio’s positions, with no minimum requirement that stocks be in any particular grid. Strictly speaking, it is possible, although quite unlikely, that a fund could occupy a style grid without having a single stock of that style type.
A picture will save further words. The diagram on this page shows American Funds’ Investment Company of America Fund. At the time of this analysis, the fund was designated as Large Value. The area covered by the green oval represents 70% of its portfolio assets, and covers nearly all of the three Large Cap grids. Several stocks that fall into the remaining 30% are Mid Cap. Overall, the fund has exposure to 6 of the 9 possible style grids, and could easily retain its Large Value status even if it owned a modicum of exposure to the final 3 small-cap grids.
Finally, every manager with whom we have ever spoken has considered overall portfolio issues when implementing a strategy. So, regarding a strategy’s top-ranked selections as the true portfolio that would exist if the managers disregarded the style policy is not realistic.
The authors’ second claim is that style boxes are not asset classes. According to them, asset classes should be obviously and readily definable entities with widely agreed-upon definitions. However, they point out, style-box grids are defined differently by different organizations. The authors also argue that asset classes should have low correlations with each other. Finally, the authors state that asset classes should have a stable membership; that is, securities should not periodically change their asset classes.
The first and third points, those of the ease of definition and stable membership, reflect the authors’ preferences, no more. Asset classes exist so as to sift a large number of securities down to a more manageable number. As long as the securities within a given asset class behave in broadly similar fashion, and the asset class has a role when building a portfolio, the terms of existence are satisfied. The authors’ further criteria are matters of taste.
The second point, that of low correlation, is more substantial. The authors are correct that asset classes should not behave similarly. The whole point of asset classes is to simplify portfolio clutter.
The problem is, asset-class behaviors change over time, enough to confound simple correlation rules. For example, in the first half of the 1990s, the Lehman Brothers Aggregate Bond Index had a 0.60 correlation with the S&P 500. In contrast, in the first half of the current decade, the Morningstar Large Value index had but a 0.39 correlation with the Small Growth index. So, what’s that about bonds and stocks are asset classes, but style grids are not?
Consider this dramatic situation. It is January 1999 and you conclude, based on a correlation study, that style definitions are unimportant. You will no longer communicate to your clients using that language. Then come the following three years of performance (see “Three Years of Performance” table, below).
Uh-oh. In the most dramatic marketplace in your clients’ memories, the single most critical element in their investment returns was how much money they had in Large Growth style stocks, as opposed to Small Value. And you didn’t clue them in.
The authors’ final contention is that the style boxes fail as a risk-management tool. This claim is backed by a study that shows only modest risk reduction between three portfolios. These portfolios were formed by: a) Investing 25% of assets into one fund from each of the four corners of the style box; b) Investing 25% of assets into each of four funds that are randomly selected from the four corners of the style box, such that the funds might be from four different grids, all from a single grid, or any combination in between; and c) Investing 100% of the portfolio into four funds that share the same style grid.
The authors then show that portfolios formed in these fashions would have had average annualized standard deviations of 18.9%, 19.6%, and 21.3%, respectively, from the years 1996–2002. They interpret this finding as suggesting that using style boxes to control risk in a portfolio yields only very modest benefits.
Once again, we must demur. By the terms of the authors’ study, portfolios of even low-correlated assets would also fail the test. Care to see? Instead of the four corners of the U.S. stock investment style grid, let’s take four types of assets that everybody would agree are diverse: a) REIT funds, b) Japan Stock funds, c) Europe stock funds, and d) Natural Resource funds. Surely, according to authors’ methodology, diversification would provide a benefit.
Not quite. We ran the numbers, using the same methodology–the same time period, the same universe of existing mutual funds, the same averaging calculations–but with the assets we specified.
There was good news for the authors: Their methodology passed the common-sense test in indicating that investors who placed every penny into a single asset were indeed at higher risk than investors who followed the other two strategies. The bad news: The methodology also indicated that it made no difference if investors diligently placed 25% of assets into each fund type, or bought four funds at random (see table below).
That’s worrisome, to say the least. The very logic that the authors offered as proof that the style boxes do not help in portfolio diversification, indicates that if your client is to invest in Japan, Europe, REITs, and natural resources, he or she needn’t worry about allocation. Just buy four funds at random, regardless of type, and your client will achieve the same results as that drudge across the street who placed 25% in each type of fund.
We don’t believe it, and neither should you.
Despite the disputative tenor of this article, we believe that we are ultimately in agreement with the authors’ basic tenets, if not their specific methodology. We both regard so-called “overlay” management–the process by which advisors monitor and control their clients’ risk exposures–as an advisor best practice. Neither of us would recommend using style boxes as rigid, mechanical constraints for hiring and firing managers. And to gauge by the authors’ rather ironic recommendation that advisors use targeted ETFs such as indexed “value” funds as portfolio tools, the authors also share our belief that style definitions and style-box defined investments are useful elements of the portfolio process.
So, we urge peace. The style boxes are undeniably useful instruments for categorizing funds, for understanding a portfolio’s positioning, and for communicating with clients. There is no reason to expect more from them. Ultimately, there is no reason for this debate. It does not advance the cause of investment analysis.
John Rekenthaler, CFA, is vice president of research and new products at Morningstar, where he oversees the company’s research methodologies, and is involved in a variety of new product development efforts. He can be reached at email@example.com.
To view the original three articles in our series on style boxes, as well as a number of online exclusives go to “Web Extras” at www.investmentadvisor.com.