May 16, 2006
STYLE ANALYSIS, AND ONE MAN'S QUEST TO BUILD BETTER INDICES
Ron Surz of PPCA Inc. says he's built a better mousetrap of indices used for benchmarking and analyzing money managers. In fact, he's just following the advice of a Nobel Laureate.
Professor William Sharpe, who shared a Nobel Prize in 1990 for his work on developing the Capital Asset Pricing Model, laid out the foundation for returns-based style analysis in a 1992 paper. The strategy is one of analyzing, say, a mutual fund's returns by regressing performance against various indexes to determine what's driving the performance. In essence, it's a quick and fairly reliable way of x-raying a portfolio to see what's going on behind the scenes.
For example, by running a returns-based style analysis on an actively managed large-cap U.S. stock fund, one might learn that the benchmark-beating results came primarily from loading up in small-cap companies. There's nothing wrong with that, of course. But if you're telling the world that your benchmark is large-cap stocks, you'll have some confused (or angry) investors if they learn that you're buying small-cap equities. Perhaps, then, a small-cap index is the better benchmark for the fund, in which case maybe the large-cap manager's performance doesn't look so impressive after all.
So it goes in the world of analyzing managers. Trying to make apples-to-apples comparisons is the bane of analysts who are forever attempting to separate alpha's wheat from beta's chaff. No easy task in the best of times. Talent, after all, isn't easily defined, and less-than-obvious when looking only at numbers. Indeed, surveying past performance offers no foolproof way for deciding that manager A has the right stuff and manager B is a pretender to the throne. That's not to say that reviewing the past is worthless. But there are limits to studying history as a short-cut to seeing the future.
That said, some argue that an effective search for talent requires going through a portfolio's holdings with a fine tooth comb, and comparing changes over time. But a holdings-based analysis isn't practical because managers' reports are infrequent, and quite often out of date. In the case of some hedge funds, you might not ever learn what it's in the portfolio. So much for timely analysis.
No wonder, then, that returns-based style analysis has proven so popular. Manager participation not required.
Indeed, since Sharpe's paper was published in the Journal of Portfolio Management in 1992, the world has embraced his technique. The reason: it provides a degree of transparency for manager evaluations that's otherwise impractical. What's more, returns-based style analysis is relatively easy to run, especially when it comes to mutual funds. With index returns in hand, you can analyze any fund simply by comparing its returns with various benchmarks over time. And since mutual fund returns are updated daily, the timeliness factor is satisfied. Some are applying returns-based analysis to hedge funds as well.
Predictably, a variety of software products catering to returns-based analysis are available, including Zephyr StyleAdvisor and Ibbotson Associates' EnCorr Attribution, to name but a few. Sharpe's theory, in short, has long been transformed into practical application.
But style analysis isn't quite what it could be, charges Surz, president of PPCA, an investment consultancy that designs its own suite of portfolio evaluation tools and indices. The critical issue is choosing the right indices, says Surz, who holds an MBA in Finance and an MS in Applied Mathematics. In other words, satisfying Sharpe's call for mutually exclusive and exhaustive indices is critical, and not every index shines by this standard. That's one reason why he came up with his own benchmarks, which are available gratis at PPCA.
By comparison, the usual choices from the big boys leave something to be desired, Surz opines. "If you read Bill's article, you'll see that he lays out clearly how the so-called style palette ought to be constructed. The operative rules [for the indices] are mutually exclusive and exhaustive."
Surz goes on to explain that there's a "strong statistical reason" for that. "If they're not mutually exclusive, it creates a problem called multi-colinearity." In other words, even though you're using two indices measure value and growth stocks, there's some overlap in what they're measuring and so some stocks show up in both indices. In which case, the information that returns-based style analysis is telling you may be of a lesser quality, if not entirely misleading, than it otherwise could be with better-designed indices.
One example: if IBM is labeled (or mislabeled) as a growth stock and a value stock at the same time, it shows up in a value and a growth index. That's a problem, and it does in fact happen, Surz warns. The solution: don't use indices that double count. But if you're using the growth and value indices for the Russell 3000 indices, Surz says, there's overlap of roughly one-third, measured by number of securities as well as by dollar value. "So a third of the companies are in both value and growth indices. That means that you have the statistical problem of mutli-colinearity if you use the Russell indexes in your style analysis."
As a result, you'll get "erroneous style profiles because those indices don't meet the criteria set out by the guy who developed the idea," he continues. "That's like writing down instructions to build the world's best car, and then you decide you want to substitute a different carburetor. The car's not going to run right. In fact, it's not going to run the way it was designed to run."
The good news is that the major index vendors are aware of the problem. Indeed, there have been some efforts of late to fix the glitch. The bad news: progress is spotty, and in some cases still nonexistent.
The new style indices from Standard & Poor's solve part of the problem, Surz admits. The new style indices replace the old S&P/Barra style benchmarks, which died a quiet death last December. But progress isn't perfect. In particular, the mutually exclusive problem is resolved, but the solution comes at the expense of introducing other challenges. "If you were to just use the Pure Style indices, you'd be well served in the mutually exclusive area," he explains. "But you won't meet the exhaustive criteria because the Pure indices throw out the stuff in the middle."
The "stuff in the middle" is purposely cut out of the new style indices from S&P. And since S&P doesn't offer a core index, this slice of the equity market is lost as it relates to the new style indices.
"The stuff in the middle is important," counsels Surz. "In my opinion, it's like taking the Oreo cookie and throwing out the middle." In effect, he asserts that the core should be treated as a distinct asset class, and made available for returns-based style analysis.
The reason is that value and growth stocks have a history of low correlation--i.e., one is usually in favor while the other languishes. The Russell 3000 Value, for instance, posts a 7.4% annualized return for the five years through yesterday. The Russell 3000 Growth, on the other hand, suffers a small annualized loss over those five years: -0.3% a year.
But while value and growth stocks are often on opposite sides of the performance spectrum, core isn't always in the middle. At least not all the time, Surz warns. Core, he says, often behaves differently than either value or growth, and so ignoring core can introduce a number of risks when analyzing returns. For example, he points out that in 2005's second quarter, core stocks fell by 0.2%, according to his calculations. Meanwhile, value and growth stocks each were up by more than 2% during that span. Overall, about one-third of the time core stocks go their own way, and so ignoring them in returns-based style analysis could deliver misleading results.
But there's a better way to analyze returns, and Surz says he has the answer: build better indices. He's done just that. His large-, mid- and small-cap indices are at once mutually exclusive when it comes to the value and growth benchmarks, and exhaustive by including all stocks by way of a core indices. "I don't throw out any companies," he boasts.
Surz offers the underlying data for both domestic and international indices, and at no charge. You can download the numbers at his site.
To be sure, Russell, S&P and the other major index vendors aren't exactly worried. Surz is a tiny fish in a big pond when it comes to the equity index business. Nonetheless, there are alternatives to the big boys, and arguably superior alternatives.
But when it comes to the old saw that the world will beat a path to your door if you build a better mousetrap, well, that's still an open question.
Posted by jp at May 16, 2006 9:57 AM
I'd be pleased to provide monthly returns on request. Also, monthly Surz style returns are already available on Zephyr, MPI, Ibbotson, Pertrac, Frontier, Sungard, Mobius, PSN, Investment Technologies, Open Finance Network, and more
Posted by: Ron Surz at May 17, 2006 9:21 AM
As a quick intro to the concept our hero does a fine job. If he'd provided monthly returns I might even be tempted to have a look. As it is, I simply don't see the point. Russell, S&P and their friends are imperfect, but they all provide much finer-grained data and their indexes are good enough for government work.
Besides, there's a long way from "could deliver" to quantification of an effect. Regression of return streams against benchmarks is inexact already. Constrain regressions, as most do, and you lose more analytical power. Throw away most of your data by using quarterlies and ignoring everything but coefficients and you're in a horseshoes-and-handgrenades ballpark where the effects of multicollinearity or core exclusion are awfully unlikely to be felt.
No criticism of our host for presenting this, though. It's always nice to see to what other folks are up.
Posted by: wcw at May 16, 2006 3:39 PM
Nicely written article.
Posted by: Scott Osborne at May 16, 2006 1:57 PM