The availability of new style software has rekindled an argument over the whether returns-based style analysis software is a good tool for choosing money managers and designing investment plan structures.
But Nobel Laureate William F. Sharpe remains firm in his defense of the widely used returns-based style analysis technique, which identifies investment styles based on portfolio returns. Ten years after inventing the method, Mr. Sharpe said he "marvels" at returns-based analysis' usefulness.
And Mr. Sharpe argues that holdings-based analysis tool, the competitive approach, has significant weaknesses of its own.
As an alternative to strictly the returns-based approach, Frank Russell Co., Tacoma, Wash., has a new style analysis product that uses artificial intelligence to recognize investment style approaches, coupled it with a returns-based style analysis product of its own and set a new floor in software pricing. Called Russell Style Classification software, it looks at the actual securities in a portfolio holdings. It's approach is typically called holdings-based or composition-based style analysis.
To gain market share on the blossoming popularity of returns-based analysis, Russell is offering both its holdings-based analysis product and its returns-based analysis product for $7,000, considered a rock-bottom price. Russell began a marketing campaign for its product in the fall but pulled back briefly to enhance its product with new reporting and graphical capabilities suggested by test users.
Russell has some support for the value of holdings-based analysis from consultants BARRA Inc., Berkeley, Calif., and Wilshire Associates, Santa Monica, Calif., both of which offer both holdings-based and returns-based analysis products.
Wilshire offers holdings-based and returns-based analysis products as part of larger system - its executive information system software, Compass, that sells for $24,000 a year to its clients and $36,000 a year to others.
BARRA's holdings-based analysis software is part of its powerful Aegis global equity portfolio risk characterization and optimization systems and its Cosmos global bond analysis systems. Both cost several times as much as BARRA's returns-based Global Style Analyzer.
Russell has opened the door for a much larger segment of the institutional investment community with its inexpensive analysis software. Typically, holdings-based analysis has been available to the very largest institutional investors or ones with big research budgets.
However, Mr. Sharpe has identified what he says are some problems with Russell's holdings-based analysis product when it comes to providing cross-border and cross-asset class analysis.
Mr. Sharpe said in interview that he first offered, on a non-exclusive basis, returns-based analysis to Russell back in 1987 during his association, since ended, with Sharpe-Russell Research, Palo Alto, Calif. To show his new method, Mr. Sharpe made a presentation of it new method at a meeting of about 50 Russell employees, about half of whom wanted to offer returns-based analysis, he said.
But Jon A. Christopherson, senior research analyst with Russell, led the charge against offering the software, Mr. Sharpe said. Since then, Mr. Christopherson has virtually called returns-based analysis "evil incarnate," said Mr. Sharpe. To the best of his recollection, said Mr. Sharpe, George Russell, Russell president, was also against it at that 1988 Russell meeting. And Mr. Russell "had more votes than anyone else." Russell had the returns-based analysis software then, said Mr. Sharpe.
Mr. Christopherson, co-inventor of Russell's new product, said that he did find flaws in returns-based analysis, and he did call it evil. He said he has since changed his mind and finds returns-based analysis useful, but Russell's holdings-based analysis product superior.
Mr. Christopherson added that he never dreamed returns-based analysis software would become so popular with pension funds. Russell now wants to regain lost market share for analysis software, but with what he called a superior product.
The claims and counter-claims consultants are making are making about analysis software will probably alarm some pension executives.
Users of returns-based analysis include more than 100 pension funds, including those of General Motors Corp., AT&T Corp., the California Public Employees' Retirement System and the Common Fund. Some of the larger plans with big research budgets, like CalPERS, also have a version of holdings-based analysis software, but many smaller plans don't.
The claims about the failings of returns-based software aren't trivial since investment style analysis is thought to influence every aspect of fund management, from selecting money managers to monitoring them. Even though not every pension fund has returns-based analysis software, many consultants are said to be using it in advising clients. Cut-rate versions of returns-based style analysis is even becoming popular with financial planners.
"The problems with the guys selling effective mix (style analysis software) is that they have taken Bill Sharpe's methodology and some things that he said and run amuck with it," said Mr. Christopherson.
"I think everybody and his brother uses it (returns-based software) because there really hasn't been any alternative" until now, said Russell's Mr. Christopherson. Returns-based analysis uses quadratic optimization of asset class returns to match a manager's return pattern and classify an investment style or style mix.
One problem with returns-based style analysis, according to Mr. Christopherson, is that some pension executives who think they are using it to monitor their money managers for radical shifts in investment style may not be doing it effectively. Users of the software wouldn't be able to catch a portfolio's radical change in style in time to make a difference using the returns-based method, according to Mr. Christopherson. That view is supported by some other consultants.
Another worry, according to Mr. Christopherson, is that some plan portfolios might not be as carefully built for style diversification as sponsors think. A plan might be making bigger or lesser bets in market sectors than is supposed because of potential misclassification of manager styles with returns-based analysis.
Returns-based analysis sometimes can't tell the difference between style attribution and random noise, according to Mr. Christopherson. Consequently, he said, it can misclassify a manager. The plan that hires the manager might get unwanted exposure to market sector and end up taking a greater bet than desired, said Mr. Christopherson.
"Returns-based analysis is based on analysis of covariance structures or correlation coefficients among manager returns and index returns. But because the analysis of one manager's returns correlates with another investment style even for five years doesn't mean that the manager has that particular style," said Mr. Christopherson.
Under returns-based analysis, a growth manager might be mistaken for a value manager and hired as such by a plan. If the growth manager reverts to a growth style as the market for values stocks rises 25%, the plan can miss a substantial part of rise in the value market, he said.
"The real issue is expectations. The reasons that you want to know the style of a manager is because you want to know what they are going to give you in the future. And it is in expectations where effective mix (the returns-based style) breaks down," said Mr. Christopherson.
Returns-based software sometimes has trouble telling the difference among alpha, factor exposure and random noise, he said.
The financial consulting firm BARRA supports some doubts about the complete adequacy of the returns-based style analysis method.
"The difference between returns-based and holding-based style analysis is like the difference between a magnifying glass and a microscope," said Paul Green, manager of strategic marketing with BARRA.
Holding-based analysis, he said, can tell an investor months sooner if a manager made a radical change in style. Returns-based analysis is slower in recognizing a style change because it looks at a long series of returns, sometimes for four or five years.
The time difference is "very important if you are a fiduciary on the hook because it gives you additional foresight into the risk managers are taking," said Mr. Green. The time can be the difference "between learning about problems in time and learning about them after a disaster has occurred," he added.
Returns-based analysis is a "fine tool" for manager selection, said Hal Reynolds, senior vice president and partner at Wilshire. But, if a manager is not well-diversified, returns-based analysis can create "spurious results with regression-based style analysis," said Mr. Reynolds.
But other consultants think the warnings about returns-based software are overstated.
Michael Henkel, a managing director with financial consulting firm Ibbotson Associates, Chicago, agrees it is harder to pick out radical shifts in investment style using returns-based software. However, he adds, it can be done "surprisingly fast for the statistical technique" used in returns-based analysis.
"In an ideal world," said Mr. Henkel, it would best to do returns-based and holdings-based analyses of portfolios.
But for an initial analysis, returns-based software is "cheaper, easier, faster and requires less data," said Mr. Henkel. For mutual funds, the holdings data is often only available twice a year, said Mr. Henkel. And it's hard to get information out of money managers more than once a quarter, he said.
Mr. Henkel said Ibbotson's style analysis software offers better technical support and better graphics than some other makes of style analysis software.
Generally agreed to be the most popular returns-based software is the StyleAdvisor software offered by Zephyr Associates, Zephyr Cove, Nev.
"I would definitely say we find it helpful," said Scott Richardson, portfolio manager for external mangers for the $5.5 billion Illinois State Board of Investment, Chicago, about StyleAdvisor.
"I think (StyleAdvisor) does a very good job for just a returns-based thing to find out if your managers are doing what they say they are doing, to see if they are in the quadrant where they should be and to see if there is any style drift over any time period," said Mr. Richardson.
While StyleAdvisor doesn't look at the actual securities in a portfolio, the Illinois board does get information online about the actual securities from its custodian, Northern Trust, Chicago.
By looking at data from Northern Trust and StyleAdvisor, Mr. Richardson says that "you get a pretty good picture of why a manager did what they did."
Returns-based StyleAdvisor is also used for the $1.2 billion profit-sharing plan of Intel Corp., Santa Clara, Calif., Intel recently conducted a manager search using StyleAdvisor, and it uses the software to update analysis of its existing money managers, said Bob McFarland, an analyst with Intel.
"It (StyleAdvisor) is nice in that I think it can present a lot of data graphically in a way that people on the committee can understand pretty easily," said Mr. McFarland.
However, Mr. McFarland said Intel might not continue subscribing to StyleAdvisor at a cost of about $10,000 a year because it doesn't plan any searches in the near future. Intel intends to look at other kinds of software for analysis, said Mr. McFarland.
There are claims being made about both kinds of analysis software that Mr. Sharpe find exaggerated.
"These are commercial issues, and people get overwrought and overstate things," said Mr. Sharpe.
He added, "Some terrible things are said in my name. If I got a dime for each mention, I would be a rich man."
But in support of the returns-based analysis method, Mr. Sharpe said, "It's not a logical truth that one (version of the analysis software) is better than the other."
He said that he has found returns-based software to work "remarkably well." Both versions of the provide information, and Mr. Sharpe said he would use both programs if he had the data.
Mr. Sharpe said he isn't sure how he would use holdings-based software "because it is very hard to make estimates security by security in a way that integrates over all the markets."
He called attention to the potential limitations in coverage of holdings-based analysis for some style analysis models that is not true of returns-based analysis. An issue that many holdings-based analysis software supporters fail to understand is that users must have an effective model, said Mr. Sharpe.
"Ultimately, the problems is that there are these asset classes out there, and ultimately what you want to know is how sensitive the portfolio is going to be in some future period - usually two to four years - to movements in these asset classes," said Mr. Sharpe.
To answer that using the holdings-based method, users need to "do it security by security and then add them up," said Mr. Sharpe.
But, he said, most of the models of individual securities are "pretty crude," unlike the sophisticated BARRA models.
To look at the sensitivity of a whole portfolio of short bonds, long bonds, foreign stocks, growth stocks and other assets classes, a model is needed for each security.
In addition, to get a picture that transcends all of the securities using the holdings-based analysis, the user needs to "mix and match (the models) exactly," said Mr. Sharpe. He doesn't know if such models exist in the commercial market.
Supporters of holdings-based software, he said, tend to put a manager in one box. He finds that approach unnecessarily "limiting."
Returns-based users, he said, tend to allow for gradations in styles. He said those classifications aren't out of necessity, but seem to be the way they are used.
In most instances, returns-based analysis won't notify you immediately of a radical shift in investment style. But, said Mr. Sharpe, Michael Markov, a programmer of returns-based analysis products who licenses a version of his own, is promising the ability to quickly identify style shifts.
Responding to criticisms of holdings-based analysis, Mr. Christopherson said holdings information can be obtained from money managers simply by asking for it, adding that online services also provide such information.
While users of holdings-based analysis tend to put a manager in one classification, the output in terms of probability numbers is the same from either analysis product.
However, to build a portfolio of managers, a plan sponsor does have to make a decision about a manager's basic style orientation.
After building a stable of managers, a plan sponsor can look at the correlation of investment styles of his managers and can tilt, for example, his passive equity component to attain appropriate overall style weighings.
Mr. Christopherson agreed that obtaining a unified model that would describe the risk and return potential of every type of security asset would be good, but it isn't needed for Russell's style analyzer.
He said the U.S. equity version of the Russell style analyzer looks only at U.S. securities. Russell has an analyzer for Canadian securities and is building others for Japan, Australia, the United Kingdom.
While the Russell version of holdings-based style analysis looks only at equities, the BARRA version looks at bonds also.
BARRA also has a global equity money and a global bond model that could be used for global holdings-based style analysis. And BARRA is building a mathematical bridge that straps its models together across asset classes and providing it with the ability to provide holdings-based style analysis across asset classes around the world.