Quants at SSgA unit trying a new spin
State Street Global Advisors Inc. is developing new quantitative models that don't go strictly by the numbers.
With techniques that incorporate Wall Street analysts’ recommendations and assessments of those analysts’ expertise, Anthony Foley, the new director of the Advanced Research Center at Boston-based SSgA, says the models he is developing represent “a new way of thinking about quantitative investing.”
Mr. Foley calls his approach “post-Bayesian” modeling, saying it goes beyond Bayesian techniques that incorporate subjective factors with objective numerical evidence to calculate probabilities.
Mr. Foley joined Boston-based SSgA in June after being the director of research at Pareto Partners in New York.
Right now, Mr. Foley is developing three new models – high-yield fixed income, regime switching and industry specific.
The high-yield model, at about two-thirds complete, is the most developed of the three.
Among its distinguishing features is a pattern recognition system, which is designed to identify situations such as when a market is overbought or oversold.
The industry-specific models will focus on European industries at first, then expand into other areas. Mr. Foley says the models will be crafted on a global and regional basis and will include industry-specific overlays, which the old industry models did not have.
The regime-switching model, designed to adjust the weightings of value stocks in a portfolio depending on the market, also is under construction, says Mr. Foley.
Regime switching and similar techniques, says Jay Tubianoso, assistant vice president of research at consultant Shields Associates in Stamford, Conn., are becoming more common among quantitative managers as they look to adapt to the changing markets.
“There’s definitely a new impetus to re-evaluate the systems on a timely basis,” he says. “The products that are able to adapt without going haywire can do very well and consistently outperform.”
On another front, Mr. Foley is looking to develop a process that weights recommendations from Wall Street market analysts. These weightings then can be incorporated into other SSgA models. The models will calculate how much weight to give a particular analyst, based on the views of two senior SSgA portfolio managers on each analyst’s expertise in a particular asset class.
“How different is that from those same two managers sitting down with paper and going through the analysts and saying, `I think we should pay more attention to this guy or this lady’?” says Mr. Foley.
The approach used in the SSgA models is rare, although other firms, such as Numeric Investors of Cambridge, Mass., employ similar practices, says Mr. Tubianoso.
Mr. Foley also hopes to build models that can gauge investors’ moods.
“The hardest thing to quantify is sentiment, because a lot of it deals with things you can’t see.”
However, he hopes to be able to estimate behavior by visible variables.
In some of his models, Mr. Foley hopes to use the Markov Chain Monte Carlo method to help gauge future activity and behavior. The process simulates a variety of potential scenarios – for example, how the stock market might perform over time – to reach an average.
“If you do that in a real careful way, the results you get will be a good approximation to the real model,” he says.
The Markov Chain method is currently in the testing phase and won’t be implemented until next year.
Neither Mr. Tubianoso nor Russell Kinnel, an analyst with Morningstar Inc. in Chicago, was aware of other firms that employed the Markov Chain technique in their models.
Mr. Foley’s non-traditional quantitative techniques come from looking outside the financial box.
“Finance is the area that’s contributed the least in terms of techniques we’re applying right now,” he says. He uses techniques stemming from fields such as medicine, engineering and physics.
Learn more about reprints and licensing for this article.
