Source : http://www.1440wallstreet.com/index.php/site/comments/rear_view_mirror_quantifiably_insane_week/ The quant implosion was the big story for us this week. Macro heads can talk endlessly about the Fed, we want to talk about what might have gone on in quantland. Michael Santoli has a solid piece in his latest Streetwise column in Barrons that provides an entree to the understanding the debacle. Statistical factor-based quant models--which weight dozens of valuation, growth and momentum variables to fashion long/short portfolios--became a crowded trade. They went haywire n recent weeks as volatility surged, leverage was cut back, heavily shorted stocks went up and statistically cheaper shares cracked. Barrons The article goes on to explain more...as leverage built at certain hedge fund complexes and position sizes increased. Certain powerful value factors such as price to cash flow became negatives as the leveraged buyout escape hatch was shut by colicky credit markets. And these strategies implicitly bet against inter-asset correlation; correlation always goes up in difficult markets. Nice work Michael, but we want to take it from there. I would like to propose a theory that goes beyond a simple unwinding of a crowded trade. In the early 90âs I worked on what was one of the first quant strategies on the Street. It ran money for large Fortune 500 pension funds and was, in part, developed with their assistance. Cutting edge stuff in the late 80âs, and still working well without many revisions or tweaks by 1993. Although the system seems archaic by todayâs standards, I will briefly explain it. It was a four factor model consisting of the following: Earnings growth versus the market rate of growth Thus companyâs earnings growth was measured against the market, and was either higher (+), in line within a band (0) or slower than market rate (-). Stock Price Momentum Either (+) or (-), and measured as last 3 months/12 months for stock divided by 3/12 for market. The stock had to be outpeforming to garner a (+) Earnings Acceleration or deceleration Relatively straightforward, resulting in either (+) of (-). The forward earnings estimate and prior three numbers were used to get the score. Historical Relative Valuation This was relatively straightforward as well, and essentially looked at the historical P/E range for a stock versus the market. If a stock had traded between a 10 and 20 P/E, and was closer to a 10 it would show favorably. This could score (+) (0) (-) So you would get a four factor score for each stock. Perhaps (0) (+) (-) (0)