By Bill Alpert | 13 July 2008
How Stock Analysis Based On Statistics And Computers Sabotaged Itself
|
Wall Street's quants have been forced back to the drawing board— Again. In 2007, many of these computer-skilled fund managers underperformed their liberal arts peers for the first time in almost a decade. Returns averaged 6% last year for the traditional long-short equity funds measured by Lipper, while quantitatively-managed counterparts gained less than 1%. The reversals for many quants were worst in a few days last summer that left the rest of the market relatively unscathed.
The quant crowd's lackluster season followed a triumphant era that began with the burst of the tech bubble in 2000. Applying statistics and computers to the analysis of stock prices and company financials, quants enjoyed seven years of steady outperformance [[Gee! Just like the Egyptians under Joseph.: normxxx]]. Money flowed in. By some measures, the assets under management of quants grew at twice the rate of traditionally-invested money. If you count "enhanced" index funds and brokers' proprietary trading desks, you can conclude that quants run upwards of a trillion dollars.
But hand-wringing was the order of business at a June 17 assembly at a Cipriani restaurant in Manhattan, hosted by Joseph Mezrich, who runs quantitative research for Nomura Securities International. Mezrich asked the assembled eggheads if their Golden Era was over. Most speakers thought not, but worried aloud that too many quants were using the same data and mathematical models to pile in and out of the same investments [[Gee. Whoever would have suspected? Do you think that the fact that they had all come from the same B schools, and read the same books, and worshiped the same gods had anything to do with it?: normxxx]]."So many people were looking at the same signals," said David M. Modest, an experienced quant who most recently ran a proprietary portfolio for JPMorgan Chase.
Quants often form portfolios based on "factors," which are generic characteristics of a stock, like: price-to-earnings, improving company profits or stock price momentum. If research shows that stocks drop when company earnings come from bookkeeping accruals instead of real cashflow, a quant might buy stocks with the lowest accruals and short those with the highest. Any broad phenomenon that seems to move stocks can become a factor. But in recent years, many quants favored value factors like price-to-book, in the spirit of a 1993 study by University of Chicago professor Eugene Fama and Dartmouth professor Kenneth French that helped establish the approach.
Value investing was a fruitful style after the excesses of the dot-com mania. From 2000 to 2007, the market awarded the longest winning streak in a quarter century to stocks that were cheap relative to their book value. But the streak ended last year, when the subprime credit crisis began.
In a credit crisis, investors shun stocks of companies they think might default, even if those stocks are already cheap. And value portfolios tend to include more than their share of companies considered likely to default (basing that likelihood on their balance sheets). In a chart (reproduced at right, below), Mezrich shows that cheap stocks have only gotten cheaper in today's subprime crisis and in the 2002 scandals that sank Enron and Tyco.
The link between valuation and default risk has clobbered quants who favor value factors. In the second half of last year, growth stocks did paradoxically well even though the overall market declined. This year, momentum stocks, with uptrending prices, have been the leaders. At the Nomura meeting, and in a timely new survey sponsored by the Research Foundation of the CFA Institute, the U.S. market's style rotation from value to momentum got much of the blame for quants' mediocre performance of late.But quants generally blame themselves for the abruptly-bad days that many suffered last summer. Markets formerly unlinked have become correlated, thanks to financial engineering— as the upper lefthand chart shows. Since around 2001, S&P 500 index options have moved in step with the spread between corporate bonds and Treasuries. One after another, asset movements have become linked: commodities, emerging-market stocks, credit-default swaps...even wheat futures.
Newly-invented derivatives and computerized hedging strategies have increased the "connectedness" of the financial system in the U.S., and indeed, the world. So when an investor suddenly sells in one market— last summer's jolt was widely attributed to Goldman Sachs' quantitative Global Alpha Fund— the selling can spread to other markets as quants respond in unanticipated ways. "We've made the financial markets incredibly complicated," Modest said [[immodestly: normxxx]].
The heightened risk from systemic repercussions has been widely debated among quants since hedge-fund veteran Richard Bookstaber made it the theme of his influential 2007 book "A Demon of Our Own Design." Bookstaber's ideas have gotten high props from Mezrich and Modest.
Last summer's sandbagging convinced many quants that their discipline has neglected its study of risk. The typical quant analyst works diligently to model stock returns, said Tony H. Elavia, an investment fund chief at New York Life, to the Nomura audience. But he then outsources the modeling of risk, which is Wall Street's term for stuff that can cause a portfolio's performance to fluctuate too widely. Risk considerations can include the covariance of a portfolio's stocks with one another [[ie, the degree to which stocks in a portfolio are NOT independnt as assumed: normxxx]] or the portfolio's unintended exposure to macroeconomic factors [[but in any case is far less susceptible to 'pure' mathematical analysis and modeling, especially at the extremes— where genuinely 'new' conditions manifest themselves: normxxx]].
Most quants subscribe to the Barra risk modeling service of Morgan Stanley spinoff MSCI (ticker: MXB). And the Barra equity risk model has been virtually unchanged for decades, say many quants, apart from its ongoing recalculation of covariances. With so much money riding on the same risk model, quants found themselves all exposed to the same stocks last summer. In the rush for the exits, the Barra models proved poor guides.Another cause for flagellation was the excessive use of leverage by quants at hedge funds and proprietary trading desks. When a quant has found a factor that produces modest investment returns, she's often tempted to magnify those returns by using short-sale proceeds and margin loans to lever up the bet ... sometimes five— to ten-fold. Leverage came in for harsh criticism among the several dozen money managers surveyed by the CFA Institute study authors, Yale finance professor Frank J. Fabozzi and researchers from the Intertek Group in Paris, Sergio Focardi and Caroline Jonas. "Everyone is greedy," said one of the study's respondents, "and they have leveraged their strategies up to the eyeballs."
Of course, stock picking has been hard for everyone in the last year, quants and traditionalists alike. In an unsettling illustration (shown above), Mezrich shows that since 2006 fewer stocks have been outperforming the market (delivering "alpha," in Street parlance), and stock performance has become more dispersed (increasing the penalty for picking the wrong stocks) [[and making the 'right' stocks less easily identified through 'traditional' statistical measures: normxxx]].
Mezrich's research tracks the performance of 55 popular equity-ranking factors, including measures of value, growth, earnings quality, profitability, analysts' estimates and stock momentum. Unhappily, the predictive power of these factors has eroded lately— approaching the lowest level in the past 20 years, except for the tech-bubble lunacy.
Twelve months of mediocre performance isn't enough data to disprove the quants' skills (or those of any other investors). Veterans like David Modest believe that quants need merely to roll up their sleeves and find return-generating factors that haven't been widely exploited [[very interesting; but the more arcane such measures that are found, the shorter their effective lifespan is likely to be and the more likely they are to prove spurious (i.e., holding true only for the data set analyzed, but failing when that dataset is enlarged with new data); ah the woes of Wall Street— and correlational analysis!: normxxx]].
And with the extreme performance of momentum stocks lately, even the tried-and-true value factors may be due for a revival. The spread between value and momentum is now more extreme than in any period in Mezrich's 35 years of data. Quants (and many other smart investors) like to bet that extreme values will revert to historical means. A Mezrich chart (at right) shows the depth of returns to a portfolio of stocks with a low price-to-book and a low price momentum (as measured by trailing year's return minus the latest month's). A turn may be coming [[or the results may be telling us something— like the 'traditional' definitions of value may be failing: normxxx]].Wall Street's science-class dropouts have found something not quite displeasing in the misfortunes of their quant friends. Quants themselves see plenty of improvements they'd like to make, but they don't think their scientific paradigm has crumbled.
"Is the jig up?" No, says Mezrich. "Is it harder?" Yes.
— — — — — — — — — — — — — — — — — — — —
E-Mail Comments To Mail@Barrons.Com
ߧ
Normxxx
______________
The contents of any third-party letters/reports above do not necessarily reflect the opinions or viewpoint of normxxx. They are provided for informational/educational purposes only.
The content of any message or post by normxxx anywhere on this site is not to be construed as constituting market or investment advice. Such is intended for educational purposes only. Individuals should always consult with their own advisors for specific investment advice.
No comments:
Post a Comment