Daily Speculations The Web Site of Victor Niederhoffer & Laurel Kenner Dedicated to the scientific method, free markets, deflating ballyhoo, creating value, and laughter;  a forum for us to use our meager abilities to make the world of specinvestments a better place.

Home

Write to us at: (address is not clickable)

2/17/2005
The Inefficient Stock Market by Robert Haugen, reviewed by Philip J. McDonnell

I must confess to a certain affinity for outspoken rebels. Robert Haugen, the author of The Inefficient Stock Market, is an academic rebel who is willing to confront the established aristocracy of academic finance. For that reason and more I enjoyed reading his book and can recommend it with a few reservations.

Haugen is a Professor of Finance at the University of California, Irvine. He is the author of several books and papers and is probably most famous for his popularized work on the January effect.

Key points of the book include:

1. The book takes issue with the standard business school beta model, where beta is used as a measure of non-diversifiable risk. Empirical evidence is presented that the return to highest deciles for beta are uniformly below the returns to the lowest beta deciles. Additionally the beta based decile rankings are substantially in the reverse order which is predicted by theory. Hence the market does not reward increased non-diversifiable risk contrary to the last 35 years of business school teaching.

2. The author takes issue with the factor models of the 1990's. In particular the Fama-French (FF) (3 or 4 factor) model is disputed. In substitute he offers a 51 factor (!) model in which he shows that the variables used in the FF model are quite weak when other correlated variables are explicitly broken out. For example the book value factor used in FF becomes weak when explicit earnings variables are used because book value is essentially an accumulation of earnings over time.

3. The major model presented in the book attempts to predict the future returns of 3000 stocks and thus, is of fundamental interest to investors. The model shows a substantial 42% annual return differential between the best decile and the worst decile over a period of more than a decade.

4. The most important factors were: 1 month excess return -.72% 12 month excess return .52 Trading volume/Market cap -.20 (2) month excess return -.11 Earnings to price .26 Return on equity .13 Book to price .39 Trading volume trend -.09 6 month excess return .19 Cash flow to price .26

No risk factors such as beta or historical volatility made the top factor list. Four price change, four value and two liquidity variables made the top list. All 51 variables were included in the regression.

5. Mean reversion in the one and two month time frame is reported, which is attributed to correction of "market error". Serial correlation in the intermediate 6 month to 2 year time frame was found which was explained as corporate exploitation of competitive advantage. Mean reversion was again found in the 2 to 5 year time frame. The book explained this as the time frame for competitors to react to any temporary competitive advantage. Furthermore the assertion is made that the markets routinely overestimate the length of time for which a competitive advantage will be sustained.

6. The author considers and effectively refutes most of the usual biases and flaws in such historical studies. Included in the list are survival bias (manually addressed), look ahead bias (3 month post publication delay and use of actual data files after 1987), data mining (the 51 variable model was the first one they tried) , counterfeiting (real time test with real money by a Norwegian fund). The author pleads guilty to data snooping which is the practice of reading other people's published reports and papers and then using variables which they found successful.

7. One of the amazing findings of this model is that the best performing deciles routinely had lower risk as measured by beta and variance. The worst performing deciles had the highest risk flying in the face of accepted CAPM theory.

My reservations include the following:

1. The model uses 51 variables fitted with ordinary least squares (OLS) regression. Because of the endemic cross correlations between financial variables I am very dubious of the quality of the confidence estimates of the individual regression coefficients. Given the more than 2500 possible cross correlations it is certain that the 99% confidence levels claimed by the author are specious. However I do find the overall results quite convincing simply because the deciles nicely line up in their expected rank order.

2. The study used Compustat data which is notoriously flawed for the purposes of historical testing including issues such as retroactive adjustments etc. After 1987 the author saved the original data releases and used only data available at the time with an appropriate lag. Although I don't trust the results before 1987, they appear to be in good agreement with the results after 1987 which are more trustworthy.

3. The book was published in 1999 and the factors must be updated before being used.

I especially enjoyed the author's irreverent tone, which is quite unbecoming one who professes in finance. He refers to the established big names in finance as the College of Cardinals and defenders of the Holy Temple of the Random Walk. His model is part of what he has termed The New Finance, which he admits is ad hoc and has little or no theory to support it, but does offer much better predictions with respect to future returns. His style makes for an easy read and motivates one to go out and seek his Super Stocks and avoid the Stupid Stocks.