[Empirical Finance Research Blog] 11/19/2009 02:26:00 PM

[Wesley R. Gray]

Investment Potential Rating: ?/10 (1 worst, 10 best)

============

Information Revelation and Expected Stock Returns

Umut Gokcen
Boston College

working paper http://www2.bc.edu/~gokcen/Umut_Gokcen_Homepage/Research_files/umut-pubinfreturns.pdf

Abstract:
If information is not perfect, theories prescribe a negative relation between information availability and expected stock returns. Using two readily available variables, price and volume, I construct a new proxy for information and test its relation to returns in the 1964-2007 period on NYSE-listed stocks. I find that information revelation predicts lower future returns, controlling for beta, size, book-to-market ratio, liquidity, and momentum. A long/short trading strategy based on sorts on the information proxy generates alphas of 3% to 4%. These alphas do not have to imply an arbitrage opportunity; they are consistent with time-varying expected returns in a rational model. 

Data Source:

The author uses price and volume for NYSE-listed stocks over the 1964-2007 period. He also uses various factors (Fama-French, momentum, liquidity, etc) from WRDS (many of these can be found at Ken French's site http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html.

Data Specification:

The semi-strong form efficient market hypothesis suggests that all public information is embedded into asset prices quickly and efficiently. At first glance, this hypothesis makes way too much sense; however, deeper inspection suggests there may be more involved in market efficiency than meets the eye. Grossman and Stiglitz (1980) bring up a simple, yet compelling argument: If market prices are efficient and information is magically incorporated into stock prices, why would anyone have incentive to do research? If I'm an active fund manager and I have to pay my analysts $100,000 a year, but they can't produce any value, what's the point of hiring them? I'll just have my trading algorithm buy me the value-weight stock market index. But if everyone has their trading algorithm buying the value-weight stock market index, who the heck is actually setting prices at the margin ensuring prices are actually reflecting fundamentals and information?

The Grossman and Stiglitz paradox has a basic point: the market can't be 100% efficient, because nobody would rationally engage in the costly price discovery efforts that are required to actually produce a 100% efficient market.

In the end, one of the basic predictions of the Grossman and Stiglitz theory is that when information is cheap, accessible, and easy to understand, prices will be very close to 100% efficient. Whereas, when information is expensive, has restricted access, and is complicated and time consuming to decipher, prices will likely stray from their "fundamental value."

This paper seeks to test a simple hypothesis related to the Grossman and Stiglisz theory, namely, expected returns should be decreasing in the amount of public information. Specifically, the author tests a strategy of short selling stocks that have recently "revealed" information and going long those stocks where information is private and/or has not caught the market's attention.

Investment Strategy:

The key aspect of this paper is empirically identifying "information revelation." The author proposes the use of the daily correlation between absolute returns and dollar volume estimated over a month. The basic idea is that when info is revealed, prices AND volume surge simultaneously, so months where the daily correlation between price and volume is high imply high information revelation, and months where daily correlation is low suggest low information revelation. The filters out market-wide movements in price and volume by using the residuals from a set of regressions and comes up with a information proxy, RHO. (see page 6 for details).

Here is the investment strategy with the highest alpha:

1. Each month do a double-sort on companies using size and RHO. 

2. Long the small-high RHO, short the small-low RHO

3. Hold for a month, rebalance.

This portfolio earns ~7% alpha a year (56bp a month) after controlling for market, size, b/m, and liquidity risk factors (see table 5). Moreover, the long only strategy for a small-high RHO portfolio earns about 1.35% alpha a month.

Implementation Issues and Remarks:
This paper finds that high information revelation stocks outperform low information revelation stocks. Not surprisingly, the results are especially powerful in smaller stocks where presumably information is disseminated less efficiently. Overall, this is an interesting quantitative factor to consider in an overall investment portfolio.

My guess is there is a quite a bit of data-mining going on in this paper, so it's important that anyone trying to implement this in the future conduct robustness tests on the recent 2004-2009 period (if any of our readers do this, please post to the comments). It would be interesting to see how this performs out-of-sample.

Investment Potential Rating: 6/10

--
Posted By Wesley R. Gray to Empirical Finance Research Blog at 11/19/2009 02:26:00 PM