A new paper by four U.S. scholars makes a contribution to the literature on factors and the modeling of stock prices.
The paper, “An Information Factor,” proposes in essence that the momentum factor isn’t what it seems to be. Ever since the publication of a 1993 paper by Jegadeesh and Titman this factor has generally been understood as the predictive power for past returns for future returns. This has always also seemed odd, because even a long streak of flipping a coin and getting “heads” does not predict that the next flip will be a head, nor (sorry contrarians) does it predict that the next flip will be a tails. So: why is there a “momentum factor” for stock prices?
What Matthew Ma, of Southern Methodist University, and his associates propose is that the momentum factor arises due to the predictive power of something else, the recent trades of informed market participants. Further, predictive power is increased if we zero in on this issue of the informed trades.
Along with Ma, the authors are: Xiumin Martin, Matthew C. Ringgenberg, and Guofu Zhou. Martin and Zhou are both affiliated with Washington University in St. Louis: Ringgenberg with the University of Utah.
They create their new factor (or their modified version of the momentum factor) this way: it is the return on a long/short portfolio created on the basis of purchases by insiders and sales by short sellers and options traders. They call this return the information factor, or INFO.
The INFO factor works, backtesting shows, and works better in predicting returns than the conventional momentum factor does. Ma, et al., suggest a mechanism: an answer to the question “why.” They propose that their factor works and their portfolio earns money, because each is “a revealed preference measure of the beliefs of skilled traders.”
Insiders, Shorts, and Options Traders
The scholarly literature has long indicated that insider purchases predict future stock market returns pretty well, but that insider sales do not. Insiders sell their companies’ stock for a lot of reasons, sometimes just to finance a family vacation or a home purchase. These sales are not tied to a belief that the stock is about to fall and this do not make it into the calculation of INFO.
Also, it is a generally accepted premise that short sellers are informed traders. This is not necessarily the same as possessing non-public information. Short sellers “earn large returns by trading soon after the release of public information” which suggests that they are skilled at processing such information as it becomes (legally) available. Likewise with options traders. Thus, the short side of the L/S portfolio behind INFO is created not from the sales of insiders but from shorts and option traders expressed informed pessimism.
In looking to the performance of INFO, Ma et al used a database consisting of stock movements from January 1996 to December 2015. This gets their portfolio an average return of 1.24% per month, which contrasts sharply with the 0.47% return of the momentum portfolio. It also beats out the market portfolio, the Fama-French size factor, and the Fama-French book-to-market factor.
The data set begins with the start of 1996 because that is when the earliest OptionMetrics data becomes available, allowing for the option traders’ contribution to the sort side of the INFO calculation.
The Sharpe ratio of INFO is 0.39, which as these authors say more than quintuples that of the momentum factor.
Final Numbers and Thoughts
The cumulative return of INFO in the 20-year-long period under study would have been 1639%. Anyone investing in the S&P 500 at the start of that period would have generated a cumulative return of only 238% by December 2015.
Since the INFO factor depends on a L/S portfolio, it is also a matter of interest whether the portfolio generates higher return as such than either of its legs would have generated alone. These authors conclude that it does. Specifically, INFO’s Sharpe ratio of 0.39 beats the ratio of its long leg alone (0.25) and beats the stuffing out of the ratio of its short leg alone (-0.07).
The article concludes: “[Our] study highlights the importance of combining positive signals with negative signals. Future research should explore the value of such information in other asset markets, such as bonds, currencies, or commodities.”