On Stocks And Machine Learning

by: Yoav Zelikovic
Summary

Can a quantitative model provide us with advantages in reaching investment decisions?

Prof. Daniel Kahneman and Prof. Tversky, pioneers in the field of behavioral economics, have provided us with insight into the pitfalls of the human thinking process.

Prof. Joel Greenblatt describes in his book "The Little Book that Beats the Market" a model that identifies stocks for successful investment.

Specifically in the field of stock investment, it appears that the human mind should recognize and acknowledge its shortcomings, "lay back" and let the computer assist us.

It is not surprising that more and more investors chose to detach themselves from the “herd“- effect and the psychological bias of investment decisions, and  base the management of their portfolios on factor investing (Smart Beta) and investment models/algorithms (Quantitative Investing).

The Major League Baseball team, the Oakland Athletics, for many years a poorly performing and unsuccessful team, surprised everybody during the 2000-2001 season. The team had a remarkable run of wins, led the Major League in performance, and broke a number of League records. Michael Lewis, in his bestselling book, "Moneyball: The Art of Winning an Unfair Game", describes the secret to this turnaround. The A’s new, young and unexperienced manager, Billy Beane, “revolutionized” the team’s roster by releasing a number of star players (despite the protests of his closest advisors) and instead, signing a number of unknown players. He chose these players by using a model/algorithm that was based on several professional, well-defined baseball parameters. This model, which completely ignored any predilections from the media, baseball fans, and all other kinds of “experts”, was successful in identifying anonymous, but very talented players, who brought about the dramatic turnaround and success of the team.

Prof. Hoffman and his colleagues from the Psychology Department of the University of Oregon, conducted in 1968 an experiment that used a well-proven and established model composed of seven radiological findings which, in various combinations, could determine the presence of stomach cancer in patients with peptic ulcer. Hundred hypothetical cases of peptic ulcer based on these findings were presented to nine expert radiologists who were asked to determine whether or not a malignancy existed and how confident they were in their diagnosis. Surprisingly, the radiologists’ responses were characterized by inconsistency and great variability not only between the answers of the different radiologists, but even between the answers of the same radiologist on identical combination of findings presented to him twice (without his knowledge). The main conclusion from this experiment was that a relatively simple radiological model was much more reliable in diagnosing stomach malignancy then the subjective judgement of expert radiologists.

These two examples, from the areas of sports and medicine, raise the question – is the so-called “Machine Learning” advantageous over human judgement? Or more bluntly – is "Artificial Intelligence", which is rapidly developing in recent years, about to replace the human mind?

Nobel Laureate Prof. Daniel Kahneman and Prof. Tversky, pioneers in the field of behavioral economics, have provided us with insight into the pitfalls of the human thinking process. These two prominent scientists found that a number of “cognitive biases/heuristics” influence the human brain and often interfere with the ability of the human being to make rational decisions. These biases include, among others, 1. "Confirmation Bias"- the tendency to give more weight to findings which confirm our previous assumptions, and ignore or give less weight to findings contradicting these assumptions; 2. "Anchoring Bias" - the tendency to overvalue or "anchor" on the first source of information obtained on a specific subject; 3. "Representational Bias"- the tendency to judge the probability of the occurrence of a specific event by finding a similar, comparable event, and assuming that the probabilities of occurrence of these two events are identical. These biases are so powerful that Prof. Kahneman himself admitted that their recognition has not helped him to overcome them in his private engagements.

What about the investment world? Can a quantitative model assist us in reaching investment decisions? Prof. Joel Greenblatt describes in his book "The Little Book that Beats the Market" a model that identifies stocks for successful investment based on concepts developed by the two leading investors, Benjamin Graham and Warren Buffett. This model identifies companies that are trading at a low enterprise value (EV) in relation to earnings before interest expenses and taxes (EBIT) (namely low EV/EBIT ratio), and at the same time achieve a high return on their capital employed (namely a high ratio of EBIT to Net Fixed Assets + Working Capital). Prof. Greenblatt performed a back test to determine his investment returns if he invested in only 25 of the highest ranking stocks by these parameters, in the two decades preceding the year of publication of his book (2006). Analysis of the test’s results yielded annual returns of over 30%. Subsequently, Prof. Greenblatt invested his clients' money in accordance with this strategy in a "blind" manner, which, indeed, has resulted in very high returns, much higher than the returns achieved by the analysts in his investment group who used their discretion to choose stocks they favored from this group of 25 stocks.

The shortcomings of “experts” in the investment field were discussed by additional researchers. The well-known investor, David Dreman, demonstrated in his book "Contrarian Investment Strategies" that investing every year between 1970 and 2010, without using human judgment, in the two lowest stock deciles (by Price/Earnings ratio) in the American stock market would have achieved an annual return of 15.2%, compared with an average annual stock market return of 11.6% for the same time period. Nobel Laureate Prof. Eugene Fama and Prof. Kenneth French, leading researchers in finance, found that a simple strategy of automatically buying stocks trading at a discount to their book value (low Price/Book ratio) in the American stock market could be a very lucrative pathway of investment. They found that investing in the lowest Price/Book decile of stocks between the years 1963 and 1990 would have achieved an annual return of 20.5% compared with a 15.9% return for the general stock market in the same time period. Moreover, they discovered that investing in the highest Price/Book deciles of stocks (which are usually most popular among analysts and investors) would have achieved only a 10.2% annual return - 5.7% less than the benchmark. Another study conducted by Elroy Dimson, Paul Marsh and Mike Staunton (and published recently in the Journal of Portfolio Management), "Factor Based Investing: The Long Term Evidence"), found that the investment strategies described above have proven to be very profitable in 23 additional stock markets around the world.

It appears that operating by human discretion/judgment when buying stocks may lead us to wrong decisions and may result in investing in unprofitable stocks. Undoubtedly, the “cognitive biases” described by Kahneman and Tversky act on and affect the decisions of even the most experienced and famed stock analysts and portfolio managers in the world.

Specifically, the “Confirmation Bias” may lead analysts to purchase stocks that are well- known, popular and “juicy”. Analysts are usually “swamped” with information and data on the companies they follow which might raise their confidential level in their analysis of these companies’ stocks. The “Anchoring Bias” will make it difficult for the analyst to sell a stock that he purchased even if he discovers that he had erred in his original analysis of this stock’s performance. The “Representational Bias” may also lead the analysts to wrong investments. The problem related to the “Representational Bias” stems from the tendency of the analyst, when investigating the history and profile of the company (growth, profitability, etc.), to assume that these parameters will repeat themselves in the future. This assumption ignores the “reversion to the mean” phenomenon which is typical for the finance market and the economic market in general. This phenomenon refers to the situation in which profitable companies become less profitable over time mainly due to the competition they face, and, vice versa, very weak companies surprise and return to high profit level. This scenario is very well summarized in Mark Twain’s statement:

"It ain't what you don't know that gets you in trouble. It's what you know for sure that just ain't so"

In summary, the ongoing debate between those who promote the computer’s gain of domination on all areas of our life and those who vehemently protect the qualities and uniqueness of the human brain is far from complete. Nevertheless, specifically in the field of stock investment, it appears that the human mind should recognize and acknowledge its shortcomings, "lay back", and let the computer guide us (and the variety of analysts and portfolio managers) in our investment activity. The big advantage of the computer over the human brain, in addition of being "bias- free", is the ability of the computer to absorb huge amounts of data over long periods of time, summarize and process them, and provide the desired result in the fastest, most accurate, and most reliable fashion.

Hence, it is not surprising that more and more investors chose to detach themselves from the “herd- effect” and the psychological bias of investment decisions, and base the management of their portfolios on factor investing (Smart Beta) and investment models/algorithms (Quantitative Investing). As we showed above, a cautious and rational purchase of a basket of stocks trading at the greatest discount to their equity value and earnings (and persisting in this method over long periods of time) is a recommended strategy of investment, has a clear economic rationale, and yields excellent long-term performance.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.