A standard, simple linear regression model can be used to identify apparent investment anomalies. Caution and common sense, however, are required to ensure that any identified anomalies have a sensible and realistic basis before they are used in decision making.
The author uses a standard linear regression model to identify a series of increasingly obscure anomalies that predict different aspects of market returns, but these anomalies are not the focus of the article. Rather, the author questions the statistical framework that is almost universally adopted to test and identify investment anomalies.
How Is This Research Useful to Practitioners?
This research is useful to practitioners because it serves as a reminder to continually question research and the underlying assumptions on which it is based. The author highlights that given the amount of data available for research, it is unsurprising that anomalies can be identified by using statistical methods.
This research is of particular relevance to junior practitioners because it demonstrates the risks of blindly generating positive results via backtests based on questionable premises.
How Did the Author Conduct This Research?
The author uses a simple linear regression model to forecast excess returns with a series of explanatory variables that have known values at the start of the period. A number of different, increasingly obscure anomalies are then tested and are proved to be statistically significant. These anomalies include the impact of the political allegiance of the US president on market returns, the impact of seasonal weather conditions on traders’ moods and stock prices, and the predictive abilities of ocean temperature anomalies (such as El Niño).
Next, the author extends his research into the heavens and identifies anomalies concerning the relative positions of the planets (in particular, Mercury and Venus and then Mars and Saturn). Finally, the importance of sunspot activity and its relationship to markets is illustrated.
Unlike in most financial literature, the details of the identified anomalies, although intriguing, are somewhat irrelevant to the underlying goal of the author. He is illustrating that an investment anomaly should pass such subjective hurdles as common sense and intuition as well as objective statistical tests.
This article is unique in the financial literature in that it leaves the reader guessing as to where the author will go next and how it will end: It is a page-turner. Fortunately, the author uses his research to make the eminently sensible point that given the amount of data available, financial and otherwise, it is likely that some datasets will appear to predict the direction of market returns.
Most analysts realize, however, that for an anomaly to be considered worthy of exploitation in a real-life investment process, it must have a sensible and intuitive financial or economic basis. Following this very basic principle would preclude most, if not all, of the anomalies discussed in this article, which shifts the research from merely interesting to informative.
As the big data evolution accelerates, there will be increased scope for erroneous conclusions to be drawn from research. The author could have emphasized this point by highlighting that a number of other potential anomalies were (presumably) tested but failed to pass the statistical tests.