I forget when I first heard the term “Big Data.” It might have been 2009 or 2010 while researching a company involved in the US mortgage industry. Around that point in time, the term seemed to cross over to the mainstream business lexicon; companies in industries as diverse as housing, finance, advertising, technology all started to tout their collection and use of Big Data. It was supposed to give them a leg up. (Just a note to all of the future CEOs out there: when everyone is trumpeting “…XYZ” as a competitive advantage, it stands to reason that “…XYZ” is not much of a competitive advantage.)
Today you can scarcely run across a company that hasn’t dipped its hands into the Big Data stream in some fashion. That includes the investment management industry. Investment funds have always looked for ways to get an “edge” on a particular investment—a data point or piece of analysis that is particularly informative and different from common opinion. Now funds are beginning to sign up for services that sift through all sorts of digital markers and patterns to glean insight: social media buzz, satellite imaging, mobile app downloads, just to name a few.
I can certainly appreciate the value of such efforts. As a stockpicker and investor, I am always looking for an advantage over the market. But perhaps some are placing too much faith in these new methods. A few weeks back a research firm called ITG downgraded its view on Harley Davidson’s stock based on data it analyzed that suggested Harley’s US retail sales declined 6% in the first quarter. The stock declined over 7% that day on the news. When Harley reported results the following week, actual US retail sales were down a measly half of one percent. In retrospect, ITG’s data seemed to be a bit of a red herring.
There’s no question that Big Data, advanced computing power and fancy algorithms will push their way further into the investment industry. It’s likely Forager can improve its own research process. But I think it’s important to view investing, and especially stockpicking, as more of a puzzle that fits many pieces together. Data is helpful, but on its own, it rarely tells the whole story.
Actually if one restricts companies to those that make a profit or give a dividend then the ASX is hardly ‘big data’. I have historic basic data (price dps eps, nta, D/E) that goes back to 1999. This includes at the moment just over 1000 companies. It does allow me to manipulate and test.
For example if one further restricts to those that have regular earnings (Dr Prices hypothesis, The Conscious Investor) then an algorithmic approach (Least squares fit and prediction, error in fit < 0.1 (10%) ) can produce earnings forecasts of similar accuracy to Analysts.
Unfortunately the sample is small. For forecasts (consensus) at June 2015 tested in December 2015: 53 examples, analysts average absolute relative error 0.21, Median 0.08, SD 0.34. Algorithm; mean relative error 0.22 median 0.09, SD 0.34. Without the restriction neither is very useful: 288 samples; Analysts SD 2.43 (median .24 max error 32.9), Algorithm 5.63, median 0.38 max 83.7.