Predicting the S&P: Voodoo Quantation, and Big Data

Dec 15, 2014

I try to be positive. I really do. But yesterday I came across a post that is so ridiculous that I can’t stay silent. The writer observes that in every year since the Chicago Board Options Exchange’s Market Volatility Index (^VIX) was created in 1986, if the VIX traded at more than 34% above its one-month low at least once during the month of December, the S&P 500 showed a gain from that point until the end of the year. (Got that? There will be a quiz on this later.) The writer claims to have found a predictor “whose results have been so overwhelmingly consistent that it is hard to ignore.”

Now, do you have any sense of just why volatility expectations would be a predictor of stock market direction? ** Or why a correlation with the index’s lowest recent value would make logical sense, let alone why “recent” should mean specifically one month? Or why the threshold for the indicator would be 34% above that low, and not 38% or 27.6825%? Or why this trend is particularly applicable to a day in December through 12/31? You don’t? Well, I don’t, either.

It is a fundamental principle of the scientific method that hypotheses should be based on observation. And then tested for validity. The “observation” part clearly happened, but I’ll bet big money that the “test” part was skipped – they just crunched through an immense number of time series until they found one that correlated with a positive return for the S&P 500 over a fairly short period of time. That approach is especially irresponsible when there is no evident reason why the correlation should exist – the writer certainly makes no effort to provide one – and when the suggestion is that people should base investment decisions on that analysis. (Yahoo! Finance found the post sufficiently profound that they posted it on their site.)

Now, why did I include “Big Data” in this blog’s subject? Because this kind of shoddy analysis is made much more possible by the existence of powerful software tools enabling people to analyze relationships within gigantic piles of data. Please consider this risk when using big data to help you make important business decisions.

Thanks for letting me vent. I will now go wipe the foam off the corners of my mouth and try to go to my happy place.

** NOTE: The VIX is an estimate of the expected volatility in the stock market over the next 30 days. Volatility is an important measure for those trying to value options and futures. However, since volatility is a notion of both upside and downside risk, it’s not considered a predictor of market direction. (Click here if you actually want to learn more.)

“Painting with Numbers” is my effort to get people to focus on making numbers understandable. I welcome your feedback and your favorite examples. Follow me on twitter at @RandallBolten.