Foot Traffic into Brick and Mortar Stores

A recent paper by six scholars affiliated with Johns Hopkins University investigates how one can make alpha by tracing the foot traffic into brick and mortar stores and engaging in a pairs trading strategy on that basis.

Let's back up. What the heck is pairs trading? It is a strategy available any time two investable assets (A and B) are related to one another in a predictable way. Consider Exxon and Chevron. When broad industry factors are the driving force, one would expect that an increase in the value of Exxon would be an increase in the value of Chevron. Their stocks would move in parallel lines. 

Sometimes, one or the other will move for an idiosyncratic reason -- say, a scandal about the CEO of one of these companies may make headlines, pressing that stock down. That may not impact the other one, or may even drive it up, as people unhappy with the CEO news but still eager to invest in crude oil sell one and buy the other.

A pairs trade is typically a trade on a "return to normal" hypothesis. If A and B have been moving in parallel for a long time, and suddenly move in opposite directions, a pairs trader may well expect that this is a brief episode and they will resume their parallel course.  

So let's get back to the Johns Hopkins study. Both Nike and Under Armour sell athletic clothing and gear at brick and mortar retail outlets. They seem like plausible candidates for pairs trading: treat a parallel movement as normalcy and bet in favor of a return to normalcy when that is disturbed.

It turns out that this works, and geolocation data helps make it work. Publicly available information (provided by the cell phones that most shoppers carry about with them) allows us to know how many people are walking into a particular retail outlet that sells Nike and/or UA products day to day. That helps us decide when their stock prices are about the vary from the historic relationship and when they are about the return.

The Johns Hopkins scholars found that geolocation information is an important variable, and that in fact pairs trading using only geolocation data yields positive returns. Machine learning and rolling analysis improves those returns. 

This trading strategy won't do anybody a lot of good until foot traffic into and out of brick and mortar stores becomes part of our national life again, but it is nice to have an academic print on it.