Is Geo-Location Information Helpful in Trading the Spread Between Under Armour and Nike? (Presentation Slides)

IDIES Annual Symposium Oct 2018

15 Pages Posted: 10 Dec 2018

See all articles by Jim Kyung-Soo Liew

Jim Kyung-Soo Liew

Johns Hopkins University - Carey Business School

Tamás Budavári

Johns Hopkins University - Department of Applied Mathematics and Statistics

Xuzhi Wang

Johns Hopkins University

Shihao Ma

Johns Hopkins University

Date Written: October 19, 2018

Abstract

In this work we examine fundamental linkages between “location data” (a.k.a. “human movement data”) and financial market equity price behavior. We concentrate our study on a popular hedge fund trading strategy known as pairs-trading. In pairs-trading, one stock is purchased while another stock is sold-short, resulting in a bet on the price spread between these two stocks. We computed the price spread between Under Armour and Nike and examined this spread behavior with respect to several variables. We also investigated the relative volume of human traffic visiting each physical store location, as proxied from anonymous cell phone geolocation traffic. The geolocation positions were recorded intraday and cover the time span from January 1, 2018 to July 17, 2018. The data set is massive, consisting of over 47.2 billion rows. Longitude and latitude coordinates are employed to create economically distinct, yet competitive, paired-physical store locations of Under Armour and Nike, respectively. We monitored the relative activities in these non-overlapping ring-fenced locations to use as a proxy for the consumers’ competing fundamental demands. In the midst of our monitoring, we gleaned the following fascinating results regarding relative location data: (1) relative location data is a statistically significantly feature at the daily frequency, but not at the intra-day frequency, (2) it is a contrarian indicator, with higher/lower relative location ratios predicting next day contraction/expansion in spreads, and (3) it is consistently in the top five features across the usual suspect of machine learning algorithms. Given our limited data set, we disclaim our findings accordingly, but these initial results show that geolocation data should be a significant factor when building professional-grade pairs-trading models.

Keywords: Pairs, Trading, Geolocation, Machine Learning, AI

Suggested Citation

Liew, Jim Kyung-Soo and Budavari, Tamas and Wang, Xuzhi and Ma, Shihao, Is Geo-Location Information Helpful in Trading the Spread Between Under Armour and Nike? (Presentation Slides) (October 19, 2018). IDIES Annual Symposium Oct 2018 . Available at SSRN: https://ssrn.com/abstract=3285883

Jim Kyung-Soo Liew (Contact Author)

Johns Hopkins University - Carey Business School ( email )

100 International Drive
Baltimore, MD 21202
United States

Tamas Budavari

Johns Hopkins University - Department of Applied Mathematics and Statistics ( email )

3400 N Charles Street
Whitehead 100
Baltimore, MD 21218
United States

Xuzhi Wang

Johns Hopkins University

Baltimore, MD 20036-1984
United States

Shihao Ma

Johns Hopkins University

Baltimore, MD 20036-1984
United States

Register to save articles to
your library

Register

Paper statistics

Downloads
2
Abstract Views
34
PlumX Metrics