If You've Got the Money, I've Got the Time: Spatio-Temporal Footprints of Spending at Sports Events on Foursquare

Notable events such as sports games have a highly visible impact both in the media and at the event location itself, but the preand post-game activities of attendees—and their spending around such events—are much less evident. In this work, we use social media footprints to analyze and interpret these broader ripples of activity: specifically, we examine the temporal, spatial, and microeconomic patterns of baseball and basketball game attendees’ activity around game events for 21 North American cities and 32 teams during 2010–2014, using mobility and price data from the location-based social network Foursquare. We identify some universal spatiotemporal patterns that hold across both sports and cities: exploiting additional information about venues’ price levels, we find that, in aggregate, users who check in later at the game come from pricier restaurants and bars. We use these insights to predict the costliness of a user’s choice of restaurant/bar, based only on a single transition to or from the stadium/arena, using the temporal properties of the transition such as the time of check-in at the game, distance traveled, and speed of transition, with the last being the most important factor. Our analysis carries broad implications for human mobility studies, mobile applications, and sports economics. Introduction The impact of event spectatorship, which can be felt in traffic congestion, the local economy (Siegfried and Zimbalist 2000), and social media (Kwak et al. 2010), is all a result of the collective behavior patterns of individual spectators around event times. These preand post-game activities carry valuable insight about the spatio-temporal dynamics of events and their microeconomic footprints, which are the result of thousands of human choices about how to spend time and money. When Benjamin Franklin wrote in his essay Advice to a Young Tradesman “remember time is money,” he was referring precisely to these trade-offs, which define our priorities, values, and ultimately our socioeconomic context. In this work, we use social media data to explore the temporal, spatial, and economic patterns of the participants at and around a particular class of events—namely, sports events in 21 North American cities during 2010–2014, covering a total of 32 baseball and basketball teams. Sports Copyright c © 2016, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. teams have a strong following, wide cultural impact, and multi-billion dollar market valuations. Despite the importance of sports events, however, there is little data-driven insight on the behavior of event attendees, such as their spending before and after the game. By complementing user mobility patterns with place–price data from the location-based social network Foursquare, we discover universal temporal and spatial patterns of behavior across both sports and many cities. These observations allow us to predict the price level of a restaurant/bar that a user visits before or after the game solely from the mobility information from that user’s single check-in at the stadium. Sports games, of course, form only a narrow slice of the kinds of events with accompanying ripples of social and economic activity, but sports are a particularly appealing domain for our analysis, for three important reasons. First, a game has a clear start time, a clear finish time, and a precise geographic location. Second, games are big events, typically bringing tens of thousands of fans to the stadium. Finally, and most importantly, baseball and basketball games are largely symmetric; we are able to aggregate an entire season’s games at a stadium (41 or 81 games, depending on the sport) by shifting each game’s start time and overlaying fan behavior from all games at a single stadium. Similar to work in behavioral economics and marketing psychology that has shown a relationship between higher spending levels and time spent at retail shops (Duncan Herrington and Capella 1995; Iyer 1989; Milliman 1982; Park, Iyer, and Smith 1989, for example), we discover that users who transition at slower speeds to and from the sports stadium tend to visit pricier Foursquare venues before/after the game, as do those who check in at the stadium later into the game. We further formulate a user-level classification task, where we show that, across cities and for both basketball and baseball, speed is a better predictor than check-in time or distance alone when predicting the spending of a user. In short, those who seem to place lower importance on time at the stadium (by moving more slowly from a pre-game venue or by being willing to miss more of the game) also seem to spend more money outside the stadium. Understanding how human mobility patterns around such events relate to microeconomic behavior is of consequence to location-based advertising and to measuring the economic impact of sports events to the local community. The Workshops of the Tenth International AAAI Conference on Web and Social Media CityLab: Technical Report WS-16-16