Estimating the Effect Of Exercising On Users’ Online Behavior

Abstract

This study aims to estimate the influence of offline activity on users’ online behavior, relying on a matching method to reduce the effect of confounding variables. We analyze activities of 850 users who are active on both Twitter and Foursquare social networks. Users’ offline activity is extracted from Foursquare posts and users’ online behavior is extracted from Twitter posts. Users’ interests, representing their online behavior, are extracted with regards to a set of topics in several subsequent time intervals. The shift of users’ interests across different time intervals is taken as a measure of user behavior change on the social network. On the other hand, we employ user check-ins at a gym or fitness center as a sign of exercise and consider it to be an offline activity. In order to find the effect of exercise on online behavior, we identify users who did not go to the gym for at least two months but did so at least nine times in the next three months. We show that shift in interest reduces significantly for users after they start exercising, which implies that the offline activity of exercising can influence how users’ interests are shaped and change on the social network over time.