Differences in online course usage and IP geolocation bias by local economic profile

Abstract

Although Massive Online Open Courses (MOOCs) have the promise to make rigorous higher education accessible to everyone, prior research has shown that registrants tend to come from backgrounds of higher socioeconomic status. In this work, I study geographically granular economic patterns in registration for HarvardX and MITx courses, and in the accuracy of identifying users' locations from their IP addresses. Using ZIP Codes identified by the MaxMind IP geolocation database, I find that per-capita registration rates correlate with economic prosperity and population density. Comparing these ZIP Codes with user-provided mailing addresses, I find evidence of bias in MaxMind geolocation: it makes greater errors, both geographically and economically, for users from more economically distressed areas; it disproportionately geolocates users to prosperous areas; and it underestimates the regressive pattern in MOOC registration. Similar economic biases may affect IP geolocation in other academic, commercial, and legal contexts.