Information Inequality in Online Education

Abstract

In this paper, we study platform solutions for improving customer engagement in online higher education by reducing informational inequality for historically under-represented groups in education such as females and workers seeking to improve their skill set. Using novel search and enrollment data from the largest online education platform in Iran, we estimate a structural model of course search and enrollment for paid courses, allowing us to recover learner belief's about courses, as well as their true preference over the characteristic space of online courses. We use machine learning methods to recover the latent characteristic space of courses, identifying which courses are substitutes via a data-driven approach. We document significant heterogeneity in how learners differing by gender and working status perceive course value, due to biased beliefs, relative to the true value. Counterfactual policy exercises suggest that the platform can increase revenue, improve consumer surplus, and reduce the gender gap in quantitative courses. Finally, we also present the problem faced by the platform from an information design perspective, and characterize the optimal signal the platform can send to learners with heterogenous priors to maximize an arbitrary objective function.