Democratizing Education or Widening Gaps? Student Demographics and Outcomes in Massive Open Online Courses

• In Event: 14.036 - Massive Open Online Course Students in Context: Equity, Motivation, and Performance

Abstract

Objectives

While Massive Open Online Courses (MOOCs) offer unprecedented access to educational providers (Pappano, 2012), preliminary findings have raised concerns that MOOCs may exacerbate preexisting socioeconomic gaps in postsecondary attainment (Christenson et al, 2013). The disappointing results of a partnership between MOOC-provider Udacity and San Jose State University led the company’s founder to question whether the medium itself was a good fit for “students from difficult neighborhoods” (Chafkin, 2013). Using observational and survey data from HarvardX courses, we address “whether the medium itself is a good fit” for less advantaged students by modeling the relationships among socioeconomic status (SES) variables and course outcomes.

Theoretical Framework

The lower and slower completion rates of college students of low socioeconomic status are often linked to the importance of parent education with respect to institutional accessibility, social integration, college cost, the complexity of the federal financial aid system, and knowledge of how to navigate institutional systems (Bettinger & Long, 2009; Bowen et al, 2009; King, 2004). MOOCs though, through their low barriers to entry (Daniel, 2012) and novel interactive context (DeBoer et al, 2014), have the potential to mitigate certain kinds of achievement disparities attributable to SES.

Conversely, differential access to high-quality technology could lead to greater performance disparities compared to traditional higher education environments.

Methods and Data
Our primary dataset includes US users for nine certificate-awarding HarvardX courses from the 2013-2014 year (n > 100,000). Examples of survey variables include age, parent education, previous familiarity with the subject, motivation for registering, and address. Survey responses are merged with data on certificate-earning, final grade, and various courseware interactions (e.g. videos watched).

Findings

We find that SES variables—while strongly correlated with enrollment—are generally weak predictors of course completion and grades. However, for enrollees of tradition college-going age or younger, we identify statistically and substantively significant advantages for higher-SES students. We estimate that the effects of SES are similar in magnitude to other SES gaps observed in postsecondary education.

As expected given previous research (Deboer et al, 2013; Christenson et al, 2013; Ho et al, 2014), we find that US HarvardX survey respondents tend to have higher SES indicators than the general US population. Using logistic regression, we find that a difference in completion rate for two enrollees who differ only by one unit on a 3-point parent education scale is approximately .25 logits, which is approximately the difference between a completion rate of 6.0% and 7.5%. A .25 logit difference is also interpretable as the approximate difference between the 62% and 68% completion rates for low-SES and middle-SES college students that Bowen et al (2009) found in their study of public four-year universities in the US.

Significance

Our findings are most consistent with the hypothesis that the MOOC medium is neither a silver bullet for mitigating socioeconomic gaps in postsecondary education nor is it especially unsuitable for educating low-SES students who choose to enroll compared to traditional four-year universities.

Authors

• John David Nadal Hansen, Harvard University

• Justin Fire Reich, Massachusetts Institute of Technology