Abstract
The recent upsurge in enrollments in computing means that student attrition has a substantial opportunity cost. Admitting a student who leaves both reduces graduation yield and prevents another equally qualified student from enrolling. Professors cannot change the background of students, but they can control many aspects of student experience in the computing major. This paper presents the results of a study to understand strongest predictors of retention in undergraduate computing based on a large-scale survey administered in 14 U.S. institutions. Although some factors have more influence for certain demographic groups, findings from this data set suggest that some teaching practices have more power for predicting retention in computing including: relevant and meaningful assignments, examples, and curriculum; faculty interaction with students; student collaboration on programming assignments; and for male students, pace and workload expectations relative to existing experience. Other interactions such as those that a student has with teaching assistants or peers in extracurricular activities seem to have less value for predicting retention. Faculty would be wise to protect their enrollment investments by inspecting course themes, assignments, and examples for student interest and ensuring that students have many opportunities to interact with faculty both in and outside of class.