Association for Public Policy Analysis & Management

As students enter community college with widely varying academic backgrounds and needs, assessments are vital in determining whether or not a student would benefit from remedial coursework and other supports (Bettinger & Long, 2005). Dysfunctional placement practices led to California Assembly Bill 705 which mandates that community colleges “maximize the probability that a student will enter and complete transfer-level coursework…” and requires schools to use multiple pieces of evidence, not just standardized test scores, to place students (AB 705, 2017). This change has led colleges to experiment with using non-cognitive measures as a means to assess academic readiness; however, the utility of these variables for predicting student outcomes in community colleges has not been tested (Thomas, Kuncel, & Credé, 2007; Ngo, Chi, & Park, 2018). This study examines the ability of a comprehensive set of non-academic measures to predict first-year community college student outcomes for a large cohort of recent students.

Specifically, we examine if self-reported non-cognitive beliefs and aptitudes could improve the targeting of support in community colleges. Almost 4,000 incoming students at two community colleges in Southern California answered a battery of questions regarding their non-cognitive behaviors and beliefs. Included in this battery are non-cognitive variables that have been the focus of much recent research, including grit, conscientiousness, academic self-efficacy, college identity, mindfulness, mindset, and teamwork. We examine if these measures improve predictions of important academic and persistence outcomes over the rich set of academic background and test scores to which schools usually have access. We approach this question in several ways, including asking if the use of non-cognitive measures could accurately predict which students will do much worse than their academic records predict, if non-cognitive variables can meaningfully improve predictions of serious academic failures, and if observed behaviors and explain the relationships we find.

We find that non-cognitive measures improve somewhat our ability to identify students who perform much worse than their academic records suggest and to identify students who are at-risk of poor academic outcomes. Non-cognitive variables increase by four percentage points students who will experience serious negative outcomes, such as being put on academic probation in the first term or not persisting to the second year. The inclusion of non-cognitive variables is particularly helpful in increasing the predictive validity of our models for white students, which raises concerns about equity. We also find that the predictive validity of our non-cognitive factors is partially mediated by help-seeking behaviors such as visits to academic counselors and tutors. Overall, our results reveal the utility of non-cognitive variables in predicting first-year student outcomes, though the accuracy of these predictions depends on demographic characteristics.