Beyond test scores: Predicting retention using application behaviors

Expert Insight

Beyond test scores: Predicting retention using application behaviors

With increased emphasis on retention and graduation, institutions are looking to Admissions to better predict student success at the beginning of the pipeline. Some institutions are now experimenting with surveys or questionnaires that measure non-cognitive variables like grit and engagement. Despite industry-wide attention, traditional non-cognitive assessments remain difficult to operationalize in admissions and have an ambiguous track record in practice.

The challenge: Operationalizing non-cognitive data

While the predictive power of non-cognitive data is clear, Enrollment Managers are concerned that non-cognitive screening tools can be gamed by applicants, are difficult to evaluate consistently, and consume too much staff time. Unlike assessments that rely on student feedback, demonstrated application behaviors are less easily manipulated. Application behaviors, like application date and frequency of contact with the institution, are already easily captured and consistently correlate with retention, particularly for at-risk student groups.

Related study: Incentivizing Behavioral Change with Aid Dollars

Application date as a predictor of retention: It is common knowledge that students who apply later in the admissions cycle yield at lower rates. Analysis of our members’ data suggests that late applicants are also less likely to retain. This trend exists at institutions across the selectivity spectrum, is independent of traditional academic quality indicators, and is especially strong among less-prepared students.

Later Applicants Retain at Lower Rates

First-Year Retention by Month of Application, Washington State University, 2004-2008 Entering Cohorts

Later applicants retain at lower rates

By including application date in a predictive retention model during the admissions process, enrollment managers are identifying students with borderline academic credentials who are disproportionately likely to succeed.

EM insight: “Screening in” borderline applicants

Actionable data insights

Washington State University was able to capitalize on this retention insight when a new president unexpectedly requested an expansion of the incoming freshman class in April, forcing a last-minute change in the enrollment plan. Faced with the need to admit more of the academically marginal prospects remaining in its pool, Washington State re-ran its academic quality analysis on the applicant pool while including application date and an index of high school quality.

Outcomes: Primarily guided by the application date insight, admissions secured 180 new enrollments. These late admits were less academically qualified based on traditional metrics, but retained at a higher rate than regular decision admits with similar entrance qualifications.

Key insight: Model retention using application behavior

Despite the rich data available, the retention impact of application behavior is rarely analyzed during the admissions process. Institutions should mine commonly observed application behaviors for correlations with student success to create institution-specific screening models. Admissions staff can use these models to identify and target at-risk students for additional support post-enrollment. They can also justify admission of otherwise at-risk students, often from under-represented groups, with greater confidence that these students will succeed.

Make financial aid work for your student success initiatives

Enrollment managers have a significant opportunity to promote student success within the functions they often control, particularly financial aid. We explored how progressive institutions turn to financial aid as additional leverage for academic and engagement interventions. Read the study

EAB asks you to accept cookies for authorization purposes, as well as to track usage data and for marketing purposes. To get more information about these cookies and the processing of your personal information, please see our Privacy Policy. Do you accept these cookies and the processing of your personal information involved?