data

Predictive Analysis of Student Data

Student Success Assessment, Evaluation, and Research Enrollment Management Orientation, Transition and Retention Technology AVP or "Number Two" Mid-Level Senior Level VP for Student Affairs
April 7, 2017 Alexis Wesaw Kevin Kruger Dr. Amelia Parnell Michelle Burke

NASPA–Student Affairs Administrators in Higher Education conducted a landscape analysis of the use of predictive analytics by student affairs professionals at higher education institutions. Most student affairs divisions are collecting student engagement data and conducting needs, process, and outcomes assessments. NASPA’s research addressed the kinds of student engagement and behavioral data that are collected within student affairs departments and the extent to which institutions are using such data in predictive analytics models. The research also addressed the factors that influence institutions’ development of data analytics projects and how various resources are employed to collect the data and conduct the analyses.

With public scrutiny over the value of higher education increasing, colleges and universities are turning to business intelligence practices to improve outcomes. In higher education, institutional performance is often centered on student enrollment and retention, with an ultimate goal of students’ timely persistence to a college degree. As a result, colleges and universities are considering how to use data to intervene proactively with students who are at risk for poor academic performance or low institutional engagement. Many institutions have adopted data analytics practices to forecast operational needs and enrollment trends, and are now applying the use of predictive analytics directly to student success initiatives.

Prior research on predictive analytics in higher education examined the prevalent uses of data and the level of support for overall institutional analytics as well as “learning analytics” related to student success (Dahlstrom, 2016; Yanosky & Arroway, 2015). Although these studies addressed institutional applications of analytics, there was limited detail regarding the factors that influenced institutional support for the use of predictive analytics to increase student retention and persistence. It also appears that a gap in the literature exists regarding how student engagement data are being used in predictive analytics initiatives. Available research suggests that early efforts related to data analytics have focused primarily on academic and learning management system variables (Arroway, Morgan, O’Keefe, & Yanosky, 2016; Dahlstrom, 2016; Yanosky & Arroway, 2015), and while student affairs professionals are called on to implement intervention strategies after at-risk students are identified, student engagement data are not often included in predictive models.

NASPA–Student Affairs Administrators in Higher Education conducted a landscape analysis of the use of predictive analytics by student affairs professionals at higher education institutions. Most student affairs divisions are collecting student engagement data and conducting needs, process, and outcomes assessments. NASPA’s research addressed the kinds of student engagement and behavioral data that are collected within student affairs departments and the extent to which institutions are using such data in predictive analytics models. The research also addressed the factors that influence institutions’ development of data analytics projects and how various resources are employed to collect the data and conduct the analyses.

NASPA interviewed professionals from student affairs, academic affairs, institutional research, information technology, and assessment at 25 colleges and universities to discuss the range of methods for utilizing data analytics to inform retention efforts. The institutions differed in size and approach; however, the interviews revealed that all are committed to student retention efforts and are using or plan to use some form of predictive analytics. Five of the institutions are in the planning phases of predictive analytics projects, 9 have implemented projects within the past 4 years, and 11 institutions have been using some form of data analytics related to student success for 5 years or more. Conversations with members of these institutions yielded several common factors regarding the use of predictive analytics, many of which pertain to the alignment and allocation of personnel and financial resources, including the following:

  • institutional commitment to increasing undergraduate retention and
  • improving enrollment management;
  • senior-level leadership encourages data-informed decision making;
  • strong partnership between campus functions, particularly information
  • technology and institutional research;
  • adequate allocation of resources for staff to effectively address the findings
  • produced from predictive models;
  • continuous training and support for personnel who collect, analyze, or
  • utilize data;
  • capacity to connect data across systems or within one system; and
  • increased accountability metrics, such as performance-based funding.

Most of the institutions in the study are still primarily focused on using academic data in predictive models. However, the range of student engagement data that could be used is much broader and could lead to deeper understanding of keys to student persistence. Although most of the institutions could do more to incorporate engagement and behavioral data into their predictive models, they are using these data in the execution of early alert systems, which are retention tactics that target at-risk students for intervention through a variety of support systems. Early alert systems utilize several types of data, including pre-enrollment variables such as high school grade point average and standardized test scores, academic variables such as mid-term grades and course attendance, motivation and self-efficacy variables such as students’ self-reports of connectedness to the institution, use of support services such as advising and tutoring, and student engagement variables such as participation in campus activities.

One challenge for many institutions is limited capacity to gather accurate student engagement and behavioral data and connect them to the student information system for inclusion in predictive models. Several institutions are strategically planning how to meet that challenge with improved data collection and less siloed data analysis. However, as institutions increase their capacity to capture and analyze student information, they will likely need to address concerns regarding data privacy and establish a process for informing students of how their information will be used. As administrators develop data-informed interventions to address students’ needs, it will be critical that such strategies are based on the experiences of all students. For example, administrators that intend to use predictive models for the purpose of identifying at-risk students will need to be careful to avoid using engagement data in ways that lead to inherent bias, particularly with regard to identifying behavior patterns for underserved or underrepresented populations. Predictive models are an attractive option for institutions that need a strategy for matching limited resources to students who are most in need. By including engagement and behavioral data in their models, institutions could strengthen the accuracy of their analyses and possibly increase the influence of support services on student retention. Several institutions in this study have had positive results from their application of predictive analytics, and other institutions similarly expect successful implementation in the next few years.