It should be easy to spot potential college drop-outs ahead of time, right? Too much partying, not enough focus on school work, or maybe a poor academic record in the past? But here’s a startling fact: new research now suggests that 44 percent of college students who leave school before finishing their degree have a GPA of 3.0 or better, and more than three fourths of students who depart are at 2.0 or above.
In a world where a degree is the price of admission for even entry-level jobs, college completion is fast becoming a moral imperative. And if we can’t improve the odds for low-income and first-generation college students, we risk excluding a generation from higher education’s promise of social and economic mobility. Higher education’s growing equity gap is complex and multifaceted, but it also reveals a troubling tradition in higher education: an over reliance on demographic data (like income and race/ethnicity) or past performance (like grades or GPA) to make assumptions about a student’s success in the future.
In other fields, the risks of making predictions based on historic averages or blunt demographic data are well established. Every financial prospectus comes with a warning that a fund’s past performance is no guarantee of its future results. Doctors consider demographics and health history, but patients expect a diagnosis rooted in their individual habits, symptoms, or lab results.
Campus-wide initiatives are often well intentioned – comparing current students with characteristics of those who dropped out to trigger “early warnings” and intervention. But reliance on the demographics or behaviors of some who leave risks codifying existing dogma in simplistic heuristics that run counter to higher education’s democratic promise. Put simply, analytics efforts that fail to consider the unique attributes of individual institutions –and students – may do more harm than good. The risk is real: at a growing number of institutions, completion rates are improving, and yet the gaps between demographic groups are growing.
Fortunately, this is changing. Faculty and administrators are beginning to follow digital breadcrumbs to identify real-time challenges and opportunities to nudge students in the right direction before they fall behind. Breakthroughs in data science now allow colleges and universities to move beyond conventional wisdom in a world where data tells stories about high performing students that struggle to find the right classes (because the institution is not offering them that term), or grapple with life and logistical challenges that often pose greater barriers than academics.
The data itself is fairly innocuous, but the hard work of integrating disparate silos of data from across campus reveals trends that provide insights into how students behave. Over time, patterns emerge. It becomes possible to predict, for instance, that if a certain student turns in her first assignment on time, she is much more likely to pass the course. If she doesn’t, a professor or an advisor can intervene before the situation becomes critical.
This doesn’t mean that historic data isn’t useful. Factors like high-school environment, GPA, and SAT scores can provide helpful indicators of risk to inform first-semester course selection and priorities. But relying only on these simple, average-based factors overlooks the potential that colleges and universities are designed to unlock – and poses the risk that demography dictates destiny when it is used to track, rather than inform, a student’s path through higher education.
Should a diverse student who received a poor mathematics grade in high school, or the prior semester, be reactively counseled out of a passion for engineering? Shouldn’t a student with the academic equivalent of a spotless medical history be informed about resources and supports available if she runs into trouble taking a course that challenges her in unforeseen ways?
The signals can be counterintuitive, which is precisely why relying on static data sacrifices the potential of the individual. Broad assumptions put institutions at risk of drawing false conclusions as they work to support students fighting for a chance to succeed. Data science is helping colleges to fill in blind spots and deliver much more personalized advice and guidance, which is essential to improving student engagement, success, and degree completion.
Of course, it isn’t all about the data. Faculty and advisors – like physicians – must interpret data through the lens of their own experienced judgment and discretion to make a final diagnosis about when and how to intervene with students. Colleges and universities must fine-tune and adapt their organizational cultures to develop the processes and expertise to responsibly tap into and utilize more dynamic data. Done well, data science can help us to understand the roadblocks to success and reduce the likelihood of a misdiagnosis for individual students. At scale, it is beginning to unlock the multi-generational benefits of college for millions of students.
Charles Thornburgh is the CEO and Co-Founder of Civitas Learning and William Serrata, Ph.D., is President of El Paso Community College.