If students struggle with a certain subject in their freshman year, are they more likely to drop out? Which students are more likely to seek out academic support, and do they graduate at higher rates than students who don’t? How much does it matter if students delay choosing a major?
Getting the answers to questions like these has never been easy, but the situation is beginning to change just at a time when student outcomes in higher education have never been more important. The reason for the change is the same one that explains which advertisements appear on an individual’s Facebook page or which movie recommendations Netflix has for you on a Friday night. Data, when collected and analyzed correctly, can uncover patterns of human behavior in new and revealing ways. Consider the influence of big data in higher education. With the help of data analytics, state education agencies and higher education institutions are trying to increase college retention and graduation rates. According to a 2016 report by the U.S. Department of Education, 60 percent of full-time students who enrolled at a four-year institution in 2008 took six years to graduate. The same study showed the average retention rate is around 80 percent, but less selective schools with open admissions policies retained 62 percent of their students between 2013 and 2014. Graduation and retention rates are key components to a school’s reputation. A typical college graduate working full time earns 54 percent more than a worker who attended some college, but has no degree. When retention and graduation rates improve, both students and institutions benefit. How can higher education keep students in school long enough to graduate? Turns out, analytics has some answers. Predictive analytics uses massive amounts of historical data to identify patterns and forecast outcomes and trends. More importantly, the data indicates places where universities can use limited resources to make the greatest impact. When colleges understand that students earning a C in a freshman math class are far more likely to drop out of school than graduate, the schools can identify those students early and offer tutoring services before it is too late. Longitudinal data, which tracks the same sample of students at different points in time, helps advisers see a clearer picture of an individual student. Why is this student, who tested well in reading, struggling with a freshman literature class? Seeing that pattern can help advisers ask students the right questions and nudge them toward success. The Southern states of Georgia, Tennessee and Kentucky fall in the bottom half of states when it comes to six-year graduation rates for first-time, full-time students who enrolled in 2007, according to The Chronicle of Higher Education. Georgia has a 54 percent graduation rate, followed by Kentucky with 49 percent and Tennessee one percent lower. (Delaware took the top spot with nearly three-quarters of its students graduating in six years, while Washington, D.C., came in last at 16 percent.) But Georgia, Tennessee and Kentucky have made significant progress from where they were a decade ago, and they continue to move forward as they harness big data to personalize the higher education experience for students, close achievement gaps and help more students graduate successfully. “All three states have long been leaders in thinking about the role of data in education,” said Brennan McMahon Parton, the Data Quality Campaign’s associate director of state policy and advocacy. “Data is a really powerful tool to inform practice and target students’ specific needs. There’s so much interest in making sure kids get to college and stay on track to graduate, and that is one of data’s big promises.” Georgia Georgia State University in Atlanta, ranked fourth in the nation in 2016 among the most innovative colleges and universities by U.S. News & World Report, raised its six-year graduation rate about 22 percentage points over a decade to 53 percent in 2013 with the help of a data analytics program and targeted advising. Nearly seven years ago, the school looked at which students attended academic advising, said Timothy Renick, Georgia State vice president for enrollment management and statistics. It was either the highly conscientious students or those required to go due to academic probation. The school realized it was spending most of its advising resources on students whose academic success was not significantly influenced by that advising. Renick knew that if they could determine which decisions and behaviors were risk factors for falling behind, they could find students with those factors who would benefit from early advising. Using software developed by the Education Advisory Board, Georgia State identified more than 800 risk factors that correlate to students dropping out early. These factors include registering for the wrong course or receiving a C in a major class. The school hired close to 50 advisers, and those advisers held more than 52,000 student meetings, which helped raise retention rates by four or five percentage points, Renick said. Georgia State has also raised its progression rate, meaning students are accumulating course work on schedule to graduate in four years. By graduating sooner, those students saved $15 million a year in tuition and fees in 2016. “We are delivering personalized attention at scale,” Renick said. “We don’t just give generic advice, but personalized advice. It’s making a big difference in our success numbers.” The initiative is especially helpful to first-generation college students, who are more likely to be undecided about their major and have less help navigating the higher education experience. “Georgia State has closed all achievement gaps. We have black, Latino and low-income students graduating at the same rate,” Renick said. “That’s due to data.” Despite successes such as Georgia State’s, the use of analytics is not without its critics. The Education Policy program at New America has released a paper, The Promise and Perils of Predictive Analytics in Higher Education by Manuela Ekowo and Iris Palmer. The authors say predictive analytics can lead to discriminatory practices as well as privacy concerns over student data. “For colleges that are just learning to use predictive analytics to make decisions, guarding against potential harms can be a struggle,” Ekowo pointed out in the paper’s introduction. “But the stakes are too high to postpone asking these hard questions.” At Georgia State, Renick points out that a student’s race is not associated with an individual student’s data. Plus, advisers are trained to avoid presenting the predictive data as something that will happen. Data is only used as one tool to help advisers identify which students are at risk and what interventions statistically help. “Regardless of the analytic, if you’re employing it in an education system, adults must be bound by a principle that they should only be used to support their learning, never to hurt them,” said the Data Quality Campaign’s Parton. “If you’re using an analytic to predict success, you need to be really transparent about what’s in that analytic and make sure they understand that this is not a conclusion, it’s a way to predict success so we can support you the best way possible.” Tennessee Middle Tennessee State University (MTSU), about 40 miles southeast of Nashville, uses the same software system as Georgia State. In 2014, MTSU began using the statistics tool, which identifies when a student is struggling in a class, for instance, and notifies an adviser who can set up the student with tutoring. In its first year, the data tool helped raise MTSU’s four-year graduation rate to 20 percent from 16 . As they try to complete their degree, many MTSU students face a number of challenges, including falling behind academically because of their poor math and reading skills. They also struggle to fit in classes around their full-time jobs, long commutes and family life. As with Georgia State, others are first-generation college students and simply don’t know how to manage college. The university hired 47 new advisers to help students cope with these challenges. Mary Losey, a sophomore music major, was identified as at-risk for losing her Tennessee HOPE scholarship during her first semester. Academic adviser Brad Baumgardner asked her to come see him. Losey confided that her family was experiencing financial hardship and had some expenses she could not cover. “I showed up in tears. He helped me fill out the paperwork to give me a small cushion,” she said. “We related immediately and talked about clarinet geek stuff. He knows the ins and outs of the College of Liberal Arts completely.” The paperwork resulted in a college microloan that alleviated some of Losey’s financial burden, allowing her to focus on school. Now, Losey is working two jobs and excelling academically. From a list of more than 200 students whom Baumgardner advises, the database at MTSU identifies which ones are on scholarship and what their GPA is. Since a certain GPA is required to keep a scholarship, correlating these two pieces of information helps advisers know who needs a pep talk. “Everybody is big on the analytics, and they are wonderful tools,” Baumgardner said. “But if you don’t have the people component, the data isn’t worth a darn. The analytics make my job easier, but we have to have the boots on the ground, people trained to make personal interventions for these kids.” At MTSU, advising through predictive analytics is a culture. From the president all the way to the cafeteria workers, every staff member is concerned with tracking students. Advising and advanced registration are two factors shown to improve student success. Students receive regular advising reminders via email, complete with a scheduling link. At campus social events, staff reminds students to meet with their advisers. At the dining hall, they ask students if they’ve registered for the fall semester. Analytics helps advisers identify individuals at risk, but it also helps identify trends that can be addressed at the institutional level. Data collected at the state level by the Tennessee Board of Regents showed that about one-third of Tennessee students who arrived on campus didn’t know what they wanted to study, and more than half of those students dropped out entirely before choosing anything, according to Tristan Denley, vice chancellor for academic affairs at the Tennessee Board of Regents. “Choosing to choose later is really an impediment to their success,” Denley said. “People are simply more committed to something when they feel the purpose of what they’re doing.” Once the Tennessee Board of Regents identified this trend, two of the system’s universities — University of Memphis and Austin Peay State University — changed their advising policies. For the last two years, incoming students have been required to meet with an adviser so they can choose a field of focus or a meta major, such as social sciences, business, science or education. To help students choose, these two Tennessee institutions use Degree Compass, a program Denley developed at Austin Peay State University that combines data on a student’s past grades with transcript and enrollment information from thousands of other students. This program then uses a recommendation system analogous to those used by Amazon or Netflix to suggest courses or fields of study based on the data. Kentucky Like Tennessee, Kentucky uses analytics at the state level to influence policy and decision-making at higher education institutions. Kentucky’s Senate Bill 1 that passed in 2009 required the Kentucky Department of Education to create common standards of career and college readiness. The bill increased the need for a comprehensive data warehouse that measured student outcomes throughout their education and professional careers. Kentucky received a federal grant to establish a data warehouse and created the Kentucky Longitudinal Data System that’s housed at the Center for Education and Workforce Statistics (KCEWS). Recently, KCEWS released the 2017 Postsecondary Feedback Report, which provides in-depth data about degrees and the subsequent employment and wages former students earn after they graduate. For example, 65 percent of education majors were still employed after seven years, and they had a median wage of just over $50,000. In comparison, 42 percent of science, technology, engineering and math majors were employed after seven years, but earned about $1,000 more than the education majors. The University of Kentucky includes that information in its academic exploration tools, allowing students to see how majors and potential future wages correlate. The information helps students make more informed decisions about their course of study. Kentucky hopes to promote this kind of evidence-based decision-making. Two years ago, the university compiled data collected through surveys, focus groups and its student record system. Leaders used the data to create a new strategic plan built on four pillars that influence student success, according to Kirsten Turner, associate provost of academic excellence operations. The pillars include academic success, financial stability, health and wellness, and community, and they’re pretty accurate, said Todd Brann, the University of Kentucky’s interim director of analytics, assessment and decision support. “As someone who looks at tables, views and code all day, I’ve yet to find a reason for a student leaving that I can’t categorize in one of those four pillars. Research for research’s sake is important, but I focus on actionable business intelligence and how we’re going to change university policies and procedures to support student success.” Brann noticed that the retention of resident students falls by 40 percent with every $5,000 of unmet financial need. The university is now in the midst of a multiyear effort to shift its scholarship awards from merit based to financial need based. Student privacy is a concern whenever states or institutions gather large amounts of data. KCEWS strips all information that could identify an individual student from the data, including the person’s name, date of birth, Social Security number and the institution they attend. The focus of the research is on the numbers, not the individual students, Brann said. University of Kentucky staff members are asked to sign a statement of responsibility that specifies that they will only use the information to complete a specific job for the student. Across campus, the departments of decision support, advanced analytics and enrollment management collaborate to use data in ways that will positively impact student outcomes. While the data infrastructure is good, Brann’s office constantly reviews data for cleanup and interpretation, as well as uses checks and balances to limit human error. “We’re trying to create a culture of evidence and evidence-based practice,” Turner said. “How do we get the campus as a whole thinking in those terms? We want to usher in a universal culture of evidence.” Information Is Power Educational data is a valuable tool that can be used by policymakers and educators to evaluate success over time. “We are excited to see other states prioritize this type of work,” said Kate Akers, executive director of the Kentucky Center for Education and Workforce Statistics. “In Kentucky, we collect and integrate data so that policymakers, the general public and institutions can make the best decisions possible.” Armed with the knowledge of how to positively influence graduation rates, progression rates, course grades, fields of study and financial aid awards, institutions are poised to better help students succeed. “The benefit and promise of education data writ large is that it’s a tool to inform conversations and actions that lead to student success,” said the Data Quality Campaign’s Parton. “A holistic picture allows adults to use information to tailor instruction to meet the needs of every student.”