Applicants for this year’s freshman class at Ithaca College didn’t have to send their standardized test scores. If they did, the scores were considered, but so were some surprising other factors — how many friends and photos they had on social media, for instance.
The same big data techniques that are transforming other industries are seeping into the college and university admissions process to help predict whether students will succeed and graduate.
“This is the kind of stuff that savvy parents, students and college counselors know about,” said Bruce Poch, dean of admission and executive director of college counseling at the Chadwick School, a private school in southern California, and former dean of admissions at Pomona College.
The point is simple: to increase graduation rates by using big data to identify the kinds of students who experience has proven are most likely to stick around.
Eric Maguire, until recently vice president for enrollment and communication at Ithaca, said using data as a part of the selection process has, in fact, already bumped up the number of students who stay after their freshman year. (Maguire is now vice president and dean of admission and financial aid at Franklin & Marshall College.)
“The question is, how do you recruit a set of students that will be successful at your school?” said Katharine Frase, vice president and chief technology officer for IBM’s unit focused on working with the public sector, which produced the data analysis program used at Ithaca.
“When a student doesn’t complete a degree, it is disruptive for everybody,” Frase said. “The student has incurred debt and the school is left with a hole in that class.”
Ithaca has been quietly collecting student social media data since 2007, when it launched a Facebook-like website for applicants called IC PEERS. The website gives applicants a chance to connect with Ithaca faculty as well as each other.
Using an IBM statistical analysis program, Yuko Mulugetta, Ithaca’s director of enrollment planning and self-styled “in-house statistician,” studied data collected from IC PEERS to see which students employing what behaviors were most likely to enroll and stay at Ithaca — how many photos they uploaded to their profiles, for instance, and how many IC PEERS friends they made.
The idea is to learn how interested a candidate is in the college, Ithaca officials said.
The Big Brother approach to using data in this way is not without its critics.
“I really didn’t think about how the school might use that information, but I guess I was already through enough of the college admissions process that I felt like all of my information was already with the schools,” said Kelly Meehan, a rising sophomore music major at Ithaca from Saratoga Springs, New York.
But Meehan worries that using IC PEERS data could put students at a disadvantage who don’t have regular access to the Internet or aren’t inclined to use social media.
Using new forms of data collection and analysis is only likely to increase, however, as universities and colleges are judged by everyone from regulators to bond-rating agencies on their ability to attract students and shepherd them to graduation.
“There’s an economic side to this that’s unnerving,” said Poch, reflecting on the time he ran admissions at Pomona. “I remember sitting down with bond raters from Standard & Poor’s and Moody’s and them asking, ‘How many applications do you get and what is your yield?’” — the proportion of accepted students who enroll, a measure of demand.
“The more demand, the higher our bond rating, and the lower our interest rates,” Poch said. “So a higher yield meant saving millions of dollars a year in interest payments.”
David Wright, chief data officer at Wichita State University, said his colleagues and counterparts talk a lot about how to get the highest yield at the lowest cost. At his school, Wright said, all potential students are assigned a probability, from zero to 100 percent, of whether they’ll enroll, based on factors such as sex, race, ethnicity, test scores, high school grades and whether they’re the first in their families to go to college. The university then focuses its recruiting dollars on reaching the ones most likely to attend.
Like many other colleges and universities, Wichita State also uses data to predict the likelihood of academic failure among enrolled students. It gives that information to academic advisers who suggest changes to a student’s schedule long before classes start in the fall, in an effort to increase the student’s likelihood of success and, ultimately, graduation, Wright said.
At Sarah Lawrence College, Tom Blum, vice president of administration, acknowledged that its use of big data is designed to increase yield rates. He added, however, that “we do want to minimize instances where we’ve admitted a student that probably wasn’t the best fit for us. How interested an applicant was is heavily correlated with the student who is going to be a good fit and stay on past the first year.”
He said: “This isn’t rocket science. Those students get our open curriculum, they get that we only have one major, they get that there’s a focus on independent study. And the students that understand those things and are excited about those things are the ones who are most likely to stay.”
“We are small enough to get away with having conversations about each applicant, we don’t use numerical formulas,” said Blum. “But we do use all of the data to cross-check the human process of building a class that is diverse and likely to show up and stay around.”