When a C Isn’t Good Enough

For freshmen taking writing composition at the University of Arizona, receiving a C at the end of the semester may no longer warrant a sigh of relief.

Instead, some of them may have to repeat the class.

Two years ago, Arizona hired Civitas, an education technology company that uses predictive analytics, to track student behavior in an effort to boost student graduation rates. One finding jumped out: students' performance in commonly required courses was linked to whether they would graduate or drop out.

For instance, students at the university who earned an A or a B in an introductory English composition and rhetoric course had a 67 percent chance of graduating, a figure calculated by predictive models of actual graduation numbers. But if they received a C in the required course, students only had a 48 percent chance to graduate -- a difference of nearly 20 percentage points.

The university had considered the course to be low risk. Most students in English composition (81 percent) persist into the next semester. But the data convinced university officials that success in the course was vital in predicting whether students were likely to graduate.

After receiving the findings over the summer, university officials decided changes had to be made. They are currently discussing whether and how curricular policy should be adjusted in light of the data, although those conversations have just begun, according to Angela Baldasare, assistant provost of institutional research at Arizona.

The possible policy changes would be made by individual schools at the university, and those possibilities range from requiring that C students take a writing competency test to providing resources for students who didn’t obtain the top two grades. Another option individual schools could adopt would be considered more dramatic: students who received a C would be required to repeat the course before taking upper-level classes.

Using data to inform policies and practices at universities is nothing new. Predictive analytics have been commonly used in higher education for the past five to seven years, said Amelia Parnell, vice president for research and policy with NASPA, Student Affairs Administrators in Higher Education.

Civitas works with scores of universities and colleges, including community colleges that have focused professional programs, and it found similar patterns beyond the University of Arizona. At El Paso Community College, for example, nursing students were much more likely to complete the program if they received an A in a foundational anatomy and physiology course -- in fact, students who receive a B or a C had a scant 35 percent likelihood of graduating.

If colleges were to start requiring students who earn Cs to repeat courses, that would be symbolic of a conceptual shift, experts said, switching the priority from students simply passing a course to students who master the material.

“What does it mean to give someone a D or a low C?” said Thomas Bailey, director of the Community College Research Center at Columbia University's Teachers College. “If the average grade in the past was a C, then it’s different than if it was a B or B-plus. You’re changing the definition of what it means to pass a course.”

One reason for this shift is grade inflation, Bailey said. Indeed, only 8 percent of students who takes the writing composition course at Arizona are graded with a D, F or withdrawal.

Vincent Del Casino, the university's vice provost for digital learning and student engagement and associate vice president for student affairs and enrollment management, said discussions surrounding the introductory writing course are nothing new.

“Our journalism school requires a C in a basic math course to be admitted,” he said. “Colleges are looking at whether students have certain grades in certain foundational classes. That’s normal.”

What is new, according to Del Casino, is that the data are being used at an institutional level to create various smaller changes unique to each college within the university.

“Nobody has dug into those relationships before on an institutional level,” Del Casino said. “If at an institutional level, we know that graduation rates start at foundational courses -- backed with data -- that’s exciting to us.”

At Arizona, 61 percent of students graduate in six years. That’s less than two-thirds of students who begin as freshmen, but it’s better than the national rate for four-year institutions. Nationally, 53 percent of students graduate within six years, according to the National Student Clearinghouse.

More Proactive, Less Reactive

“The most critical piece of predictive analytics is to provide intervention,” said Parnell.

That means contacting the student during the semester instead of waiting until end-of-semester grades. And the earlier the better.

For example, Parnell encouraged universities to give the student’s adviser data about how many times she logs into her course management system and grades; to instruct the adviser to reach out to the student with resources; or to create an automated early alert system, which reaches out to students through text messaging or email when they’ve failed to turn in an assignment.

Andrew Koch, executive vice president and chief academic leadership and innovation officer at the John N. Gardner Institute, likened predictive analytics in academia to a game of bowling.

“I’m going to predict who’s going to bowl a 250 or better based on analyzing what happened in the first three frames, I’m going to have a very sound prediction,” Koch said.

But a prediction is just that, and nothing more. To actually encourage students to succeed in the class -- or if the goal is to bowl a 250 or better instead of just predicting the score -- that’s another story.

“Then I’m going to want to intervene in the first three frames to make sure things are going well, that the technique is right, etc.,” Koch said.

At Arizona, that advice hasn’t fallen on deaf ears.

This semester, professors in some departments at the university are calculating and posting midterm grades for students. The hope is that if students stay informed about their grades, they’ll be more likely to take steps to raise them.

“The important thing for us is that we can now use our data to more quickly identify students who aren’t doing well in composition and get them the support they need to improve their writing skills and their grade in the class,” said Baldasare. “We know that improving their grade in composition will serve them well regardless of the requirements in their particular major.”

Helping or Creating Barriers?

The use of predictive modeling isn’t a guaranteed solution. Although experts agree that it’s useful for institutions to have the data, they also brought up concerns about how the information is used.

“One question I have for universities: Is your policy about creating more barriers, or creating resources for students?” said Amber Garrison Duncan, strategy director at the Lumina Foundation

Since Arizona’s policies are currently being molded, it’s difficult to find data that can answer her question. But when it comes to requiring students to repeat a course even if they earned a C, federal data suggests that underrepresented and minority students may be the ones who are most likely to be affected.

“But if your policies are saying that you must get a B before you proceed onward, well, what do we know from analyzing the data?” Koch said. “It’s that first-generation, low-income, historically underrepresented students are the least likely to do well in these gateway courses.”

Indeed, the university's graduation rates for students from minority groups are lower than the rates for white students.

In the 2008 freshman cohort at Arizona, 62 percent of white students graduated within six years. In contrast, 54 percent of Latino students and 45 percent of black students graduated within the same time frame.

In addition, low-income and minority group students are more likely to be first-generation college students, which means they tend to have fewer resources to help them complete classes with high marks. Requiring these students to repeat a course with a C could inadvertently create one more hurdle for the school’s most vulnerable students.

Who Makes the Decisions?

Another concern: giving administrators greater discretion in curricular policy while simultaneously giving faculty members less.

“When you start mapping out the curriculum, you’re talking about something that gets into academic design. Traditionally you leave it to an academic department to map out curriculum and set standards,” said Phil Hill, a market analyst who studies technology in higher education and co-publisher of the “e-Literate” blog.

In deciding how to use the data, administrators at Arizona are not directly in discussion with faculty. Instead, they've invited deans and associate deans from each college to review the data; they are invited to "bring whatever key staff they deem appropriate" to the meetings, Baldasare said.

"It’s a first step in what we expect to become an expanded conversation with colleges about how we can use predictive analytics to support student success," she said.

Civitas, too, sees potential in the data. 

"The University of Arizona is taking the first steps in thinking through the appropriate, responsible use of these insights to remove previously unknown barriers to student success and graduation. These specific course findings will empower faculty and advisers to identify students most in need and provide the timely support services to those students during the term,” Laura Malcolm, senior vice president of outcomes and strategy for Civitas, said in a written statement. (Note: This paragraph has been added to a previous version of the article to include a comment from Civitas.)

Even so, an expert on writing courses is worried about what a possible curriculum shift might trigger.

A policy requiring a C in a gateway class could “put an inordinate focus on grade performance, which puts a lot of pressure on faculty,” said Doug Hesse, executive director of writing at the University of Denver and president of the National Council of Teachers of English. “It’s often the case in writing faculty that there are adjunct and grad assistants who don’t have protections of tenure. It’s pretty tempting to say, ‘Whatever, I’ll give the A or the B.’”

This debate, one with multiple strands, implications and consequences, is one with which everyone in higher education is familiar. They've got the data. How do they use it?