If Canadian universities were private corporations, their customer turnover rates would be ruinous. Up to 20 per cent of students quit university, never to return to any postsecondary program, while 20 to 50 per cent drop out of the program they had initially chosen.
Of course, students are not customers, and universities are not businesses. Yet in many ways, they face similar challenges. Like businesses, universities invest in marketing their wares and retaining students. Like customers, students sample the product, stay loyal to a brand, or look elsewhere. Deciding to drop out, however, has lifelong consequences – for society and for students.
"For students, leaving is a failure," said Harvey Weingarten, the executive director of the Higher Education Quality Council of Ontario, an Ontario-government-funded think tank that has researched many programs aimed at helping students to graduate. "There is a loss of confidence, there is a psychological cost of failure."
And there are tangible costs, too. University and college graduates have higher lifetime earnings, are in better health and have closer ties to their communities than high-school grads.
Until recently, however, universities appeared content to live with the situation. As a Globe and Mail analysis reveals, graduation rates across the country have not improved over the past decade. But now, postsecondary institutions are turning to big data analysis to help them find the students most at risk of dropping out. Financial pressures are forcing them to take action.
After a decade's explosive growth in enrolment – matched by increases in overall public funding – universities will soon face a harsh reversal of fortune. In all but the most optimistic scenarios, Canada's population of 17-to-24-year-olds will decline over the next 10 years. This demographic shift will be matched by smaller government grants. And if provincial governments base a portion of those grants on how well universities perform, there could be further erosion of the schools' revenues.
So, every student who can be helped to the finish line will help to pull Canadian universities back from the brink of a financial reckoning.
Key to increasing graduation rates is "predictive analytics," used in fields from health care to newspapers to financial services. Drawing on demographic and behavioural data, predictive analytics tries to forecast future behaviour. Eventually, as in health care, the hope is to find those most at risk, and deliver individualized interventions before the person drops out.
With predictive analytics, "instead of reaching out to 4,000 students in a [high-risk] category, I am going to reach out to the 14 who have the highest risk factors," said Mark Milliron, co-founder of Civitas Learning Inc., one of the largest companies offering such services in the United States. "That is optimizing scarce resources."
It is clear that Canadian universities must do something to move the needle on graduation. The Globe's analysis reveals that while the rates increase and decrease from year to year, the percentage of students who graduate from various programs shows little change over the long term.
To gather the numbers, The Globe collected publicly available provincial graduation rates and requested them from other individual institutions.
The responses show Canada has no simple answer to the question, "How many students who start a university program graduate?"
Ontario and British Columbia, home to more than two-thirds of Canada's undergraduate student population, calculate and publicize graduation rates for individual universities. The Maritime provinces – Nova Scotia, New Brunswick and Prince Edward Island – make public only the graduation rates of recent years.
The Globe also requested information from individual institutions across the Prairies and Quebec, but the responses covered too few years to be useful.
What we do know is that approximately three-quarters of students in B.C. and Ontario who are studying social sciences graduate from that program. In computer science, on the other hand, the odds of completing a degree can be the same as flipping a coin, and rise no higher than 68 per cent on average.
Program graduation rates don't reflect students who drop out of one program but finish a degree at another institution. Once these "temporary leavers" are counted, the overall graduation rate is estimated to increase to between 80 and 90 per cent.
Universities must work to track, understand and decrease their dropout rates, Dr. Weingarten said. His organization is currently working with six universities and colleges to test what makes underrepresented groups drop out and what kinds of interventions encourage them to stay.
"You accepted a student into your institution because you believed they could succeed, they would grow, thrive and develop. When it doesn't work, you have an obligation to figure out what went wrong here," he said.
Who drops out?
Experts know which students have the odds stacked against them.
A young man who enrolls in university but whose parents did not attend either college or university and who squeaked through high school with less-than-stellar marks is at high risk. If he has taken time off before enrolling, his risk increases yet again, partly because he is less likely to get involved in extracurricular activities. Income matters, but it's unclear how much. If he has children? Good luck with that.
On the other hand, the best predictor of whether a student will graduate is her high-school marks – a reflection of how well prepared she is for postsecondary studies.
But such demographically based risk assessments are not perfect. Women leave in smaller numbers, but may decide to drop out even when they have reasonably good marks. High-achieving students who show they are bored are at risk of quitting.
Identifying and intervening in those marginal cases could quickly improve graduation rates.
That's the pitch Civitas Learning makes. It's far from the only company in the predictive analytics space: IBM's Watson Analytics, Tableau Software Inc. and Microsoft Corp. are all pitching their services to the university market, including to Canadian institutions.
The way universities have been collecting data on graduation is like doing an autopsy, Mr. Milliron said. He'd like to change that.
"Instead of looking at the students who were there before, [we] shift to getting the data during the operation, where it saves the patient, or even getting to the patient before they end up on the table," he said.
Civitas tracks how often students log onto virtual discussion boards, their extracurricular activities, how often and how quickly they drop courses, and compares this behaviour with their background and grade-point average
In the United States, they have 275 educational clients, but have not yet expanded to Canada. Canadian universities are talking to various data analysis companies, but for now, many are developing their own tools.
The University of British Columbia (UBC), for example, is running a pilot project to try to connect the academic data the school has about an incoming student with their participation in extracurricular programs, their university grades, and even their post-graduation employment.
Already, the university implemented preorientation sessions for international students after they noticed that, as is the case at many Canadian universities, students in this group were less likely to graduate than domestic students.
"We had $25-million of lost revenue because there was a 10-point [percentage] difference," said Gage Averill, the university's dean of arts. "It was an eye opener to look at all the frenetic energy we were putting into developing revenues elsewhere, when in fact $25-million was sitting on the table if we could help these students."
The gap between the graduation rates of international and domestic students has now narrowed and even reversed in some areas.
As with all experiments, success is far from certain. Universities have to adopt the tech mantra: Some of the time, an experiment will fail.
Susan McCahan, the vice-provost of innovations in undergraduate education at the University of Toronto, remembers an experiment she ran several years ago when she was a senior administrator in the faculty of engineering.
The faculty had previously allowed students who scored between 45 and 50 in their first term to continue taking classes with their cohort as long as they retook the courses they had failed.
But when she looked at their graduation rates, Dr. McCahan discovered the vast majority in this group were failing to finish the program.
"We said 'Let's not just keep moving these students along, when there is so little chance that they are going to be successful,'" she said.
So students who did poorly in first term were told they would have to repeat first year, but had the option of enrolling in a new program called "Refresh." It included academic, as well as professional and personal development courses.
But when the department compared the first-year success rate of students who had participated in Refresh to those who chose to try first year again without taking the enrichment course, there was not much difference.
"For some students, it was life-changing," Dr. McCahan. "But the numbers were meh."
Now, Dr. McCahan is leading a university-wide effort to link all the school's data to generate ideas about what kinds of indicators, alone and in combination, best predict graduation rates.
A long experiment
The promise of short-term, data-driven interventions is that regardless of students' backgrounds, universities can do something – right now – to help them graduate.
While no school can reach back into the past to change history, if researchers start early enough in a student's life, they can possibly change the future.
That is exactly what one of Canada's longest-running experiments in postsecondary enrolment and graduation has attempted to do. Back in 2004, the Future to Discover project picked thousands of New Brunswick high-school students, all from lower-income families, and randomly assigned them to three types of university-preparation programs.
One group took an academic skills and career development class every two months for four years of high school; another was guaranteed an $8,000 contribution toward the cost of postsecondary education. The third group had the benefit of both.
Twelve years after he first stepped into one of those after-school classes, Matt Wheaton says they made a difference.
"I ended up at university. Whether that had anything to do with the program, I don't know. But it opened my eyes to a realm of possibility beyond high school."
Mr. Wheaton now works as an admissions counsellor at Crandall University, a Christian liberal arts school in Moncton, N.B., where he also earned his bachelor's degree.
"I meet a lot of kids like me who don't think they are smart enough, or good enough. It's become my job to do the same kind of thing, to say, 'There are possibilities beyond what you are thinking,'" he said.
The study's final results – released this past summer – reflect Mr. Wheaton's personal experience.
Participating in the workshops led more students whose parents had not attended college or university to enroll in university. It increased college graduation rates, although – contrary to Mr. Wheaton's experience – the workshops had little impact on graduation rates from university.
Still, "the results are encouraging," said Reuben Ford, a Vancouver-based research director at the Social Research and Demonstration Corp., the national social policy research centre that ran the study. "It means that these kinds of interventions that try to help students who would not go to postsecondary can be effective.
"It doesn't mean that they are not going to face hurdles – but they don't drop out any quicker than their non-disadvantaged peers," he said.
What makes Future to Discover different from today's data-driven interventions is that it borrowed from the most successful early-childhood-education programs, involving the students' families as well.
Mr. Wheaton's mom attended almost all of the classes with him.
"It was a parent-and-kid thing. All of us had our parents there; it was an odd dynamic," Mr. Wheaton recalled.
"But sometimes high schoolers don't tell their parents what happens at school, so it was a good call."
In the United States, approximately 30 states now link a portion of their education grants to college outcomes, following a model introduced in Tennessee years ago, hoping that universities would be prodded to move the needle on graduation quickly and efficiently.
But because one of the key predictors of graduation rates is admittance marks from high schools, linking public money to graduation rates can have the perverse effect of channelling funds to selective universities, said Alan Harrison, the provost of Queen's University. Queen's attracts high-achieving high-school students and has the highest graduation rate in Ontario.
"I always tell the [Ontario government to] be careful about the changes you're making. There is a lot to be said about the status quo," he said. The "worry is that if we start to reward retention, we will reward who gets admitted."
Others say that if they are designed carefully, incentives to improve graduation rates can be a nudge in the right direction.
"If you look at the funding formula that Tennessee uses, they give the institution more money per student if that student comes from a disadvantaged family," Dr. Weingarten said.
"You could say for this university – 'X' – we expect a 90-per-cent graduation rate because we know the kind of students they take. But for university 'Y', we will reward them equally if they have a 75-per-cent graduation rate, because they are taking a more difficult cohort." Canadian provinces are already thinking about how to encourage higher graduation rates from their universities.
British Columbia ties some of its funding to how many students graduate in occupations in demand in the labour market. Within a few years, Ontario will also link some of its money to performance standards. Most likely, that will include graduation rates.
No matter how thinly and carefully the data are sliced, no university should forget that education can be transformative in ways that predictive analytics can't foresee.
"Our students are more than data sets," Dr. Averill says. "The really transformative things that happen to students when we are moving them from a somewhat unformed high-school kid who doesn't know what he or she wants, to someone who might be starting an NGO in Chad – that is not predictable. … We need to understand that even as we are trying to do good work to help students along."
With files from reporter Mahnoor Yawar