Every September, Statistics Canada publishes data on “average tuition fees”. It’s a standard date on the back-to-school media calendar, where everyone gets to freak out about the cost of education. And we all take it for granted that the data StatsCan publishes is “true”. But there are some… subtleties… to the data that are worth pointing out.
Statistics Canada collects data on tuition from individual institutions through a survey called the Tuition and Living Accommodation Survey (TLAC). For each field of study at each institution, TLAC asks for “lower” and “upper” fees separately for Canadian and foreign students, for both graduate and undergraduate students. Now, in provinces where the “upper” and “lower” figure are the same (eg. Newfoundland), it’s pretty simple to translate lower/upper to “average”. In Quebec and Nova Scotia, where “upper” and “lower” are functionally equivalent to “in-province” and “out-of-province”, averages can be worked out simply by cross-referencing to PSIS enrolment data, and weighting the numbers according to place of student origin. Everywhere else, it’s a total mess. In Ontario, significant variation between “upper” and “lower” numbers are the norm, even inside the institution (for instance, with different tuition levels for different years of study). Somehow, StatsCan uses some kind of enrolment weighting to produce an average, but how the weights are derived is a mystery. Finally, in a couple of provinces where there are differences between the “lower” and “upper” figures, StatsCan chooses to use the “lower” figure as an average. (No, I have absolutely no idea why).
But the tuition data is squeaky clean compared to the mess that is StatsCan’s data on ancillary fees. Institutions fill in the ancillary fee part of the questionnaire every year, but usually without much reference to what was reported the year before. Since StatsCan doesn’t have the staff to thoroughly check the information, institutional figures swing pretty wildly up and down from one year to the next, even though everyone knows perfectly well ancillary fees only ever go in one direction.
Another complication is that “average” is a central tendency – it is affected not just by posted prices, but also by year-to-year shifts in enrolments. As students switch from cheaper to more expensive programs (e.g. out of humanities and into professional programs), average tuition rises. As student populations grow more quickly in the more expensive provinces (e.g. Ontario) than in cheaper ones (e.g. Quebec, Newfoundland), then again average tuitions rise – even if all fees stayed exactly the same. Both of these things are in fact happening, and are small but noticeable contributors to the “higher tuition” phenomenon.
A final complicating factor: the data on tuition and the data on enrolment by which it’s weighted come from completely different years. Tuition is up-to-the-minute: the 2014-15 data will be from the summer of 2014; the enrolment data by which it is weighted will be 2012-3. And, to make things even weirder, when StatsCan presents the ’14-15 data next year as a baseline against which to measure the ’15-16 data, it will be on the basis of revised figures weighted by an entirely different year’s enrolment data (2013-4).
In short, using SatsCan tuition data is a lot like eating sausages: they’re easier to digest if you don’t know how they’re made.