Every once in awhile, some policy-maker or journalist gets their knickers in a twist about dropout rates. And whenever that happens, people start looking for data. Which, in this case, basically doesn’t exist.
Institutions have their own non-completion data, but since lots of people switch institutions for one reason or another a non-complete doesn’t equal a “dropout”. Our national unit-record system – Statistics Canada’s Post-Secondary Student Information System – is supposed to be able to solve this precise problem, but for reasons too boring to relate is unable to do so, except in Atlantic Canada (see my colleague Ross Finnie’s report on this, here). Quebec, Alberta, and BC all have provincial/regional data systems in place to look at this, but to my knowledge none have done so.
There is, however, one data source that allows people to track dropouts over time; namely, the Labour Force Survey (LFS). If you simply take all Canadians between the ages of 25-34 who are: a) not in school; and, b) indicated that they have “some postsecondary education”, and divide them by the total number of Canadians aged 25-34 who are: a) not in school; and, b) either have a degree, or indicate that they have some postsecondary, then what you get is an estimate of the percentage of 25-34 year-olds who are dropouts from the postsecondary system.
This method obviously doesn’t provide precise year-to-year estimates of dropouts; some of the 34 year-olds in the numerator might have dropped-out as many as 15 years previously, for instance. And due to the nature of the data source, one can’t derive separate rates for universities and colleges. But it does provide a sense of general trends over time in non-completion.
So, what does the evidence say? Check out the figure below.
Figure 1: System-wide Non-completion Rates, 1980 to 2010
Source: Labour Force Survey
It turns out that the percentage of former postsecondary students in the 25-34 year old population who dropped-out of school has declined steadily over the past three decades, reaching an all-time low of 9.4% in 2010. This is almost a 50% decline from the 1990 value of 17.1%. While a definitional change means that pre-1990 data is not directly comparable to post-1990 data (hence the permanent downward shift indicated by a dotted line in the Figure above), it appears that the decline began sometime around 1986 – prior to that date, the dropout rate hovered fairly steadily at just under 25%.
A wise man once said that if a result is big enough, even a really bad experiment will pick-up that result. It’s the same thing here: using LFS to look at dropouts is, at best, a third – or fourth – best option. But the result is unequivocal: dropouts have been going down for nearly three decades.