A couple of weeks ago, the Labour Market Information Council released a whack of material, produced by Ross Finnie and his Education Policy Research Initiative, on graduate labour market outcomes using Statistics Canada’s new Education and Labour Market Longitudinal Platform (ELMLP).  The material included a paper, a couple of briefs on earnings by gender and international students, and a nifty online widget that lets you play with the data yourself. The data contains a few surprises, though nothing that radically shakes up much of what we knew about graduate labour markets. Nevertheless, the work represents something of a technical breakthrough in the study of outcomes and points to a bunch of interesting new research questions.

First, what is the ELMLP?  Well, as I described back here, it’s basically an administrative data linkage between the Post-Secondary Student Information System (which has all sorts of information on students and graduates) and what’s called T1FF, which is every tax return filed in the country.  This allows us to look at graduate outcomes more or less continuously, rather than every five or so years when Statscan gets around to doing a “National Graduates Survey”.

(Statscan is still doing National Graduates Surveys, by the way.  It’s not entirely clear why, given that the only thing anyone is going to use it for – now that the ELMLP exists and Statscan eliminated the value of the NGS as a time series by messing with the time interval between graduation and the survey – is information on student debt.  And that information is available through a few access to information requests. I’d suggest cutting it so the agency might produce useful data on something like the social origins of the student body that could help monitor access, or even just produce usable enrolment statistics in less than 40 months.  But let’s face it, this is Statscan and they’d probably just blow the money on another Mushroom Census, so I suppose we should just let sleeping dogs lie. And OK, technically, the decision to kill a graduates’ survey and replace it with something useful with respect to measuring access is an ESDC decision, not a Statscan decision.  Still.)

For the most part, the new data is used to permit more frequent observations of variables that were already tracked.  So, there’s not much in here we didn’t know about the gap in salaries between credentials and fields of study; it’s just that we can see it every year instead of having to wait and it’s much easier to follow individual cohorts in a longitudinal fashion.  Further:

• We’ve known for a long time how gaps in male and female pay grew over time, but this data shows how this happens, at every level of education;
• We’ve known for a long time about the gap between say, Engineering and Humanities graduates, but now we can see how the gap in starting salaries between the two widened a bit in 2011 and 2012 (though it has stayed relatively stable since then). 
•  We’ve known for a long time how post-study salaries generally grow fairly rapidly in the first five years after graduation, but this data fills things with in with a lot more detail. 

If there was one set of salary numbers that surprised me, it was those for Science PhDs, whose results were godawful (on average, it takes them three years after graduation to hit $60,000)

That said, there is one big shock in the data and it is the stonking huge gap in salaries between Canadian students and international students who choose to stay in Canada (free idea for a future research project: an important possible use of the ELMLP is figuring out how many and which kinds of students stay in Canada).  Depending on the level of study, the gap in salaries is between 20-40% at graduation, favouring Canadian students. After that, the gap shrinks over time but is still distinctly noticeable five years after graduation.  Canada has long known about the gap between native-born and immigrants, holding education constant, but has attributed this to the idea that “well, they were mostly educated abroad, and so their skills aren’t necessarily quite as relevant to our labour market – if only they were educated here, that gap would close in a snap!”  This data should refute this: even when foreigners study here before proceeding to the labour market, our terribly opaque labour markets still bias outcomes enormously towards those with established family connections (another free idea for a research project: when NGS comes out do this same analysis, but use the survey’s ethnicity question as a filter).

If there is a problem with this kind of research, it’s the same one that affects all post-graduate pathways work – namely, that these data sets exclude the incomes of those students who take further education, which at the Bachelor’s Degree and college Diploma levels include over a third of all graduates.  For instance, there is a whole phenomenon of university grads going to take programs at colleges (and vice-versa) which goes unexamined here.  The long-term performance of humanities and social science graduates is almost certainly under-played because the existing methodology does not permit an analysis of what happens to graduates who go on to business or law school (something similar plays out with science students and health careers, I think).  The issue is that you need a fairly long data series to make these kinds of analyses work and I gather there are really only five years worth of data in the ELMLP at present.  I hope that in a few years time Finnie and his team will be able to do a ten-year analysis, which includes some of this information.  Because as it is, I think we’re truncating the analysis to over-represent the weaker students in many fields of study.

 (Bon long week-end.  See you back here on Tuesday.)