Are the returns to university education falling?

[Updated to fix coding errors in some of my graphs.]

Prices are determined by supply and demand. The supply of university-educated Canadians, relative to the supply of college or high school graduates, is increasing. University participation rates have been climbing for decades (Source: CAUT):

Canada's points based immigration system, which prioritizes those with valuable skills, further increases the supply of highly educated Canadians. Unless the demand for university graduates rises along with the supply, ECON 1000 reasoning suggests that rising education levels will, eventually, erode educated workers' earnings advantage.

There are a number of excellent papers on the returns to education in Canada, and how these have evolved over time over time, such as Burbidge, Magee and Robb, Boudarbat, Lemieux and Riddell, 2010,  Ren and Shannon (2017, gated), and Bourbeau, Lefebvre and Merrigan.

The most up-to-date analysis I could find was Foley and Green, 2015, which uses Labour Force Survey data to calculate the earnings gap between more educated Canadians and high school graduates up to 2013. They found that, particularly for younger men, the wage gap between more educated workers and high school graduates has gradually narrowed since the year 2000. The figure below is taken from their paper, and summarizes some of their findings. The thick solid lines, say, show the difference between the average log wages of a bachelor's degree holder and the average log wages of a high school graduate. That difference is rising, for men, until 2000, and then it starts to fall. For women the difference in log wages between bachelor's degree holders and high school graduates is larger, and more stable.

The downward trend in male university graduates' wage advantage from the late 1990s to 2013 is clear, but it is not large - and it only takes the wage premium for bachelor's degree holders back to where it was in the 1980s. This trend does not, for me, provide a satisfactory explanation why pundits are alleging that campus life isn't what it used to be, and that the jobs available to university graduates are not exactly what they had in mind. While the trends documented by Foley and Green are consistent with the concerns expressed by university-skeptics, they just aren't that large or dramatic. [Update] Moreover, when I take 2018 Labour Force Survey data, and attempt to replicate and extend Foley and Green's findings, I no longer find that clear downwards trend. The average gap in log wages between 25-34 year old bachelor's degree holders and high school graduates was 0.24 in January, 1998 - and almost exactly the same in January, 2018. 

Yet even if education remains a good investment, it is also a very risky one. The figure below shows kernel density distributions of hourly wages for male high school graduates and bachelor's degree holders in 1998 and 2018. These are simply descriptions of raw, underlying, hourly wage data, comparable to the data Foley and Green used, weighted to take account of the Labour Force Survey's sampling design and, for the 1998 data, adjusted for inflation using the Consumer Price Index. (Note that self-employed individuals are excluded from the distribution.) 

[Updated: the above image is correct. The small image below is the original one I posted, which actually compares people with some post-secondary education to graduate degree holders]

 

The kernel density distributions make it immediately apparent that wage premia, which focus on differences in average wages, miss much of the story. There is much greater wage inequality among bachelor's degree holders now than there was in 1998, and also among high school graduates. While the top-earning bachelor's degree holders are doing just fine, the lowest-earning degree holders are seeing their wage advantage over high school graduates shrink. [Updated] The key questions for anyone thinking about pursuing post-secondary education are "where would I be on that high school wage distribution?" and "where would I be on the bachelor's degree wage distribution?" This determines, at an individual level, the returns to education - and it's something economists don't know a great deal about.

The picture for women is slightly different. Women with high school degrees now seem to be somewhat more concentrated in low wage jobs than they were in 1998, while the wage distribution for women with bachelor's degrees appears to have shifted to the right, as well as becoming more spread out. The declining wage premium for bachelor's degree holders seems, to some extent, to be a guy thing. [Updated] The female wage distribution, especially for high school graduates, is bounded at the bottom by the minimum wage. Increases in the minimum wage, as happened in Ontario in January, 2018, would be expected to shift the wage distribution upwards.  This could explain why the estimated average gap in log wages between high school graduates and bachelor's degree holders actually falls between 1998 and 2018.

[Updated, the above picture is correct; the small picture below is the one originally posted, which actually shows people with some post secondary v. people with a post-graduate degree]

It has always been the case that some university degree holders don't do any better than high school graduates. But two things have changed. First, there are more bachelor's degree holders now than in, say, 1998, so that bottom chunk of the bachelor's degree wage distribution represents more people. Second, university tuition has increased, so a given wage now represents a lower return on investment. Bachelor's degree holders in the bottom of the wage distribution are completely justified in asking, "what went wrong?"

 Here economists can only give partial answers. There is very good work by Finnie and Frenette (2003), Ostrovsky and Frenette (2014), and Frenette and Frank (2016) on returns to field of study. This literature, however, tells a story about how not why. That computer science or commerce graduates earn more than psychology or fine arts graduates explains how variation in the earnings distribution is generated. But it does not explain why people choose low-paying majors over high-paying ones. Is it because they are unable to gain admission into a program of study with better job prospects? Are they ill informed? Or do they enjoy the non-pecuniary benefits of studying psychology or fine arts? The field of study literature also does not explain why employers choose to hire who they do. For example, why might an employer prefer a commerce major when marketing is, to a large extent, just applied social psychology? How much of employers' willingness to pay more for, say, engineering graduates reflects those graduates' gender, rather than (or as well as) their specific knowledge? 

The literature is also largely silent on what aspects of university education create value. Here it is interesting to look at a couple more diagrams, which add the wage distributions for college graduates onto those of high school and university graduates, for men:

[Updated: original diagram is small picture.]

And for women:

 Universities typically have stricter entrance requirements than colleges, and the courses take much longer to complete.  Yet many men graduating from university earn less than the typical college graduate. 

What is it that universities do that generates earnings premiums? Yes, I know, some mix of signalling ability and building human capital. But what mix? And what is it that universities do that generates that human capital? Does listening to me lecture magically confer upon my students some earnings advantage? Or do students learn more from their peers than their profs? Or perhaps much of the value of university comes from building social capital, as argued by Mara, Davis and Schmidt?

Please share your thoughts in the comments.

Comments

[Majromax]

If the returns to university come from signaling, peer-learning, or social capital, then an increasing share of university enrollment might erode the institution's returns. For all of these causes, the advantage comes mostly from a student being part of a semi-elite group, distinguished ostensibly by ability but practically by class as well.

But with an increasing participation rate, university education becomes more like Grade 13-17 than something qualitatively different. This trend might even be enhanced by universities themselves, if they define their missions as primarily teaching and focus on hiring teaching faculty on a contract basis rather than tenure-track professors that can develop a durable and distinct university culture.

In the meantime, some individuals entering the workforce after high school might be doing so because of particularly attractive opportunities (like a family business). That could plausibly give us the nearly-bimodal distributions seen in the third figure, where the high-school workforce consists of people who didn't have the ability/resources to go to university plus those for whom working was more attractive despite the opportunity to attend university.

[Coady]

I agree that focusing on changes in average earnings miss a lot of the information. But I'm not sure the density plots are as worrisome as they seem. There is more wage inequality among BA holders today than there was in the past. However, the change in the distribution seems to come from moving mass from the middle of the distribution to the upper tail of the distribution.

--The red (1998) and orange (2018) lines are right on top of each other at up to around $20 per hour. (The probability of being a lower wage BA holder is about the same in 1997 and 2018.)

--The orange line from 2018 has fewer people earning $20 to $40 per hour. (The probability of being a middle wage BA holder is lower in 2018.)

--But the 2018 distribution has more people earning $40 to $80 per hour. (The probability of being a high wage BA holder is higher in 2018.)

If university degrees are lottery tickets, then the 2018 lottery is riskier than the 1997 lottery. But the 2018 lottery is riskier because it has more upside. Generally, more risk seems bad. But...the pattern in the graph seems like the good kind of increase in risk to me. Perhaps it's misleading to think about wage returns in isolation? Tuition costs have changed, people have student debt, some of the biggest earners are the winners in non-wage space (self-employed or entrepreneurs).

[Frances Woolley]

Majromax: "In the meantime, some individuals entering the workforce after high school might be doing so because of particularly attractive opportunities (like a family business)."

I don't think we know nearly enough about the role of family background in general, and social class in particular, in the returns to education. This particularly drives me nuts when it comes to discussions of the returns to degrees in the humanities, and people trot out some, say, Michael Eisner type figure, and say "look what you can do with a theatre degree!" It's more like "look what you can do with a theatre degree and a whole whack of other things going for you!"

"This trend might even be enhanced by universities themselves, if they define their missions as primarily teaching and focus on hiring teaching faculty on a contract basis rather than tenure-track professors that can develop a durable and distinct university culture."

But do you think tenure-track professors are still fulfilling that role, at least with respect to undergrads? How many go to convocation to see their students graduate, for example? How many spend any amount of time on campus with their office door open? In the economics department of one high profile Canadian university, which will remain nameless, some profs don't even hold office hours. Then again, how many would drop by if they did?

[Majromax]

> But do you think tenure-track professors are still fulfilling that role, at least with respect to undergrads?

There might be a chicken-and-egg problem. I have experience as a PhD student and then (briefly) postdoc/adjunct in the math faculty of a southern Ontario university, and in my experience the... not quite curriculum, but perhaps environment? of the first two years was generic.

If a student's first exposure to calculus (to pick on a class I taught and TA'd for) is of a large lecture hall plus tutorial sessions (or TA office hours in a dedicated tutorial centre) that mostly focus on homework help, then students in the new environment are shown that the purpose of a university education is to attend lectures and get acceptable grades.

Such an environment is not conducive towards any of the soft-causes of the university premium. Big lectures (which are then often coordinated at a higher level still to keep a half-dozen sections interchangeable) are not conducive to peer learning. Similarly, the large cohorts don't encourage the tight social bonds for the social capital, and high pass rates don't even turn the course into a strong signal.

By the time a typical student passes through this gauntlet to upper-year courses that are (usually) smaller and taught by more senior faculty, they're no longer looking for individual treatment even if it's on offer. Those few students who do -- particularly those that seek out independent studies or summer research assistantships -- mark themselves as good candidates to be groomed for graduate school. It's there (in my experience) that the small-group dynamic really comes to the fore.

[Frances Woolley]

Coady: "Perhaps it's misleading to think about wage returns in isolation?"

This is an excellent point. On my personal facebook page, some friends have raised the issue of housing prices. If a university degree isn't enough to get you to the point where you can buy a home, what's the point?

Or put another way, suppose you have a choice between
- a $200,000 50-acre rural property which you can use as a base for hunting, subsistence farming, raising organic ducks for sale at the local farmers market, etc - no education required
- a university degree, which only provides job opportunities in the city, where housing for a family costs $1 million plus.
Suddenly a university degree that gets you into a $40/hour job doesn't seem like such a fantastic deal.

On your general point about whether the Bachelor's degree lottery has gotten better or worse - the high school earnings curve has moved to the right, so even if the Bachelor's degree lottery is the same, it's worse *relative to the alternatives*. And that's what matters for an economist.

[Coady]

Frances: Yes. This is a good point about relative to alternatives. But it's awfully tricky to think about these relative comparisons for an individual worker.

Both of the distributions -- HS and BA -- changed by different amounts in different places. If people are assumed to occupy the rank/position in both distributions, then the change in the return to BA vs HS varies people. You could have a return to BA that is growing for workers at the 70th percentile but shrinking for people at the 20th percentile. Or something.

But then...it's probably wrong to assume that people will hold the same rank in their respective distributions. A person who would land in the upper part of the BA distribution, might land in a lower part of the HS distribution. You could have a comparative advantage in one sector rather than the other. Roy models and all that.

Also: I keep having to resist using the word "college". In the US, people mostly use "college" the way Canadians use the word "university". I think I've finally assimilated in at least this one dimension.

I think it's tricky to sort out what this means for an individual worker. It's tricky to

The upper part of the BA wage distribution moved to the right more than the upper part of the HS distribution.

[Jacques René Giguère]

Thanks Frances. I already used it in a class this morning.

[Frances Woolley]

Coady: "A person who would land in the upper part of the BA distribution, might land in a lower part of the HS distribution."

Or not. This is crucial. The mapping of the BA distribution to the counter-factual "what if had chosen HS instead" distribution. Anyone reading know any literature on this?

The college/uni distinction is actually quite interesting - Canada has a much bigger college sector than most other countries, partly because of CEGEPs. Number of college grads is increasing too, rate of increase is smaller than bachelor's or post-grad.

Jacques René: Delighted to hear that!

[Phil Koop]

"... worse *relative to the alternatives* ... that's what matters for an economist."

I don't think that point is in dispute. The problem is that no matter how long I stare at that first graph, I can't see your conclusion.

A bunch of mass has shifted from left of the mode to the right on the high school plots, sure, whereas on the bachelor's plots, it has shifted from the centre to the right. So I guess you could say that the *relative* risk of a very bad outcome has improved for high school and deteriorated for bachelors. But conversely, the size of the rightward shift is bigger for bachelors than for high school, so the relative risk of a very good outcome has improved for bachelors - relative high school outcomes don't *dominate* bachelors, whereas absolute bachelor's outcomes do continue to dominate high school. So I think that Coady's point is valid, it is doubtful the prospective bachelor's holder should view the change in risk from 1998 to 2018 as a disimprovement.

I mean, unless you are making a convex utility argument which you haven't mentioned yet.

[Linda Welling]

My daughter, who skipped university (at least so far), raised another point. Both her parents are academics, and she sees that as a very undesirable jog: long hours, lots of guilt, commitment...She sees her generation as focusing more on work-life balance; she wants the closer connection with individuals she does not see in teaching a class of > 100, or working in an office.

[Frances Woolley]

Phil Koop: "relative risk of a very good outcome has improved for bachelors ... I think that Coady's point is valid"

I'm going to show this picture to a group of colleagues at a talk tomorrow. It'll be interesting in what they see. You and Coady may well have a better spatial sense than I do. I was assuming Foley and Green's results for 2013 were in that 2018 diagram, and they might not be. Also I was assuming that the changes in returns to education were the same whether I looked at logged earnings or untransformed earnings, which of course may not be a valid assumption.

Linda: interesting. It's going to fascinating to see how her generation's labour force participation plays out. I had a lot of illusions that my kids, having seen my life, don't have.

[Frank Restly]

Frances,

Do real interest rates play a role?

Given that to remain profitable in a competitive environment while paying significant positive real interest rates , a debt financed company must find productivity improvements through technological advancements.

Those technological advancements often require a higher degree of education to research and develop.

As real interest rates fall / go negative, productivity improvements are no longer required to remain profitable.

Of course there is the chicken and egg problem - do productivity improvements drive real interest rates higher irrespective of central bank actions or do central banks "coddle" companies so that they never have to become more productive to maintain profitability?

[Frances Woolley]

Frank - "As real interest rates fall / go negative, productivity improvements are no longer required to remain profitable"

That's an interesting idea - though in a market with the possibility of entry, wouldn't there still be an incentive for potential entrants to look for ways to innovate and undercut, and for established firms to look to innovative and improve productivity in order to prevent being outflanked by competitors?

Coady and Majromax - I think you were right about what you were seeing in the distributions.

[Frank Restly]

Frances,

"That's an interesting idea - though in a market with the possibility of entry, wouldn't there still be an incentive for potential entrants to look for ways to innovate and undercut, and for established firms to look to innovative and improve productivity in order to prevent being outflanked by competitors?"

To a point yes. But there is also the case to be made that technological innovation takes time to develop and startup companies (especially those with high up front financing requirements) must rely on a significant amount of debt when they first get started.

It would be out of the norm for a new company to build a factory and finance it strictly with the sale of equity shares.
The reason is that debt claims are often collateralized with an underlying asset.
If the debt associated with a factory can't be supported, the owner of the debt gets the factory or it's liquidation value. Not so with equity shareholders.
This is doubly the case if that factory was employing some new untried manufacturing technique.