Higher Education Propels Population Growth (Probably)

Spinoff Benefits, R&D, Easier Retooling, Cutting-Age Amenities, and Attracting Baby-Making Youngsters Will Do The Trick

I have argued several times recently that universities can propel population growth, indeed may even be one of the few reliable sources of population growth. The reason I think universities are reliable is that they constantly generate new marketable or institution-creating ideas. Patenting, spinoff research hubs, annealing social institutions, workforce development, all serve to make a local area not just economically stronger today, but more resilient to economic changes. Pittsburgh’s recovery in recent years, such as it is, can largely be attributed to a growing skills-and-tech hub fueled in part by the presence of Carnegie-Mellon. Of course, fracking has a role to play too, and more on that in a bit, but natural resources like that can’t be “created” by policymakers or philanthropic billionaires. Universities can be.

Today’s post will be about the association between higher education and population growth, especially graduate education.

As I’m going to show, a comparatively large population of graduate students has a pretty impressive association with population growth.

Where Do Students Live?

Let’s start with the geography of higher education. Here’s the percent of county-level population enrolled in universities by county:

The first thing you should notice is how spikey enrollment is. Some counties have really high higher-ed enrollment, some have virtually none. Basically, you can see a geography of where colleges are.

But this metric includes a huge range of different institutions with a wide range of quality. In my opinion, and the opinion of most experts on the subject, the spillover benefits from universities aren’t likely to accrue without some kind of research function, or advanced course of study, or a program that keeps students around and academically engaged for longer. A reasonable proxy for this would be the graduate student population.

Where Do Grad Students Live?

This map is colored to the same scale as the all higher-ed map. First of all, you should notice how thin the overall distribution of graduate students is: the grad student population is something like 4–4.2 million people; the undergrad student population is something like 19–20 million.

Second of all, if you looked at that map and thought, “Huh, grad students are more concentrated than undergrads” you were correct! We can use several metrics. The coefficient of variation in undergrad population share is 76%, versus 86% for grad students, indicating that the prevalence of graduate students varies much more widely (especially on the low end) than that of undergrads.

We can also look at concentration at the county level. The top 5 grad-student counties contain 10% of grad students, while the top-5 undergrad counties contain just 9% of undergrads. For top 10 counties, it’s 15.4% and 13.4%. For top 25, it’s 26% and 22%. All the way out to the top 1,500 counties, graduate students are more concentrated than undergrads. Finally, a basic Herfindahl-Hirschman test shows an HHI of 0.417 for undergrads vs 0.516 for grad students, indicating graduate students are more concentrated in a few counties than undergrads. And I wager if I exclude state universities, the concentration rises even more, suggesting that what diffusion does exist is in large part a product of policy choices.

So all higher education is concentrated around universities, but graduate school the most widely economically influential component of higher ed, is even more concentrated than undergrad.

So let’s look at population growth across these counties.

Counties With Graduate Schools Grow Faster

The chart below will be hard for some folks to read, so let me explain it. I have ordered every county by its population growth rate from 2010 to 2015. I have then made an unweighted average of the population share of graduate students for 100-county blocks. So the first point would represent the average graduate student share for the 100 counties with the most steeply negative population change from 2010–2015. The 2nd point drops the absolute worst county (lookin’ at you, Blaine County, Oklahoma), and adds the 101st worst-changing county, in this case Custer County, Idaho. This progresses throughout the series, so that the last point shows you the average share of graduate students for the 100 fastest-growing counties.

In case there are quibbles about the use of 100-county blocks, I also tested 30, 50, 200, and 500 county blocks, and there was no difference in the general conclusion reached.

To read the graph, know that the Y axis indicates the graduate-student % of county population, while further out along the X axis indicates faster population growth.

The first thing to note, and the absolute most important thing, is that the line slopes upward: faster population growth is associated with more graduate students.

Now, my choice to average large blocks obscures a lot of variation. That’s intentional: I want to communicate the broad trend without bogging the reader down in a regression table with gazillions of controls.

I’d also note that the steep decline you see in the middle of the series is very tricksy. The points from 1300–1900 have virtually no difference in population growth rates; they are all very similar, as they are all clustered in the near-zero range (around 1650 is when growth becomes positive). Don’t read too much into these mid-series changes. The crucial thing is to look at the consistency on the far left side of the graph (rare graduate students + slow growth), and the very high levels on the right side (many graduate students + faster growth).

Critics will note that the line declines at the end. The reason is oil and fracking. Oil counties have seen incredible growth despite little or no higher ed presence. But that growth will fade soon enough. If I exclude Texas and North Dakota from the analysis, that tail-end decline is reduced by about half. The remaining high-growth low-grad-student counties are virtually all either military bases, fracking sites outside of TX and ND, or fast-growing exurbs of large cities, so not necessarily a model for independent, medium-sized cities or towns.

Furthermore, it should be noted that this grad-school-growth association applies as much in coastal as in heartland states.

Here’s a scatterplot of Alabama counties, population growth vs. the grad student share:

That one negative outlier is Macon County, home of Tuskegee and the Tuskegee institute. I’m not sure why it has declined so much, especially given its university, but that steep of a decline usually indicates either a boundary change, major employer loss, or something like that.

Or take Kentucky:

Or take New York:

This relationship holds across just about every state. California:

Now, let me respond to two possible criticisms here:

Isn’t This Just a Proxy for Density?

To some extent maybe, but I don’t think so overall. It’s true that if I graph out population growth by density, I get a similar result as I got for graduate schools. However, that graph is screwed up by the fact that density has incredibly huge variation. That is to say, the distribution of graduate-student share of the population is far more even than the distribution of density, and the number of very high-density counties is far smaller than the number of very high-graduate-student counties, thus the idiosyncracies of one or two counties (looking at you, New York City counties) destroy the whole time series in very short order. The result is that when, for funzies, I run a regression of density + grad share on population growth, the regression on the whole isn’t super meaningful (just like a 0.1 R-squared, so yeah not impressive at all), but grad share is about 3 times as statistically significant as density, and moving 1 standard deviation along the graduate-share spectrum has about 3 times the effect of 1 standard deviation move along the density spectrum. Now, to be clear, such a simple regression is pretty much one step away from a steaming pile of crap in terms of how it informs us about specific questions, but it does give us at least suggestive reason to think that maybe we (and by we I mean PhD-holding researchers who have time, energy, skills, and financial support to do complete, published research; decisively not me) should be looking at least as much at how universities drive population trends as how density patterns may.

Plus, I think “skills, research, education, and institutions” are more likely to be a meaningful indicator than “people in tight spaces.” That is, when we think about the alleged benefits of density, many of them basically relate to things universities due in a supercharged version: foster exchange of ideas, create communities of diverse services and common interests, spin off community institutions, encourage entrepreneurship, etc. Universities often even create their own density, as many students live in university housing that is usually not an exurban McMansion. There are more additional reasons I’ll explain below for thinking graduate schools explain growth, but I’ll save the goodies for the end.

Isn’t This Just the Student Loan Bubble?

To some extent, yes. But note that I’m not charting change in grad students. I’m not saying, “places that expanded grad schools grew.” Oh no, I’m just saying having grad students was associated with growth. I used average grad students across the 2010–2014 period. I have not back-cast to previous periods due to data constraints, but I hope somebody else will try to do so.

It is possible that these effects have been made more pronounced by the rapid inflation in student loans driven in part by very low interest rates. But the reality is that in many of my high-grad-school counties, population growth exceeded the total graduate school population. Charlottesville grew by 3,100 people; the entire grad population is just 4,000. Williamsburg, VA grew by 1,400 with just 820 grad students. Johnson County, Iowa grew by 13,3000 people with just 7000 grad students total. If we limit to change, it’s unlikely to me that UVA’s graduate program had 0 students in 2000; in most cases these programs have probably only grown by a small fraction of total enrollment. So whatever is explaining this growth, it is not just a loan-fueled boom in enrollment.

So Why Do Grad Schools Create Population Growth?

First of all, astute readers will not that by saying “create” I am alleging causality that I have not, strictly speaking, proven. Very true. To test my theory, we need to charter a couple mid-sized universities in towns without them, and see if those towns see growth improvements relative to peers.

Oh, come on. What person thinks that wouldn’t happen? Really? Who in their right mind says, “Nah, dropping a $150 million endowed university in a county of 60,000 residents probably wouldn’t impact population growth in a positive way of the following two decades.” It obviously would. I don’t need to demonstrate that causality because asserting that crowding out would be so massive as to counteract thousands of students and hundreds of staff and a binge of new construction and investment is certainly the more epistemically unwarranted claim.

But how do grad schools drive growth? That is a very good question, and I don’t have a perfect answer. Here are some ideas that come to mind:

1. They may boost the birth rate. As it happens graduate students are older and more likely to be married than undergrads. Shocker, I know. Their birth rate is probably lower than demographic peers without education, but nonetheless the influx of young people almost certainly boosts the non-age-adjusted birth rate. Indeed, counties with high grad school populations have higher birth rates, even controlling for density as a proxy of rurality. When people have years of advanced schooling that involves applied research and semi-employed conditions as grad school often does, and then have a spouse and child(ren), they are more likely to put down roots. They find a church they like, they find a social network, they get used to a style of living, they find a nice housing situation, they get familiar with the transportation… they get habituated and comfortable. Because graduate schools attract people right at that crucial “habituation window” in their lives, local areas may benefit from families sticking around for at least a few extra years on average.
2. They may extend time spend by undergrads. To the extent undergrads roll into a local grad program, graduate programs may help “retain” undergrads. The longer those undergrads stick around, the deeper their roots become, and voila, you’re back at item #1.
3. They may create spinoff businesses that create employment. University hospitals. Seminary publishing houses and missionary boards. Policy research institutions. Applied research companies. Tech start-ups. The number of ways a university’s knowledge-creation mission can create valuable spin-off business is large. This feature is likely fairly unique to graduate schools, as graduate schools perform more research than undergrad institutions, and vastly more than workforce-training-oriented schools like community colleges. To get this benefit, you need good grad programs.
4. They may train workers in sufficient volumes that firms are attracted to the area, creating employment. This is the classic workforce-training-economic-development model. I am somewhat skeptical of this as firm relocations and expansions are far rarer than individual-level migration, but it is nonetheless possible, and certainly happens in some cases.
5. Universities may drive the creation of local amenities, institutions, or norms that improve quality of life. If the local university crowd supports nice restaurants, a vibrant downtown, a liquid real estate market, community activities that engage locals, etc, then it may be that other people move to the area for those amenities, regardless of association with the university. Graduate schools would share this trait with undergraduate institutions, but as grad students have more income, are more settled in life, have gotten past their low-standards undergrad days, and are generally more likely to want to put down roots, they may have more of an effect than undergrads.

There may be other factors as well, but these all come to mind as probably forces at work that are also, conveniently, more-or-less empirically testable!

Why Write About This Now?

Richard Florida just wrote a piece on Medium (his first it seems; welcome to Medium, Rich!) about the great divide in the knowledge economy, and how it influences the election. It’s good and worth reading. His basic point is that the high concentration of knowledge workers in a few urban centers dilutes their political power in our political system. His suggested solution is a devolution of powers to cities.

I am 1,000% on board with a huge leap in Federalism, as are many conservatives, provided we devolve pieces of governance to localities and states. It’s incredibly encouraging to see a leading progressive, urbanist voice call for increased Federalism, and I hope my Team Red allies will get on board and seize the moment to devolve large chunks of revenue-collection and spending-authority to state and local governments.

But there’s another answer here. We spend about $77 billion for the Department of Education every year, of which ~$30-$40 billion goes towards subsidizing demand for higher education, in the form of loan subsidies, grants, etc.

We spend under $10 billion on subsidizing supply of higher education. That is, the federal government bids up demand more than it pushes up supply. Aside from the statistically curious effect this could have on tuition (a hotly debated question I won’t get into here), it also has severe regional effects. Subsidizing demand for education is subsidizing regions with educational institutions. Every Pell grant is a subsidy for the local economy where the student attends.

We would be well-served by the Department of Education investigating the viability of increasing the supply of universities. A good university sufficient to revitalize a mid-sized town can be launched for well under $500 million. For no more than we currently spend on subsidizing demand, we could found 60–80 good universities each year in struggling areas, areas where high tuition bills are stripping families of their financial assets and communities of their youth.

We could drive employment, population growth, and greatly widen the geographic franchise of the knowledge economy by chartering new universities in small cities.

The model would be simple: the Federal government, in partnership with state and local government, and maybe some corporations or philanthropists, puts up $500 million dollars to charter a university. That university is then made into a private school, required to balance its own budget, perhaps with some special rules covering them for the first 30 years or something regarding tuition, access to enrollment, standards, aid offerings, etc. The idea here is that you turn over control to locals, and force the university to stand or fall based on its ability to actually function as an institution. Maybe you say that the Federal government reserves the right to appoint some share of the board members; that’s fine. The point, however, is budgetary and mission independence. We can’t have the new university in a random town have all its functions determined by directives in DC about How To Fix Small Towns. Let locals decide how their knowledge-factory should work.

Some of these universities will fail. More to the point, I suspect a large number of existing universities would have to change policies to face the new competition.

But to Richard Florida’s point, this kind of policy would fundamentally re-orient the geography of knowledge, bringing wide swathes of the country into contact with previously isolated fellow Americans. Call me an optimist, but I can’t help thinking that fear will have less hold on peoples’ politics when their towns start growing again, their young peoples dreams can be satisfied nearer to home, their ambitious locals see their businesses reach larger scale faster, and their local diners play host to a wider range of viewpoints around the debating-table. I don’t think this policy would change the partisan balance at all; I just think it would make both parties better.

Check out my Podcast about the history of American migration.

If you like this post and want to see more research like it, I’d love for you to share it on Twitter or Facebook. Or, just as valuable for me, you can click the recommend button at the bottom of the page. Thanks!

Follow me on Twitter to keep up with what I’m writing and reading. Follow my Medium Collection at In a State of Migration if you want updates when I write new posts. And if you’re writing about migration too, feel free to submit a post to the collection!

I’m a native of Wilmore, Kentucky, a graduate of Transylvania University, and also the George Washington University’s Elliott School. My real job is as an economist at USDA’s Foreign Agricultural Service, where I analyze and forecast cotton market conditions. I’m married to a kickass Kentucky woman named Ruth.

My posts are not endorsed by and do not in any way represent the opinions of the United States government or any branch, department, agency, or division of it. My writing represents exclusively my own opinions. I did not receive any financial support or remuneration from any party for this research.