Who Is the City-Statey-est Of Them All?
Spoiler: It’s Mongolia. But for rich countries, it’s Chile or Denmark.
UPDATE: I’ve written a second post comparing U.S. state population concentration. Check it out!
I’ve recently criticized the work of both Ross Douthat and Matt Bruenig for adopting international comparisons that aren’t actually very good comparisons. While the specific critiques varied whether I was looking at Singapore or the Nordics, the core critique came down to both writers ignoring the fundamental problem of size, and how size alters the economic, political, and institutional assets available to a country. Well, many readers fairly accused me of telling a just-so story, that no matter what country was picked, I would find a way to criticize it as incomparable.
On the one hand, this is true: I don’t think it is prudent for policymakers to do copy-paste on laws passed in other places. I think that local idiosyncracies make identifying optimal policy a very difficult information problem. This is true in the U.S. for states as well, and increasingly relevant as cut-and-paste-policymaking becomes more in vogue on the left and the right. I think policymakers have a duty to fret the details and craft policies to the specific needs of their communities, and usually that means the exact technical workings of a policy adopted from another area will need to be changed. The changes may be very small or very big, depending on the degree and kind of differences in other relevant circumstances.
But on the other hand, okay, fair point: if we just poo-poo other countries’ way of doing things and claim that American is so exceptional that we have nothing to learn, then we’ll be stuck with undesirable status quo policies forever. This was basically Steve Waldman’s critique of my pieces, as he reads me as making two kinds of claims I did not intend to make: (1) institutional change is effectively impossible, (2) there is very little or nothing to learn or emulate from other countries. I don’t want to take time arguing whether I may have communicated these views or not, rather I just want to be clear I did not intend to communicate these views. The first one is a very thorny but theoretically solveable empirical problem I don’t want to deal with today. The second one, however, is interesting. What can we learn from other countries?
Well, presumably a lot, but before we start importing Swedish policy, we should take a moment to understand some basic facts about Sweden, or any other country we’re studying. One set of facts that seems very relevant is the degree of population concentration in a given country. Population concentration has a huge influence on numerous economic and political outcomes, so, if a country’s internal distribution of population changes, so will the economic and political options available to it.
We can take an extreme example. Singapore will never need to debate farm subsidies. But while many people may see farm subsidies as purely wasteful, there probably is a legitimate state role at least for rural insurance products. Singapore will never need to debate rural mail delivery or rural service delivery of any kind: and this service delivery is costly. Likewise, Singapore will never need to debate separation of powers between different levels of government. There’s functionally just one level to the state.
In a country with a large, dispersed population, these issues will be much more salient and important for economic wellbeing. Likewise, a small city state will have an economy heavily dependent on just a few economic sectors: a big, diverse economy has to respond to a much wider range of sectoral and industrial policy needs.
This isn’t the sole determinant of policy comparability by any means. But it is one that seems oft-forgotten, that to my knowledge does not have readily-available comparative statistics, and conveniently is also right up my population-economics alley. So let’s take a look at how countries vary in terms of population concentration.
How Big Is That City?
We can start by just asking a simple question. What share of a country’s population resides in its largest city?
The map above reveals some interesting features. First, small countries are often a dark color. Kuwait, Uruguay, Panama, the Baltics, Israel, Djibouti, Greece, Gabon, Republic of Congo, Puerto Rico, etc, they all show up in dark colors. Large, populous countries like China, India, Ethiopia, the United States, Germany, or Poland, meanwhile, tend to be lighter-colored.
But then we get large, dark-colored countries like Argentina, Peru, Mongolia, Japan, and to some extent Australia or Canada. The United States ranks 134th of my 150 countries in terms of the largest city’s share of national population.
But that’s just one metric of concentration. We can also look at the rural share of the population. Large amounts of rurality almost certainly change the relevant economic, political, and institutional conditions for countries.
Now having both the largest city share and the rural share, we could combine them in an index to get a very, very simple measure of population concentration.
Okay, cool. The U.S. is in the more-concentrated half, while a lot of poorer countries are less concentrated. That seems neat, I guess.
But is that really useful? Yes, we are unlike South Sudan. But population concentration is not the biggest reason why, and this method doesn’t help us distinguish between potential peers very well.
To do better, we need more detailed data.
But Really, How Big IS That City?
The basic problem when we try to talk about comparative population concentration is that different countries track this data in different ways. So, for example, if I told you that NYC is more populous than Tokyo, you might reasonably be skeptical, since NYC has just 8.5 million people and Tokyo has 13.5. But maybe I meant the New York metro area, which has 20 million people! Well, 20 million is indeed bigger than 13.5 million. But shouldn’t I compare apples to apples, and compare to the Tokyo metro area, which is about 36 million people?
And are those apples really apples? The city proper of Lahore, Pakistan has about 18 million people in it, while the whole metro area is just 19 million. That can’t possibly be right, unless the city proper is defined very generously. There must be a different Pakistani definition of the city proper than how Americans would define the city proper.
Metro areas are fraught as well: how countries define metro areas can vary! Maybe Lahore city proper really does have 18 million people, in which case surely the metro area has more than 1 million additional people, if we defined it in a more conventionally American or European way.
Many researchers have taken a stab at this problem, but the reality is nobody has great data, and the most rigorous estimates are all very hard to update and so usually well out of date. Even developed countries in Europe often lack readily interpretable, comparable, timely data by agglomerative geographies like a metro area.
So if we want to estimate the degree of concentration in population across countries, it’s actually quite hard. I have done my best, but will readily admit my methods are flawed. I used UN data as my main source and, when “urban agglomeration” level data was unavailable, I attempted to develop metro-area definitions on my own, or use outside estimates. I use data ranging from 2004–2016, so this isn’t a single consistent year estimate. And despite my best efforts, for some countries I simply could not find sufficiently thorough metro area designations, and had to use city proper definitions.
But all those caveats aside, let’s look at some key statistics. For example, we could measure the degree of “population monopoly” in a given country. I use a Herfindahl-Hirschman index, which is a standard tool used to measure market concentration for anti-trust cases. In this case, I use the share of the urban population in my sample of cities to estimate the degree of urban concentration. Here’s a map, with darker colors indicating more concentrated urban population:
So, this is getting interesting now. The U.S. shows up as having very low concentration of the urban population, comparable to China, India, or Brazil; maybe also Russia, Japan, or Germany. Meanwhile, other fairly big countries Australia, Canada, Argentina, South Africa, Peru, France, the Nordics, or the UK are much more concentrated.
For reference, the way this works is that if a country has one or two really big cities that are a large share of urban population, they show up as very concentrated. So what the dark countries show is that one or two cities dominate the urban cluster.
But hold on. How complete are my city estimates? Do they actually match estimated total urban population?
Ah. Hahaha. Seriously? No, of course not:
As you can see, for Mexico and the US, my estimated urban population exceeds the World Bank estimate, but shortfalls are more common, in some cases big shortfalls. The red bars indicate the data is for “city proper,” the orange indicates it’s a mixture of city proper and urban agglomeration data, the mustard color indicates pure urban agglomeration data. There’s no extremely strong correlation between these different data reporting levels and the degree of shortfall, which is frustrating.
However, we can make a few assumptions. First, a large shortfall in estimated urban population probably means the World Bank is estimating urban people who don’t reside in the huge metropolises included in the database. This likely means exurban areas not included in the UN database definitions, or else smaller cities dropped due to being insufficiently important. It’s also possible some countries have fundamentally incomplete datasets, but, again, we can probably assume they failed to include smaller rather than larger cities.
In other words, to the extent coverage is incomplete, we should assume that the inclusion of uncounted urban populations would tend to reduce estimated concentration, whereas, for Mexico and the US, removing our over-counted people should tend to somewhat increase estimated concentration.
If we assume this is a simple ratio, we can adjust our HHI estimates, and arrive at this new estimate of country-level data coverage-adjusted HHI:
If you flip back and forth between this and the previous map, you’ll see that the most prominent effect is that several European countries appear less concentrated now. The United States looks somewhat more comparable to much of Europe, though still less concentrated, and the Nordics remain very concentrated, as are the countries of western South America.
But we need more adjustments. As I noted, some of these reflect urban agglomerations, and some reflect just municipalities. Now, of course, the definition of a municipality or an urban agglomeration varies across the sample of countries. But, broadly speaking, we should assume that “urban agglomeration” will always show up as more concentrated than “city proper”, because an urban agglomeration by definition involves several cities. Thus, if a dataset for a country is based on the city proper definition, it will tend to understate concentration relative to a metro area definition. I assume that the “mixed” countries are the neutral median, and put in a concentration-boosting ratio for city-proper countries, and a concentration-reducing ratio for urban agglomeration countries. Adjusted for data reporting level, here’s the output:
Again, you can flip back and forth comparing the maps… or I can just show you a bar graph. Let’s do that; it’s easier:
A few things should stand out.
First, no matter the adjustment, very few countries are greater than a score of 50 on my indexed HHI, whereas a true “pure” city state like Singapore will always score 100.
Second, no matter the adjustment, the US is always one of the lowest-concentration countries, along with China, India, Brazil, Germany, and Japan. We have a very diversified metropolitan ecology, as do those countries.
Third, I’ve highlighted Nordic (purple) and Anglo (orange) countries. Notice that all of the Nordics are much more concentrated than the United States, as are all of the Anglo countries! That one was surprising to me, as I expected large countries like Australia and Canada to be much more comparable to the US. As it is, in terms of population concentration, Poland is more American than Canada.
But hold on: the concentration of the urban population is just one variable. We need more than that. What about the rural population? What about adjustments for the sheer size of a country, or the absolute number of major metro areas?
Adjusting for the rural population is not that hard. There are three ways we could do this. One is that we could add “rural” as a city in the urban city list and see how that changes concentration: but that would show very rural countries as having very concentrated populations, so isn’t right. Alternatively, we could add “rural clusters,” where we simply add a huge number of small cities of population 10,000, or 1,000, or 100, or just 1, until we’ve equaled the rural population. That’d be nifty, but will understate concentration, because many rural people live in “dense rurality” despite having no city, and also introduces a whole new set of systematic biases concerning how large a unit I introduce. Rather, I’ll use a third option: the more rural a country, the less “concentrated” it should show up as, so I’ll multiply my final concentration index by the World-Bank-Estimated urban share of the population, showing more-urban countries as more-concentrated, but doing so in a consistent, easily estimable way that introduces relatively few major biases that could swamp the measured city distribution.
I’m done with maps for a bit, so here’s that bar graph again, with a section added now for the “rurality-adjusted” index:
Every country sees their concentration estimates fall, but by varying amounts. Overall, our core conclusions don’t change, though: nobody is even close to as city-state-ish as a real city state, and none of the Nordics or Anglo countries are as diffuse as the United States. Our peer list is still Brazil, China, Germany, and India, with maybe Russia, Bangladesh, and Japan added in, for funzies. Sidenote on Bangladesh: there may be some data problems there.
Okay, now we get to another question. Maybe the absolute concentration of cities doesn’t matter so much. Maybe what really matters is just the number of important cities. There are two key ways we might measure importance: share of population residing in a city, and the raw population of a city.
So for every city with over 2% of a nation’s population, I’ll add a very small value to the total score. For every city with over 5%, I’ll add another increment. For every city over 10%, yet another increment. And then again, for every city over 500,000, I’ll add a subtract value to each country’s total score. And again for each 1 million+ and 5 million+ city. In other words, having cities that are a large share of population indicates more concentration, but having a large raw number of cities indicates less concentration. We can then normalize so Singapore is still 100%. We also have to add a fixed factor because, with a subtraction term China and India go negative, and I want only positive values.
There are some shifts here or there but, again, our key conclusions remain intact: our best population concentration peers seem to be India, China, Brazil, Bangladesh, Germany, Russia, maybe Poland.
The next factor gets political. I mean, literally, it’s about the political structure of a country. Specifically, is the largest city the capital? This matters to me, because it seems likely that a largest city-capital pairing for one city helps concentrate power, national attention, and resources more than just being the largest city would. So I’ll add a fixed factor for countries where the capital is the largest city, in this case, just a flat index increase of 3.
No huge changes, but, since Washington, DC is not the U.S.’ largest city, it does increase the gap between us and key comparison countries like the UK, or any of the Nordics. Notably, Australia and Canada do not have largest-city-capitals.
Having arrived at an index value that accounts for (1) the mix of city sizes among cities, (2) the rurality of the population, (3) the sheer number of influential cities, (4) basic data quality issues, and (5) the confluence of political and population-based influence, I think this final index is a useful tool.
Let me be clear. This index is not some magic bullet to perfectly describe everything about a country. It is not the Universal Indicator of Policy Comparability. But if two countries are far apart on this index, it should at least give us a moment’s hesitation in assuming policies will translate easily. We should at least do some checking to see if, for example, that other country actually centrally administers its single-payer healthcare program or not (Sweden and Canada manage much of it provincially or regionally, for example).
Here’s a map of my final index values:
And, as a fun sidenote, my most concentrated countries are indeed Mongolia and Peru. Not kidding here. Both results surprised me given that both countries are fairly large and have big rural populations and, in Peru’s case, my impression was that there were a good number of meaningfully sized cities. But it turns out that, in Peru, Lima metro area alone is almost 30% of the population, and then the other cities are pretty small by comparison; and Lima is, of course, also the capital. In Mongolia, Ulaanbaatar metro area is over half of the nation’s population!
So. If you want to know what country is the most city-state-ish, I would have to answer… it’s Mongolia.
Readers will of course note that I have no variable for raw land area and no variable for “rural population is composed of nomads.” These two factors probably matter, but I didn’t want to just do a pure, arbitrary size penalty, and I had no good data on nomadism by country.
But now this should make us curious about something else. Do countries vary in the spatial distribution of their government administration?
How You Structure the Government Matters
The chart below shows what percent of GDP is collected in taxes by each country, broken out by level of government, 2010–2015 average, sorted by total tax collections:
You can scroll down and see that there are substantive differences in how countries collect taxes. But let’s zoom in on state and local taxes:
The chart above shows the OECD countries by the share of GDP collected in state and local taxes. As you can see, every Nordic except Norway is above the OECD average and above the US. Norway’s low share is largely thanks to high national revenues from oil, so is in some sense the exception that proves the rule. We are situated between two of our concentration-peers, Japan and Germany. The only Anglo country above us is Canada.
The reason many of these countries are above us is that many of them finance healthcare through local taxation, with central government aid. Sometimes localities also do the things they do in the US (police, school, etc), but sometimes some of those jobs are more heavily supported or provided by central governments. The point is, a very consistent feature of the Nordic model, our nearest-Anglophone-peer Canada, and some of our concentration peers is lots of government being funded locally.
But I’ve argued before that the better comparison to the US is not France or Sweden, but the EU. So let’s look at EU data.
UPDATE: Several readers have complained this is just such a silly comparison as I criticize. Bear with me, I’m aware of that concern. Suffice to say, I am well aware that Brussels is not DC. I am well aware of the differences in where policy is administered and how. My argument here is not that the EU is fundamentally like the US in political terms, but rather that there are valid lessons we can learn from a similar-size, similar-wealth, similar-population, similar-culture continent with which we have a long history of exchange and interaction, and that viewing the EU as an entity can help us conceptualize some of those useful lessons, and help us identify some not-so-useful lessons. Am I saying we should become the EU? No, absolutely not! But in the grand scheme of things, I do think the US is more like the EU than it is like Singapore or Iceland.
The United States has a vastly larger Federal government than the EU, which makes sense, given that the EU is a few decades old at most, and the US has been around for centuries. Likewise, while the EU does have freedom of movement, goods, and capital, internal barriers remain somewhat stronger than in the US, and within-EU cultural and linguistic differences are of course vastly larger than in the United States. So it totally makes sense for the US to have a larger Federal government than the EU. Should it be 19 times the size of the EU’s? I dunno: maybe the EU’s central government is below optimal levels, but it seems plausible that having 19 percent of the earnings of every person in a 300-million-plus person Federation getting redistributed very far from their home, eyes, or votes would create political strains.
At the other end, we can see that EU states do substantially more at the local level, particularly in the Nordics where much healthcare is locally run. So again, we see at the Federal end and at the local end that despite its vaunted tradition of Federalism, the United States is more centralized than Europe is.
Now, where we get the weirdness is in state/nation comparisons. U.S. states aren’t quite E.U. member states. But E.U. sub-state regions aren’t quite U.S. states either.
In terms of population, U.S. states are much smaller than EU member states, and roughly comparable to EU NUTS 1, which loosely corresponds to something like “groups of EU subnational states,” though the exact definition varies by EU country. If we use U.S. counties, they are themselves much smaller than the smallest EU NUTS 3 unit. So we don’t have a comparable geography to use here.
On the other hand, we can look at land area:
Here, we see that U.S. states are actually substantially larger than EU member states on average, and that EU NUTS 3 and U.S. counties are very comparable in area.
My main concern here is how to compare EU members and U.S. states. And what I’m seeing is that EU member states are substantially more populous, but also, on average, smaller. That suggests the EU is just fundamentally a much denser place than the US, which goes back to our basic comparability problem, but if we accept that density difference and want to go ahead comparing, then we should ignore population generally (which is derived from density), and just compare land area. And by land area, EU member states and US states are indeed fairly comparable.
And if you were to compare EU member states and US states, then it would be reasonable to add subnational EU governments in with EU localities, as those subnational governments, as we saw from the NUTS regions, are more comparable to US localities than to US states. If we group that way, our tax graph becomes:
And here we can see that what’s going on is that EU member states have way bigger local governments, way bigger state-equivalent governments, and a way-smaller Federal government. If we want to become more Nordic/European, we would reduce the size of the Federal government while increasing state and local governments by a larger amount.
Now, I know many readers will argue that I’m comparing apples and oranges because the function of EU members is different from US states; after all, they have militaries! Shouldn’t I allocate spending across the same functions? If healthcare spending is at the local level in the EU, shouldn’t I compare EU locals to wherever healthcare spending occurs in the US?
No! Because the whole question is “what type of policy reform is best.” Or, similarly, “How could we best implement better policy?” or even, to come full circle, “What is the right lesson to learn by comparison with the European Union?”
To do that, you can’t just assume that, oh, of course it doesn’t matter what level of government does something. No, it does matter! And you should pay attention to how other countries do this stuff! And when governing hundreds of millions of people, it is best to let smaller geographies have large amounts of autonomy.
Should the US be as extreme as the EU and have just 1% of GDP go to the Federal government? Absolutely not. We are far more united and consolidated as a people than Europeans ever will be. But, on the margin, if the goal is to ape European welfare systems, it may be advisable for advocates to also ape some of the deep structures that make those systems possible: distribution of responsibility, funding, and to some extent policymaking to more local levels. Without such distribution, policymaking will just escalate into a more-and-more rancorous sectional debate.
This weekend, Ross Douthat had another column, arguing we should “break up” the monopoly power of big cities. I agree with many of his policy proposals like phasing out the state and local tax deduction (I’d include the real estate tax deduction as well!), promoting greater diversification of university geography, and broadly challenging the notion that cities are these magical growth engines that we should all want to be like.
However, Douthat is wrong on one essential empirical question. The United States does not have an urban monopoly. Unlike many countries around the world, we have an exceptionally diverse ecology of population centers. No small cluster of cities actually dominates our country. There is no monopoly to be broken up as there is in, say, Iceland or Mongolia.
Where there is an odd monopoly, however, is in how much of our government is done at the Federal level. That monopoly may be worth breaking up, pushing more fiscal authority and responsibility onto the states. And lest people think this is just some conservative small-government ploy: just watch what happens when Texas tries to cut their state-financed component of Social Security in my future more-federalized America. See how long it stays a red state.
The way to bring about responsible policymaking is to impose responsibility. We are testing this theory right now on the Republican coalition and President Trump and, thus far, the GOP has abandoned an 8-year-long quest that defined the party after their proposal had a harsh encounter with reality. With more devolved fiscal authority, foolhardy proposals left and right at the state level would meet the same fate.
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.