The Urban Inversion Is Over
And It Was Exaggerated to Begin With
Update: I’ve got a follow-up post here.
Today we got county estimates of population for 2016. Virtually all of the coverage is going to compare 2016 to 2015. That’s good and useful. The truth is, I don’t have tons to add to that coverage. I think Jed Kolko pretty well has it covered over at 538 (or at his personal blog). I’ll quote him:
Basically, suburbanization continues apace. Urban cores that thought they were facing a new generation of urban demographic supremacy should be rethinking any policy, budget, or planning choices made on that assumption. If your budget forecast was anything other than “Current strength is likely to weaken in coming years,” then you made a predictable (and predicted!) forecasting error. I’m being a bit harsh here but this is important: the question of whether the strength of urban cores during the Great Recession was a new epoch in American economic geography likely to continue for a long time, vs. whether it was a passing, cyclical phenomenon, was an empirical one, with brass-tacks policy and budgeting consequences. Team Suburbs has won the empirical argument. We can debate why all day long, but what matters for responsible planners and policymakers is as much what as what, and the what at this point is clear: most urban areas are likely to return to similar demographic conditions as in the mid-2000s. Not identical, of course. There will be differences. But for the vast majority of people, the urban cores are going to continue to be of diminishing demographic relevance, and most urban cores are going to experience demographic difficulties.
But there’s more to the story than that.
Census Makes Forecasting Errors Too
No estimate is perfect. Any professional forecaster has normal, reasonable sources of error and we don’t fault them for it: that’s just the name of the game. Census is a pretty good forecaster given the task assigned to them. But it’s a monumentally big task, the data is very limited, and, despite their efforts, revisions can be large.
So while most press is going to be on how the New 2015 Estimate compares to the New 2016 Estimate, I want to focus on how the New 2015 Estimate compares to the Old 2015 Estimate.
I did this when we got state estimates as well. I also used those estimates for states to try and forecast what the county estimates would eventually say, and gave this broad prediction:
On twitter as well, and in a follow-up post, I made the claim that we could use these state estimates to calibrate our expectations regarding counties.
Turns out, I was right. Here’s 2015 revisions expressed as a percent of old estimate 2015 population, by county density:
As you can see, every density class was cut, but the cuts were sharpest for the densest counties. This trend appears again on a much smaller (essentially negligible) scale in 2013 and 2014 as well. The point is: Census is going back and realizing that some of their prior-year forecasts were too exuberant. So not only are 2016 vs. 2015 growth rates worsening for dense cities on a current-estimates-basis, but 2015 population was already cut versus the last round of estimates!
Mapping Revisions to Population
We can also look at where cuts to 2015 population were made. For example, here’s the percent change to 2015 population made in the 2016 revisions:
This is pretty noisy because small changes in small counties can show up as a very big percentage change. So instead, here’s a map that shows dots giving the size and direction of any county with a larger than 500 resident revision:
As you can see, red dots far outnumber blue dots, and red dots are disproportionately located in coastal, urban areas around the Bay Area, Los Angeles, New York, Boston, and DC. In other words, past Census estimates fundamentally over-estimated how many people lived in those counties. Which means go back and revise your view of punditry on those areas: they had less demographic success than originally estimated!
What Caused These Revisions?
As I did for states, I can go back and compare the source of differences. To start with, we can look for just pure volatility: disregarding the sign of a change, what components of population change saw the most county-level revisions? Here’s the sum of the absolute values of county-level revisions for each component for each year, going back to 2010:
As you can see, international migration is the biggest component, but major domestic migration revisions persist all the way back to 2011 too! Major birth/death revisions also occur in 2014/2015.
We can also explore how these vary by density. For the charts below, I will show for each density group a time series of the cumulative revisions for a given component, divided by the old cumulative total for that Year-Group-Component set. So if 2014 2000+ ppl/sq mi Births were previously estimated at 100,000, and are now 99,000, it would show -1% in that year. If, then, 2015 births had been forecast at 105,000, and actually came in at 103,000, I would divide (99k + 103k) / (100k+105k), then normalize at 0%. So it’s a rolling total of the cumulative effect of all prior revisions for a given group. Here’s births:
As you can see, on net, the 2015 revisions cut the cumulative number of births in mid-to-high density places, while births in the most dense places (2000+) and the least dense places (Under 50, 50–100) rose the most. I don’t know what to make of that and it’s a very noisy revision series, but neat nonetheless!
In 2014 and 2015, we can see that Census has revised up their death estimates for the most rural areas, while revising down for the most urban areas. In other words: rural deaths are occurring at a faster pace than Census previously expected, while urban deaths are occurring at a slower pace.
International migration shows fairly consistent revisions. This is no shocker, because the source of these revisions is a methodology change at Census. Nonetheless, it appears that the sharpest drops in international net migration were in the most dense (2000+) counties, but also in some low-density ones (50–100). Upper-mid-density counties saw the smallest cut in international migration.
For domestic migration I use absolute figures, not rates, because the sum of domestic net migration must always be equal to zero, and its sign is frequently negative for large groups of counties, screwing up % change figures. As you can see, the revisions made indicate that migration into the densest counties has been lower than previously estimated, while migration into some mid-density and rural counties has been substantially higher than previously estimated.
These Census population figures re-affirm the view that strong urban population performance was tied to some cyclical component of the late 2000s recession. During the recession, urban centers did well. As prosperity has returned, they’re emptying out again, and Census keeps revising their figures to make that emptying more and more severe. By looking at revisions, we can re-calibrate our assessment of the accuracy of prior claims by commentators, and see who was right, and who wasn’t.
As much as this will sound like pugnacious coup-counting, I want to be clear: error is normal. All forecasts have errors. Being right one time doesn’t make you right the next, either. But commentators have a duty to track their own forecasts and call BS on themselves when their views are revealed to be BS. I do that fairly aggressively for myself. This matters, because most municipal governments have a limited ability to forecast population trends for themselves, and so rely on freely available commentary. If providers of that commentary do not police themselves, they risk inducing local governments into making reckless investments and squandering taxpayer money. All of that to say: Can we please stop with the dense-urban-area-comeback stories?
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.