Where Has Population Fallen?
Most Counties, But Not Where Most People Live
A number of discussions recently have led me to wonder about population decline, not least my recent thinking about Atlantic City. I realized as I was pondering the issue that I actually didn’t have a clear idea off the top of my head what the geography of population decline actually looked like. What regions have experienced population declines? What cities? Does anybody know?
The simple way to check this would just be to download a time series of counties and sort by some change over time. That should be easy. Counties are clearly-defined, Census reports on them every decade, and the Intercensal program reports on them every year since 1969. Easy-peasy, right?
No. Although you can download Intercensal data from BEA easily enough, getting those old decennial Censuses is not easy. The Census Bureau provides a long time series of every county from 1790 to 1990, but it doesn’t line up neatly with Census’ downloadable data for the 2000 and 2010 Census, or the American Community Survey, because the counties are not FIPS coded in the historic time series, and many currently non-existent counties are in the list, and the county names aren’t always formatted the same (“La Salle County” versus “LaSalle County” for example), etc. The result is that connecting Census 1790–1990 time series data for all counties to available Census 2000/2010 data is sort of a pain.
Now, of course, the IPUMS database could address this. They could give you county population in a time series format. But as a rule, I try to avoid using microdata, and I also have a concern for data access and availability: I know how to use IPUMS, but most people don’t. A simple time series really ought to exist.
I have taken the time to build that time series for every state except Alaska, where region definitions have changed too darn much to make heads or tails of anything. Anybody who wants it, just shoot me an email. As a sidenote, the Census 1790–1990 document includes tons of notes on county boundary changes which help explain lots of oddities in the data. I’ve stripped out those notes.
But once we’ve finished the way-harder-than-it-should-have-been assembly of a time series, we can actually look at what places have experienced population decline.
Also, before I go any further, huge shout-out to the support team at Datawrapper. I’ve really put them through their paces with this project, and they have responded with friendliness and what I can only call truly German efficiency. Their service is worth paying for. And no, they didn’t pay me to say that.
Jefferson’s American Promise Died in 1940
Depopulation is Concentrated in the Most Rural Areas
The map below color-codes each county based on its 2015 population as a percent of its maximum Census-reported population. I include the 2015 intercensal estimate as “present day,” but note that I do not include the 1969–2014 intercensal estimates, so there may be counties that actually peaked in 1983, or 1997, or 2008, etc. I’m just showing which Census they peaked in, or else a peak in the most recent data.
The map below color-codes counties by when they hit peak population, from 1920–2015. Also, all population peaks in 1920 or earlier are color-coded identically, in the very lightest tan shade.
The same clear story arises from the light areas in both maps.
Depopulation is concentrated, both in severity and in how long it’s been since peak population, in the Great Plains, the Cotton Belt, and Appalachia.
We hear about some of these, like Appalachia. The explanation for Appalachia’s lost population is variously offered, with good cause, as being the decline of coal, the declining relative economic profitability of small-scale farming, demographic trends favoring urbanization, and the decline of American industry. These all seem like pretty solid explanations to me.
When we turn to the cotton belt, the explanations do not seem to fly as thick and fast. The decline of… large-scale agriculture? Wait, no, that didn’t happen. Oh, right, the decline of labor-intensive cotton-harvesting. That’s what happened. Then there’s civil rights issues and the “Great Migration” northwards. We might also note that, curiously, these are also the most rural southern counties, so this may just be demographic trends favoring urbanization again.
In the Great Plains, however, we face a problem. The cotton belt and Appalachia have out-migration, but also social immobility, very low employment, poor social services, etc. That’s not the case in the Great Plains. the Plains states routinely show up as having low unemployment rates for decades on end, high rates of social mobility, well-functioning and accountable governments, etc. Plus, these areas have seen an extraordinary agricultural boom, especially over the last 20 years. Nor are the Plains states infertile; indeed, controlling for race and age, they tend to be some of the higher-fertility states, with middling levels of mortality. So why the great decline in the Plains? Here, we must simply say that it’s urbanization: these rural counties emptied out as cities offered education, affordable housing, less strenuous jobs, and by the 1930s, relatively clean living conditions.
So those are the areas of decline. But let’s look at this another way.
Who Lives Amidst Decline?
More People Than You’d Think
The maps above make clear that decline-counties are common. It turns out, very common. Just 1,162 counties, of 3,140 I measured, had peak population in 2015. That means that 63% of counties had experienced some level of population decline as of 2015.
But what about in terms of actual population?
As of 2015, 86 million people lived in population-decline counties. That’s 27% of the population. Fairly remarkable. Now to be clear, that’s not how many people live in counties where population is currently falling. In Manhattan, for example, population has recently been rising, but remains well below peak, so I call it a population-decline county.
But, of course a smaller share of the population lives in off-peak counties. By definition, these are counties with weak population growth. That’s just a basic mathematical feature. Sure, 27% of the population living in a place haunted by the ghosts of a grander past may seem like a lot, but is it really all that much?
Very fair question! We can look at these same counties in 1940. In 1940, the counties I classify as population-decline counties in 2015 had 73 million residents, amounting to 56% of national population. Meanwhile, the currently-peak counties have risen from a 1940 population of 58 million, to a 2015 population of 235 million.
Impressive, most impressive.
The areas in decline now can in many cases look back a generation or two, or maybe three, and see a time when they were as or more populous, and, crucially, where they were more important. Whatever the raw numbers, they were a much bigger share of national population.
Take Atlantic City. In 1930, it peaked at 66,000 residents, which was 0.054% of national population. By 2010, it has fallen to 39,000 residents, a 40% decline. But in percent of population terms, it has fallen by 75% to just 0.013%. So it’s not just that Atlantic City, or Appalachia, or Cleveland, has fewer people. It’s that even the people they have matter less in comparison to the rest of the nation.
Who’s the Biggest Loser?
Because, Be Honest, That’s What You’re Here For
I know that what you want is to know what areas have lost the most people. The map below shows the counties with losses of over 100,000 people since 1940. By this measure, the biggest loser is St. Louis city (which, confusingly, is a county-level unit). Another big loser is Baltimore city. Manhattan also makes a strong showing, as does Philadelphia county.
The map of “large population losses” is not quite the same as the map of population declines as a percentage from peak. Eastern seaboard counties loom much larger, as does Pennsylvania, while the Great Plains nearly vanishes from the map of decline. This is in many ways a bias of county size: Great Plains losses in total are substantial, but divided among numerous counties. Meanwhile, eastern losses are from cities.
And some cities like Manhattan, Baltimore, Philadelphia, and St. Louis share a common trait: they are essentially “city-counties.” They are geographically small, and in some cases, actually seceded from surrounding counties, such as in the case of Baltimore and St. Louis. As such, they have suffered more than larger counties from suburbanization. In essence, these large losses partially reflect inconsistency in what it means to be a “county.” If I consolidate Baltimore city and county, then there is actually a 438,000 person population increase since 1940. For St. Louis, there’s a 228,000 person increase. Now, for both cities, there is a decline since 1970. But I want to show you the graphs because, when you find a gem like this, you don’t waste it:
Look at that. Look at that! I’m not making an academic point here, I’m marveling at the beauty of probability. Baltimore and St. Louis have uncannily similar population histories, from their city/county split, to the level of population, to major trends, etc. Since 2000, they’ve diverged a bit, but seriously, I just love when I find something like that, so I wanted to show it to you.
But I want to talk about some different places. See, it turns out that there are actually relatively few counties that are at their minimum population since 1900 as of 2015. Just 396, in fact. I’ve mapped them below:
These are almost all small, rural counties. The only minimum-population counties to have over 100,000 people in 2015 are St. Louis city, and Schuykill County in Pennsylvania, which has declined with the loss of the coal industry and manufacturing jobs, to a 2015 population of 144,000 from a peak in 1930 of 235,000. The next-largest minimum-population county is Washington County, Mississippi, with 48,000 people, from a 1960 high of 79,000. Curiously, Washington County is also quite urban, with 33,000 of its residents living in the county seat of Greenville. Its continued prominence may be because the city is the site of one of the 2 bridges over the Mississippi between I-20 at Vicksburg and I-40 at Memphis.
Now, I’ll tell you what’s even rarer among these counties than big cities: temporary periods of recovery. It’s rare for any of these counties to show periods of deviation from decline. And when they do, it is almost always because nearby cities began to spill over into their area. The one exception I found was Elk County, Pennsylvania, and I actually have to explanation for that. The county’s population grew at a good clip from 1930 to 1980, despite the major towns declining. So this was a growth of rural population. Which, yeah, very weird.
When we look at the counties with the steepest population losses, we find some fairly extraordinary figures. Issaquena County, Mississippi, for example, lost 90% of its population from 1890 to 2015. But then again, it only had 12,000 people in 1890 anyways. Jewell County, Kansas lost 85% of its residents, but again, under 20,000 people. These are small potatoes.
What we want is a county with a fairly substantial population that also lost a huge share of that population. And we will find our dead-ringer for that in McDowell County, West Virginia.
In 1950, after demobilization and in the midst of a growing economy with high demand for McDowell County’s abundant coal reserves, population levels reached a new high. But it would be downhill from there. Despite a very brief plateau in the 1970s, the county’s population would fall, and fall, and fall, and fall. By 2001, the public infrastructure was at least as bad as it was in the 1960s when JFK singled out the county for commentary on poverty.
Indeed, let’s keep in mind, during the period when McDowell County was booming, it was one of the counties used to justify the War on Poverty. This is a bit odd since, ya know, people were moving there for jobs just a decade earlier.
Today, McDowell County’s population is about 20,000 people. It’s fallen from about 5% of West Virginia’s population in the 1920s, to barely 1% today. Where once the county was genuinely dense, with almost 100,000 people, today it is a fairly low-density, rural county.
In 2001, the school system had to be taken over by the state due to the decrepit physical facilities and poor student performance. Major new teacher training programs were implemented and a number of schools were closed altogether. Boosters for the new programs point out that the school system is now locally-managed again and claim that new programs are working. Maybe so. But McDowell County’s population is continuing its slide downwards.
If you want a candidate for “most severely depopulated place in America,” I would say McDowell County, West Virginia is your best bet.
But Are They Worse Off?
Population Decline Doesn’t Always Mean Stayers Suffer
You might think McDowell County was experiencing a unique degree of suffering. But it’s not at all clear from the data. McDowell County has been sitting at about 50% of the average national level of income (in nominal terms) since the 1970s. In other words, in nominal terms, incomes in McDowell County have kept pace with incomes nationally. They’re low, but in relative terms no lower than in the past. Of course, if more people purchase nationally-priced goods than in the past, like student debt or iPhones, this relative parity may still lead to an increase in inequality of consumption, as the nominal size of the relative gap rises. This seems plausible, so it’s possible that more nationally connected markets could make McDowell residents feel poorer despite being at exactly the same relative position they’ve been at for a long time.
More generally, it’s not clear that those who remain in McDowell are worse off as a result of population loss. They are certainly worse off as a result of lost jobs, which means lost income and thus lost tax revenues and lost real estate demand and lost consumer base for businesses, etc. But lost population? That’s not clear. In fact, if the school reforms do work out, the lost population may turn out to be a blessing in disguise as streamlined schools are better able to provide instruction. We’ll see how that turns out.
But the key here is to distinguish between lost total income and lost average income. Total income in McDowell has fallen. Nominal and relative average incomes have not. I haven’t done “real” incomes because inflation-adjusting localities is tricky, especially geographically isolated localities with huge shocks that may persistently alter local price levels.
If total income fell, as a result of reduced coal output, and yet population stayed the same or rose, then average income would necessarily fall. In practice, this means that, if McDowell’s young people had stuck around when there were no jobs, they would have had to either suppress wages to get employment, depend on social and family networks of support, apply for welfare, or else be entrepreneurs. Wage suppression could attract investment in the long run, but is constrained by the minimum wage and unionization. Welfare is very unlikely to fully replace lost income. Entrepreneurship could drive growth, but is, well, very hard, especially for resource-constrained individuals and the poorly-educated. So finally, we come to social and family support: the remaining employed people are asked to support a growing network of people.
For the record, this does happen. But if average income statistics are to be believed, then the average McDowell resident should be no worse off, in material terms, than they were 30 years ago, when populations were more than double as high.
Conclusion
It would be handy if the Census Bureau put out a new time series table for every county’s Census records, and even handier if they could also link it to intercensal estimates, or at least provide a FIPS code. In lieu of such convenience, I’ve done my best to create the data. Anybody who wants it, just ask.
Population decline has occurred in the majority of U.S. counties, and those declined counties included over half the population of the United States 60 years ago. Today, fully a quarter of Americans live in counties below their peak population. However, that number can exaggerate the prevalence of ongoing decline: only 396 counties hit their minimum population in 2015, and they accounted for just 4 million people in 2015, versus about 7 million in 1940: from 5.3% of the nation to 1.3%.
Population decline has been concentrated in small, rural counties, former coal-producing areas, and urban centers. Of those categories, urban centers have shown some recent strength in some cases, while rural areas and coal-counties have been far less likely to show any sign of recovery.
Even so, it’s not clear that depopulation itself has done any harm to those who remain behind. Depopulated Great Plains counties have among the lowest unemployment and highest economic mobility of anywhere in the United States. Meanwhile, even in Appalachia, the most depopulated areas don’t seem to have gotten any poorer as a result of depopulation, even if they haven’t gotten richer.
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 graduate of the George Washington University’s Elliott School with an MA in International Trade and Investment Policy, and an economist at USDA’s Foreign Agricultural Service. I like to learn about migration, the cotton industry, airplanes, trade policy, space, Africa, and faith. 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. More’s the pity.
