To better illustrate that, I've put together a map that shows just when each county's population peaked, according to the decennial Census. The darker the color, the further back in time the population peaked. The yellow counties are as populous as ever; orange counties peaked in the later part of the 20th century; red counties peaked in the early part of the 20th century; and by the time you get into the purple counties, they peaked back in the 19th century. As you can see, it's not a random spray; instead, there are waves that steadily march across the country.
Below the fold, there's an interactive version of the map that you can pan and zoom and mouse over particular counties to see the specific year. There's also additional discussion of what the patterns that you're seeing might mean.
The first pattern that you're probably noticing is that the counties that peaked at some point in the past tend to be the most rural counties, and, for the most part, counties where the fewest people live. That's borne out by the actual population data: people used to be distributed more evenly throughout the country, when it was a more agrarian nation. At this point, more than three-quarters of the U.S. population lives in an urban area, but as recently as 1910, a true majority of the population was rural. There's also a very robust statistical relationship (a correlation coefficient of 0.46) between what year the county's population peaked, and the current percentage of the people in that county who live in an urban setting. In other words, the more urban a county is, the more likely it is that its population is currently larger than ever ... and the more rural it is, the likelier it is that its heyday, population-wise, was longer ago.
However, there's a noticeable, though much weaker, relationship (with a correlation coefficient of 0.18) between the year of a county's population peak, and its total current population. That's because there are a number of the nation's most populous counties, almost exclusively in the Northeast and Rust Belt, where the population nevertheless peaked several decades ago (most typically in 1970, which was something of a tipping point as freeway construction and suburbanization led to depopulation of many cities).
Cook Co., IL (Chicago); Cuyahoga Co., OH (Cleveland); and Wayne Co., MI (Detroit) all peaked that year, while New York Co., NY (Manhattan); Kings Co., NY (Brooklyn); Suffolk Co., MA (Boston); and Washington, DC, peaked a little earlier in 1950. That isn't to say that their metropolitan areas, as a whole, have plateaued; the suburban counties that surround them are larger than ever. And in most cases, with post-recession growth being heavily concentrated in cities (even more so than in suburbs), these urban counties have all rebounded from hitting bottom (usually around 1990) and are on track to hit new peaks in 2020 or 2030.
The newer cities of the Sun Belt and west coast, by contrast, are currently at their peaks ... and so, too, are pretty much any county that's even remotely within commuting distance of any major city. Instead, with the exception of the old-school cities, the counties that are off their population peaks are remote counties where the main economic activity was agriculture (all across the Great Plains, spilling over as far as Montana and Texas, but also in the Black Belt across the Deep South, and in the bottom lands near the Mississippi River) or mining (all across the spine of Appalachia from Kentucky into Pennsylvania, but also randomly dotting the Rockies and Sierras, and in Michigan's Upper Peninsula and Minnesota's Iron Range). There's certainly less mining now than there was many decades ago, and probably fewer acres under cultivation as well ... but more importantly, farming and mining are simply less labor-intensive activities than they used to be, and don't need as many people present to do it.
You can also notice different gradations across the Great Plains, as you move from east to west, seemingly following the frontier as it got pushed steadily westward over the decades. The rural parts of Iowa and Missouri peaked in the late 1800s; the peak hit Kansas and Nebraska around 1900 and 1910; it washed over the Dakotas in the '20s and '30s; and it finally washed over eastern Montana and west Texas in the '30s and '40s. You can actually see the pattern of emptiness giving way to expansion, and then boom turning into bust, repeating itself, until finally there wasn't anywhere new left to move to.
This being a political website, though, you might be wondering: what's the political application of all this? Well, it speaks to the nature of the demographic change that's slowly remaking the country in a more Democratic-friendly way; many of the reddest parts of the country are also the places where their best days are behind them, population-wise. You've probably noticed some superficial commonalities between this map and the standard county-level red/blue map of presidential elections, if you focus on where the reddest areas are (across the High Plains down into Texas, and along the Appalachian arc) ... or maybe more accurately, a map showing the trend in presidential voting over the last few decades (in this case, the change in Partisan Voting Index from 1988 to 2012, as seen right below), which puts Appalachia and west Texas into particularly sharp relief.
Still, you'll notice a lot of other areas don't correspond very well: the Deep South's Black Belt may have peaked long ago in terms of population, but it's also a Democratic stronghold and even getting bluer as turnout improves there. Conversely, Utah is one of the fastest-growing parts of the country, and remains blood red. That's borne out when you look at the statistical relationship: when you correlate what year a county's population peaked with the percentage of its votes that went to Obama in 2012, the coefficient barely even registers at 0.11; in other words, if you're sitting in a county where the population peaked long ago, it's just as possible that, for instance, you're in a dark blue Rust Belt city as you are in the fields of west Texas.
But more importantly, the counties where the population peaked long ago simply aren't a dominant part of the overall political equation. Even though they account for half of the nation's total number of counties, they only account for about 68 million of the nation's residents (less than a quarter of the nation's population). Most of the growth in Democratic votes is in dynamic, still-growing counties ... but, then, so too is most of the growth in Republican votes.
There's still a good deal of polarization in where the parties' votes come from—the Democratic votes are disproportionately coming more and more in urban counties, while the Republicans are running up the score in sprawling exurbs that you may not even be familiar with (places like Montgomery Co., TX, and Forsyth Co., GA). We'll talk more about the other half of the equation—these growing counties—in next week's installment.
[Methodology: If you're wondering what tools I used to create this map, it's the terrific Open Heat Map, a free service where you can upload any geo-coded data set (either a .csv or a Google Doc) and it'll spit out a map of it. (If you're using county-level data like this, you'll need a FIPS code for each county.) The population data comes from Wikipedia, where some anonymous hero has compiled a table in each county's page that tracks its population over every single Census; all of that information is ultimately accessible at the Census's site, some of which I used to randomly double-check to make sure that the Wikipedia data is accurate, though pre-2000 Censuses are only available in more-difficult-to-work-with pdf form.]