The demographic underpinnings behind America's blue shift, illustrated with interactive maps

Net racial change since 1990

Net racial change since 1990

If you've been around for a few decades' worth of presidential elections, you probably have a strong sense of which parts of the country are trending toward or away from the two political parties. California has gone from swing state to blue state; Virginia has gone from red state to swing state; Tennessee has gone from swing state to red state; West Virginia didn't even bother to pause at swing state en route to switching from blue state to red state.

But why would that happen, in a country with nationalized campaigns, run mostly on nationwide media? Shouldn't the swing from election to election, from place to place, be pretty uniform? Well, no: The population of each state, and the characteristics of the people living there, constantly change. And knowing that different categories of people—whether it's based on race, or education, or religion, or marital status—are considerably more likely to vote a particular way, then it stands to reason that as the mix of people changes from place to place, so too will the way that place votes.

Several weeks ago, I wrote a post based around an interactive map that looked at how the presidential vote had changed at the county level over the last two decades, not in terms of percentage change like usual, but in terms of the raw number of votes. This showed how the changing geographical pattern of votes—huge numeric gains for the Democrats in the nation's most populous counties, while smaller gains for the Republicans were spread out across the nation's rural and exurban areas—is a huge boost for Democrats' chances in presidential elections but also (thanks to the increased consolidation of more and more Democrats in fewer and fewer places) makes control of the House more difficult.

While it looked cool, I later realized that it was missing an important component: the "why" aspect, explaining who moved into or out of those various places (or, just as importantly, aged into the electorate or died out of it). It occurred to me that I could use the exact same method, looking at the net change in, say, white residents versus non-white residents, or college-educated residents versus non-college-educated residents, over the same two-decade period. Most likely, it would show that the places that had tremendous growth in non-white residents or college-educated residents would be the same places that showed tremendous growth in Democratic votes. Did it? Follow over the fold to find out ...

The short answer is, yes, of course it did—although much more noticeably so with race than it did with education. The first map that we'll look at, right below, looks at these changes in racial composition. The counties that show up in blue are the ones where the gain in non-white residents between the 1990 and 2010 censuses outstripped the gain in white residents, while the counties that show up in red are the ones where the gain in white residents in that 20-year period outstripped the gain in non-white residents.

It's an interactive map, so you can mouse over the map to see the data associated with each county; you can also pan and zoom around the map to get a better look. As you look around, you may notice that Los Angeles County, California, has by far the largest net gain, at 2,736,499 people. (As well it should, considering that it's also by far the nation's most populous county, larger than most states.)

What exactly does that number mean, though? Let's break the process down. In 1990, Los Angeles Co. had 8,863,164 people; that had grown to 9,818,605 in 2010, for a total gain of 955,441. The number of whites (or, more accurately, non-Hispanic whites) in 1990 was 3,618,850, but that actually fell by 2010, to 2,728,321: a decrease of 890,529.

There were 5,244,314 non-white residents in 1990—broken down further to 934,776 African-Americans, 29,159 Native Americans, 907,810 Asians, 21,327 people who checked "some other" but not Hispanic, and 3,351,242 Hispanics of any race. By 2010, the number of non-white residents had grown to 7,090,284 (broken down into 815,086 African-Americans, 18,886 Native Americans, 1,325,671 Asians, 22,464 Hawaiians and Pacific Islanders, 25,367 "some others," 194,921 people checking 2 or more races, and 4,687,889 Hispanics of any race—and yes, they did add categories during that twenty-year period). That's an increase in non-white residents of 1,845,970. The final step, though, is to net them against each other: usually that'd be a subtraction problem, but since the white change is a negative number, the net is even bigger than the non-white gain: a net of 2,736,499 in favor of non-white residents.

So, does that match the net gain in Democratic votes that I calculated a few weeks ago? Yes, it does; Los Angeles Co. was also the biggest gainer there, too. In the period between 1988 and 2012, it gained 845k Democratic votes and lost 354k GOP votes, for a Democratic net of 1,198,934. Obviously that's less than the net gain in non-white residents, but bear in mind that the pool of voters is much smaller than the pool of residents; many of those new non-white residents are under 18, some of those residents aren't citizens, and any rate, you need to factor in the multitudes of adult citizens who still can't be bothered to vote.

Here are the 25 largest net gains in non-white residents:

County	Descrip.	Total popu. change	White change	Non-white change	Net change
Los Angeles, CA	Los Angeles	+ 955,441	- 890,529	+ 1,845,970	+ 2,736,499
Harris, TX	Houston	+ 1,274,260	- 178,467	+ 1,452,727	+1,631,194
Cook, IL	Chicago	+ 89,608	- 637,276	+ 726,884	+ 1,364,610
Dallas, TX	Dallas	+ 515,329	- 330,403	+ 845,732	+ 1,176,315
Orange, CA	LA area	+ 599,676	- 226,002	+ 825,678	+ 1,051,680
San Bernardino, CA	LA area	+ 616,830	- 184,515	+ 801,345	+ 985,860
Miami-Dade, FL	Miami	+ 559,341	- 202,056	+ 761,397	+ 963,453
Queens, NY	Queens	+ 279,124	- 320,830	+ 599,954	+ 920,784
San Diego, CA	San Diego	+ 597,297	- 133,234	+ 730,531	+ 863,765
Broward, FL	Ft. Lauderdale	+ 492,578	- 179,528	+ 672,106	+ 851,634
Riverside, CA	LA area	+ 1,019,228	+ 114,928	+ 904,300	+ 789,372
Santa Clara, CA	San Jose	+ 284,065	- 242,965	+ 527,030	+ 769,995
Alameda, CA	Oakland	+ 231,089	- 165,458	+ 396,547	+ 562,005
Maricopa, AZ	Phoenix	+ 1,695,016	+ 602,979	+ 1,092,037	+ 489,058
Bexar, TX	San Antonio	+ 529,379	+ 22,974	+ 506,405	+ 483,431
Prince George's, MD	DC area	+ 134,512	- 174,237	+ 308,389	+ 482,626
Tarrant, TX	Ft. Worth	+ 638,931	+ 79,863	+ 559,068	+ 479,205
Philadelphia, PA	Philadelphia	- 59,571	- 263,254	+ 203,683	+ 466,937
Clark, NV	Las Vegas	+ 1,209,810	+ 377,080	+ 832,730	+ 455,650
Sacramento, CA	Sacramento	+ 377,569	- 34,766	+ 412,335	+ 447,101
King, WA	Seattle	+ 423,930	- 3,890	+ 427,820	+ 431,710
Nassau, NY	Long Isl.	+ 52,184	- 186,594	+ 238,778	+ 425,372
Bronx, NY	Bronx	+ 181,319	- 121,294	+ 302,613	+ 423,907
Orange, FL	Orlando	+ 468,465	+ 30,210	+ 438,255	+ 408,045
Hidalgo, TX	McAllen	+ 391,224	+ 6,294	+ 384,930	+ 378,636

If you refer to the table of the 25 counties with the largest Dem gains, there's a lot of overlap. Fifteen counties show up on both lists. In fact, for the most part, this list overlaps with a simple list of the nation's most populous counties (indicating both the sheer size of the overall national non-white gain, but also how concentrated that gain is in a few very populous places).

Of the 38 counties that had a million or more people in the 2010 census, 23 of them are on the list above. (The only two with less than a million residents that made the cut are Prince George's County, Maryland, a suburb of Washington, DC, that has transitioned from in the last few decades from a mixed population to being mostly black, and Hidalgo County, Texas, in the Rio Grande Valley, which was mostly Hispanic even in 1990 but has added almost exclusively only new Hispanic residents.)

One other detail that might leap out at you is that in a majority of these counties, the number of white residents actually dropped from 1990 to 2010; in other words, more left or died in that period than were born or moved in. Without digging much deeper into Census data it isn't clear how much of that is natural decrease and how much of that is white flight (or for that matter, how much is white flight to the county next door and how much is to a totally different part of the country). In a few cases, like Philadelphia or Chicago's Cook County, the overall population barely changed at all, and the net gains are largely based on white subtraction.

Now let's look at the 25 largest net gains in white residents:

County	Descrip.	Total popu. change	White change	Non-white change	Net change
Douglas, CO	Denver area	+ 225,074	+ 185,981	+ 39,123	- 146,828
Utah, UT	Provo	+ 252,974	+ 185,652	+ 67,322	- 118,330
Ada, ID	Boise	+ 186,590	+ 144,212	+ 42,378	- 101,834
Hamilton, IN	Indpls. area	+ 165,633	+ 131,124	+ 34,509	- 96,615
Washington, DC	Washington	- 5,177	+ 43,333	- 48,510	- 91,483
St. Charles, MO	St. Louis area	+ 147,578	+ 117,373	+ 30,205	- 87,168
Delaware, OH	Columbus area	+ 107,285	+ 89,299	+ 17,986	- 71,313
St. Johns, FL	St. Augustine	+ 106,210	+ 88,202	+ 18,008	- 70,194
Williamson, TN	Nashville area	+ 102,161	+ 84,250	+ 17,911	- 66,339
Clark, WA	Vancouver	+ 187,310	+ 126,241	+ 61,069	- 65,172
Forsyth, GA	Atlanta area	+ 131,428	+ 97,686	+ 33,742	- 63,944
Warren, OH	Cincinnati area	+ 98,784	+ 79,181	+ 19,603	- 59,578
Davis, UT	SLC area	+ 118,538	+ 88,851	+ 29,687	- 59,164
Horry, SC	Myrtle Beach	+ 125,238	+ 91,812	+ 33,426	- 58,386
Montgomery, TX	Houston area	+ 273,545	+ 165,175	+ 108,370	- 56,805
Larimer, CO	Ft. Collins	+ 113,494	+ 83,834	+ 29,660	- 54,174
Livingston, MI	Detroit area	+ 65,322	+ 59,686	+ 5,636	- 54,050
Washington, UT	St. George	+ 89,555	+ 71,617	+ 17,938	- 53,679
Baldwin, AL	Mobile area	+ 83,985	+ 68,400	+ 15,585	- 52,815
Deschutes, OR	Bend	+ 82,775	+ 67,167	+ 15,608	- 51,559
Kootenai, ID	Coeur d'Alene	+ 68,699	+ 59,737	+ 8,962	- 50,775
Cherokee, GA	Atlanta area	+ 124,412	+ 87,353	+ 36,789	- 50,564
Placer, CA	Sacramento area	+ 175,636	+ 112,693	+ 62,943	- 49,750
Yavapai, AZ	Prescott	+ 103,319	+ 74,577	+ 28,742	- 45,835
Union, NC	Charlotte area	+ 117,081	+ 80,474	+ 36,607	- 43,867

Again, you'll notice a number of overlaps with the parallel list of counties with the biggest net gains in Republican votes. For instance, the number-two county gainer here, Utah County, Utah (the location of Provo, and of Brigham Young Univ.), was also the one with the largest net gain in GOP votes. Some of the nation's most archetypal exurban communities—counties located 20 or 30 miles away from large cities that had small populations in 1990 and had filled with new middle-class subdivision development by 2010—show up in both lists, too, like Montgomery Co., Texas, Douglas Co., Colorado, and Williamson Co., Tennessee.

There is also a number of western counties here, on the list of biggest net gains for white residents, that didn't make the cut for the list for biggest GOP gains; the common thread tying them together is their reputation as destinations for people white-flighting it out of southern California (not all of whom are conservative, though these still tend to be pretty red-leaning places). That includes Boise, Idaho, and Placer County in Sacramento's exurbs (which has picked up rapidly-diversifying Orange County's mantle as California's conservative stronghold) in particular, but also places further afield like Bend, Oregon, Prescott, Arizona, and northern Idaho's Panhandle. It also includes, surprisingly, politically swingy Clark County across the river from Portland, Oregon, where the white growth is a mix of California refugees, Oregonians seeking an income-tax-free haven, and evangelical Protestant Russian immigrants.

Finally, there's one entry on the list that stands out like a sore thumb: one of the bluest jurisdictions in the country, Washington, DC. This one is pure gentrification; the population of Washington stayed basically flat over the two decades, but one-tenth of the city's population turned over in that period, with whites replacing non-whites. (Scroll down to the table below showing net gains in college-educated persons, and note how the numbers very closely match.) Interestingly, Washington managed to get even bluer during this same period, with Michael Dukakis getting 83 percent in 1988, while Barack Obama got 91 percent here in 2012.

There's one other important similarity between the list of biggest white gainers and biggest GOP gainers: These are counties at the periphery of the nation's metropolitan areas, or, in some cases, medium-sized towns that stand alone. In addition, these are much smaller counties, population-wise, than the non-white/Dem gainers. It's not as lopsided; while only about 900 of the nation's 3,000-plus counties had a net gain in Democratic votes, over half of the nation's counties had a net non-white gain. The much larger numbers in the non-white table mirror how growth in non-whites, nationwide, far outstripped white growth. The nation's white population grew from 188,123,888 to 196,817,522 from 1990 to 2010, a gain of 8,693,664, while the nation's non-white population grew from 60,578,936 to 111,928,016, a gain of 51,349,050. That's a net non-white gain of 42,655,386!

Now, let's look at a totally different variable, the change in levels of education. Here's the interactive map, with the same color scheme, although here, the blue counties are the ones with a net gain in college-educated residents, and the red counties are the ones with a net gain in residents without a college degree (though, as I'll discuss shortly, the red and blue can be a little misleading in some places; education doesn't correlate as strongly with voting behavior as race):

This is done according to the same procedure: Find the change, from 1990 to 2010, in the number of persons over age 25 with a four-year degree or better. Then find the change in the number of persons over age 25 without a bachelor's degree, and, finally, find the net difference between the two. To start, here's the list of the counties with the biggest net gain in college-educated residents:

County	Descrip.	Total popu. above 25 change	Bachelor's or more change	Non-bach. change	Net change
Cook, IL	Chicago	+ 121,906	+ 381,887	- 259,981	+ 641,868
Los Angeles, CA	Los Angeles	+ 786,899	+ 593,164	+ 193,735	+ 399,429
New York, NY	Manhattan	+ 71,443	+ 202,775	- 131,332	+ 334,107
Kings, NY	Brooklyn	+ 155,311	+ 222,609	- 67,298	+ 289,907
Middlesex, MA	Cambridge	+ 77,882	+ 169,731	- 91,489	+ 260,860
Wayne, MI	Detroit	- 117,298	+ 63,578	- 180,876	+ 244,454
Allegheny, PA	Pittsburgh	- 70,485	+ 82,272	- 152,757	+ 235,029
Santa Clara, CA	San Jose	+ 187,067	+ 208,296	- 21,229	+ 229,525
King, WA	Seattle	+ 281,763	+ 253,327	+28,436	+ 224,891
Nassau, NY	Long Isl.	+ 23,408	+ 106,240	- 82,832	+ 189,072
Cuyahoga, OH	Cleveland	- 62,612	+ 58,690	- 121,302	+ 179,992
Philadelphia, PA	Philadelphia	- 59,262	+ 58,067	- 117,329	+ 175,396
San Francisco, CA	San Francisco	+ 72,548	+ 123,878	- 51,330	+ 175,208
Queens, NY	Queens	+ 177,400	+ 172,226	+ 5,174	+ 167,052
Alameda, CA	Oakland	+ 156,656	+ 159,390	- 2,734	+ 162,124
Hennepin, MN	Minneapolis	+ 76,342	+ 118,966	- 42,624	+ 161,590
Oakland, MI	Detroit area	+ 105,167	+ 131,047	- 25,880	+ 156,927
San Diego, CA	San Diego	+ 388,443	+ 268,964	+ 119,479	+ 149,485
Bergen, NJ	NYC area	+ 41,647	+ 93,108	- 51,461	+ 144,569
Orange, CA	LA area	+ 401,795	+ 268,784	+ 133,011	+ 135,773
Washington, DC	Washington	- 9,674	+ 60,228	- 69,902	+ 130,130
St. Louis, MO	St. Louis area	+ 13,048	+ 70,149	- 57,101	+ 127,250
Baltimore city, MD	Baltimore	- 68,771	+ 28,672	- 97,443	+ 126,115
Fulton, GA	Atlanta	+ 164,106	+ 144,741	+ 19,365	+ 125,376
Fairfax, VA	DC area	+ 169,359	+ 147,168	+ 22,191	+ 124,977

Again, you'll see a lot of the same familiar faces here, the nation's most populous counties, the ones containing the nation's major cities and economic engines. In fact, this table may say more about the nation's economy—and the "Two Americas"-ification that's going on—than it does about its politics. While college-educated persons used to be more evenly distributed throughout the country, they're clustering increasingly in the metropolitan areas, where, not coincidentally, the well-paying jobs that need an advanced education increasingly are. Factor in the multiplier effect, and you've got the recipe for metropolitan areas further increasing their economic hegemony, while the corresponding "brain drain" exacerbates the isolation of rural areas.

One striking detail about the table of biggest gains in college-educated residents is that in many of the counties, the number of adult residents without college degrees actually fell from 1990 to 2010. As you'd expect, that's the case in Rust Belt cities that had a strong manufacturing reputation, and where the older blue-collar workers have died or retired elsewhere and many of the younger ones have sought out better employment opportunities elsewhere. At the same time, those same places, despite their overall population loss, have still managed to gain college-educated population; you'd expect that in, say, Pittsburgh, which has become a research hub in a number of high-tech areas, but it's even happening in the allegedly-wrecked Detroit area too, which should give us some hope it might re-imagine itself along Pittsburgh's lines too.

But that decline in non-college-educated residents is also happening in places without a once-strong manufacturing base. It's happening in places like Manhattan and San Francisco that are expensive enough that a college diploma is something of a minimum ticket to entry if one wants to afford to live there. It's even happening in places like Minneapolis and Philadelphia that don't fit in either the glittering financial center or post-industrial hellscape archetypes.

Finally, let's look at the list of the top 25 counties for net gains in residents without college degrees:

County	Descrip.	Total popu. above 25 change	Bachelor's or more change	Non-bach. change	Net change
Clark, NV	Las Vegas	+ 755,542	+ 202,073	+ 553,469	- 351,396
Maricopa, AZ	Phoenix	+ 1,031,385	+ 392,403	+ 638,982	- 246,579
Riverside, CA	LA area	+ 559,709	+ 157,706	+ 402,003	- 244,297
Harris, TX	Houston	+ 721,013	+ 238,599	+ 482,414	- 243,815
San Bernardino, CA	LA area	+ 356,585	+ 94,000	+ 262,585	- 168,585
Hidalgo, TX	McAllen	+ 201,643	+ 37,568	+ 164,075	- 126,507
Kern, CA	Bakersfield	+ 156,959	+ 27,540	+ 129,419	- 101,879
Dallas, TX	Dallas	+ 284,377	+ 98,728	+ 185,649	- 86,921
Tarrant, TX	Ft. Worth	+ 360,375	+ 138,280	+ 222,095	- 83,815
Pinal, AZ	Phoenix area	+ 145,962	+ 33,037	+ 112,925	- 79,888
Bexar, TX	San Antonio	+ 310,117	+ 118,201	+ 191,916	- 73,715
Fresno, CA	Fresno	+ 143,622	+ 38,923	+ 104,699	- 65,776
Lee, FL	Ft. Myers	+ 193,899	+ 67,894	+ 126,005	- 58,111
Gwinnett, GA	Atlanta area	+ 264,547	+ 103,964	+ 160,583	- 56,619
Mohave, AZ	Kingman	+ 77,081	+ 10,304	+ 66,777	- 56,473
Polk, FL	Lakeland	+ 126,103	+ 36,542	+ 89,561	- 53,019
Cameron, TX	Brownsville	+ 83,606	+ 15,751	+ 67,855	- 52,104
Marion, FL	Ocala	+ 101,961	+ 25,013	+ 76,948	- 51,935
Osceola, FL	Orlando area	+ 94,402	+ 22,213	+ 72,189	- 49,976
El Paso, TX	El Paso	+ 123,054	+ 36,935	+ 86,119	- 49,184
San Joaquin, CA	Stockton	+ 113,809	+ 32,390	+ 81,419	- 49,029
Tulare, CA	Visalia	+ 65,158	+ 10,496	+ 54,662	- 44,166
St. Lucie, FL	Ft. Pierce	+ 83,349	+ 19,867	+ 63,482	- 43,615
Stanislaus, CA	Modesto	+ 86,370	+ 21,381	+ 64,989	- 43,608
Lake, FL	Orlando area	+ 99,702	+ 28,618	+ 71,084	- 42,466

This table—unlike the list of the counties with the largest net gains in white residents—has no overlap at all with the list of the biggest net gains in GOP votes. Instead, these counties are mostly politically swingy, though in many of them there's enough variation in performance between presidential and non-presidential elections that they're characterized by a lot of falloff in off years. There are a few (like Hidalgo Co. and El Paso Co. in Texas) that are so heavily Hispanic that they are solidly blue, and some of the counties in Florida are white enough that they're still pretty solidly in the GOP column.

These counties are entirely in the Sun Belt states, spanning from California to Florida, but there are two different trends at work here. One is counties where there has been a lot of Hispanic growth, not just in California but in Texas and Arizona as well. Some of these counties have moved pretty sharply in the Democratic direction (especially Clark Co., Nevada, which contains 2/3s of that state's population and is almost single-handedly responsible for Nevada's shift to blue-tinted swing state). Others, though—thanks to the combination of many Hispanics either not being citizens or not having aged into the electorate, and low turnout even among adult citizens—we're going to be waiting a decade or more before we start seeing many dividends. (In particular, that applies to the counties in Texas.)

The other trend is growth in downscale retirement areas, primarily in Florida (though perhaps also describing Mohave Co., Arizona, where the retirement destination of Lake Havasu City is). Unlike tonier Florida destinations like Naples or Sarasota or Palm Beach (which tend to attract more affluent northeastern Yankee or Jewish retirees), the counties on this table tend to attract a more blue-collar Midwestern retiree base, who are less likely to have college degrees (partly because they were more likely to have had manufacturing jobs, but also in large part simply because college wasn't as commonplace a thing to do back in the mid-20th century as it is now).

The one county here that doesn't fit easily in either category is Gwinnett County, Georgia, the suburbs to the northeast of Atlanta. This was a middle-class white area when it was first developed (back when it was Newt Gingrich's home turf when he served in the House), but it's one of the most rapidly diversifying places in the country as its white residents decamp for exurbs further north, and they're replaced by an interesting mix of African-Americans moving out from Atlanta and Hispanics and Asians moving in from a variety of other countries.

This last table points to why education isn't a straightforward predictor of voting patterns. As far as education is concerned, Democratic preferences form kind of a U-shaped curve: The most heavily educated segments tend to be likely to vote Democratic (people with advanced degrees aren't necessarily wealthy members of the 1 percent; probably the biggest group is public school teachers with M.Eds), but then the least educated segments of the population also tend to be likely to vote Democratic (who tend to be people of color and/or immigrants, who are voting based on their economic interests or because the GOP's perceived disdain for them). Instead, it's the people with middling levels of education—two-year degrees, or "some college"—that tend to be the Republicans' strongest group. Unfortunately, the method I've used in this article can't pinpoint the "some college" crowd.

There is any number of other variables I could have plugged in, that would have yielded equally interesting results: Marital status, for instance, is a particularly strong predictor of voting behavior (in other words, places that have gained a lot of single adults are also places likely to have gained a lot of Democratic votes). But race and education are two of the most important variables, and they do a lot to explain the "why" behind Democratic gains in certain parts of the country. It's not just that the Democrats suddenly decided to "try harder" in, say, Virginia, or developed better messaging that mysteriously changed a lot of once-Republican minds; it's that the underlying population has changed in ways that let Democrats be more competitive there.