A million news stories may have already dislodged it from your memories, but think back to September, to the special election that happened in North Carolina’s 9th Congressional District. In a court-ordered re-do of the 2018 election—after absentee ballot fraud by a Republican operative called the original election into dispute—the Republican nominee, Dan Bishop, only very narrowly held onto this longtime GOP stronghold. Bishop defeated Democratic nominee Dan McCready by around 2 points, though this was actually less close than the 2018 result, where McCready lost by only a fraction of a point. McCready was hampered a bit by a mix of lower turnout and a worse performance in the rural parts of the district.
But even if you were paying attention to the race in the 9th, you might have still missed that there was another special election that same night in a nearby district, North Carolina’s 3rd. (This was brought about by the death of longtime Republican Rep. Walter Jones.) Unlike the race in the 9th, though, the special election in the 3rd wasn’t especially close. The Republican nominee, Greg Murphy, won by over 24 points, 62-37. And that wasn’t because the Democrats nominated some random no-name: The nominee was Allen Thomas, the former mayor of Greenville, one of the main population centers of the district.
One possible difference is that, while the parties spent millions of dollars on the race in the 9th, there was almost no outside investment in the 3rd. That’s not because, as some might worry, the DCCC decided to sandbag Thomas; he’s not a progressive firebrand, and in fact has a more insider-ish resume than McCready. More likely, they simply polled the districts, liked what they saw in the 9th, didn’t see encouraging results in the 3rd, and spent accordingly. And the results weren’t surprising, based on those districts’ baselines from the last presidential elections: While they started in a relatively similar place in 2012 (Mitt Romney won the 3rd 59-41and the 9th 55-44), in 2016, the 9th took a small move toward becoming more of a swing district (Donald Trump won there 54-43), while the 3rd got much redder (growing to a 61-37 margin—not coincidentally, almost exactly the same as the special election result).
What explains this difference? More than anything, it’s the educational difference between the mostly rural 3rd and the mostly suburban 9th. In the 3rd, 25% of residents who are 25 or older have a college degree; that number is 36% in 9th. That may not seem like a huge difference at first glance, but most congressional districts are closely grouped around the national average of 33%; very few districts have a number above 40% or below 20%. If you break the nation’s congressional districts into quintiles, that’s enough to put the 9th in the above-average quintile and the 3rd in the below-average quintile. (It also helps that the 9th is more racially diverse than the 3rd: 67% of residents in the 3rd are non-Latino white, while in 9th it’s 59%.)
That's a pattern that we’ve seen replicated throughout the country.
Race and education by congressional district, by quintiles
Above is the main graphic that we’ll be discussing. (You can see a larger version as well, if you’d like to be able to zoom in.) This graphic makes those quintiles that I mentioned more explicit; it puts every congressional district in the country into one of 25 buckets, based on which quintile it’s in on two different axes. The leftmost column is the nation’s (approximately) 87 congressional districts with the highest percentage of non-Latino white residents; the rightmost column is the quintile with the whitest districts. And the topmost row is the nation’s districts with the highest percentage of persons 25 years or older with bachelor’s degrees or more; the bottom row contains the districts with the least college-educated population.
Placement by quintile is based on the U.S. Census Bureau’s recently released 2018 one-year American Community Survey; here is the full set of data from that release. For instance, to be in the leftmost diversity row, the population of a district would have to be no more than 38.3% non-Hispanic white; to be in the rightmost row, at least 81.2% of its population would need to be white.
And to be in the topmost education row, at least 41.2% of a district’s adult population would need to have bachelor’s degrees or more; to be in the bottom row, no more than 23% of its adults would have bachelor’s degrees or more. (This doesn’t always exactly work out to 87 districts in each tier, i.e., 435 divided by 5. The goal is to have each tier represent as close to 20% of the population as possible, so while congressional districts have similar populations, there’s some variation, and the number in a particular tier may be, say, 85 or 89.)
If you’re a regular Daily Kos Elections reader, this may be familiar to you: I wrote a similar article two years ago in which I unveiled this concept, put congressional districts in their appropriate buckets, and tried to generalize where Democrats’ best chances in the 2018 election were. That in turn was based on a more extensive piece looking, at the county level, at the 2016 election results and trends going back for decades beyond that, showing that places that are both very diverse and very well-educated are the ones moving most rapidly into the Democratic column, and that places that are both very white and the least educated are the ones moving most quickly toward Republicans.
The 2018 elections largely matched my analysis. As you can see from the graphic, most Democratic gains came in the most-educated rows, the most-diverse columns, and, in some cases, districts that are in both of those, concentrated in the upper left part of the graphic. On the graphic, Democratic pickups are denoted in dark blue, while districts retained by the Democrats are light blue. (Districts that stayed in Republican hands are dark red, while the two open seats in Minnesota that the Republicans picked up are light red.)
For example, of the 42 seats that the Democrats picked up, 18, or nearly half, are in the top 20% of the nation’s most-educated districts. If you expand to the top two quintiles for education, that covers 29 districts, or nearly three-quarters of the Democrats’ pickups.
Racial diversity accounts for a slightly less profound shift: The top two quintiles for non-white population account for 14 of the Democrats’ 42 pickups. That’s largely because Democrats already held so many seats in those quintiles, thanks in large part to the creation of minority-majority seats under the Voting Rights Act. (If you’re wondering about the number 42, keep in mind that the Republicans picked up two previously Democratic-held seats in Minnesota. That takes you to the “40 net pickups” number that you’re used to seeing in reference to 2018. As you might expect, those two rural seats are in the lower-right quadrant of much-below-average diversity and average or below-average education.)
Instead, most of the Democrats’ most notable pickups in 2018 were in suburban areas that until recently were Republican strongholds, but where Republican performance collapsed in the 2016 presidential election over Donald Trump-related dismay. That trickled down to the congressional level in the 2018 midterm. The best-known example is, of course, Orange County, California, one-time epicenter of the Reagan Revolution, where the GOP delegation was wiped out entirely in 2018. But it also includes the suburbs of other Sun Belt cities such as Houston, Dallas, and Atlanta, and other longtime Republican hotbeds such as the suburbs west of Chicago.
In my 2017 version of this story, in fact, the most vulnerable Republican-held seat in the country, based purely on this analysis, was California’s 39th Congressional District, then held by Ed Royce. It was the only GOP-held seat in both the “very high diversity” and “very high education” bucket. Royce retired, and Democrat Gil Cisneros won it 52-48, despite what many observers viewed as a somewhat lackadaisical campaign.
Interestingly, using 2018 census numbers, that bucket now includes one other pickup, in Florida’s 27th District. (The 27th was in the “very high diversity,” “high education” bucket previously; it’s probably not so much a matter of a huge influx of Ph.D.s into Miami as much as that it narrowly missed the cutoff before and fairly normal fluctuations pushed it barely over the line.)
Also, since 2017, another Republican-held district has moved into that bucket! Texas’s 22nd District, in Fort Bend County in Houston’s suburbs, was previously in the “high diversity,” “very high education” bucket. This is genuinely one of the most rapidly diversifying districts in the country, though, as middle-class Asians and Latinos flock to this area’s mushrooming suburbs. Much as Royce bailed out last time, the 22nd’s current Republican representative, Pete Olson, sees what’s coming and is retiring in 2020. Olson won by a narrow 51-47 in 2018 after having never really faced a competitive race before, and this open seat is now one of the Democrats’ top pickup possibilities in 2020.
In fact, the “high diversity,” “very high education” bucket was a very good predictor of Democratic success generally. It contained some of the archetypal affluent suburban pickups that Democrats pulled off in 2018 (Texas’s 7th and 32nd districts, and Orange County’s 45th and 48th districts). It also included a couple of the Democrats’ closest misses in 2018: Georgia’s 7th District (where the Republican won by a fraction of a percent) and Texas’s 24th (where the Republican won by only 3). Those guys know what’s coming, too: Rob Woodall in the 7th and Kenny Marchant in the Dallas-area 24th are both retiring in 2020, and those districts join the 22nd as top-tier pickup opportunities.
The demographic focus is a different approach than the way you usually see congressional districts discussed, which tends to focus on the presidential numbers in districts as a means of assessing how likely a district is to change hands. Focusing more on demographic change is a way of looking over the horizon at where districts are going, rather than at where they are now. It makes it possible to spot up-and-coming opportunities, such as Georgia’s 7th and Texas’s 24th and other Texas suburban districts that are starting to come into play, such as the 10th and 31st.
Keep in mind, of course, that not every Democratic pickup comes from that corner of the chart. There’s still room for winning races based on the time-honored traditions of being a skillful retail politician and having the good fortune to run against an unusually execrable opponent. That would largely explain some of the Democratic successes in the low-diversity, low-education quadrant, such as the pickups in Iowa’s 1st District and New York’s 22nd District (where the Democratic candidates ran against Rod Blum and Claudia Tenney, respectively).
There’s one other consideration in those districts, too (and in Maine’s 2nd District, which was an open seat in 2018). These are fairly rural districts with a long-standing Democratic tradition, even though they’ve slid to the right in this decade. So there are still a number of voters in those districts amenable to voting for the right Democrat at the House level, even as Democratic prospects have gotten worse lately at the presidential level in those areas.
Let’s discuss one last thing: Suppose you’re a statistical purist who doesn’t like the bucketing technique. Forcing items into quintiles when many of them are closely bunched might overemphasize narrow differences while underemphasizing outliers. For instance, a district with a population that’s 22% college-educated might more closely resemble a district that’s 23% college-educated, but the cutoff line puts the 22% district in the same bucket as the district that’s 9% college-educated.
Race and education by congressional district, as a scatterplot
For you, here’s an alternative way of visualizing the information: a traditional scatterplot! Here each district is plotted as a dot using the actual values. We’ve retained faint lines showing the quintiles, though, which helps you see that the middle three quintiles are pretty closely grouped together, especially on the vertical axis depicting education. The unfortunate side effect of the scatterplot, though, is that it’s too cluttered to show the individual district names. (You might cross-reference it with the above quintiles chart, though, to at least get a sense of where a particular district that you’re interested in would be found on the scatterplot.) And if you need to zoom in for a closer look, here is a larger version of the same scatterplot graphic.
