One of the most common intramural arguments that Democrats seem to have is which voters to reach out to in order to try to win close elections: Do you try to turn out “base voters”—usually meant to mean young people and/or people of color—who don’t usually vote at high rates but are likely to vote Democratic if they do vote, which means convincing them to vote at all? Or do you try to win over “swing voters” who tend to alternate between voting Democratic or Republican—where you don’t have to convince them to vote, but you do have to convince them to vote Democratic?
As with many long-winded arguments, though, it’s one that turns out to be something of a false choice. Doing one doesn’t preclude you from doing the other as well. And based on new research, doing one will help strengthen your position, but to make really big gains, you need to do a lot of both.
And a good example of big gains is the 2018 election, where Democrats gained a net 41 seats in the House. That’s not as big as Republican gains in 2010, but it’s still the largest one-year Democratic gain in the House that most of us have seen in our lifetimes. It’ll be difficult to expand much further in the House in 2020, but there are still a White House and a Senate that need to be won back in 2020, and the lessons of 2018 will be helpful.
This week, Yair Ghitza of Catalist, the Democratic data-gathering firm (they maintain the vast depository of information known as the “voter file” that you sometimes hear political professionals speaking of in reverent tones), put out a deep dive into how those 2018 victories came about. As you may have guessed, Democrats substantially expanded the electorate for 2018 compared with previous midterms; the composition of the electorate, in terms of the races and ages of those who showed up, was a more Democratic-friendly composition, thanks to an influx of people who don’t usually vote in midterms.
But Ghitza also points out some details that may challenge the conventional wisdom about 2018: Democratic gains in 2018, relative to 2016, were actually the largest in rural areas, not suburban areas, and the bulk of the impact in 2018 came from winning back the votes of 2016 Trump voters, not from the newly activated voters.
There’s a wealth of information in Ghitza’s writeup, and it’s worth reading the whole thing. However, I’ll zoom in on three of the key points:
THE 2018 ELECTORATE WAS SIGNIFICANTLY YOUNGER AND LESS WHITE THAN 2014
One of Ghitza’s key takeaways is that, as he puts it, “the composition of the 2018 electorate resembled recent Presidential electorates much more than recent midterms.” In other words, midterms, almost always, have lower turnout than presidential elections; as important as midterms may be, they don’t command the same level of attention in the media as presidential elections, and the tens of millions of people in this country who don’t follow politics closely tend to tune midterms out. That means midterm participation tends to skew toward the most reliable categories of voters, who tend, not coincidentally, to be older and whiter than average.
In other words, you get a pattern where gradually over time, the electorate becomes less white, as the country as a whole becomes less white. However, it’s a jagged pattern, bouncing up and down every two years, because the electorate gets whiter in each midterm, compared with the presidential election before it (because it’s a smaller pool of the electorate), but then gets even less white in the following presidential election (compared not just to the midterm, but to the previous presidential election).
But, somewhat remarkably, 2018 broke that pattern: the electorate was just as white in 2018 as in 2016, and for that matter, actually less non-college white in 2018 than in 2016. The following table illustrates the pattern, showing the percentage that each group made up of the total electorate (so each column should total up to 100):
|
2018 |
2016 |
2014 |
2012 |
2010 |
2008 |
2006 |
| NON-COLLEGE WHITE |
47 |
48 |
52 |
50 |
53 |
52 |
56 |
| COLLEGE WHITE |
28 |
27 |
28 |
26 |
26 |
25 |
26 |
| BLACK |
12 |
12 |
11 |
13 |
11 |
12 |
9 |
| LATINO |
8 |
9 |
6 |
7 |
6 |
7 |
5 |
| ASIAN/OTHER |
5 |
5 |
4 |
5 |
4 |
4 |
4 |
If you look at the non-college white row, reading right to left, you can see the pattern: the midterm electorate in 2010, for instance, was 53 percent non-college white, but that fell to 50 percent in 2012 and then bounced back up to 52 in 2014. What happened in 2018, though, is that the number kept falling from 2016 (to 47, from 48), instead of bouncing back up. Also, you’ll notice in the black row, that the black percentage of the electorate bounces up in each presidential election and down in midterms (like 11 in 2010, up to 13 in 2012, then down to 11 in 2014), but that also didn’t happen in 2018: the black percentage stayed steady at 12 from 2016.
In retrospect, it’s pretty amazing that the Democrats managed to gain as many seats as they did in 2006, given how large the non-college white portion of the electorate was back then! But keep in mind that it was a much less racially and educationally polarized electorate even just a decade ago, with rural non-college whites much more willing to elect a Blue Dog Democrat to the House then than they are now.
You can see a similar pattern if you look at the age composition of the electorate over the same years. While there isn’t an overall gradual trend toward a younger electorate the way there is toward a less white electorate (and there shouldn’t be, because there’s no demographic change reason for that; if anything, the average age within the electorate should be getting a bit older, as life expectancies get longer), the 2018 midterm saw much less of a collapse in 18-29 voters relative to 2016 than we saw between 2008 and 2010, or between 2012 and 2014.
|
2018 |
2016 |
2014 |
2012 |
2010 |
2008 |
2006 |
| 18-29 |
11 |
14 |
8 |
14 |
9 |
16 |
9 |
| 30-44 |
21 |
22 |
18 |
23 |
20 |
24 |
23 |
| 45-64 |
39 |
39 |
43 |
40 |
45 |
40 |
44 |
| 65+ |
29 |
25 |
31 |
23 |
26 |
20 |
24 |
We may never see youth turnout in a midterm that matches youth turnout in 2008 (when Barack Obama was on the ballot for the first time), but the good news is that youth turnout in 2018 fell to only 11 percent of the electorate from 14 in 2016, compared with a collapse to 8 percent in 2014 from 14 percent in 2012. Likewise, the spike in senior citizen turnout only went from 25 in 2016 to 29 percent in 2018, compared with a gain from 23 in 2014 to 31 in 2014.
Keep in mind that the “spike” here isn’t caused so much by a rush of elderly voters showing up unexpectedly, but by subtraction; seniors make up a bigger percentage of the electorate because they keep showing up while young voters drop out. So in 2018, when the senior share was smaller, it’s because a lot fewer young voters dropped out for the midterm than usual.
THE BIGGEST DEMOCRATIC GAINS IN 2018 WERE IN RURAL AREAS
Ghitza’s claim that the biggest Democratic gains were in rural areas bumps up against the dominant narrative of the 2018 election, that it was a revolt of the affluent suburbs against the Republican party. However, as he points out, the reason it feels that way is because the affluent suburbs are where many of the most high-profile marginal seats that the Democrats picked up are located (for instance, Orange County, California, or the suburbs of Atlanta and Dallas). But those CDs were already reasonably close, in terms of their presidential vote in 2016, so the winning Democratic candidates there didn’t have to improve much on the totals that Hillary Clinton got in those places in order to get over the 50 percent mark.
Instead, the biggest numeric gains, margin-wise, tended to be in rural areas; you can see this in the map at the top of the article, where the darkest shades of blue reflect the biggest swings, and are located in places like rural Minnesota and West Virginia. The reason that these swings didn’t result in pickups that affected the total number of House seat pickups is because that either these seats were already in Democratic hands—in the cases of, for instance, Minnesota’s 7th district, Oregon’s 4th district, or Wisconsin’s 3rd district—or because the Democratic candidate started out in too deep a hole from the presidential baseline, and a substantial overperformance still left them short of 50 percent—in the cases of, say, Iowa’s 4th district, Montana’s at-large seat, or West Virginia’s 3rd district.
In fact, one potential problem with this kind of analysis may be that these are races with some unique quirks about them, such as strong, long-time Democratic incumbents who either won by huge margins in a closely divided district (like Peter DeFazio in OR-04 or Ron Kind in WI-03), or by a narrow margin in a dark-red district (like Collin Peterson in MN-07), a dark-red open seat with a compelling Democratic candidate (Richard Ojeda in WV-03), or Republican incumbents with terrible reputations who barely managed to hang on in dark-red districts (Steve King in IA-04 or Greg Gianforte in MT-AL).
It might be more interesting to see a map comparing 2018 House results with 2014 House results (rather than the 2016 presidential results) to try and smooth out those quirks. Regardless, though, the strong overperformances here still have a positive message for 2020, that these mostly white rural areas with a lot of voters willing to ticket-split aren’t lost for good. And that even if we don’t actually “win” those areas in the presidential vote in 2020, Democrats can still potentially bank a lot of votes there from former Trump voters, which can be instrumental in turning, for instance, Iowa and Wisconsin back into the blue column in 2020.
VOTE CHANGERS, NOT NEW VOTERS, HAD THE BIGGEST IMPACT IN 2018
This may be the most difficult claim by Ghitza for people to wrap their heads around, not just because it contrasts with the way that many Democrats have taken as an article of faith that the way to win going forward is by ramping up young voter and non-white voter turnout, but even because it seems to contradict his earlier data about how well Democrats improved turnout from those groups in 2018. But the math seems to work out; I’ll explain. It may help to look at the table from his article, which I’m including here:
Catalist breaks voters down into three categories: people who voted in both 2016 and 2018, people who voted only in 2016, and people who only voted in 2018. It turns out that people who voted only in 2018 are considerably more Democratic-leaning than the ones who voted only in 2016; the 2018-only crowd broke 60 percent for Democrats and 39 percent for Republicans.
But the 2016-only crowd was also pretty Democrat-friendly (they went 50 percent for Democrats and 42 percent for Republicans); it’s the people who vote every time who are the ones you have to watch out for. And moreover, there simply aren’t that many 2018-only voters in the grand scheme of things; there are 14.4 million of them (which works out to only 13 percent of the total 2018 electorate, when you include them with the 99 million both-times voters).
The good news, though, is that there was a pretty big swing within the much larger group of people who voted in both: they broke 47-47 in the 2016 vote, but broke 52 Democratic and 47 Republican in the 2018 House vote. Getting 5% of 99 million people to change their minds about something is a pretty big deal! So Catalist’s calculation is that those vote-switchers were responsible for 89% of the total Democratic gain in 2018, while the new voters (who broke “only” 60-39 in the Democrats’ favor) were responsible for 11% of the total gain.
That’s a very interesting finding, but it’s still mostly academic; it doesn’t mean we should go all-in on persuasion and scrap our plans for pushing turnout. You still need a hell of a lot of both! Considering how many of those House pickups in 2018 were won by only a few thousand or even a few hundred votes, those seats wouldn’t be in the blue column today if it weren’t for the newly activated voters as well as the vote-switchers.
And it’s certainly not a zero-sum game; a broad, optimistic message that appeals to infrequent voters can just as easily appeal to swing voters, or vice versa, rather than one pushing the other away. (See, for instance, Obama 2008.) Beyond that, a good campaign can do the walking-and-chewing-gum of orienting its urban GOTV efforts toward turnout and its rural GOTV efforts toward persuasion, or, for instance, differentiate by running turnout-themed ads in urban media markets and more persuasion-themed ads in rural media markets.
