Yair writes:
The 2022 election defied conventional wisdom and historical trends. In a typical midterm election year with one-party control of the presidency, House and Senate, the incumbent party would expect major losses. Instead, Democrats re-elected every incumbent senator and expanded their Senate majority by a seat, won the overwhelming majority of heavily contested gubernatorial elections, gained control of 4 state legislative chambers, and only narrowly lost the U.S. House. . . .
Unlike other recent midterm years, our analysis shows a stark contrast between the electorate in areas with one or more highly contested House, Senate or gubernatorial races versus those with less contested races. . . .
Their key findings:
Gen Z and Millennial voters had exceptional levels of turnout, with young voters in heavily contested states exceeding their 2018 turnout by 6% among those who were eligible in both elections. Further, 65% of voters between the ages of 18 and 29 supported Democrats, cementing their role as a key part of a winning coalition for the party. While young voters were historically evenly split between the parties, they are increasingly voting for Democrats.
Extreme “MAGA” Republicans underperformed. . . . Candidates who were outspoken election deniers did 1 to 4 points worse than other Republicans, contributing to their losses in important close races. Of course, election denial is one of many extreme positions associated with “MAGA” Republicans, so this analysis likely reflects relatively extreme stances on other issues, including abortion rights . . .
Women voters pushed Democrats over the top in heavily contested races, where abortion rights were often their top issue. After Republican-appointed justices on the Supreme Court overturned abortion rights, a disproportionate number of women voters registered to vote in states with highly contested elections. At the same time, polls showed Democratic women and men indicating they were more engaged in the election. While relative turnout by gender remained largely stable, Democratic performance improved over 2020 among women in highly contested races, going from 55% to 57% support.
Democrats largely retained their winning 2020 coalition in heavily contested races, with some exceptions. Turnout and support among voters by race, education, gender, and other demographic factors remained relatively stable in heavily contested races. . . .
Democratic support among young voters is partly due to the diversity of this group, as America becomes more diverse over time. But that is not the whole story. Democratic support was higher among young voters of color, both nationally (78%) and in highly contested races (also 78%).9 But support among young white voters rose between 2018 (53% national, 52% highly contested races) and 2022 (58% nationally, 57% highly contested races). This 5-6 point support change is notable, indicating a broad base of Democratic support among young voters across the country. . . .
By any historical standard, 2022 turnout was high. Nationally it did not match 2018’s record-breaking turnout of 118 million votes, but it did reach 111 million ballots cast. . . . However, these national turnout numbers mask important differences at the state and congressional level: namely, that turnout matched or even exceeded 2018 turnout in the most highly contested elections in the country. . . . in these heavily contested races with higher turnout, Democratic candidates generally prevailed. . . . While some of these turnout trends are driven by population increases, it mostly reflects the high turnout environment that has been consistent from the 2018 election onward. In highly contested elections — where voters know the race could be decided by a small number of votes and campaigns invest resources into engaging voters — turnout often matched the historic “Blue Wave” election in 2018. . . .
Campaigns and voter registration groups invest significant resources in identifying, registering and mobilizing new voters as they seek to grow their coalitions. The high turnout era has been marked by millions of new voters entering — and staying in — the electorate. . . .
Lots more numbers and graphs at the above link.
> In a typical midterm election year with one-party control of the presidency, House and Senate, the incumbent party would expect major losses.
I know that there are historical tends for this, and there’s much speculation about the underlying cauality…but I have never fully understood how that putative causal mechanism plays out.
For example, I have a hard time figuring out how the explanations listed for how relatively well Dems performed in 2022 would be somehow in turn be casually linked to it being a mid-term election. How would, say, underperformance of MAGA Republicans be tied to it being a mid-term election after a period of Dems controlling the executive branch?
Joshua:
It’s the pendulum. There’s a tendency for parties in power to go beyond public opinion. There’s a logic to this: when you’re in power is when you have the opportunity to make policy. But then the swing voters want some moderation. 2022 was different from earlier midterms in that voters were reminded that the Republicans control a key branch of government, and by getting what they wanted on abortion, the Republican party went beyond public opinion. But it’s not like the Democrats had any sort of landslide: abortion’s only one issue, and on the economy it was reasonable for swing voters to feel that Biden and the Democrats had gone a bit too far to the left.
Andrew –
OK – I see the logic there, but it still feels a little post hoc-ish to me. I tend to think that at the national level, voting is less issue-based and more identity-based. Is there data that really show that abortion, say, explains voting trends in 2022, as opposed to a kind of post hoc reasoning? How do we really know that there isn’t a sample size aspect popping up here.
I Googled and found this article and thought it’s pretty good. It discusses, among other issues, just how strong the historical trends are.
https://fivethirtyeight.com/features/why-the-presidents-party-almost-always-has-a-bad-midterm/
The “differential turnout” and hatred points from that article seem pretty on the mark to me. Seems to me that hatred is a powerful motivator for voting and hatred grows with the feeling of having recently lost and being out of power whereas having won recently creates kind of complacency. I’m not sure how much an overreach by the parties in power explains the hatred component. Some, for sure, but I happen to think that much of the stronger aspects of party affiliation are more hardwired identity orientation than issues based. But then again, it’s really the swing voters who switch parties one election to the next where much of this plays out and indeed, maybe they are more issues-based.
Maybe the real problem comes in trying to come up with a generalized rule.
Joshua:
Most voters are not swing voters. But some are. Take a look at the first graph in the above-linked report. Since 1980, the Democratic share of the congressional vote has fluctuated between 46% and 56%. That’s the range where there are swing voters—enough to change party control of the legislature, even though only 10% of voters.
The “midterm penalty” is one of the most consistent patterns in American politics, going way back to the nineteenth century. There have been almost too many explanations proposed for it–more than one might have some truth to it, but they can’t all be true.
Personally, my favorite explanation is one that doesn’t seem to get a lot of press. The idea is that in a presidential election year, the victorious presidential candidate has a “coattail effect,” and carries a lot of congressional candidates to victory. Some of those winning candidates turn out to be pretty weak. Either they’re just not very appealing to their own constituents, or they’re in very marginal districts, or both. In the next election, they no longer have presidential coattails to cling to, and some of them go down to defeat. That is, the midterm penalty is just a “loss of coattails” effect.
This would explain why the midterm penalty is more pronounced on the House side than on the Senate side. It also explains why, if anything, a party that has an especially good presidential year often gets clobbered in the Senate six rather than two years later. (Look, for example, at 2008 and 2014.)
In 2020, note that Biden didn’t really have coattails. The Democrats lost a fair number of House seats even though their candidate won the popular vote by 4.5 points, which is pretty much unprecedented. My theory will then also explain why there wasn’t much of a midterm penalty in 2022.
Kind of disconcerting that being an extreme election denier only has a penalty of 1-4%.
Well, yes and no. The problem is that the left keeps getting scarier too.
By the way, this graph is what it looks like when you have a real “regression discontinuity design” effect to detect.
I just finished reading the Catalist report. It’s not bad, but it could have been a lot better. Instead of putting elections into just two categories (“heavily contested” and “other”), I’d like to see some scatterplots that show, for example, Democratic shift 2020–2022 versus Democratic vote-share in 2020.
One of the main points in the report is that Democrats did especially well in heavily contested elections in 2022. And how do they determine whether an election was heavily contested? They follow the Cook Report. One problem is that some elections counted as heavily contested only because Democrats were already showing signs of doing well there. In places where Democrats failed to show signs of strength, elections that might have been heavily contested didn’t get categorized that way. For example, if Democrats had polled better in the gubernatorial elections in Texas, Florida, and Ohio, Cook could and would have rated those elections as lean or toss-up. There’s a bit of a “chicken-or-the-egg” problem here.
The one time the report does show an actual scatterplot (Fig. 8), I find the conclusion pretty questionable. They’re inferring Democratic-support change among white college-educated voters from a plot of change by census block versus percentage of the census block’s population that consists of white college-educated voters. But, obviously, there are a lot of other ways census blocks will differ when you’re comparing those that have a large percentage of white college-educated population versus those with a small percentage.
The report also should have done a better job of emphasizing which data comes straight out of voter files, and which comes out of exit polls and modelling. Obviously, the latter are considerably less solid. (How exactly do they know that 62% of Gen-Z/Millennials voted Democratic in 2022 House elections? It’s not obvious, and it’s not obvious what the error range ought to be.)
Finally, the report entirely ignores the third-party vote. When shifts of two or three percentage points matter, that’s not really justifiable.