Evan Scrimshaw: Midterms and Math, and Why The Midterm Penalty Is Dead
(Evan Scrimshaw returns to our blog for this excellent data-backed piece on why the conventional wisdom may not hold in 2022. Follow him at @EScrimshaw)
Why did Democrats do really well in 2018, gaining 40 House seats and gaining control of a chamber many had predicted would be red controlled for a decade after the 2010 disaster? It's a question with two common answers, and which answer you believe is key to your theory of politics - or, at least, what kind of year 2022 will be.
The first theory is that it was a midterm, and the out party always does well in a midterm, so Democratic gains were merely an extension of that fact. This theory has the benefit of history behind it, kind of, and facts, again, kind of. More accurately, a defence of this position has the benefit of having a lot of true statements you can say… it's just that a lot of it isn't actually relevant to 2022 anymore.
The second theory is that 2018 was about turnout, and that Democrats did well because the power of educated white voters was amplified from 2016 to 2018, and that Democrats did better with the rural white voters that turned out in a midterm year - meaning that the decline in rural white turnout was concentrated in Trump supporters. It's a theory that 2020 backs up in many ways, because we saw that Trump appeal again bring out many of those 2016-but-not-2018 Trump voters, and many more new non-degree white voters, and they overwhelmingly broke for the President. It's also backed up by Georgia, where without Trump on the ballot, the share of the electorate who were white without a degree fell from 36% to 34%, despite Trump still being in office, controlling the newscycle, and holding that big rally the day before.
The first theory is held by many in the chattering classes, many of the same people who shared a belief that Democrats needed a "miracle" to win Georgia (yes, I'm still salty about that Mike DeBonis tweet, why do you ask), and their response is summed up by Nathaniel Rakich's blithe listing of midterm defeats going back 100 years. The problem with that tweet is that what happened in 2012 is barely relevant to 2022 politics anymore, let alone 1922, and if Rakich, who I respect greatly, seriously wants to suggest it does, he needs more than nonsense tweets to prove his argument. The problem for people making his argument is that it ignores two basic facts - polarization, and history.
We'll start with history, and the fundamental fact that for most of American history, most people could see themselves as voting for either party. Coalitions shifted all the time, and even into the 2000s, conservative Democrats could win ruby red states and Congressional Districts at the same time as those same voters would choose Republicans up ticket. Democrats held a majority of the Tennessee House Delegation as recently as 2010, before losing three of their members in the 2010 midterms, and Democrats could run the right candidates to win landslides in Wyoming and Tennessee as recently as 2006. Of the 36 states which elected Governors in 2006, 16 of them voted for a different party than they had done so in 2004. In 2018, that number fell to 8. The last two Presidential years have seen exactly one state split their ticket for President and Senate, a far cry from the 6 who split their tickets in the halcyon year of 2012 alone.
There were a lot of swing voters in the past because there were a lot of white people either willing to change their vote up and down the ballot, or split their tickets, because there was not a fundamental difference between what Reagan could do with Tip O'Neill or without him there. Things changed on the margins, but the differences between parties was fairly subtle, and therefore voters were a lot more volatile. Now, the differences couldn't be starker, as the GOP has tried to roll back decades of social progress and Democrats have started to tentatively venture out into being a real left wing party. In the past, a wealthy, white person could freely vote for the GOP, secure that Roe was settled law and not seeing a fundamental difference between Bill Clinton and George Bush on gay rights. When Democrats go about signing onto things like the Defence Of Marriage Act, it's hard to be super compelling that you must vote for the Democrats to protect your social liberalism.
Whereas in the 90s there was a settled consensus on issues like abortion and gay rights, now there is a chasm between the parties, with the GOP an active threat to both a women's right to choose and a gay couple's right to marriage. Roe and Obergefell are on the table because of the Trump Presidency, and it shouldn't be surprising that the voters whose social liberalism took a back seat to tax cuts aren't willing to trade social rights of themselves and their friends for a few thousand dollars anymore.
All of this is, of course, just more of the same rhetorical flourishes that I find so laughable when the Doomers partake, so let's get to the actual math. Why did Democrats do better in 2018 than 2020? Because the electorate got more Republican between those two years - it got more non-degree white, and the Hispanic population got Trumpier. The share of the electorate that was white, and didn't have a degree, was 40% in 2018, down from the 44% of 2016. In 2020, however, it spiked back to 43%, coming at the expense of non-white voters (27% of the electorate in 2018, 26% in 2020) and whites with a degree (33% in 2018, 31% in 2020). The opinion of Trump didn't change for whites with a degree between those two elections - Republicans lost them by 4% in 2018 and 6% in 2020, per the Fox exits both times, but where there was real slippage was with whites without a degree - where Democrats lost them by 25% three months ago, but only by 20% in 2018. As white, non-degree turnout surged, the GOP's margins with that group also surged, making it pretty clear that those surge voters were Trump supporters.
Hispanics were the other problem area for Democrats, but south Texas tells a story of voters not so much changing their views, but new voters breaking heavily to the GOP. In all those heavily Hispanic counties in the Rio Grande and nearby, Democrats got their usual number of votes - the GOP just got a ton more than they ever have before. It fits the pattern of low-propensity voters turning out and breaking for the GOP, but if your electoral success is based on low propensity voters, then you're in a lot of trouble when they don't turn out.
In 2014, Cory Gardner and Rick Scott both underran Mitt Romney's vote share with white voters from two years previous in Obama states, but both won where Romney didn't. Mitt got 61% of the white vote in Florida and 54% of the white vote in Colorado, but Scott won with only 58% and Gardner only got 50% on route to victories. How is this possible? In 2014, non-white Democrats didn't turn out to vote, and therefore smaller margin leads with white voters weren't overran by huge Black and Hispanic margins as they had been two years previous. The story of 2014 wasn't some huge number of Obama-GOP voters, but of Republicans turning out and Democrats not. Those very same 2014 midterms featured a 35% white college electorate, which fell to 30% in 2016. The problem for Democrats was that in 2014, whites with degrees still voted for the Republican party overwhelmingly. Now they don't.
Look at Ohio, and you'll see what happened clear as day. Why did Mike DeWine win by 4% in 2018 but Donald Trump win by 8% in 2020? The share of the electorate that was white without a degree went from 52% to 55%, and Trump stretched DeWine's 23% margin with those voters to a 29% win, on the backs of his surge voters. Whites with a degree were constant - DeWine won them by 3%, and Trump exactly matched that. Wisconsin saw a very similar pattern - non-degree whites made up 58% of the electorate in 2020, up from 54% in the midterms. Trump really improved on Scott Walker's results in the southwest of the state because of his surge voters, but Biden did a bit better than Evers in Green Bay, so the surge wasn't enough to cost Biden the state. Even in Georgia, the Trump surge voters showed up in November, but they didn't show up again two months later - an inauspicious start to the 2022 cycle for Republicans.
The answer to why there would be a Midterm Penalty for Democrats this year is a hard one. It requires one of three things - a high number of swing voters, suburban reversion, or Hispanic non-reversion - and in some cases, two of those three at minimum. If you want to start talking about national environments in the R+9 range, you'd need all three. The high number of swing voters argument has already been addressed, so we can swiftly move along there. Suburban reversion? I can't rule it out, but there's a ton of reasons to doubt it, from the fact that Ossoff did better than Biden did in Gwinnett two weeks ago to the fact that this is happening everywhere in the English speaking world, but I've done this rant many times before so I'll spare you it again. And, on the Hispanic question, there's this proof, courtesy of Whitfield County Georgia, of Hispanic areas reverting back to their 2018 partisanships as soon as Trump is off the ballot.
Is this argument airtight that 2022 will be better than 2020 was for Democrats? Of course not. Nobody is making that sort of facile argument, but it does seem like people are penciling in a GOP-friendly 2022 but then adding in the same rote, trite "but nothing is certain" disclaimers I spent an entire year adding to my 2020 work - but, of course, I never believed Trump would actually win. I just wrote it because that's what you do, as it feels many are already doing about 2022.
The Midterm Penalty was a very real thing for a very long time, a real part of American politics that explains a lot of the past. But what it isn't is meaningful moving forward. I wrote for this site the morning after Georgia that "people only remember the "what" of history without remembering the "why" of history," and it is bearing itself to be true again in 2022 discourse. There are very real reasons to think Democrats will have a good 2022 - a more educated electorate, a Senate map where our Hispanic exposure is limited and the GOP has huge amounts of danger if their rural turnout drops, and House lines that are likely to be advantageous to Democrats in California compared to current results.
What's the argument for why it'll be a bad year? Blithe nothingness about how voters used to be a lot less partisan and a lot more willing to split tickets, as if that day will suddenly return? A prayer that Trump-style rural turnout will be a reality without him on the ballot, which we've already seen doesn't hold up in Georgia? If you just default to 2022 being a 1994 style wipeout of the in-power party, that's your right, but that was a different era. To believe that is possible, you'd also have to believe that the GOP could win Cobb County by 13% while losing statewide in Georgia, which is so patently absurd as to prove why the past is only marginally useful in these conversations.
The first piece I wrote for this site referred to a "final, twisted form of American exceptionalism," referring to the failure of Americans to understand global politics and how that ignorance leads to people drawing the wrong conclusions about why things happen in US politics. I was wrong, then - not about that ignorance being dangerous, or about it being a twisted form of American exceptionalism. I was wrong about it being the final form. Turns out that many in the United States don't want to actually understand their history or what caused it, content to remember the past without any understanding of it. It has already led many to saying that Republicans would win in Georgia, and it will continue so long as American exceptionalism causes so many to focus on the "what" of history and not the "why."