In the 2022 midterms, special elections were one of the klaxons alerting us to a changed political environment following the Supreme Court's decision to overturn Roe v. Wade in June. As a result of the decision in Dobbs v. Jackson Women's Health Organization, Democrats turned in a historic performance: They not only held the Senate but also gained a seat, and their losses in the House were far smaller than nearly all predictions.
And yet, when looking at the average Democratic performance in special elections, the overall picture barely budged with the inclusion of post-Dobbs specials, in part because there were only a handful of such races. But that aggregate number was still spot on. It indicated a slightly Republican environment, and Republicans indeed won the House popular vote by a small amount, about 3 percentage points. In other words, after Dobbs, everything changed, and yet nothing did. So what happened?
One of these is not like the others
It turns out 2022 was not a normal midterm year—at least, not normal compared with the last several midterms. Specifically, Democrats performed well (on average) in states that held competitive statewide elections. But elsewhere, it was just a fair-to-middling year, and there was a red surge in a handful of states home to a quarter of the nation's population, including Florida. Overall, though, when we’re toting up wins and losses, it’s the competitive races that we notice most. And so the cycle as a whole was a strikingly good one for Democrats.
Every cycle, and especially in midterms, different states have different political environments. But they're usually more-or-less randomly distributed among competitive and uncompetitive states alike.
If we dig into the past a bit, we can see that this was true in the prior three midterm elections. We can calculate the political environment in each state by comparing the margins of governor's races to Hillary Clinton's margin in 2016, and adjusting for Clinton's national popular vote win of 1.8 points. (We use gubernatorial races rather than Senate contests because Senate seats are up every six years rather than every four.)
We can see in the graphic at the top of this article that, on average, the political environment in competitive races for governor was nearly the same as the House popular vote in each of 2010, 2014, and 2018. But this was emphatically not the case in 2022: In competitive races, it looked like a return of the 2018 blue wave, a stark contrast to the GOP win in the House popular vote.
If we take a closer look at the states with no competitive statewide race for governor in 2022 (and add in Senate races for additional data, as we’re only making calculations for one cycle in this figure), we see that three of the four most populous states—New York, California, and Florida—had, on average, a more Republican environment than we saw in the 2010 midterms. That year, of course, a massive red wave washed ashore, and Republicans won the House vote by almost 7 points. (The remainder of the uncompetitive states, meanwhile, averaged out to a neutral environment.)
There's one more twist to the story, though. If we go back to the post-Dobbs special elections and calculate a political environment based on those numbers—either starting after the date the Dobbs decision leaked (May 2) or the decision was handed down (June 24)—those calculations are right in line with the observed numbers calculated from competitive statewide races. See the dark orange points in the graph below.
So that is the paradox of the special elections in 2022. They seemed to be warning us both that Dobbs had scrambled the entire picture and that the overall political environment had changed very little—and both things turned out to be true.
How can this be?
It could be a coincidence since we don't have very many post-Dobbs elections to go off of. (This is part of why we weren't making hard, number-based predictions in the summer of 2022.) If not, though, what is the link between special elections and competitive midterm elections in particular?
Digression: The tricky business of finding a trend in special election data
Did Dobbs really have an effect on special election results?
We have to tread carefully here to make sure we're seeing something that's really there and that we’re not just working backward from conclusions. There were, after all, only five special elections after the official release of the Dobbs opinion. We’ve seen in the past that five specials in a row can come in unusually high or low, just by chance. However, the Dobbs decision leaked well ahead of its official release.
First question: Is it reasonable to use the Dobbs leak as a breakpoint? We can use moving medians to check. For the number of points used in the moving median, five is pure noise and more than 20 covers too much time. The graph below shows moving medians of 10, 15, and 20 points:
In all three lines, we do indeed see a bump up on or about May 3, 2022. (There also might have been a downward bump around July 13, 2021, although there's no clear event to precipitate it other than a general lack of unicorns and flying cars that leads to the end of a honeymoon phase for a new president.)
Second question: Was there a real change between pre- and post-Dobbs results? The graph below shows lines indicating the averages over given time periods defined by events and the changes found above. We have two choices for a pre-Dobbs value (green lines in the figure below): an average from the beginning of the cycle through April 7, 2022, (the last pre-Dobbs leak special), or an average from July 13, 2021, to April 7, 2022. We also have two potential post-Dobbs numbers (blue lines in the figure below): the average following the early May Dobbs leak and the average following the late June Dobbs ruling.
Fortunately, there's not much ambiguity here: No matter which endpoints you choose, there's a sizable change, ranging from 9 to 14 points. In part, that’s because U.S. House races, of which there were six post-Dobbs leak, have less variation compared to state legislative races (meaning you need fewer of them to get a good read on the political environment). We can therefore be pretty confident we saw a real change with respect to numbers after the Dobbs leak. With only five elections post-Dobbs decision, however, there’s less certainty.
One last thing to check: Is there another variable besides time that can account for the changing values? We saw this in 2018, when what appeared to be sinking Democrat fortunes wasn’t at all—rather, a different mix of states started holding special elections. In the case of 2022, nothing else is immediately evident that could account for the observed changes.
We can conclude, then, that there was indeed a real change in special election results after the Supreme Court gutted half a century of precedent protecting abortion rights.
The salience of abortion
Now that we feel fairly confident the specials were genuinely capturing a signal and not noise, we can turn to the next question: What did post-Dobbs specials and competitive November elections—but not safe November elections—have in common? The obvious answer is that abortion rights were foremost in the minds of voters in the former, but not the latter. But is there evidence to confirm this?
In states like New York, California, and Florida, there wasn’t much doubt as to the outcome of major statewide elections. One might conclude, therefore, that there was no reason for voters to believe that access to abortion could be affected by the midterms. The opposite, however, was true in states like Michigan.
And we do have some evidence for this. For instance, in Michigan, 45% of voters said abortion was their most important issue. In Pennsylvania, it was 37%. In Florida, it was just 24%. Clinton barely lost all three states in 2016; in 2022, the Michigan and Pennsylvania governor's races were Democratic double-digit blowouts, while Florida was a double-digit Republican win.
What about the post-Dobbs special elections? Here, our evidence for the salience of abortion comes both from their timing and from the campaigns themselves. Five special elections happened the day after the Dobbs leak, May 3. Abortion was certainly the top story of the day. Then there were four more elections before the Dobbs decision was handed down; the effect of the Dobbs leak on these elections is more uncertain.
The next five specials (which were in fact the final specials of the cycle) were all for U.S. House seats. The first, in Nebraska’s 1st District, flew under the radar but was shockingly close. It took place just four days after Dobbs was handed down, so it’s reasonable to assume abortion was a key motivation for voters.
The last four specials, however, were more closely contested. And because of both Nebraska’s tight result and the thumping defeat of a referendum that would have allowed Republicans to restrict abortion rights in Kansas, they were more closely watched. In New York’s 19th District, for example, Democrat Pat Ryan heavily emphasized abortion rights in his upset of Republican Marc Molinaro.
Alaska Democrat Mary Peltola, meanwhile, won on a platform that emphasized abortion rights. Her pithy campaign slogan: “Pro-jobs, pro-fish, pro-family, and pro-choice.” The Democrats in the other two elections were, in each case, also supportive of abortion rights, though they did not center the issue in quite the same way. But on balance, it seems fair to say that in the post-Dobbs specials, abortion was at least an important issue, if not the central issue.
Alternatively, however, abortion may simply be more important to the sorts of people who vote in special elections. We don’t have any evidence to support this idea, however. In fact, quite the opposite: Older voters, who tend to be more consistent voters, are the least supportive of abortion rights.
Could this sort of thing happen again?
So now we know that special election results can tell us more than what the general political environment looks like. In the case of 2022, because of what elections were held when, we can see a relationship between a group of special elections and a group of November elections. But it’s tricky to figure this sort of thing out, and it’s best done after the elections are over. We also should be very cautious because looking at a subgroup of special elections necessarily gives us a smaller sample size, so only very dramatic changes can be identified.
What about 2024? Might we see a repeat, where specials show great performances by Democrats (as they in fact are), only to see Democrats perform well in the November general elections in just certain states? For that to happen, we’d need another split in salience: something that is important to special election voters and a subset of general election voters but not all general election voters. Or we’d need a sudden and dramatic change in the electoral environment after special elections have concluded—an October surprise the likes of which we've probably never seen.
But in a presidential cycle, such shifts are much more difficult to achieve, as politics are so heavily nationalized. This implies issue salience does not vary much by state, and even dramatic events seem to move votes only at the margins. After all, only a handful of states have been deemed to be competitive at the presidential level in recent decades, and yet voters still turn out at rates far exceeding the midterm elections even in the safest states. Much more likely is that 2024 ends up broadly in sync with the special election results, just as in prior years.
Special elections are closely correlated with the House popular vote: Still true, six years later
Democrats are crushing it in special elections this year. That's unambiguously good news