You end this helpful piece with the following statement "This means, yes, special elections are less predictive of outcomes than before — but that does not mean they are useless or that we should ignore them."
But I'm curious, given the strong correlation in '16, '18, '20 and '22 between special elections and general election outcomes, is it not noteworthy that this didn't hold in '23 and '24?
Should Jodi Ernst be getting worried. Between these 2 special elections and the Selzer poll before the election, I wonder if Iowa has an unusually large split between highly engaged and disengaged voters.
Great piece. One observation: In 2022, PA-SD-36 was uncontested. The GOP incumbent got 92K votes. Interestingly, when it was last contested in 2018 (under slightly different lines), total turnout was not a *lot* higher—105K.
50K is low in comparison of course but seems high to me for a special!
Thanks! Any idea what the raw 2024 Pres turnout was in SD-36? We can back into a turnout estimate for 2025 using that (assuming the SD turnout was same as PA, at 77%, which is crude but the best we have).
This overstates the impact. Clinton won the popular vote by 2.1% in 2016. Trump won by 1.5% in 2024. So Trump did 3.6% better relative to the country in 2024. So if Democrats are doing 10% better than Trump in these specials, it is not 1% worse than in 2017. It is really about 4.6% worse relative to the country.
Democrats are doing well so far, but not close to 2017 yet.
I don't think the gap is quite so big as you're suggesting. Dems are also doing 7 pts better than Biden's 2020 numbers this year. Since he won nationally by more than 4 points, that's D+11 so far. Compared to Harris, it would be about D+9, and versus Clinton in 2017, D+13. So yeah, somewhat behind, but these to me are all in the same realm.
If the margin relative to national numbers is 4% worse than in 2017 it is not quite the same ballpark. I just thought the comment was a little too strong on the special election performance (still very small sample). I also worry that state races are more variable than federal races. We'll probably get positive relative results in Florida tomorrow so that will change this somewhat, but I also discount numbers in less competitive races. They are just not replicable in higher salience and higher turnout elections. The liberal candidate should be favored tomorrow in WI, but I would be shocked (and pleasantly surprised) if the margin approached 10%.
You end this helpful piece with the following statement "This means, yes, special elections are less predictive of outcomes than before — but that does not mean they are useless or that we should ignore them."
But I'm curious, given the strong correlation in '16, '18, '20 and '22 between special elections and general election outcomes, is it not noteworthy that this didn't hold in '23 and '24?
And have you looked at this data:
https://substack.com/@smartelections/p-152549741
Is there a reason the California elections are not listed on the table? Is it because of them being a top 4 primary instead of one-on-one contests?
Should Jodi Ernst be getting worried. Between these 2 special elections and the Selzer poll before the election, I wonder if Iowa has an unusually large split between highly engaged and disengaged voters.
Great. What does it really change? The democrats when in power aren’t fixing things. They are just trying to return to “normal” which is long gone
I wonder if candidate quality has to do with all of these
Good read. And an intriguing point
Great piece. One observation: In 2022, PA-SD-36 was uncontested. The GOP incumbent got 92K votes. Interestingly, when it was last contested in 2018 (under slightly different lines), total turnout was not a *lot* higher—105K.
50K is low in comparison of course but seems high to me for a special!
Thanks! Any idea what the raw 2024 Pres turnout was in SD-36? We can back into a turnout estimate for 2025 using that (assuming the SD turnout was same as PA, at 77%, which is crude but the best we have).
Yes, we have those figures... somewhere on Jeff Singer's hard drive. 😁 He's on vacation right now, but will share when he's back.
This overstates the impact. Clinton won the popular vote by 2.1% in 2016. Trump won by 1.5% in 2024. So Trump did 3.6% better relative to the country in 2024. So if Democrats are doing 10% better than Trump in these specials, it is not 1% worse than in 2017. It is really about 4.6% worse relative to the country.
Democrats are doing well so far, but not close to 2017 yet.
I don't think the gap is quite so big as you're suggesting. Dems are also doing 7 pts better than Biden's 2020 numbers this year. Since he won nationally by more than 4 points, that's D+11 so far. Compared to Harris, it would be about D+9, and versus Clinton in 2017, D+13. So yeah, somewhat behind, but these to me are all in the same realm.
If the margin relative to national numbers is 4% worse than in 2017 it is not quite the same ballpark. I just thought the comment was a little too strong on the special election performance (still very small sample). I also worry that state races are more variable than federal races. We'll probably get positive relative results in Florida tomorrow so that will change this somewhat, but I also discount numbers in less competitive races. They are just not replicable in higher salience and higher turnout elections. The liberal candidate should be favored tomorrow in WI, but I would be shocked (and pleasantly surprised) if the margin approached 10%.
I think you mean next week, not tomorrow!
Not sure if you've read this analysis from my colleague Daniel Donner, but taken in aggregate, the specials have correlated closely with the House popular vote ... except for last year: https://www.dailykos.com/stories/2023/10/31/2194553/-Special-elections-are-closely-correlated-with-the-House-popular-vote-Still-true-six-years-later