"We think this difference between the presidential and congressional results more likely points to Trump overperforming expectations, rather than downballot Republicans underperforming expectations."
Still reading but this is really encouraging. There seems to be just a three-point Magic Trump Factor. It *might* have applied to Congressional GOP in 2016 or 2020, but apparently didn't even help them in 2024 and almost certainly won't help in 2026, nor with JD Vance/Marquito Rubio/whoever atop the ticket in the future.
It's kinda wild that the same person has the two least popular presidencies in the mass polling era and started with a three-point bonus, but I guess being off-the-charts polarizing will do that to you.
Has to be spot on for David Valadao. Still, he'll likely lose to any decent Dem come 2026. Hope to see analysis for the Senate seats, like, how does Cornyn rate, since he's getting challenged by another Republican.
You explain this to some degree but I’m still a bit curious as to the decision to model candidate skill as opposed to two-way WAR (maybe just because I’ve only ever seen split ticket’s two-way version). I’ve typically assumed that the dichotomy between the two candidates is important enough to warrant modeling it in that way (that the two candidates’ separate skills are less important than the way they play off each other). But did this version come out statistically as sound as it would have if you used the two-candidate WAR? In terms of a single candidate’s skill being predictive of vote share
We model individual candidate skill for two reasons:
1) it allows us to replace individual candidates with simulated hypothetical candidates
2) it's more akin to how we think about candidate skill
Re: point 2, say representative A is a "highly skilled" candidate and won their most recent election narrowly against another "highly skilled" opponent and that representative B is a "highly skilled" candidate who won their most recent election in a blowout against a "low skilled" opponent. Say also they get redistricted such that they face one another in the next election. Even though representative B won in a blowout, modeling the individual skill directly allows us the model to infer that the matchup between A/B will be close.
I read through the analysis and skimmed the data. Two questions:
1. Do you expect mid term results (assuming they're not rigged by the repig dictatorship) to shift or skew differently b/c there is no presidential election?
2. do your data indicate that politics have become more rigid/inelastic and we're heading to a situation where control of the house will be dictated by a vanishingly small number of races, whether incumbents or fresh candidates are running for either party?
1) We may publish a piece exploring the model covariates in more detail (if it doesn't make sense for SIN, I may put together an independent writeup at the Data Diary) --- but yes, we observe that the outparty benefits in midterm elections. I believe Elliott also has a piece looking at the Democrat's overperformance in special elections this year.
2) Yes --- with the exception of lagged presidential vote, most (not all!) of the model's predictors have shrunken towards zero.
I really like authors who go out of thier way to give full credit to co-authors. Now I will read ans enjoy your approach to data analysis.
Impressive analysis!
(Minor note: in the first table, Valadao and Golden are incorrectly listed as 37% and 35%.)
Senate WARPs?
"We think this difference between the presidential and congressional results more likely points to Trump overperforming expectations, rather than downballot Republicans underperforming expectations."
Still reading but this is really encouraging. There seems to be just a three-point Magic Trump Factor. It *might* have applied to Congressional GOP in 2016 or 2020, but apparently didn't even help them in 2024 and almost certainly won't help in 2026, nor with JD Vance/Marquito Rubio/whoever atop the ticket in the future.
It's kinda wild that the same person has the two least popular presidencies in the mass polling era and started with a three-point bonus, but I guess being off-the-charts polarizing will do that to you.
Has to be spot on for David Valadao. Still, he'll likely lose to any decent Dem come 2026. Hope to see analysis for the Senate seats, like, how does Cornyn rate, since he's getting challenged by another Republican.
Plus a decent chance his district gets gerrymandered to be a near Dem lock in an anti-Trump 2026 election.
You explain this to some degree but I’m still a bit curious as to the decision to model candidate skill as opposed to two-way WAR (maybe just because I’ve only ever seen split ticket’s two-way version). I’ve typically assumed that the dichotomy between the two candidates is important enough to warrant modeling it in that way (that the two candidates’ separate skills are less important than the way they play off each other). But did this version come out statistically as sound as it would have if you used the two-candidate WAR? In terms of a single candidate’s skill being predictive of vote share
We model individual candidate skill for two reasons:
1) it allows us to replace individual candidates with simulated hypothetical candidates
2) it's more akin to how we think about candidate skill
Re: point 2, say representative A is a "highly skilled" candidate and won their most recent election narrowly against another "highly skilled" opponent and that representative B is a "highly skilled" candidate who won their most recent election in a blowout against a "low skilled" opponent. Say also they get redistricted such that they face one another in the next election. Even though representative B won in a blowout, modeling the individual skill directly allows us the model to infer that the matchup between A/B will be close.
I read through the analysis and skimmed the data. Two questions:
1. Do you expect mid term results (assuming they're not rigged by the repig dictatorship) to shift or skew differently b/c there is no presidential election?
2. do your data indicate that politics have become more rigid/inelastic and we're heading to a situation where control of the house will be dictated by a vanishingly small number of races, whether incumbents or fresh candidates are running for either party?
1) We may publish a piece exploring the model covariates in more detail (if it doesn't make sense for SIN, I may put together an independent writeup at the Data Diary) --- but yes, we observe that the outparty benefits in midterm elections. I believe Elliott also has a piece looking at the Democrat's overperformance in special elections this year.
2) Yes --- with the exception of lagged presidential vote, most (not all!) of the model's predictors have shrunken towards zero.
Is this being done in collaboration with the people over a Split Ticket? I know they released a WAR analysis towards the end of last year.