OH-Sen 2022: So you’re telling me there’s a chance

[S019]

Favorability numbers are a useless data point these days.

[wbrocks67]

Favorability numbers are a useless data point these days.

Not necessarily - they're good for seeing if attacks are working on particular candidates or not. Vance may full well win simply because he's a Rep. But the story this entire cycle - from the summer to now - has been his unpopularity with voters.

[S019]

Favorability numbers are a useless data point these days.

Not necessarily - they're good for seeing if attacks are working on particular candidates or not. Vance may full well win simply because he's a Rep. But the story this entire cycle - from the summer to now - has been his unpopularity with voters.

People can and do vote for candidates that they don’t like (especially if they dislike both, which is where a lot of Ohio voters will find themselves (and in such a case, they tend to stick to their party)). I was looking back at the 2018 polls and Cordray was consistently hitting 48%, it’s quite clear that yet again lots of voters who disapprove of an Ohio Republican (hmmm where have we seen this before) will vote for one.

[Unelectable Bystander]

2020 NBC exit polls (this poll):

46% under age 50 (48%). Note this is even less realistic when considering it’s a midterm

39% R / 31%D / 30%I (37% R / 33%D / 26% I / 4.5% not sure). Why would the share of republicans drop in a midterm?

40% conservative / 21% liberal / 39% moderate (37% conservative / 21% liberal / 37% moderate / 5% not sure)

32% white evangelical Christians (26%)

34% under 50k income (36% under 40k income)

Not to mention that the issues don’t track with the recent student debt polls in the state or that NBC had 44% saying abortion should generally be illegal in 2020. Also, conducted by college students?

https://en.wikipedia.org/wiki/Central_limit_theorem

I am aware of the CLT definition but can you explain it’s applicability here? I’m assuming you mean that certain traits of a sample can be biased while the sample is fine as a whole. But isn’t the whole point of weighting a sample to account for the fact that you think your original sample is inaccurate? I’m not sure that you can look at those numbers from the weighted sample and say they will resemble the Ohio electorate with a straight face.

Weighting is about representation, not measurement. The central limit theorem applies to both. To take party ID as an example, you could credibly argue self-ID Republicans are underrepresented here based on the 2020 exit poll. The central limit theorem would posit that, assuming the sample is not structurally biased, this underrepresentation would be corrected by, potentially, the self-ID independent subsample leaning more Republican than it possibly should. This is a very clean example, and doesn't take into account intersection with other demographics that could also have a measurement error that potentially corrects a representation error, or vice versa. And of course the margin of error complicates all of this as well. This is why "unskewing" polls is a fool's exercise and generally discouraged and why most internal pollsters use subsamples only directionally.

Interesting, my two follow-up questions would be:

Would this still apply to features that are very strongly associated with voting behavior? For example, if the true proportion of self ID liberal and urban black voters this November in Georgia is 15% (arbitrary number) and a poll reports this group as 12%, is it possible for the sample to contain enough variation to still be accurate while underestimating a group by 3% where everybody in the group has basically a 100% chance of voting democrat? Or the converse, if a certain republican demographic is underrepresented, how much pro-R variation can you get from liberal and urban black voters?

Would you then say that we should be skeptical of polls with “perfect” cross tabs, such as ones with 8 different factors and the proportion looks right for each variable, because they have perhaps massaged their sample too much?

[wbrocks67]

This all seems overdramatic to me; the main reason is most likely the fact that Ryan is a prolific fundraiser while Vance is not. There's a reason GOP groups are blowing their wads in this race - bc Vance is broke. Ryan doesn't need outside spending. His money is going way further, and he also would be attacked anyway for "taking national Dem money" etc

If Ryan wins or loses, it's not going to be because of a few more dollars from Senate Majority PAC.

https://twitter.com/HenryJGomez/status/1579525077825380353

[Pollster]

Interesting, my two follow-up questions would be:

Would this still apply to features that are very strongly associated with voting behavior? For example, if the true proportion of self ID liberal and urban black voters this November in Georgia is 15% (arbitrary number) and a poll reports this group as 12%, is it possible for the sample to contain enough variation to still be accurate while underestimating a group by 3% where everybody in the group has basically a 100% chance of voting democrat? Or the converse, if a certain republican demographic is underrepresented, how much pro-R variation can you get from liberal and urban black voters?

Would you then say that we should be skeptical of polls with “perfect” cross tabs, such as ones with 8 different factors and the proportion looks right for each variable, because they have perhaps massaged their sample too much?

Short answers: yes and yes.

Long answer: for your first question, the margin of error and statistical noise alone prettymuch guarantee that the answer is yes. Do keep in mind that the margin of error on subsamples is greater than the sample overall (sometimes reaching double digits) - that is OK and to be expected - and makes situations like the one you're describing all the more likely. The answer to your second question is yes for mostly the same reasons. No poll is perfect - that is also OK and to be expected - and if one looks like it is, there was probably excessive weighting/data cutting done to make it so, to the point that the design effect is so large that the poll is useless.

[MT Treasurer]

In a D+0.5 to R+4 year there is no way a Democrat is going to
come that close nevermind win in a state that is R+2 to R+9

https://en.wikipedia.org/wiki/2016_United_States_Senate_election_in_Missouri

Elections are about actual people, not numbers.

You of all people are saying this? Wow maybe Ryan will win!

Not sure why you singled me out (although I appreciate living rent-free in yet another person's head — you’re the third one today, actually), but I’ve never been a believer in uniform/linear shift, nor do I believe that elections are predetermined by PVI. My position is that in the vast majority of cases, it can be assumed that both sides will run a reasonably competent campaign, which will then usually lead to underlying 'fundamentals' such as state lean/trend, demographic composition of the electorate, and overall national environment tipping the scales one way or the other.

This, however, is not one of those cases. Here, we have a largely dysfunctional campaign and an extremely inept and entirely unappealing candidate who is reliant on outside groups to do the job/messaging for him — his own campaign is practically non-existent! We had a somewhat similar case in OH-SEN 2016, with the lopsided advantage being in favor of Republicans in that race, so it’s not like there is no recent precedent for this. It is entirely reasonable to argue that the underlying 'fundamentals' are too strong for campaign quality to make the difference in terms of binary outcome in the Ryan vs. Vance race, but I’d certainly urge people to move away from repetitive, simplistic, and frankly lazy approaches to elections like "NV is inherently inelastic/Titanium Lean D because it’s urban and always narrowly voted D in the past" or "OH is a Trump +8 state, nothing else even matters." I say this as someone who believes that Vance will win (even if by a very embarrassing margin, with Ryan leading for most of the night).

[Amenhotep Bakari-Sellers]

In a D+0.5 to R+4 year there is no way a Democrat is going to
come that close nevermind win in a state that is R+2 to R+9

https://en.wikipedia.org/wiki/2016_United_States_Senate_election_in_Missouri

Elections are about actual people, not numbers.

You of all people are saying this? Wow maybe Ryan will win!

Not sure why you singled me out (although I appreciate living rent-free in yet another person's head — you’re the third one today, actually), but I’ve never been a believer in uniform/linear shift, nor do I believe that elections are predetermined by PVI. My position is that in the vast majority of cases, it can be assumed that both sides will run a reasonably competent campaign, which will then usually lead to underlying 'fundamentals' such as state lean/trend, demographic composition of the electorate, and overall national environment tipping the scales one way or the other.

This, however, is not one of those cases. Here, we have a largely dysfunctional campaign and an extremely inept and entirely unappealing candidate who is reliant on outside groups to do the job/messaging for him — his own campaign is practically non-existent! We had a somewhat similar case in OH-SEN 2016, with the lopsided advantage being in favor of Republicans in that race, so it’s not like there is no recent precedent for this. It is entirely reasonable to argue that the underlying 'fundamentals' are too strong for campaign quality to make the difference in terms of binary outcome in the Ryan vs. Vance race, but I’d certainly urge people to move away from repetitive, simplistic, and frankly lazy approaches to elections like "NV is inherently inelastic/Titanium Lean D because it’s urban and always narrowly voted D in the past" or "OH is a Trump +8 state, nothing else even matters." I say this as someone who believes that Vance will win (even if by a very embarrassing margin, with Ryan leading for most of the night).

LAXALT underpolled Sisolak in 2018 he was leading Sisolak most of the polls and lost but the only poll that has Vance ahead was Trafalgar and Emerson the other polls had Ryan ahead and DeWine isn't winning by 20 he is winning by six did Rs not learn what happened in red KY in 2019 Beshear won

As I have said Rs forgot we won AK and NY 19

I can come up with the same argument that many users put out and cross examine because that's what I Law Students do cross examination and I am in law school

We are on the eve a month before Election night it's a 52/45 map anyways we're gonnaa net WI, PA and OH, NC and UT are too close to call if we don't win Ut we're gonna win SD or FL or LA

The Rs we're plus 6 on the GCB in 2010)2014 so they are underpolling where they were in other midterm anyways

[Kahane's Grave Is A Gender-Neutral Bathroom]

In a D+0.5 to R+4 year there is no way a Democrat is going to
come that close nevermind win in a state that is R+2 to R+9

https://en.wikipedia.org/wiki/2016_United_States_Senate_election_in_Missouri

Elections are about actual people, not numbers.

If I could change one election result...

[It’s so Joever]

In a D+0.5 to R+4 year there is no way a Democrat is going to
come that close nevermind win in a state that is R+2 to R+9

https://en.wikipedia.org/wiki/2016_United_States_Senate_election_in_Missouri

Elections are about actual people, not numbers.

You of all people are saying this? Wow maybe Ryan will win!

Not sure why you singled me out (although I appreciate living rent-free in yet another person's head — you’re the third one today, actually), but I’ve never been a believer in uniform/linear shift, nor do I believe that elections are predetermined by PVI. My position is that in the vast majority of cases, it can be assumed that both sides will run a reasonably competent campaign, which will then usually lead to underlying 'fundamentals' such as state lean/trend, demographic composition of the electorate, and overall national environment tipping the scales one way or the other.

This, however, is not one of those cases. Here, we have a largely dysfunctional campaign and an extremely inept and entirely unappealing candidate who is reliant on outside groups to do the job/messaging for him — his own campaign is practically non-existent! We had a somewhat similar case in OH-SEN 2016, with the lopsided advantage being in favor of Republicans in that race, so it’s not like there is no recent precedent for this. It is entirely reasonable to argue that the underlying 'fundamentals' are too strong for campaign quality to make the difference in terms of binary outcome in the Ryan vs. Vance race, but I’d certainly urge people to move away from repetitive, simplistic, and frankly lazy approaches to elections like "NV is inherently inelastic/Titanium Lean D because it’s urban and always narrowly voted D in the past" or "OH is a Trump +8 state, nothing else even matters." I say this as someone who believes that Vance will win (even if by a very embarrassing margin, with Ryan leading for most of the night).

lol I don’t actually dislike you, you just have a reputation for having a…robotic view of the electorate. Anyways I don’t entirely disagree, patterns are meant to be broken, but I don’t think Vance’s campaign has been as bad as people are making it out to be.

[S019]

It’s entirely reasonable for Democrats to not want to be burned by Ohio for a third straight cycle, I don’t understand what is so unreasonable about it. Ohio polling is also so bad that until it actually gets it right, I won’t take it seriously (and no this isn’t the same as “all polling is trash.”) Those who think this race is Likely/Safe R (like myself) are having flashbacks to Ohio polls in October showing Biden up and don’t believe that pollsters have figured out how to poll Ohio, until they do, we have every right not to take then at face value.

[Orwell]

I don't think Ryan will win, but it should be a Lean R rating not a likely R rating.

[Ferguson97]

In a D+0.5 to R+4 year there is no way a Democrat is going to
come that close nevermind win in a state that is R+2 to R+9

https://en.wikipedia.org/wiki/2016_United_States_Senate_election_in_Missouri

Elections are about actual people, not numbers.

...The 2016 Missouri Senate election is your argument against the idea that Tim Ryan has a serious shot in November?

[Orwell]

Honestly, if you look at Vance on paper he seems like a really solid + tier candidate:

- Veteran
- Successful in Business
- Great Biography
- Above Average Name Recognition

But when you put all this into practice... Vance is pretty terrible because he's just not likeable. No one wants to have a beer with JD Vance.

[wbrocks67]

Vance comes out in support of nationwide abortion ban

https://twitter.com/Acyn/status/1579613029523951617

[Senator Incitatus]

I heard that Ryan just said Biden shouldn't run in 2024. Can anyone watching confirm? That would contradict other accounts that he's just playing for the donors rather than the voters (e.g., by criticizing Vance for campaigning with DeSantis) and would be an extremely bold move.

[wbrocks67]

and then Vance apparently said -

yikes

https://twitter.com/Thorongil16/status/1579614436222210048

[wbrocks67]

i'm sorry, what??

https://twitter.com/Acyn/status/1579616217954148352

[wbrocks67]

https://twitter.com/atrupar/status/1579616933003300866

[politicallefty]

If there's one thing that restrains any optimism I may have for this race, it is the relative competence of the Ohio Republican Party. They know how to win as much as the Florida Democratic Party knows how to lose.

[SnowLabrador]

If there's one thing that restrains any optimism I may have for this race, it is the relative competence of the Ohio Republican Party. They know how to win as much as the Florida Democratic Party knows how to lose.

I mean, they lost to Sherrod Brown in 2018. They're likely to beat him in 2024, admittedly. But I think a better analogy is the Florida Republicans, who organize like there's an election every week.

Back to this race, while Vance's responses are horrific, they're not going to cost him politically. Nobody who supported abortion rights or gay marriage was ever going to vote for Vance. He'll still win comfortably because it's Ohio and we can't have nice things.

[Chancellor Tanterterg]

If there's one thing that restrains any optimism I may have for this race, it is the relative competence of the Ohio Republican Party. They know how to win as much as the Florida Democratic Party knows how to lose.

Not really.  That used to be true from like 1994-2004, but they’re basically just a generic Trumpy state party at this point.  OTOH, the Ohio Democratic Party has been a dumpster fire for decades.

[wbrocks67]

Good strategy for the whole "Dems are soft on crime" narrative.

https://twitter.com/Acyn/status/1579619841685991426

[wbrocks67]

https://twitter.com/Acyn/status/1579622218388344832

[new_patomic]

Pick your own adventure spin