AL-Monmouth: It's complicated

[Sorenroy]

To all the people complaining: you realize these things happen on a continuum, right? Statistically it is much more meaningful to specify under which conditions which candidate has what chance to win than it is to say 'R+5' or 'D+4' without anything else.

Like it is just straight lazy and bad methodologically to give one top line score with no real details of the assumptions being made to arrive at that score. I applaud Monmouth and SurveyMonkey for actually speaking in reasonable terms statistically in an effort to improve the standing of real, mathematically rigorous polls. Polls like these will bring the polling industry back.

And no, this is not a 'hedge their bets' strategy. If the electorate is a traditional midterm one and Jones wins by 4, this poll is wrong regardless of if it had Jones+4 in some alternate scenario. This poll is, in essence, a total view of the landscape that allows the reader the leverage to make turnout assumptions and provides the model to produce a top line score from said turnout assumptions.

Can't stand the hate literally the best polls I've seen are getting from people with zero understanding of basic statistics.

I understand that, but at the same time, they should have the guts to pick one model as the "most likely", at least so people can look at it at a glance to get a sense of the race. I'm fine if they have other turnout models in the full report.

RCP  is running with this poll being a tie for what it’s worth. I don’t know if that’s because they’re using the 2017 turn out model or if they’re using an aggregate of all three numbers.

[Pericles]

To all the people complaining: you realize these things happen on a continuum, right? Statistically it is much more meaningful to specify under which conditions which candidate has what chance to win than it is to say 'R+5' or 'D+4' without anything else.

Like it is just straight lazy and bad methodologically to give one top line score with no real details of the assumptions being made to arrive at that score. I applaud Monmouth and SurveyMonkey for actually speaking in reasonable terms statistically in an effort to improve the standing of real, mathematically rigorous polls. Polls like these will bring the polling industry back.

And no, this is not a 'hedge their bets' strategy. If the electorate is a traditional midterm one and Jones wins by 4, this poll is wrong regardless of if it had Jones+4 in some alternate scenario. This poll is, in essence, a total view of the landscape that allows the reader the leverage to make turnout assumptions and provides the model to produce a top line score from said turnout assumptions.

Can't stand the hate literally the best polls I've seen are getting from people with zero understanding of basic statistics.

I understand that, but at the same time, they should have the guts to pick one model as the "most likely", at least so people can look at it at a glance to get a sense of the race. I'm fine if they have other turnout models in the full report.

RCP  is running with this poll being a tie for what it’s worth. I don’t know if that’s because they’re using the 2017 turn out model or if they’re using an aggregate of all three numbers.

They're using the 2017 turnout model.

[Co-Chair Bagel23]

Spoiler alert: Moore will win.

Stop kidding yourselves.

[BudgieForce]

Spoiler alert: Moore will win.

Stop kidding yourselves.

But what if he doesnt though?

[Co-Chair Bagel23]

Spoiler alert: Moore will win.

Stop kidding yourselves.

But what if he doesnt though?

I would rejoice at being wrong on this, post that I was wrong in all caps, and even put up a jill stein and paula jean swearingen poster for a week on my sig if I am wrong. That's my confidence level, but I still hope that I am really really really wrong on this one.

[IceSpear]

New gold standard!

[Brittain33]

I gave Monmouth crap in the headline but I think they handled this very well.

[Pollster]

Hopefully the statistical accuracy does not encourage future publication of results in the way that Monmouth published this one - will make polling significantly less accessible to the public and will mess up news reporting on polling even more - which is something I did not think possible.