Post

Is any of what happened here relevant to the unexpected divergence between Michigan Democratic Senate primary polling results and the actual outcome of the Michigan Democratic Senate primary?

no, that does not seem to matter

It depends upon whether they think a Dem will win regardless. Perhaps not in this case. If you're a Dem in Wyoming or a Rep in Massachusetts, your only chance to have your vote matter is to vote in the other primary.

correction: Marquette published yesterday I was just too busy coding to see it. law.marquette.edu/facultyblog/...

... (down to individual drop boxes) are separately batched, processed, retained. Should be adequate data - even by limited sampling - to nail down the late swing component of polling "error".

more US elections seem to have later changes (in WI, Crowley & Rodriguez were fairly late, but Barnes was after absentee ballots went out. Platner was fairly late. etc). Are pollsters taking that into account in their models and error rates?

Basically, I worry that we miss interesting dynamics when we classify all racists with a progressive/socialist insurgent as two-lane races. E.g.: afaict, Kat Abughazaleh would have won her race had she only been facing bona fide moderates and not a candidate with *some* progressive credibility

I think that's a pretty accurate read, coming from the district.

it's called the limitation of data driven prediction. I knew Hong was going to lose on July 30th, but that was more based years of experience

Conservatives are allowed to vote in Dem primaries. Surely this had an effect in Michigan and Minnesota. They'd rather rat fuck than participate in their own primaries.

That's a legitimate concern. A failure to disclose generates one type of bias and disclosure generates another. I wonder if this has led to some of the noise in recent polling where the pool of respondents has changed more than the pollsters are accounting for.

And in practical sense, that's why Fox's polling *may* identify as Beacon Research, or Shaw to respondents. Or they may even identify to respondents as another subcontractor who are the folks actually handling data collection for Beacon and/or Shaw for them -- all of which are very typical.

You’re referring to the dems missing the bullet, yes?

Whats your best estimate of the percentage of the 2026 general electorate that should be young adults? And does that match most polls?

I wonder how does the lack of a younger-voter spike in WI vs. previous primaries compare to what happened in other primaries this cycle (and how predictable this is vs. just reverting back to "should have just used the historical data")

Weighting to voter turnout in past midterm primaries, which is what @gelliottmorris.com finds works best, would capture the effect of students being out of school. Weighting to almost anything else wouldn't. Seems important.

In multi-candidate races, should polllsters ask for 2nd choices, just in case the 1st drops out and then they can report those new results?