In this final stretch of the 2024 campaign, we’d like to do things a little differently, and share our Fully Unweighted, Completely Transparent (F.U.C.T.) data. Below you’ll find our toplines, crosstabs, and raw data.Why raw data, you ask? Think of this as March Madness. As pollsters and...
www.thebullfinchgroup.com
I assumed it was RV and LV, but it looks like they just did two different LV samples maybe? Both +8
This poll is just different from anything I've seen. They aren't weighting their respondents. Rather, they set targets for the demographics they wanted, and kept calling people until the quotas were filled (there was some margin of error). It's not clear what they did with excess respondents -- i.e. if they already had all the white men they wanted, if they reached another white man, did they just drop it out? -- but anyway, the idea is to avoid trying to weight respondents. I'm not sure it's as different in practice as they try to make it sound; there's still some weighting going on through the quota approach, but I guess it could reduce some sort of errors. As for details/observations:
1. the topline reported results appear to be RVs only. They do ask some LV screen questions, apparently with the idea that people can download the data and apply their own LV screen. But the +8 in MI appears to be RVs.
2. The poll isn't nearly as good for Kamala in WI and PA. It's as close to tied as possible in PA, and she's up 3 in WI. That said, if you put them all together, the aggregate result is Kamala +4 (and this is using 600 voters from each state, not weighting by population), which is really good. Except it's somewhat hard to compare because it's different.
3. Interesting to look at the "recalled vote" questions. In MI, for instance, the sample recalled voting for Biden 47-44. That's pretty close to what it was. It's a slightly D skewed sample, but very slight. PA appears to be slightly skewed R (it reported Biden/Trump tied, when Biden actually won by a point) and WI looks on the nose.
But they also asked about recalled 2016 vote. Here, Trump led in sample by 6-8 in all three states. So that could mean a few things. One is that it simply reflects the problems with recalled vote weighting -- namely, that people don't report their past votes correctly. If so, that's good for Kamala since she's been doing better in the non-recalled weighted surveys.
4. The poll in all three states slightly oversampled the 50-64 demographic, relative to the target. That's Trump's strongest demo, though the oversampling was 1% in each case and not really all that significant. They also slightly undersampled 18-29, as would be expected.
5. As for race, this could provide a bit of explanatory power for the Michigan outlier. They were looking for 80% white; they got 78%. In WI, they were looking for 86%; they got 88. In PA, they were looking for 80, and got 81.
All told, I think it's a pretty good poll for Kamala, but the MI result does seem to be an outlier. Still, +4 overall across the three states seems good.
Link?
