Are polls wrong this midterm because of COVID? If a 1000-person telephone poll—the pollster keeps dialing more numbers until they hit target sample size. But what if some demographic is MISSING & died in pandemic? This is NOT captured in current polling methods. #ElectionTwitter 
2) We epidemiologists deal with these issues all the time. We are demographic pollsters for public health & study causality risk factors. Hence epidemiologists know a thing or two about polls and the weighting methods pollsters use. ➡️Your “weighting” can be off if people missing
3) Polling surveys can try to “weight” for demographics— but that assumes their presumed demographic % of sub populations is accurate. US census in 2019-2020 was extremely messy amid pandemic—data came mostly before end of 2020. As the top graph shows—R’s died more in 2021-2022.
4) In the world of pirates 🏴☠️—the analogous phrase they use is “Dead men tell no tales”—similarly, those who died in the pandemic don’t come out to VOTE either. I discussed this DEATH BIAS in the old debates on herd immunity — you’re not winning if you’re culling your herd.
https://twitter.com/drericding/status/1443564127713124352
5) Also, we know telephone pollsters have horrible time getting people to pick up phones these days—especially to answer cell calls. There is a huge PHONE PICKUP RESPONSE BIAS, where those who respond on a polling call are not at all similar to likely voters. Not fixed either.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
