Pollster:
Yes, that one was a screamer of a miss.
I agree the November poll was a huge outlier, but the poll had 9% other while it appears actual voters only went about 1% other and only 2% of the 9% were for âDonât knowâ or âSomeone elseâ. While I doubt all of the other in the poll went for Trump but probably a much more significant % than went for Harris even compared to the poll or actual results.
But even the September poll was right on for Harris but missed Trump by 9%.
Seems like she thinks polling doesnât work.
She seems to have made a friend.
1 Like
I wonder if polls are leading indicators or lagging indicators. They are trying to be used as leading indicators but between the time of the poll and elections the individual may change their view so it clearly is not a good leading indicator before you even factor the difficulty designing and carrying out a good poll. If you take it as a lagging indicator then it should best be used for course correction rather than predictive power.
Maybe a poll is like a thermometer. Just gets the âat the momentâ information and should be used with a lot of other information.
ETA: Maybe polling doesnât work now because unlike social media it canât keep real time information. Polling doesnât install âcookiesâ to follow you around or keep your data. They just donât have enough information beyond the questions they ask to make a good prediction.
I think 1/2 the problem is the questions that are asked. The other half is a failure to understand who exactly is being polled.
This is the crosstabs, btw:
https://www.documentcloud.org/documents/25318672-nov-2024-iowa-poll-crosstabs?responsive=1&title=1?embed=true&responsive=false&sidebar=false
[145 pages]
Here are the top level results:
https://www.documentcloud.org/documents/25329719-nov-2024-iowa-poll-questionnaire-with-weighted-and-unweighted-responses?responsive=1&title=1?embed=true&responsive=false&sidebar=false
I looked at a few of the cross-tabs, and they did look odd. Just looks like a bad sample.