This is a great post, Kristen, thank you. You educate a great many people with simplicity and clarity around the challenges pollsters face. While I'm not a pollster, I've worked with several over the past several decades in some three dozen US House and Senate races, and you ignore the fact that a small number of pollsters manufacture surveys favorable to one party over another to help candidates with fundraising and party support, or recruit certain people to run. Smart people see through it with a simple look at the crosstabs and demographics. I've personally experienced it. And guess what - the media highlights those "surprising" polls that show unusual support for candidates (especially Democrats). I've seen polls that over/undersample certain party IDs or conduct surveys on Friday and Saturday nights that tend to favor D's. Many Americans sense this and, as a result, reject polling. Woke University polls are among the worst. Elon University smartly gave up their polling when they're results consistently favored Democrats, and misled the public on the status of elections. If only other schools had such integrity. And then there's the WaPo/ABC poll in Wisconsin just before the 2020 election that had Biden beating Trump by 17 points. It turned out, a few days later, to be .6 percent. Damage done. It would be nice to see the polling industry police their own and get rid of these bad actors.
Assuming the RNCs 800 minimum primary vote requirement is to have a large enough sample to trust the polls results, why even call registered democrats at all? If it’s to collect data on how non likely republican voters feel, doesn’t that data need to be sufficiently large as well?
It seems like the weighing technique is just used to circumvent the spirit of the RNCs minimum because pollsters believe they can get accurate data with smaller sample sizes? If that’s true then why isn’t the RNC listening to the pollsters that work for them in regards to required sample sizes?
My head’s spinning, but it seems that this technique purposely skews the data and then later on un-skews it. If the goal of the RNC is to have accurate polling data, it seems like skipping those two steps would result in something more accurate.
This is a great post, Kristen, thank you. You educate a great many people with simplicity and clarity around the challenges pollsters face. While I'm not a pollster, I've worked with several over the past several decades in some three dozen US House and Senate races, and you ignore the fact that a small number of pollsters manufacture surveys favorable to one party over another to help candidates with fundraising and party support, or recruit certain people to run. Smart people see through it with a simple look at the crosstabs and demographics. I've personally experienced it. And guess what - the media highlights those "surprising" polls that show unusual support for candidates (especially Democrats). I've seen polls that over/undersample certain party IDs or conduct surveys on Friday and Saturday nights that tend to favor D's. Many Americans sense this and, as a result, reject polling. Woke University polls are among the worst. Elon University smartly gave up their polling when they're results consistently favored Democrats, and misled the public on the status of elections. If only other schools had such integrity. And then there's the WaPo/ABC poll in Wisconsin just before the 2020 election that had Biden beating Trump by 17 points. It turned out, a few days later, to be .6 percent. Damage done. It would be nice to see the polling industry police their own and get rid of these bad actors.
Assuming the RNCs 800 minimum primary vote requirement is to have a large enough sample to trust the polls results, why even call registered democrats at all? If it’s to collect data on how non likely republican voters feel, doesn’t that data need to be sufficiently large as well?
It seems like the weighing technique is just used to circumvent the spirit of the RNCs minimum because pollsters believe they can get accurate data with smaller sample sizes? If that’s true then why isn’t the RNC listening to the pollsters that work for them in regards to required sample sizes?
My head’s spinning, but it seems that this technique purposely skews the data and then later on un-skews it. If the goal of the RNC is to have accurate polling data, it seems like skipping those two steps would result in something more accurate.
Very interesting piece as always!