I just can't help myself. It's time, yet again, for another chapter of "What can @surveyUSA get wrong about the California electorate??" [with gifs]

If you missed it, here was the thread last month regarding their awful San Diego survey: https://twitter.com/CA_120/status/1301020691928547329?s=20

And let's quickly go back to that poll in San Diego, and the @surveyusa response to it (which you can find here: https://www.pbmonthly.net/news/story/2020-09-03/criticisms-of-new-mayoral-poll-prompt-defense-from-polling-company). Two of our beefs were that they had a whopping 98% of their voters deemed "likely" and they didn't do any weighting.

Their incredulous response to the 98% likely was "We’ve seen this for years and years" like it is normal, and on the weighting issue they basically slammed all pollsters and idea of weighting by saying "We don’t want to put our thumb on the scale." Then in this poll they did!

In this one 88% of voters are deemed "likely" and they weight it to..."US Census targets for gender, age, race..." OMG! If you're going to weight, that is NOT how you do it. Weighting should be done to the voter file and the shares of subgroups in past similar elections.

To start, they have their share of the total voters that is Republican and Independent basically flipped. Everyone knows there are more Ind than Reps in CA, but they have 29% Republican (they are 24%) and 20% Independent (they are 29%). Dems are low, but just by 3 points.

Then we get into the likely voter weighting and it is just as bad, or worse. In partisanship, they have Republicans at 31%, but @Political_Data has them at 26%, and they have Independents at just 17% but they should be between 25-26%. Maybe they should stay off the weights...

Then they have Latinos at 32% of likely voters, but they should be around 24%.

Doing the math, even if turnout was sky high at 88% statewide, for Latinos to be 32% of the electorate, that would mean 6 million Latino voted, but there are only 5.6 million Latinos registered!

Finally, courtesy of someone much better at this than me, a closer look shows that they have 7% of the Trump voters as being... not registered to vote.

And, with that, we close another chapter on "what can @surveyUSA get wrong about the California electorate??" [with gifs]