There’s a weird/complicated issue here that you don’t have with other splits like men/women and white/black: people self-identify as Hispanic, and that self-identification changes over time. In particular, Hispanics are shifting their classification from Hispanic->White over time, which obviously doesn’t happen at any large scale for men/women and white/black splits.
If the choice to self-identify as white is non-random, which it almost certainly is, you can see “trends” in Hispanic voting which aren’t trends at all.
For example, picture a starting group of 100,000 Hispanics in 2016 who split 70/30 Dem/Rep. 20% of those individuals self-identify as white in 2020, and those who do are 90/10 Dem/Rep. Assume that no individual voter changed their party preference from 2016 to 2020. What you’re left with are 20,000 new white voters that are split 90/10 Dem/Rep, and 80,000 Hispanic voters that are split 65/35 Dem/Rep.
Holy shit! The Hispanic vote went from 70/30 Dem/Rep in 2016 to 65/35 Dem/Rep in 2020. What are Dems doing to lose Hispanic voters? Absolutely nothing - what you’re observing is just the effect of individuals choosing whether to self-identify as Hispanic.
I’m obviously not an expert, but my guess is that there are systematic differences between Hispanics who retain their self-identified Hispanic status and those who shift towards identifying as white.
Huh, I didn’t realized that Center for American Greatness was literally Trump’s Super PAC rather than being a technically-unaffiliated conservative website. I don’t know why Nate both wrote and feels beholden to rules that flag polls coming from campaigns but not Super PACs as being partisan and thus de-emphasized, but that seems like a substantial and fairly obvious oversight that should have been corrected a while ago and for which there’s no excuse not to correct now. I can understand his reluctance to relegate Rasmussen and Trafalgar based on his existing criteria, but excluding a Super PAC from being partisan because it’s not technically a campaign is really stupid.
In 2008 and 2012, when Nate wrote a lot more about his developing methodology, he noted that LV screens tend to produce outcomes about 1 point more favorable to Republicans than RV screens, maybe a little more, a reflection of the fact that the olds are much more reliable voters. He did it so he could appropriately mix LV and RV polls in the same model. This year, I’ve seen on several occasions instances where Biden fared better in the LV screen than in the RV one, which used to be really rare.
Wasn’t that what Rasmussen claimed he was doing the whole time when he binked some election everyone else was wrong about? Seems like I remember something about that, him saying he only polled LV and never RV like other polling outfits.
I do think in a polling situation most people prefer not to come off as ignorant and/or non caring, so being saying they are undecided is likely the shield they grab on to.
I do wonder how various polls handle respondents who are non committal to any choice. Are they steered to an undecided choice? Are they just automatically marked undecided? Why don’t we have a 87 minute documentary on Netflix titled “Undecided”?
Not since 2008 have I ever seen a pollster never do LV screens, at least not that I can recall. Typically you will see RV-only polls in the spring and earlier, and then they transition to doing LV screens in the summer, when they’re more reliable.
