Really valuable and careful work here examining how to ask gender on surveys. Importantly, more inclusive question design/response options does *not* cause noticeable signs of backlash in survey behavior (even though 56% of Americans in 2018 say they oppose nonbinary options).
In an experiment, compared to traditional single q on male/female, 2-step q (sex assigned at birth + current gender identity) better captured nonbinary respondents and did not affect 1) question nonresponse, 2) survey breakoff, or 3) views toward the surveyor (i.e. Pew).
Ultimately the post concludes that "asking about sex in addition to gender wasn't necessary." Key takeaway is validating the move from...
- original q: "Are you male or female?"
- to new gender q: "Do you describe yourself as a man, a woman, or in some other way?"
One broader thought: this is the kind of survey methods work that is crucial (underpins the data in so much of social science research) but that is too often lacking and undervalued.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Analyses of 2020 polling error often find that error was largest in the most Republican states, but it may not be too informative. The problem is that polling avg's in the most R (& most D) states are worse off in terms of 1) volume of polls behind the avg and 2) quality of polls
These graphs show that both volume and quality of polls in final 2 weeks is greatest in most competitive states and that it declines as a state gets more red/blue.
To give an example... according to 538's final day poll avg's, North Dakota had the worst polling error of 2020 (off true margin by 16 pts).
Here's what ND polls look like in final 2 weeks (4 polls from same D- pollster). Is this an adequate/fair test of general poll quality?
There's been a lot of informative work on subgroup voting patterns in 2020. One thing I wonder about: how consistent is the story across surveys? I've been digging into crosstabs to get a sense of within-pollster, 2016-20 changes across 8 pollsters. Threading some results here...
A few things:
-why looking within pollster is key dataforprogress.org/blog/5/02/bide…
-crosstab data is here docs.google.com/spreadsheets/d…
-there's a lot of variation in how pollsters report crosstabs
-graphs show diff in Biden's subgroup margin - Clinton's (& legend shows poll's overall shift)
First by gender--movement generally in Dem direction, but shifts among women are much greater. Net gender gap increases are 11, 9, 3, 2, 9, 7, -8, 14.
Random thing I've been wondering about lately: based on past elections, should we care about how Independents break? (Are their voting patterns a good indicator of who wins overall?) ANES doesn't always give too large samples of Ind.'s who voted, but overall trend points to yes.
Correlation between two-party Democratic president vote shares among pure Independents (from ANES surveys) and among all Americans (election returns) is 0.87.
Same strong correlation holds for Electoral College vote share too.
2020 seems like it might fit well with this pattern given strong Biden leads among Independents in recent polls (YouGov +11, CNN +22, Morning Consult +13, NY Times +9) and overall strength in the election. (Compare that to 2016 where Clinton lost among Independents.)
New from me @UpshotNYT: survey experiment I conducted gauging voter reactions to greater leftist rhetoric + policies coming from Dem candidates this primary season. Progressive turn solidifies Dem support, but there's backlash among Independents. Thread: nytimes.com/2019/09/30/ups…
A dominant theme of the first few Democratic debates was the leftward turn taken by candidates on a national stage, and what were the possible implications of this embrace of various progressive policies -- see some example headlines here:
Political science research (andrewbenjaminhall.com/Hall_APSR.pdf, nyu.edu/projects/polit…) suggests such extremism would be punished by voters, and that's informed speculation on reactions to current Dem rhetoric. But we still lack good contemporary evidence on how reactions play out.