Nate Cohn

Nate Cohn @Nate_Cohn

@dcg1114 if that's the L2 breakdown, not party ID, that would be easy to compare. but it may just be party ID — PolitiTweet.org

Nate Cohn @Nate_Cohn

Kind of odd, given that they're using a voter file for their sample -- and the voter file contains good information on partisanship in Georgia (IE: whether you last voted in a d/r primary) https://t.co/FCkfu52Hjx — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins ultimately, i'm just asking whether you're surprised by the numbers cited by trafalgar or whether they comport with your expectations — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins how many records would you need for 1000 completes — PolitiTweet.org

Nate Cohn @Nate_Cohn

@BlueCheese608 we call people off of lists of registered voters, so we know whether they're registered dem/rep or whether they've voted in dem/rep primaries in advanced — PolitiTweet.org

Nate Cohn @Nate_Cohn

I didn't dwell at length on Oct '19 in the newsletter, but at a gut-level it's one of the reasons I think '20 error could be transient https://t.co/a6QcnLgloa — PolitiTweet.org

Nate Cohn @Nate_Cohn

I didn't mention this in the piece, but our Oct '19 results do play a pretty real role in my gut-level instinct about whether 2020 error could be transient https://t.co/caFByLKMAj — PolitiTweet.org

Nate Cohn @Nate_Cohn

In this poll, white Democrats were about 5 percent likelier to respond than Republicans. That's quite a bit less than fall '20, when Democrats were 20 percent likelier to respond. It's more like Oct '19, when Ds were 6 percent likelier -- and our results were better for Trump — PolitiTweet.org

Nate Cohn @Nate_Cohn

@qdbui sure felt like it could happen! — PolitiTweet.org

Nate Cohn @Nate_Cohn

The response rates are unusual, though it is consistent with the method described -- 75k records, worked until they reach their target (however hard that may be). I am nonetheless surprised there aren't cases where they exhaust their initial records or finish early — PolitiTweet.org

Nate Cohn @Nate_Cohn

On first read, a few things surprise me --Live text methodology (expensive) --No weighting (though could be quota sampling or something, given crosstab stability) --A claim of a 4:1 D:R response rate ratio in '20 (not clear how seriously to take this; ours were more like 1.2:1) — PolitiTweet.org

Nate Cohn @Nate_Cohn

Plenty of interesting tidbits in this interview of Trafalgar by @lxeagle17 https://t.co/Cg89XlFKLv — PolitiTweet.org

Nate Cohn @Nate_Cohn

You can sign up for the newsletter here https://t.co/vfLG2cPxwD — PolitiTweet.org

Nate Cohn @Nate_Cohn

In today's newsletter, a nautically themed report on whether you can trust the polls -- including relatively good news on response rates by party in the latest Times/Siena survey https://t.co/df8DoQykon — PolitiTweet.org

Nate Cohn @Nate_Cohn

@tbonier so what's your term for a GA poll with 55% college grads https://t.co/TKXO0G1ZAP https://t.co/g5IUFHkiAl — PolitiTweet.org

Nate Cohn @Nate_Cohn

@DrBillPeris @BrendanNyhan @ElectProject @protctdemocracy i'm entirely agnostic on whether it's fair or not. entirely. but clearly VRA + FPTP + apportionment by total population is, on paper, a pretty favorable set up for latino representation v. PR — PolitiTweet.org

Nate Cohn @Nate_Cohn

@BrendanNyhan @ElectProject @protctdemocracy anything that moves off of population-based representation toward proportional representation among actual voters will hurt latino representation, much like the moves to apportion based on noncitizens (or perhaps even more so) — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @carlosodio: Our takeaways are consistent with the new NYT/Siena: in most states, Latino support for Dems has neither declined since 202… — PolitiTweet.org

Nate Cohn @Nate_Cohn

@doug_rivers across / i'm not calculating the design effect assuming stratification. for reference -- the 1.3 is just the longer-term, average design effect in nyt/siena polls, not something from this survey — PolitiTweet.org

Nate Cohn @Nate_Cohn

@doug_rivers i'm just referring to the longer term, average deff across times/siena polls — PolitiTweet.org

Nate Cohn @Nate_Cohn

@jon_m_rob @doug_rivers @b_schaffner @shirokuriwaki @RuthIgielnik and on the quota question -- we are doing stratified quota sampling on race x party x region — PolitiTweet.org

Nate Cohn @Nate_Cohn

@jon_m_rob @doug_rivers @b_schaffner @shirokuriwaki @RuthIgielnik and to your point on response rate adjustment: there's no way we could regularly weight on a dozen categories with n < 1000 polls and keep our mean design effect around 1.3 without having a very clean sample on the front-end — PolitiTweet.org

Nate Cohn @Nate_Cohn

@jon_m_rob @doug_rivers @b_schaffner @shirokuriwaki @RuthIgielnik this survey does have an additional step of weighting compared to usual though, and that's the separate weighting of self-reported hispanics/non-hispanics to allow us to weight on national origin, spanish households etc — PolitiTweet.org

Nate Cohn @Nate_Cohn

@jon_m_rob @doug_rivers @b_schaffner @shirokuriwaki @RuthIgielnik we almost never have a design effect over 1.5, and that's true for this sample without the huge oversample of modeled hispanic voters (which is even larger than self-report figures, given that not all modeled = real) — PolitiTweet.org

Nate Cohn @Nate_Cohn

Pew found Dems up by 25 points among Hispanic voters, 53-28. We were D+24, 56-32. And among all RVs, Pew and NYT/Siena both found D+2 (this is a cxn from prior and now deleted post) — PolitiTweet.org

Nate Cohn @Nate_Cohn

Pew found Dems up by 25 points among Hispanic voters, 53-28. We were D+26, 56-32. And among all RVs, Pew and NYT/Siena both found D+2 — PolitiTweet.org

Nate Cohn @Nate_Cohn

One interesting coincidence: last month, Pew also did a midterm poll with a Hispanic oversample. Despite big methodological differences, we have virtually identical numbers overall and among Hispanics -- helping a smooth transition for @RuthIgielnik https://t.co/jkd7uvOF51 — PolitiTweet.org

Nate Cohn @Nate_Cohn

The newsletter this AM has a little bit of the backstory behind the Hispanic oversample and how it fits into our polling plan this cycle, along with a bonus map of our respondents. You can subscribe here https://t.co/tcKfAoXBSl https://t.co/xDySzs38nm — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @BenjySarlin: “How are we not winning by 10, we have polling leads on all these economic and policy issues” is a longtime D complaint th… — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @jon_m_rob: Latino voters support for Democrats is 63% of the two party vote in this poll, almost exactly what our @Catalist_US estimate… — PolitiTweet.org