Nate Cohn

Nate Cohn @Nate_Cohn

OTOH, this remedy could put big chunks of Seattle, Queens and Detroit and wherever into the 'suburbs.' And for me that's kind of facially unacceptable, at least for this kind of generalized purpose. In this context, I'd settle for calling San Antonio urban. — PolitiTweet.org

Nate Cohn @Nate_Cohn

The way around it is to use a density/development based definition, which would make almost all the Sun Belt metros 'suburban.' I think this vision has merit and it also correlates well with vote choice. — PolitiTweet.org

Nate Cohn @Nate_Cohn

The most coherent critique, IMO, is that this is a very industrial-northeast focused definition of a suburb. There are many Sun Belt metros where the city-limits are basically coterminous with the whole urbanized area, so we would have few no 'san antonio suburbs' — PolitiTweet.org

Nate Cohn @Nate_Cohn

Of course, every definition has shortcomings. And depending on the purpose, another may be more appropriate. But if I get one single, national standard definition of a 'suburb' for polling, I think something like this is most consistent with how it's used politically — PolitiTweet.org

Nate Cohn @Nate_Cohn

Of course, every definition has shortcomings. And depending on the purpose, another may be more appropriate. But I get one single, national standard definition of a 'suburb' for our polling, I think something like this is most consistent with how it's used politically — PolitiTweet.org

Nate Cohn @Nate_Cohn

A secondary opportunity is to see if there's a way to grow our suburban definition to include post-2010 suburbs (possibly based on new ACS density data) or tiny non-urbanized rich enclaves (tough, as the pop density is quite low). I'd guess this isn't worth it or workable — PolitiTweet.org

Nate Cohn @Nate_Cohn

So that's anything in 'orange' here that isn't part of the central cities of the MSA (like Philly, NYC, Newark, etc.) https://t.co/2RvJug37Eu — PolitiTweet.org

Nate Cohn @Nate_Cohn

The biggest opportunity for improvement would be to see if there's a smart way to make the Alpharetta/Waukesha type 'central cities' count as suburbs, based on some kind of population floor. I'd guess it won't have a huge effect, but worth seeing if there's a non-arbitrary option — PolitiTweet.org

Nate Cohn @Nate_Cohn

If we leave our definition of suburbs here--census-defined urbanized areas outside of any central city in any MSAs--then the last NYT/Siena result for suburbs is Biden 50, Trump 34. — PolitiTweet.org

Nate Cohn @Nate_Cohn

I'm fairly satisfied with this end of the definition, though there are some high growth suburbs in the Sun Belt sprawling into 2010 'rural' areas and there are rich mansion zones that aren't urbanized by census standards, tho very few people there — PolitiTweet.org

Nate Cohn @Nate_Cohn

Another option is some kind of population density floor, where a block group outside of the central city in an MSA county would be suburban if its population density is greater than x/sq mi. That said, the urbanized area definition is close to many plausible density floors — PolitiTweet.org

Nate Cohn @Nate_Cohn

Here, the census does offers a possible solution: 'urbanized areas,' where the census generally considers non-urbanized areas to be 'rural.' I think this does a pretty good if imperfect and maybe outdated job of pulling out rural areas from MSA counties https://t.co/nwHSTM00ko — PolitiTweet.org

Nate Cohn @Nate_Cohn

On the other end of the spectrum is the challenge of delineating suburban and rural. It's easy enough to say the non-metropolitan counties are rural. But there are many metropolitan counties that are pretty rural, and there are certainly indisputably rural parts of metro counties — PolitiTweet.org

Nate Cohn @Nate_Cohn

One option here would be to limit ourselves only to central cities of a certain size (hoping to, say, keep SF and OAK but lose Alpharetta or Waukesha in ATL/MIL). This might also help with the small MSA problem. Still worth investigating. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Though easy enough, this definition of 'urban' does leave something to be desired: there are plenty of 'central cities' that I'd say are part of the suburbs, like Alpharetta, GA. There are also many tiny MSAs: Do we really want Dothan, AL as urban? https://t.co/pDtvunCfDk — PolitiTweet.org

Nate Cohn @Nate_Cohn

So this certainly helps identify areas that aren't suburban, and in our last national poll, voters who live in the 'central city' of a metropolitan area (and for, say, the NY-Newark-Jersey City MSA, that's anyone within city limits of any three), Biden led 64 to 24 percent. — PolitiTweet.org

Nate Cohn @Nate_Cohn

The best definition may vary on the purpose. But colloquially, I think that when people refer to the suburbs of 'x' city, they pretty explicitly do not mean any voters within central city 'x', regardless of density, and there are good political reasons to care about city limits. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Our polls use a variety of different geographical variables in different contexts: census block group density, made up regions like 'philly suburbs', census-defined metro areas, or the NCHS urban-rural scheme. None obviously offer a consistent definition of 'suburb' — PolitiTweet.org

Nate Cohn @Nate_Cohn

Thanks to the news this week, I spent some time over the last few days grappling with a question that many, many people have thought about before me: 'what's a suburb?' and what's our poll result for 'suburban voters?' — PolitiTweet.org

Nate Cohn @Nate_Cohn

Feeling left out. But our battleground polls were D+1, same as October, and they were better for GOP than the exits among 2016 voters. — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @PatrickRuffini: New from our monthly survey: Biden +15 (and +13 in a ballot with third party candidates) https://t.co/AfLKZouDGN — PolitiTweet.org

Nate Cohn @Nate_Cohn

@RachelBitecofer @jdcollens @maristpoll it's not a voter file sample and it's not weighted on education — PolitiTweet.org

Nate Cohn @Nate_Cohn

@LeeMiringoff @maristpoll I think it is likely that education, on its own, is insufficient to ensure a representative sample. What I don't understand is why it is either a) unnecessary or b) disadvantageous. If it correlates with both vote choice and nonresponse, why *shoudn't* we weight on it? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Redistrict i asked and, yeah, it's geography in place of education, not in addition to — PolitiTweet.org

Nate Cohn @Nate_Cohn

Anyway I find this all very hard to understand. The NBC/Marist poll is, AFAIK, the only major live interview telephone poll at this point that's not weighting by education in some way or another, and I can't explain why after reading 4000 words on the matter — PolitiTweet.org

Nate Cohn @Nate_Cohn

And I would sure think they would want to consider that data: the NBC/Marist poll was the most biased poll toward Democrats of any survey that conducted more than 6 polls in 2018. Do these solutions fix McCaskill/Donnelly leads in final '18 polls? What about Evers+10 in Oct? — PolitiTweet.org

Nate Cohn @Nate_Cohn

Separately, it is odd to depend so much on this test given the huge amount of data that NBC/Marist has to work with! They have 9 final polls from '18, more in previous years. Would they have fared better with education weights? That's testable and far more generalizable. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Third, it's still just one data point, even if it's generalizable and perfect. Polls are noisy and so is the effect of education. Sometimes, weighting by education makes a poll less or more accurate, just by chance. The reduction in *bias* is only evident over many polls — PolitiTweet.org

Nate Cohn @Nate_Cohn

Second, is that the Kentucky gov election, whatever its merits, may not be the most representative race. After all, we're talking about an election a Dem won in Kentucky. Even if the test was perfect, why assume KYGOV test is generalizable given that it's facially so unusual? — PolitiTweet.org

Nate Cohn @Nate_Cohn

Second, is that the Kentucky gov election, whatever its merits, may not be the most representative race. After all, we're talking about an election a Dem won in *KY* in Kentucky. Even test was perfect, why assume KYGOV is generalizable given that it's facially so unusual? — PolitiTweet.org