Nate Cohn

Nate Cohn @Nate_Cohn

@nataliemj10 @kwcollins i say borderline ideological because the conversations have a feature that's very common to ideological discussions: even very smart anti-rounding people have a very hard time understanding and then acknowledging that there are real tradeoffs — PolitiTweet.org

Nate Cohn @Nate_Cohn

@emilythorson @kwcollins i see the twitteratti running with what i know to be misinterpretation of rounded (or twice rounded in cases of differences) in times/siena polls all the time — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins i'm agnostic on the tradeoff. we round like everyone else; the downside of precision is fairly straightforward. but i see a lot of pollsters treating this like a borderline ideological question and seem blind to the idea that there's a real trade off — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins all of the options are 'false' -- there's 'false' precision or there's an intentional loss of precision due to rounding. the question is which falsehood is more dangerous for the consumer, one that yields misinterpretation or one that risks overconfidence. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins and my argument is that's irrelevant if data consumers notice and interpret the distinction — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins as long as people interpret D+1 and "tie" differently, or a "no change" v "D+2 shift" differently, the loss of precision is relevant — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins i think the estimate should be sufficiently precise that readers/public don't interpret it differently than it actually is. the confidence interval is irrelevant in that formulation, imo, even if it ought to be relevant — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins the good case, imo, is that people look at 47.23-48.23 and think it's as precise as like the latest numbers from the s&p 500. that's a serious case, to me. idk whether it outweighs damage from imprecise findings and downstream misinterpretation, but it's a good case — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins i think there's a good case for rounding but idk if i think that's it. i don't see the 'rounding' deterring those overly precise interpretations. what i do see is that a loss of precision yielding different and often worse interpretations under a lot of circumstances — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins if you take the rounding, everyone on this website says the RV/LV gap is 2x what it is. there's a real trade off on this stuff — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins i think this is more debatable than most. like, what's worse -- releasing RV 45-45, LV 48-43 or RV 45.49 - 44.51 (or 45.5-44.5); LV 47.51, 43.49 (or 47.5-43.5) — PolitiTweet.org

Nate Cohn @Nate_Cohn

it's an interesting question. i think i'd rather be oz at this point, but i can come up with a decent case either way https://t.co/Y1bBJEsTwi — PolitiTweet.org

Nate Cohn @Nate_Cohn

you can subscribe (and see two tidbits in the most recent post, which is visible on the page) here https://t.co/vfLG2cPxwD — PolitiTweet.org

Nate Cohn @Nate_Cohn

you can subscribe (and see the most recent newsletter, which contains two such tidbits here) https://t.co/vfLG2cxoiv — PolitiTweet.org

Nate Cohn @Nate_Cohn

newsletter subscribers already know about at least two polling projects in the works https://t.co/82U1ZOZyAn — PolitiTweet.org

Nate Cohn @Nate_Cohn

afaik, McMullen is an independent who pledges not to caucus with a party, so that seems consistent to me https://t.co/VEYvn2z3gl — PolitiTweet.org

Nate Cohn @Nate_Cohn

Quick question for the crowd: who are viable candidates in '22 who will not caucus with the party that they're listed as on ballot (ie: Sanders (I, VT) or King (I, ME) would count, as they caucus D, but they're not up this cycle) — PolitiTweet.org

Nate Cohn @Nate_Cohn

Big gap between registered and 'likely' voters in this Marquette poll. Likely voter screens are both challenging and noisy, so worth keeping your eye on both numbers in this case https://t.co/0iSIOLuf7k — PolitiTweet.org

Nate Cohn @Nate_Cohn

@jon_m_rob @tbonier 1800s would have been fun — PolitiTweet.org

Nate Cohn @Nate_Cohn

@baseballot @LoganDobson well the model does know about the RGV shift -- it just doesn't take it as seriously as a model trained only on recent data, for ex. if you think this is an R+2 CD, not sure you need to know the rest here to have a very different answer than 538 / understand the DCCC — PolitiTweet.org

Nate Cohn @Nate_Cohn

@baseballot @LoganDobson well the model does know about the RGV shift -- it just doesn't take it as seriously as a model trained only on recent data, for ex. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@baseballot yeah, that's where i figured. it's not where i'd have it, personally. but i don't mean to fight about it -- i mainly think that this is a case where modeling choices are unusually important and yours probably represents the best case for Ds — PolitiTweet.org

Nate Cohn @Nate_Cohn

@baseballot is this because of your partisanship measure in a traditionally dem district? other model specifications would have this as a pretty plainly GOP favored CD (if we turned on the needle rn, this would start at like R+8) — PolitiTweet.org

Nate Cohn @Nate_Cohn

@baseballot is this because of your partisanship measure in a traditionally dem district? other model specifications would have this as a pretty plainly GOP favored CD — PolitiTweet.org

Nate Cohn @Nate_Cohn

@RossRichendrfer wouldn't have been so fun if it wasn't true! — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @UmpScorecards: Umpire: Lance Barrett Final: Mariners 4, Blue Jays 0 #SeaUsRise // #NextLevel #SEAvsTOR // #TORvsSEA #Postseason More s… — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @declanwalsh: After Secret U.S. Talks Fail, a Hidden War in Africa Rapidly Escalates https://t.co/QMyGSRzUIh — PolitiTweet.org

Nate Cohn @Nate_Cohn

@baseballot Wouldn’t say there are none and I like to think the jays fans there at least have a soft spot for Seattle. But it’s a de facto national team at this point — PolitiTweet.org

Nate Cohn @Nate_Cohn

@baseballot definitely the blue jays — PolitiTweet.org

Nate Cohn @Nate_Cohn

the popularist approach https://t.co/tPVNxDxOMT — PolitiTweet.org