Nate Cohn

Nate Cohn @Nate_Cohn

@neeratanden yes, and the biden team will need to evaluate whether they want to play more defense on it as it emerges as something more like the central attack, as opposed to one of many things being thrown at a wall — PolitiTweet.org

Nate Cohn @Nate_Cohn

And I'll wait to offer thoughts on it until after the microdata is out. One thing I'll be interested in is the share of the electorate that's 2016 non-voting, which could be quite large with a poll weighted this way — PolitiTweet.org

Nate Cohn @Nate_Cohn

The USC tracking poll is out. Whatever else you think of it, I can't help but note that they don't have the crosstab by race anymore https://t.co/XBPcqNqgCT — PolitiTweet.org

Nate Cohn @Nate_Cohn

I see many of you think Biden has responded to this, which is literally true. But this isn't a high school debate, where biden 'responds' simply by contradicting it. the RNC spent several days elevating a claim; a couple of interviews doesn't undo the damage, if there was damage — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Eggymceggerson3 to the contrary, you're following it too closely! this isn't a high school debate, where biden has 'responded' simply by contradicting it. the RNC has spent several days elevating a claim; a couple of lines in interviews does not undo the damage, if there was damage — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Eggymceggerson3 to the contrary, you're following it too closely! this isn't a high school debate, where if biden has 'responded' simply by contradicting it. the RNC has spent several days elevating a claim; a couple of lines in interviews does not undo the damage, if there was damage — PolitiTweet.org

Nate Cohn @Nate_Cohn

Usually, the argument against responding is that you don't want to elevate the criticism, esp if it's not credible. OTOH, the RNC just spent four days suggesting that you support something you oppose, so maybe it's already high profile. It'll be interesting to see the call here — PolitiTweet.org

Nate Cohn @Nate_Cohn

Now that team Trump has showed its hand here, it'll be interesting to see how Biden reacts. Do you go on air defending against the 'defund police' attack, for ex? Biden's been served well by inaction, but that's in part because Team Trump hasn't done much requiring a response — PolitiTweet.org

Nate Cohn @Nate_Cohn

The Trump team has been pretty slow to resolve on a clear line of attack on Biden. There was Ukraine. There was establishment Joe, sleepy Joe. Now Biden, stealth radical, has won out. It's a tough sell in some important respects, but it aligns with the news and energizes GOP vote — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @capitalweather: Meteorologists are in awe. The National Hurricane Center predicted where #Laura would make landfall within 0.6 miles 87… — PolitiTweet.org

Nate Cohn @Nate_Cohn

@mlow29 no, opposite — PolitiTweet.org

Nate Cohn @Nate_Cohn

(as an aside, this is why those polls about 'how your neighbor will vote' may not be about social desirability bias. the median voter is a clinton voter; the median neighborhood voted trump) https://t.co/bSPJ7chBCO — PolitiTweet.org

Nate Cohn @Nate_Cohn

It's not like it's useful in other contexts. In game 2 of an playoff series, no one's like "well, at this point in the second quarter, we were were trailing by 10 and won by 2; now we're trailing by 5, thus we're on track to win by 7! — PolitiTweet.org

Nate Cohn @Nate_Cohn

I understand why people would compare Biden's current polling to Clinton's final polling, but I'll never really get the comparison between Biden's current polling and Clinton's polling at the same date — PolitiTweet.org

Nate Cohn @Nate_Cohn

@Redistrict suppose that biden's up 8.5 pts w LVs nationwide on 9/14. where would you land then? — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @jaycordeira: NHC's 3.5-day forecast for #Laura landfall was off by ~1 km. wow. https://t.co/RCnBnYQVSm https://t.co/f1wmsQynAZ — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @ineedja_kadeeja: A few weeks ago, residents of Lake Charles were demanding that a confederate monument at the city court house be remov… — PolitiTweet.org

Nate Cohn @Nate_Cohn

@ArmenianBacon @TimFargus @gelliottmorris @BrendanNyhan @On_Politike a weather forecaster actually simulates the future, rather than simply modeling the past the applying it to the present — PolitiTweet.org

Nate Cohn @Nate_Cohn

@TimFargus @gelliottmorris @BrendanNyhan @On_Politike nope. don't call it either — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @BrendanNyhan @On_Politike in my world, you don't call this a forecast and you don't call it a prediction https://t.co/AYZ5DO4bpC — PolitiTweet.org

Nate Cohn @Nate_Cohn

@BrendanNyhan @gelliottmorris @On_Politike if you *are* invested in selling your model of N=18 as predictive forecasting, then i think you probably do need to more or less make stuff up. there are smarter and dumber ways to make things up, of course. but is any of it principled, exactly? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@BrendanNyhan @gelliottmorris @On_Politike for me, part of the answer is that this shouldn't be framed as predictive forecasting, but instead as something more like descriptive analytics: a friendly tool to help get a sense of things ("if polls are as good as in the past, then...") https://t.co/qKHVElpEwL — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike and to the extent you appear to have any doubts about the representativeness of the past, like polarization, you think you can address it with a more sophisticated model rather than by acknowledging the deeper uncertainty posed by a changing world — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike broadly speaking, i think the answer to those two questions is 'yeah, kinda' and 'maybe.' i think that's the fivethirtyeight answer, too, more or less. i think your answer is basically yes and yes. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike it is different, and yet the fundamental challenge to the forecaster is similar: can i really assume that the future will behave like the past, and does the past contain the range of possibilities for the future? — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike a model of the relationship between war outcomes and GDP between 1700-1945 might be useful! it's going to be pretty confused by vietnam and korea though. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike and that's not to say that small n social science models aren't *useful*! you just don't have a model you can expect to be predictive — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike i mean, i'm plenty use to social science! but i don't think it's reasonable given the amount of data here, no. — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike as far as i'm concerned, there is *no* single cycle measurement that's terribly useful to analyzing a model national elections. it's like judging a model of football based on the outcome in one game — PolitiTweet.org

Nate Cohn @Nate_Cohn

@gelliottmorris @On_Politike but, uh, yeah, you should *definitely* have some kind of uncertainty beyond what you've observed in our N=18 or whatever dataset if you want a robust model — PolitiTweet.org