Nate Cohn

Nate Cohn @Nate_Cohn

Either way, some of the systematic polling errors in recent years, or polls that look like (or may be) 'herding', i… https://t.co/2aPNa5wsM0 — PolitiTweet.org

Nate Cohn @Nate_Cohn

And some kinds of model-based polling estimates--like those that pool data across states or over time--could be eve… https://t.co/m4NJAGuWdh — PolitiTweet.org

Nate Cohn @Nate_Cohn

(or, i should say, an individual pollster might make a very different choice on bias-variance trade off in model selection then a poll aggregator would want that pollster to make) — PolitiTweet.org

Nate Cohn @Nate_Cohn

In a model-based world, poll aggregation gets tricky and the incentives don't line up. An individual pollster would… https://t.co/NS6y1nlkIU — PolitiTweet.org

Nate Cohn @Nate_Cohn

For a poll aggregator, there are some additional consequences. Poll aggregation makes most sense in a design based… https://t.co/v0LT72ImZy — PolitiTweet.org

Nate Cohn @Nate_Cohn

In the model-based view, rake-weighting is just a really horrible form of modeling; may as well use the good kind w… https://t.co/1lz7rGeBdZ — PolitiTweet.org

Nate Cohn @Nate_Cohn

The truth is probably in between. A good design substantially reduces the burden on your model; a good design does… https://t.co/DtXgOv5UiA — PolitiTweet.org

Nate Cohn @Nate_Cohn

The model-based view says design doesn't do it anymore; that's why they all use extensive weighting. Instead, polls work because of the model: we basically just make inferences about the population based on an unrepresentative sample by controlling for demog. used for weighting — PolitiTweet.org

Nate Cohn @Nate_Cohn

The design-based view is the classical view. Here, a probability sample = a good poll. If everyone has a known chance of selection, then the results should fall within a theoretical margin of error due to sampling — PolitiTweet.org

Nate Cohn @Nate_Cohn

Without wading into this specific debate, the feud over MRP gets at a pretty fundamental question about survey research right now: do polls work because they have a design that yields a representative sample, or do polls work because they have a good model to correct error? — PolitiTweet.org

Nate Cohn @Nate_Cohn

RT @JohnAnzo: It is caucus day and @Nate_Cohn just made my head explode https://t.co/mipeLJfUcj — PolitiTweet.org

Nate Cohn @Nate_Cohn

@PatrickRuffini it's all simultaneous afaik — PolitiTweet.org

Nate Cohn @Nate_Cohn

And this isn't a prediction, but I can imagine scenarios where this whole mess contributes to an end to Iowa's first-in-the-nation-caucus status or at least leads them to switch to a standard popular vote, like the IA GOP caucus — PolitiTweet.org

Nate Cohn @Nate_Cohn

@PatrickRuffini almost like it's wrong — PolitiTweet.org

Nate Cohn @Nate_Cohn

As an aside, the NYT--and most news orgs--will project a winner based on the state delegate equivalency results. Doesn't mean the other metrics will be ignored, and all will be available, but it'll be SDEs for the purpose of the coveted projection 'check-mark' — PolitiTweet.org

Nate Cohn @Nate_Cohn

There will be four tonight https://t.co/MDieei6bpw — PolitiTweet.org

Nate Cohn @Nate_Cohn

Happy caucus day, and welcome to the 2020 Democratic presidential nominating season https://t.co/BZ0c7hnxv3 — PolitiTweet.org

Nate Cohn @Nate_Cohn

My strong prior is that this cost Bernie the 2016 caucus. Assuming we've matched up the data right, Sanders probably lost a point or two on SDEs compared to the final alignment results. — PolitiTweet.org

Nate Cohn @Nate_Cohn

I would note, BTW, that this data is highly consistent with the electorate in our poll, which is younger and better… https://t.co/nx85CvIvn9 — PolitiTweet.org

Nate Cohn @Nate_Cohn

Here, 2016 caucus goers per 2020 SDE by college grads. A clear pattern here again https://t.co/y17Wkv0xw9 — PolitiTweet.org

Nate Cohn @Nate_Cohn

Here, for instance, the voters per SDE by pct under age 24 in a precinct. In precincts with many young voters, 200… https://t.co/JTP5nApjYO — PolitiTweet.org

Nate Cohn @Nate_Cohn

This is true in our poll, as you know. It's also plainly true in the '16 turnout by precinct. Young/well educ. prec… https://t.co/seyhp4zHqr — PolitiTweet.org

Nate Cohn @Nate_Cohn

In particular, a candidate who fares well among young and well-educated voters could do much worse in SDEs than fin… https://t.co/6rfpLY3ZLt — PolitiTweet.org

Nate Cohn @Nate_Cohn

Basically, SDEs are allocated based on average Dem vote in '16 and '18. If the caucus turnout is different than avg Dem vote in '16/18, then there's the potential for a SDE-Second alignment split — PolitiTweet.org

Nate Cohn @Nate_Cohn

The split between the final alignment and state delegate equivalency, however, is unique to Iowa and worth some discussion. — PolitiTweet.org

Nate Cohn @Nate_Cohn

Some of the splits are obvious: second choices could break one way, shifting the result from first to final alignment. And the tabulated vote counts and pledged delegates have split before (NV '08, for ex.) and could again under some fairly unlikely and arcane circumstances — PolitiTweet.org

Nate Cohn @Nate_Cohn

Most of you know that there are *four* different sets of results this year. They could all split. https://t.co/CZTc4mb0cg — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins @micahcohen it was also my understanding--and i'll concede i'm not sure about this--that all yougov polls use MRP, and that they then rake their survey data back to their estimates so that they could have crosstabs etc — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins @micahcohen if they are, that strikes me as a fundamental difference. i don't immediately see a huge difference if they're just taking the same 3 day sample and estimating w mrp v. raking — PolitiTweet.org

Nate Cohn @Nate_Cohn

@kwcollins @micahcohen are they pooling over time? — PolitiTweet.org