So it’s now: okay the polls were off but we will weight the polls better next time. Next time!!! The undercount weren’t “shy” or necessarily suffering from social desirability. Some of them think of the pollsters as the cultural enemy. Good luck modeling that void with weights.
-
Show this thread
-
zeynep tufekci Retweeted Sadie Gurman
The problems outlined here, especially non-random low trust of pollsters, are not fixable by demographic weights. Or any method I can think of—besides already knowing the answer. Not amount of talent or effort can model this. Sometimes one can guess right. Sometimes not.https://twitter.com/sgurman/status/1324150511323500546 …
zeynep tufekci added,
Sadie GurmanVerified account @sgurmanSome believe a distrust of institutions is more pervasive than anticipated across many voter groups, and that it leads conservative voters to avoid participating in polls in disproportionate numbers. The story everyone wants to read by @aaronzitner https://www.wsj.com/articles/what-went-wrong-with-the-polls-this-year-11604536409 …14 replies 20 retweets 141 likesShow this thread -
zeynep tufekci Retweeted Nate Silver
Nice denominator except that you barely need polls to call about 40 of those states. Of the remaining 10, the polls are off by large numbers systematically in practically all of them. (Some got the coin toss right! Very reassuring!)https://twitter.com/NateSilver538/status/1324379322740867073 …
zeynep tufekci added,
Nate SilverVerified account @NateSilver538And a lot of it is because of the blue shift in late-counted ballots. If we go to bed at 1am on Wednesday and polls have called 48/50 states correctly and Biden's won the popular vote by 5.3 points or something, I don't think there's a "polls blew it again!" narrative.Show this thread10 replies 27 retweets 316 likesShow this thread -
zeynep tufekci Retweeted Kevin Robillard 🇺🇸
Empirical proof for a possibility I had mentioned in my articles. The increasing certainty people felt over Biden's presidency might well have been one of the factors that cost Democrats the Senate.https://twitter.com/Robillard/status/1325823456680808450 …
zeynep tufekci added,
Kevin Robillard 🇺🇸Verified account @RobillardA key point: Over the final month, GOP internal polling showed voters in red states started to believe Biden was likely to win the presidency. As they did, their willingness to vote for Democratic Senate candidates down-ballot plummeted. https://www.huffpost.com/entry/how-democrats-senate-dreams-crumbled_n_5fa8914fc5b623bfac5164a9?d6f … pic.twitter.com/TR1Pq90S2MShow this thread8 replies 66 retweets 188 likesShow this thread -
Cunningham just conceded the NC Senate. Look, the things wrong with our polls are not fixable by any known method. Pollsters may guess the post-hoc weights right here and there, but the uncertainty is giant and structural. What then is another question. First, let’s face reality.pic.twitter.com/FyOnafGMbL
8 replies 18 retweets 113 likesShow this thread -
Collins won ME by almost 10%. The most damning part of these misses is they’re in the same systematic direction as the 2016 misses—so you know the pollsters tried to correct for it. They still missed big. We can’t model and weight dark matter electorate with these response rates.pic.twitter.com/9F93mdYvDY
40 replies 112 retweets 787 likesShow this thread -
I like all the "why were the polls were so systematically wrong *again*" stories coming out, but here's the part that keeps getting skipped. The problems people are identifying are all plausible AND not fixable by any known method like weighting. https://www.nytimes.com/2020/11/10/upshot/polls-what-went-wrong.html …pic.twitter.com/CgA65TFg1V
3 replies 15 retweets 69 likesShow this thread -
I'm just going to put an excerpt here from my NYT piece on why polls and forecasts were even less trustworthy in 2020 I published *before* the election. I'm not making post-hoc claims here. (Current text updated a bit now to reflect that's what happened) https://www.nytimes.com/2020/11/01/opinion/election-forecasts-modeling-flaws.html …pic.twitter.com/txR41T6G7I
4 replies 7 retweets 59 likesShow this thread -
zeynep tufekci Retweeted nathanjurgenson
I think we have a lot of this going on. Also it’s hard for people who think of themselves as quants to admit they are now more in pundit category—feeding the horserace but without a superior empirical basis. The polls aren’t fixable by any known method. Let’s talk about that.https://twitter.com/nathanjurgenson/status/1326564758795870209 …
zeynep tufekci added,
13 replies 8 retweets 80 likesShow this thread -
Replying to @zeynep
Predictive power of polling is probably greater when subjects don't know they are participating in a poll. It can't be about surveys or questionnaires: polling will have to be more about revealed preference than intention. As per Dr House: "Patients always lie".
1 reply 0 retweets 0 likes
-
-
Replying to @zeynep
Sure - but you get the point. We'll have to create data-mining models from other data sets, than just what people say they're going to do.
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.