Even if we were on some uninterested mission to evaluate presidential forecast models, that would be a stupid way to do it. We can't model well both because we can't poll well and because they are too rare to give us a chance to test the models. Sooner we accept this, the better.
-
This Tweet is unavailable.
-
-
Replying to @zeynep @mattyglesias
It would help if the pollsters were transparent about their weighting. Then interested observers could at least understand why polls differ in a given cycle. But point taken - when you don't have a sufficient sample size or a stable sample, each new data point is an outlier
1 reply 3 retweets 8 likes -
Replying to @daggaroosta @mattyglesias
We just can't do this. Not with 3% response rates, and not when pollsters are seen as part of the polarization. I just don't see how we can argue the model is great despite unreliable data and no sensible way to validate it! Come on, folks. Writing is on the wall. 2016, 2020.
1 reply 5 retweets 59 likes -
This Tweet is unavailable.
-
Replying to @mattyglesias @daggaroosta
And it distorts the campaigns & us. I don't want to go back to vaporware pundit narratives either, but at least they didn't come with an aura of scientism. If we're gonna do astrology, let's at least put it in the horoscope section. Not everything lends to well-modeled forecasts.
5 replies 7 retweets 78 likes -
but why does this have to be so binary? it's not astrology and it's not weather forecasting - both are straw men. it's a hard-to-forecast social science process. I agree gets too much attention given the challenges & uncertainty underestimated but that doesn't mean it's astrology
1 reply 0 retweets 10 likes -
I did not say all models in social science are astrology. But it's hard to defend the presidential modeling as it is. So many major issues, including we're in a pandemic! We should have just said that, especially this year, stop refreshing that forecast. Too much uncertainty.
2 replies 1 retweet 11 likes -
Replying to @zeynep @BrendanNyhan and
the uncertainty was baked into the model itself. you can quibble with whether it had enough, but it does seem like throwing up your hands and saying "the future is unknowable" is not that great a plan
1 reply 0 retweets 1 like -
Replying to @fortenforge @zeynep and
no, the uncertainty wasn't baked into the model, quite the opposite. The MoE for most polls is ~3% and the collective polls for swing states were off by more than 5%. False precision + false accuracy = false narratives that wreak havoc on efficiency of effort
2 replies 1 retweet 3 likes
Yep. We're at GIGO stage of models for events too rate to validate or check.
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.