Trump will win again

philostam said:
How is that a worry when answering a poll on the internet/over phone? :grinning:

I assumed polls are private and/or annomymus.

I also don't think Trump supporters
are particularly shy anyway.

I still don't believe it's a thing, but I don't have other explanation at the moment.
Click to expand...

I never really got the lying to pollsters thing either...

Unless it was a coordinated effort to give democrats false confidence.
 
Last edited:
philostam said:
How is that a worry when answering a poll on the internet/over phone? :grinning:

I assumed polls are private and/or annomymus.

I also don't think Trump supporters
are particularly shy anyway.

I still don't believe it's a thing, but I don't have other explanation at the moment.
Click to expand...

If I were to take a wild guess, I'd say it's because Trump supporters skew older and aren't likely to answer online polls. Might be more likely to answer phone polls but I'm not sure what the relative turnout is for each.

Then again some polls are conducted entirely through phones, so...
 
Overconfidence in the models in 2016 was a real vulnerability of the left and that could’ve been taken political advantage of.

Anyway, the pollsters are getting better. The effects from lying were either small enough, or the models robust enough to incorporate them, that they didn’t work. But, the pollsters are still off so maybe it’s something that’s gotten worse since 2016.
 
:wink:
The news folk say the counters have until November 12th to finish? That length of time is going to feel like a blessed eternity.

:m051:
:m058:
 
Chickensoup said:
Overconfidence in the models in 2016 was a real vulnerability of the left and that could’ve been taken political advantage of.

Anyway, the pollsters are getting better. The effects from lying were either small enough, or the models robust enough to incorporate them, that they didn’t work. But, the pollsters are still off so maybe it’s something that’s gotten worse since 2016.
Click to expand...

To me it seems that polls are not getting better.

It's the job of sites like 538 to sort trough all of the polls, weight them correctly and make an overall prediction.

So 538 did a good job, despite polls being off. They gave Biden 90% of the probability - and that 90% included scenarios like the one we had now (polls being off). They did a good job even in 2016, to be honest.

I've been following them all trough election process and they were pretty much spot on. Credit where it's due. They have some serious statisticians and quants.

That's why I was surprised how some people actually predicted a Trump win. Like it could happen (still can, perhaps) but it was undoubtedly the less likely outcome. I think it was more a hope/wish than an unbiased prediction.
 
philostam said:
To me it seems that polls are not getting better.

It's the job of sites like 538 to sort trough all of the polls, weight them correctly and make an overall prediction.

So 538 did a good job, despite polls being off. They gave Biden 90% of the probability - and that 90% included scenarios like the one we had now (polls being off). They did a good job even in 2016, to be honest.

I've been following them all trough election process and they were pretty much spot on. Credit where it's due. They have some serious statisticians and quants.

That's why I was surprised how some people actually predicted a Trump win. Like it could happen (still can, perhaps) but it was undoubtedly the less likely outcome. I think it was more a hope/wish than an unbiased prediction.
Click to expand...

Yes, I agree. When I say pollsters, I mean people like Nate Silver and 538. Not the physical polling through interviews, etc. The business of political polling which now includes statistical modeling.

Personally, I did not like Nate Silver’s explanation for the 2016 f* up. Though I agree he addressed it (partially) in the modeling in 2020.

I think in 2016, most models did not accurately incorporate a change in people’s opinions after James Comey cast doubt on “her emails!” days before the election.

It would have been impossible to do so. Which is why when you look at something like the vaccine trials going on right now, anytime an anomaly comes up in a patient, the whole process stops. Until the drug companies can isolate and measure what’s happened.

No one could’ve done that in 2016 and that’s what 538 should’ve been upfront about. Not be all like, we’re totally 99% accurate. They took a randomized poll after the Comey announcement, but that’s insufficient to gauge things.

Trump voters who changed their minds were like, you don’t know how I think. Die hard Trump voters were like, see I knew you “coastal elites” were wrong this whole time. So, Trump’s campaign leans into the “they’re wrong, they’re out of touch, only you know what’s what - nudge nudge, wink wink” narrative and messaging.

Also, leading up to the 2020 election, they tried to replicate what happened in 2016 with the whole October surprise and “his emails!!!” bit. In case anyone was wondering what that was all about.

As for for how there could people predicting a Trump win, I mean, both sides have the same models...

I didn’t really follow the 538 models this time, though I agree they did a good job playing out possible paths and all that. They are now updating their simulation models using fatter distribution tails, so that’ll help improve the models further.
 
You must log in or register to reply here.
Back
Top