Why Kamala’s Campaign Is Spiraling Into Desperation Mode

[iluvatar5150]

You felt the need to post that? Interesting.

You felt the need to post another thread that requires signing up to read beyond the ridiculous headline?

 

[Hammster]

I am not going to create an account to a site that is sure to send an endless supply of emails.

Okay.

 

[Hans Blaster]

On Nate Silver’s model, 60% is pretty close to a toss-up. I’m not going to parse the html again to read another of Matt Walsh’s takes that, if the headline is any indication, is just as inane as the others I’ve read.

Agreed. There is little to no difference in these models between 40% and 60% likelihood of winning as they are fairly sensitive to "corrections" applied to poll averages. I am quite skeptical of models that claim to predict the outcomes of singular events, like who will win tomorrow's Monday Night Football game. Statistical models of collective things, like how many US voters will vote for Donald Trump is different. Measuring errors and correcting bias in data collection are normal things.

For anyone who doesn't know, in these election models they select random samples of test elections with the percentages for each state based on the measured voter polling preferences to create a random election in each state. Half will be above the average and the "spread" will depend on the measured statistical polling margins of error. A whole bunch of "simulated elections" and then count the percentage of simulated elections that goes to each way. That is the number they report as the 60%, etc.

Where things go quite astray is that Silver has decided to model the estimated final poll numbers. Since this "Trump lead" started circulating online, I have seen several people commenting on the "model adjustments". One that apparently relevant now is the "convention bounce adjustment'. In the old days, when more people watched live TV and partisan polarization wasn't as strong, pollsters usually saw a candidate go up a few points in the polls after their convention. Usually this "bump" would go back to "normal" afterwards. Silver has built "corrections" for the "convention bump". From what I have read online, with only minor polling changes, the "model probabilities" shifted because now there is a (negative) correction for the assumed "bump" in polling from the convention (there really wasn't) and so it shifted from ~60% Harris to ~60% Trump. If there are no more changes in actual polling, it will flip back. Neither for any actual events, just because of assumed correction in a model.

I'm not saying it should really be 60% Harris, or anything, but rather that these models are generally problematic and full of assumptions masquerading as "scientific statistical models". Track the polling averages if you like, but if professional campaign managers are "panicking" about a shift in Nate Silver's model (they aren't) they are truly incompetent.

 

[iluvatar5150]

Agreed. There is little to no difference in these models between 40% and 60% likelihood of winning as they are fairly sensitive to "corrections" applied to poll averages. I am quite skeptical of models that claim to predict the outcomes of singular events, like who will win tomorrow's Monday Night Football game. Statistical models of collective things, like how many US voters will vote for Donald Trump is different. Measuring errors and correcting bias in data collection are normal things.

For anyone who doesn't know, in these election models they select random samples of test elections with the percentages for each state based on the measured voter polling preferences to create a random election in each state. Half will be above the average and the "spread" will depend on the measured statistical polling margins of error. A whole bunch of "simulated elections" and then count the percentage of simulated elections that goes to each way. That is the number they report as the 60%, etc.

Where things go quite astray is that Silver has decided to model the estimated final poll numbers. Since this "Trump lead" started circulating online, I have seen several people commenting on the "model adjustments". One that apparently relevant now is the "convention bounce adjustment'. In the old days, when more people watched live TV and partisan polarization wasn't as strong, pollsters usually saw a candidate go up a few points in the polls after their convention. Usually this "bump" would go back to "normal" afterwards. Silver has built "corrections" for the "convention bump". From what I have read online, with only minor polling changes, the "model probabilities" shifted because now there is a (negative) correction for the assumed "bump" in polling from the convention (there really wasn't) and so it shifted from ~60% Harris to ~60% Trump. If there are no more changes in actual polling, it will flip back. Neither for any actual events, just because of assumed correction in a model.

I'm not saying it should really be 60% Harris, or anything, but rather that these models are generally problematic and full of assumptions masquerading as "scientific statistical models". Track the polling averages if you like, but if professional campaign managers are "panicking" about a shift in Nate Silver's model (they aren't) they are truly incompetent.

IMO, one of the better things about Silver is that he tends to be pretty open about how things work under the hood. Whether or not his model is correct, it’s interesting to learn about how it’s done.

 

[Hans Blaster]

IMO, one of the better things about Silver is that he tends to be pretty open about how things work under the hood. Whether or not his model is correct, it’s interesting to learn about how it’s done.

Openness is nice, but frankly these "predictive models" are likely just junk. Silver or other modelers include effects for "fundamentals" and other things that "explain" election performance. They also assume the variations are independent from state to state. They are not. The reality is that most of the close states are going to shift one way or the other together.

 

[iluvatar5150]

They also assume the variations are independent from state to state. They are not. The reality is that most of the close states are going to shift one way or the other together.

You sure? I’m not going to go dig it up, but I could’ve sworn I’d read him talking about this sort of correlation in the past.

 

[Hans Blaster]

You sure? I’m not going to go dig it up, but I could’ve sworn I’d read him talking about this sort of correlation in the past.

You could be right. I've not seen anything about them using correlation in sampling before.

 

[Hazelelponi]

Openness is nice, but frankly these "predictive models" are likely just junk. Silver or other modelers include effects for "fundamentals" and other things that "explain" election performance. They also assume the variations are independent from state to state. They are not. The reality is that most of the close states are going to shift one way or the other together.

The betting market was accurate in 2016 and 2020 for predicting if I remember correctly.

 

[Pommer]

Agreed. There is little to no difference in these models between 40% and 60% likelihood of winning as they are fairly sensitive to "corrections" applied to poll averages. I am quite skeptical of models that claim to predict the outcomes of singular events, like who will win tomorrow's Monday Night Football game. Statistical models of collective things, like how many US voters will vote for Donald Trump is different. Measuring errors and correcting bias in data collection are normal things.

For anyone who doesn't know, in these election models they select random samples of test elections with the percentages for each state based on the measured voter polling preferences to create a random election in each state. Half will be above the average and the "spread" will depend on the measured statistical polling margins of error. A whole bunch of "simulated elections" and then count the percentage of simulated elections that goes to each way. That is the number they report as the 60%, etc.

Where things go quite astray is that Silver has decided to model the estimated final poll numbers. Since this "Trump lead" started circulating online, I have seen several people commenting on the "model adjustments". One that apparently relevant now is the "convention bounce adjustment'. In the old days, when more people watched live TV and partisan polarization wasn't as strong, pollsters usually saw a candidate go up a few points in the polls after their convention. Usually this "bump" would go back to "normal" afterwards. Silver has built "corrections" for the "convention bump". From what I have read online, with only minor polling changes, the "model probabilities" shifted because now there is a (negative) correction for the assumed "bump" in polling from the convention (there really wasn't) and so it shifted from ~60% Harris to ~60% Trump. If there are no more changes in actual polling, it will flip back. Neither for any actual events, just because of assumed correction in a model.

I'm not saying it should really be 60% Harris, or anything, but rather that these models are generally problematic and full of assumptions masquerading as "scientific statistical models". Track the polling averages if you like, but if professional campaign managers are "panicking" about a shift in Nate Silver's model (they aren't) they are truly incompetent.

I see a lifetime‘s worth of government grants to do studies on how to get a quarter to flip heads 75% of the time, and successful politicians who’ll benefit mightly with such “knowledge”?

 

[Hans Blaster]

The betting market was accurate in 2016 and 2020 for predicting if I remember correctly.

I see no reason to trust your memory, nor any betting market. This thread is not about gambling markets, but rather so-called "predictive models".

 

[Servus]

Historian Allan Lichtman, the guy who predicted nine out of the last ten election winners, including Trump in 2016 and Biden in 2020, says Kamala is going to win. If he says she's going to when then I think she's going to win. I'm actually looking forward to the Kamala Harris presidency. Not because I think she'll be any good, but just the opposite. I've got my popcorn ready for the clown show.

Me too. The downside of enjoying the clown show though is how much damage it's going inflict.

 

[Hazelelponi]

I see no reason to trust your memory, nor any betting market. This thread is not about gambling markets, but rather so-called "predictive models".

Some predictive models use the betting markets in their analysis.

I thought that was along thread lines. I'll excuse myself then.

Good night

 

[Pommer]

Me too. The downside of enjoying the clown show though is how much damage it's going inflict.

It’ll be something to have been seen, but I think we’ll be able to steer out of this slide.
All of us are mostly good people, and we all know that about each other, and trust that.

 

[Pommer]

Some predictive models use the betting markets in their analysis.

I thought that was along thread lines. I'll excuse myself then.

Good night

People bet on everything, a vote is a bet that one understands enough to have picked the very best candidate to run our country.
And people will bet on how smart/stupid that choice is.

 

[Servus]

There are unusual circumstances in this election. There's a former president who's running for a second non-consecutive term, which only happened once before 131 years ago. (Although in that case Cleveland ran aginst incumbant Harrison a second time. In this case Trump will now be running against the incumbent's VP for a second non-consecutive term. So that now makes this a first time in American history event).

Then there's the other presidential candidate who started their campaign only 3.5 months before the election after the sudden withdrawal of the incumbent.

 

[Servus]

It’ll be something to have been seen, but I think we’ll be able to steer out of this slide.
All of us are mostly good people, and we all know that about each other, and trust that.

What I should have said is all the continuing damage. Even worse financial hardships, crime, illegal immigration, continued indoctrinating young kids in gender and sexual orientation fluidity etc. I'm expecting a Harris presidency to be a 2.0 version of the Biden administration along those lines.

 

[Servus]

A Kamala presidency would be hell on earth. I don't expect us to make it if she's elected, not as a nation like before.

Maybe there's people happy for the downfall of America. But it won't be anything good.

I'm here for the trying to save us part only.

Yep there's millions who are salivating over the prospect of the downward spiral continuing.

So we'll have to see if enough American voters have had enough of it already.

Or it will take four more years of even worse until they've had enough.

Or perhaps we're already past the point of no return.

 

[Pommer]

What I should have said is all the continuing damage. Even worse financial hardships, crime, illegal immigration, continued indoctrinating young kids in gender and sexual orientation fluidity etc. I'm expecting a Harris presidency to be a 2.0 version of the Biden administration along those lines.

A Presidential Administration works In the areas that it identifies as needing some work to have the best society that we can and the American people are free enough to disagree with as much of those things as they want to tolerate or use as a reason to vote that administration out.

If the ex-incumbent secures enough votes in the right places he could win the Electoral College and who knows, maybe even the popular vote for the very first time?

 

[Hazelelponi]

Yep there's millions who are salivating over the prospect of the downward spiral continuing.

So we'll have to see if enough American voters have had enough of it already.

Or it will take four more years of even worse until they've had enough.

Or perhaps we're already past the point of no return.

In 4 more years we will be past the point of no return. This is it. We save it now or we sink...

I don't think after 4 years of Kamala we could ever pull it back.

I deleted my comment by the way... No point to it. Just wanted people to care about our future is all .. stupid dream of mine is all.

 

[Nithavela]

In 4 more years we will be past the point of no return. This is it. We save it now or we sink...

Funny, that's what you guys say every 4 years. Both sides.