On one hand, Trump being unhappy is good.
On the other hand, this might lead to him trying increasingly stupid things to get his numbers rise, up to starting a war.
9 Likes
It is a word if you don’t misspell it.
To your point, however, aside form being manipulable the polls don’t reflect how a president is elected in the U.S. and should probably have a larger margin of error than they let on.
5 Likes
I remember the liberal web site Daily Kos predicting, on election day 2016, that, based on polls, Hillary Clinton would win 323 electoral votes to trump’s 215. You can see that prediction on their archived web page (captured on 11/8/16 at 12:14 pm), here. A little confusingly (because the electoral vote totals are a tiny bit different), they also have a graph that shows that Clinton has a 92% chance of winning.
I don’t know what they based their poll numbers on (probably an aggregate). I do remember that DKos showed Clinton having a much greater chance of winning (in the 80% range IIRC) in the week before election day. Then trump won. So much for the polls. That ended my interest in what polls said.
And so much for polls, especially when the first caucuses won’t take place until February 2020, over 6 months from now.
2 Likes
THIS.
National polls don’t mean squat. Let me see polls in states that voted trump, then I’ll have some hope.
To be fair, that was prior to Comey’s surprise on the Friday before election day. By Monday, the 538 poll was a dead heat.
3 Likes
The cynic in me believes this is already in the cards: It’ll be an “Intervention” in Venezuela, or some kind of punitive / pre-emptive strike against the Iranian nuclear facilities.
Or both.
When was the last time US voters failed to re-elect a President who led us into a war during his first term?
Edit to answer my own question:
Bush Sr. led the US (and a multi-national coalition) into the First Gulf War, basically the liberation of Kuwait.
I guess his “mistake” was that it was over and done before the next election.
4 Likes
I think both Obama and Trump won elections in part as a repudiation of foreign intervention (not that that always works out so well). I sorta feel like the general public really has no appetite for it anymore. I think if Trump started a war it would turn even more people against him at this point. I know a lot of democrats got on board with Iraq, but I also feel like they all felt burned by the whole thing. Obama’s “hope” campaign was in part a repudiation of Bush and endless war. Trump is a dumpster fire, but he was also a repudiation of Bush era republicans…
5 Likes
Yeah, I think it’s interesting that a Fox News poll has Bernie neck-and-neck with Biden against trump. Everyone else shows him about 5% ahead of Bernie in head to head matchups.
1 Like
No, it assumed that errors in the polls were correlated, whereas most other models did not.
1 Like
No, most people treated the polls like someone who hadn’t had a statistics class (or worse yet had one, forgot what all the words mean, but did use the words in mis-explaining it to others). The few that did actually treat them like probabilistic models said things like “Trump’s chance of winning is about the same as losing a game of Russian roulette”.
For example: Election Update: Is The Presidential Race Tightening? | FiveThirtyEight
Which is not only more accurate then adding a bunch of stuff up and getting 99%, it also describes things in terms of events that are unlikely, but clearly happen from time to time.
5 Likes
I’m wondering if the people who seem to think that polls showed a 99+% chance of Hillary winning were actually misreading statements like “99% of polls show Clinton leading Trump”? Because that’s totally plausible but it’s still a very important distinction.
To use the Russian Roulette analogy, 99+% of statisticians would say the odds of surviving are in your favor, but that’s not the same as saying you have a 99+% chance of surviving a round of Russian Roulette.
7 Likes
4 Likes
Uh, no. No it wasn’t. Doing stats well is hard, but well done stats aren’t just bullshit someone plucked out of their fundamental orifice.
4 Likes
Its more that no poll is predictive, and this far out from the election in question polls don’t do such a great job of capturing real support level. None the less how people would actually like to vote. Its fairly well established that the results of early polling like this are mostly driven by name recognition. To the point that you can get very similar results by just polling name recognition. The other major contributor is approval/disapproval of the incumbent.
So practically what this is showing is that as many people are familiar with Bernie as Biden. And that few enough people actually support Trump, that almost anyone could potentially beat him.
And plenty of those battleground or contestable states are just as locked in. GOP gerrymandering, voter suppression and other nonsense intended to maintain solid GOP control and reliable results in the nationals despite rising support for the DNC.
Almost every swing or battleground state we talk about is a state that’s swung leftward rather than shifted towards the GOP.
From what I’ve read that was more down to internal polling than assumptions that Democrats win in Michigan by default. Clinton used an Obama campaign derived polling and turn out contractor that specializes in telling campaigns where to target resources. Apparently they seriously fucked up their internal polling and missed some problems in specific parts of specific states. They advocated pulling resources from particular states to focus that money on the problem areas they had identified. And based on that inaccurate, and incomplete info, the turnout end of it failed to get enough people to the polls in spots that turned out to be really important. And didn’t do nearly enough in whole states.
Basically they were driving turnout where they didn’t need it. While all those surprise Trump districts were left to their own devices. Dropping top level campaign activities in a state like Michigan isn’t itself a bad idea. Provided you offset it with additional turnout efforts. That sort of shifting, targeted resource use was a Hallmark of Obama’s campaigns and was considered a revolutionary use of data in campaigning. But if your data can’t accurately tell you where they need turnout instead of stump speeches. You fuck yourself.
These numbers, percentage chance to win. Are not the polls. They’re probability predictions based on the polls. Polls are not chances to win, they’re counts of who a representative sample of the population plans to vote for. X number of people support Clinton, y number of people Trump.
And in large part the polls were plenty accurate. Most polls in most places match the actual vote numbers for what they were sampling. And where they did not it was less that they were wrong, than that they were outdated. Many states do not see regular district level polls in all areas. So there were districts where the last poll conducted was weeks before election day. And whole states that were polled pretty irregularly leading to sudden changes in numbers.
So we got an accurate snap shot of how many people were going to vote for each candidate, but in large part we could not see where those votes were. Or at least not specifically enough to accurately predict the outcome.
Remember that Trump lost on raw votes, by really close to the margin predicted by most venues. Yet he won the election by taking very specific districts scattered around just a few states. Even in those districts and the states he surprisingly took the margins were not big. In some cases less than a percentage point. A few hundred thousand votes in three states as the math geeks kept pointing out. And it almost seemed accidental, they certainly didn’t think they had won. And looking at that campaign it’s pretty hard to say they were even attempting to do that. It looks very much like their shit just happened to stick a little better in certain places, and by happenstance or other factors a couple of them turned out to be the perfect spots to cobble together an insanely narrow win. The exact same number and proportion of votes, in slightly different places and it would have been a loss.
Bernie is weirdly popular with a certain subset of very conservative voters and I do not understand it.
Fox and the rest of the right are also boosting Bernie because socialism scary. And I think they’re assuming he’d be easy to beat. I think he’d be easier to beat than some other candidates, but that doesn’t mean it would be a simple thing over all.
So Fox has something of an active interest in promoting the impression that Bernie is the front runner. No matter what’s actually going on.
5 Likes
That was my read. The worst thing for them in the world would be for Trump to be well ahead in the polls.
6 Likes
2 Likes
That was essentially one guy’s statistical model producing an outlier result, and even he gave Trump around 1-in-100 odds, not 1-in-1000. That same article notes
Other polls were less bullish on a Clinton victory. FiveThirtyEight said the likelihood Ms Clinton would win was 65 per cent, while the New York Times upshot said there was a 85 per cent chance.
So “Clinton has a 99.9% chance of winning this thing” was hardly a mainstream position among pollsters and statisticians.
5 Likes
There’s been a bit of scuttlebutt recently about GOP internal polls looking really bad for Trump and the party getting worried about it and or having less tolerance for his horseshit as a result. As well as Trump being enraged about it. Here and there “sources say” “a staffer reports” sort of thing.
So you got this Fox poll, then some of Trump’s own internal campaign polling leaked right around the same time. So this could be more about pushing Trump, than pushing his voters.
1 Like
