This race has peeled off the veneer in American politics. When you look at the demographics, even older white women are more inclined to at least consider a Trump presidency. Clinton is particularly dominating the female vote between the ages of 18 and 45 (roughly). Once you get women who are in their 50s and older the support for Clinton drops off somewhat dramatically.
Couple this data together with other demographic stats and it’s clear that Trump is surviving because much of white America, particularly older white men, sees its grip on power slipping – it’s not the country they grew up in. I don’t think all of these people are robe-wearing bigots, either. Rather it’s perhaps just a fundamental discomfort about the idea that this country’s demographic Rubicon is about to be crossed.
I’ve seen some progressives saying in recent days that economics are not a factor and that it’s purely about race, which I think is incorrect as well. I think that economic insecurity is a factor. Trump supporters might be doing well statistically in terms of salary and wealth but they’re still insecure. They worry about the lack of stable jobs. In the factory days there were jobs aplenty for white men without college degrees or with lower levels of education. Now? Not so much. And the financial crisis made retirees anxious and cynical about the future of their retirement.
Economic insecurity has a habit of feeding racism, particularly when used by opportunistic racists. The politician who tells you that the reason your job isn’t secure is because those people are stealing American jobs can always rely on a certain percentage of the populace to buy into that argument out of fear and anxiety, even if those people aren’t particularly racist themselves and the argument has little factual basis.
At this point polls aren’t moving the lines by more than a smidgen each update, in general. It would take several polls back to back with landslide numbers to get Hillary to 95% or so, and it would take several polls back to back with even numbers to get her below 70, I think. Neither seems very likely, but we’ll see.
Looks like we’re finally seeing the convergence I was asking about a little while back. Guess 538’s model has finally decided that the ‘fundamentals’ are mostly baked in.
Actually, there is an article posted yesterday evening explaining that it takes a lot to move the forecast from 90% to 95% but much less data to move the forcast back towards a close race. The models are actually designed to become more responsive to new data as the election approaches (which is why the NowCast always jumped around a lot). So only one or two good polls with a strong Trump showing might push Clinton’s chances down considerably, whereas five more polls showing Clinton in the lead doesn’t really tell us much we (and the models) don’t already know.
(On the other hand the polls themselves tend to move less as the election approaches, and as RTFirefly points out, more is already “baked in” to the results.)
It’s simply a variation of the ‘plexiglass principle’ from SABRmetic research showing itself.
The further away from the baseline something is, the harder it will become to move away from the baseline and the easier to move it toward the baseline.
By way of comparison, Sam Wang has Clinton as 97% to win.
Betting markets generally have it 85-15. Betting markets can often be much more accurate than the experts but in this case I don’t think so. In binary decisions there’s a lower limit to how far the odds will typically go, and the Trump numbers are likely being pumped up by a lot of folks making bets based on emotion rather than logic.
In truth, I think the likely number is somewhere between 538 and Sam Wang. Perhaps 93-7. It will be interesting to see how Iowa turns out. Iowa is a difference of opinion between the two; Sam Wang has had it pink for basically the entire election, but 538 believes it likelier to go Clinton.
No, this is something different. If you read the 538 piece, it has to do with the increased difficulty of approaching an asymptote (100% chance of winning) as you get closer, rather than the difficulty of getting farther from the baseline.
A lot of the blue in Texas has to do with the fucking moron Trump being the red choice. I doubt there would be a lot of bluing up if there was a viable Republican candidate (I realize that “viable Republican candidate” is an oxymoron on this board, but still).
As has been pointed out, white non-Hispanics make up less than half the population of Texas, and that demographic shift is likely to continue. The fact that Texas is still red is probably more due to white non-Hispanics being more likely to vote than minorities.
No high quality Iowa polls since before the Sex Tape. Nate is making Iowa adjustments based on national trends since then, Sam is going by last good state polls showing Trump. A good state poll should converge them.
Well, this is why I find Wang’s refusal to consider national polls will tend to lead to obviously incorrect results. I’m not saying his model is wrong, exactly - Wang is trying to keep things VERY simple, on the belief that simple is likely to be right.
But the absence of a national poll adjustment, while it significantly complicates things, is clearly wrong in a Bayesian sense. Generally this probably won’t matter, but it could if no one polls Iowa (which I assume someone will.)
[QUOTE=Duke of Rat]
A lot of the blue in Texas has to do with the fucking moron Trump being the red choice.
[/QUOTE]
Well of course that is true. It’s true of almost every state. Trump is the main reason Florida is leaning blue. He’s the main reason Virginia isn’t really even a swing state anymore. He’s the reason Georgia and Alaska have a distinctly blue tinge to them.
But you know where there is no blue at all? West Virginia, Idaho, and Alabama. So there’s clearly something about Texas that allows the blue to start seeping in, that is not true of WV, ID, or AL. To my mind, it’s obvious what that is; big cities. Texas has big, rich, cosmopolitan cities - Dallas/Fort Worth, Houston, Austin, San Antonio and El Paso are all larger than any city in some of those dead-red states and in some cases have more people than whole states. Just going by population in city limits, Texas has at least fifty cities larger than West Virginia’s largest city. City voters are disproportionately Democratic voters.
This isn’t to say Houston is as liberal as Boston, it’s not. But it’s more liberal than the sticks, and has great diversity, and that’s not good for Republicans.
Yeah, but you can also look at the shift by county and the blue shift is ( at least in my county) way out of proportion to any shift in minority population. I think a lot more white non-minorities think Trump is a moron as opposed to a sea change in demographics due to population make up.
It seems this is what it would be if various generally Republican states shifted toward Clinton…where she was down previously in this same election cycle. That’s obviously not a demographic shift in who is eligible to vote over a matter of months. It’s arguable that Trump is further energizing Hispanic voters against himself lately, but doesn’t seem the simplest explanation, since specifically Hispanic-related issues and controversies have tended to fade in the general election. What’s been the focus lately is mainly Trump appearing even more personally unsuited to the presidency. So the simplest explanation for shrinkage in margin in states where he’s ahead is that it’s more of a parallel shift, or even more concentrated among whites, against him because he’s just too much of an idiot and lout to be an alternative to Clinton even for those who dislike her.
My county has ~17% non-white population. Obama got ~11% of the vote in 2012, IOW he didn’t even get proportionately all the non-white vote. The 538 map shows a + ~25% blue shift. Even if every non-white in the county voted for Clinton, I don’t think it would be close to enough to account for this gap. That’s why I think that traditionally Republican white voters are (if the predictions of 538 pan out - yet to be seen) accounting for the difference because they cannot vote for Trump the Imbecile. Not because they are still voting red and there has been some unseen surge in non-white voters to the county - there are only ~ +/- 20K people in the county, we’d notice if some extra 4000 voting age Hispanics showed up.
The “incorrectness” presupposes either that (a) the national shift applies to all states evenly, not necessarily true, or (b) if you apply each national poll to each state by making by-state demographic adjustments - I think Nate does this - you end up with pretty small sample sizes and chasing noise.
I don’t think it’s a matter of “wrong” (Bayesian sense or otherwise) but that Sam’s results either lag the national mood (if there’s an actual trend) or ignore the noise (if the national polls are just noisy). And we just can’t tell at the moment for Iowa.
Presumably this question could be approached statistically ; you could look at historical data and check if fluctuations in national polls consistently lead to changes in state polls. You could also see which states have the most and least correlation and adjust your forecast accordingly. I am guessing Nate has done this work and that just ignoring the national polls isn’t the optimal approach.
Well, Wang is a neuroscientist. I don’t know if he has the time and resources to devote to election models that 538 would have. He may have chosen a simpler model simply out of convenience.