Two and a half hours after final poll closures the networks finally decided to call Michigan for Sanders moments ago. It is close, 50-48% or some 22K vote spread. I guess the trade argument worked, though I for one have never been convinced of the connection between that and job losses especially this far removed and HRC actually voted against the only such deal to come up when she was Senator. OH and PA have similar demographics so this might go on for a while longer, though she is still way ahead even in pledged delegates. She also trounced Sanders in Mississippi as expected.
The political scientist in me is interested in how the prognosticators so badly failed on this one. I think the most recent Michigan poll I saw had Clinton up 55-42% and as recently as this evening 538 was giving her better than 99% odds and I don’t recall Nate Silver ever getting this kind of thing wrong.
I am not sure if it’s ever been shown that he is doing anything other than aggregating polls by other people.
It takes those aggregated polls and bends them by factors he thinks are likely to shift the results. Things like $$$, endorsements, etc.
Modelling works great when you have many similar past elections to base the data off of, but not so great when you don’t.
Nate’s been consistently wrong about this election, and I think it’s because it’s safe to say this election has been very, very different than the big successes Nate’s had in the past. All those factors that made it more likely someone would win in years past, like big endorsements or raising lots of dollars from corporate donors, either aren’t helping or are actually hurting.
There’s an economic populist insurgency that exists now that didn’t exist in 2012 or 2008, taking place in both political parties. Until Nate’s modelling can account for it, it’s likely Michigan won’t be the last state to shock or surprise.
Like I said I’m pretty sure this is the first time this year he did not accurately predict the winner.
When it’s a few days before an election, it’s hard to be wrong unless the polls are way off (and only Michigan seems to be that way so far). Silver has been seriously discounting Trump for a long time – but now admits he has been wrong.
It’s been more in the commentary of “Trump won’t win the nomination” than specific state predictions, though.
And the “polls-plus” model, in which party endorsements provide extra weight to a candidate’s chances, doesn’t seem to be the best model for this crazy year.
And place the degree of uncertainty front and center. They have always been quite clear that polling in the primaries is nowhere near as robust, and therefore reliable, as polling in the general. The problem is that there are too few participants, which makes things volatile, especially in a multi-candidate race. The statements from last summer on TRump really weren’t based in anything other than an observation that he meets none of the criteria usually required of a candidate. In recent weeks, they have been fairly straightforward about the degree to which the assumptions upon which the model is based are proving ephemeral, and what that might mean for assumptions for November.
I personally like Silver, and I’m glad he’s made data “cool” again (or something like that). But while I know ryepower12 and I may disagree sometimes, rye is definitely right about the problems of quantitative modelling. It works in the vast majority of cases when things are operating as normal. It doesn’t work when there’s a possible realignment or other unusual political shifts going on.
Increasingly, the latter looks like what’s going on in 2016.
And the head of our poli sci department really liked Nate Silver and even assigned us 538 posts to read, and this was before he really caught on nationally. Then again, I predicted the EC correctly last year so I am not sure how much modeling is needed to figure out which state goes to which candidate.
His modeling works best for a static and polarized electorate as we’ve had the past six presidential elections, which is itself an anomaly
in our history. Winning the shrinking center matters in those kinds of races, it matters less where you have candidates making class based arguments with substantial crossover appeal.
…most recently here, these models depend upon the premise of stable electorates. This doesn’t take nuances or intangibles into consideration. For example, there is the matter of organized and effective GOTV.
In addition there are emotional issues such as enthusiasm for a given candidate or intense dislike of same. In the context of the Democratic Presidential nomination environment these things skew results.
Further complicating matters is the fact that, with the advent of caller ID and cellphone-only households, it is more difficult for pollsters to get representative samples.
Finally, in most areas – and I include the bulk of the Commonwealth – operational grassroots Democratic organizations are nonexistent. This provides traction for insurgent campaigns.
Hard Line Party Democrats are still insisting that running their hand picked candidate who runs a weak campaign and does poorly with white suburban males is the right choice to run against a self described outsider businessman who runs on a populist agenda, or at least, an agenda against the power of the state house/party in power.
that the Berners could hand the GOP the Whitehouse by not supporting the nominee, just as the Berwickers handed the governorship to Baker.
I really cannot imagine Bernie not wholeheartedly supporting Hillary if she is the nominee.
Every Berwick supporter I know worked their ass off for Coakley in the general. Cohn, Harmony Wu, Robert Fitzpatrick who was a Grossman backer did the same. She lost because she was a lousy candidate who offered no concrete ideas or agenda to govern the state. She lost because she spent more time indicting artists and caregivers than corrupt politicians. She lost because she was fucking awful.
Berwick had nothing to do with it, he and his supporters worked their asses off for her. And the fact that party hacks like you continue to disparage them despite their hard work is proof that you take them for granted and they should join a new party that would actually value their time and ideas.
It was the Capuano supporters who gave the Senate seat to Scott Brown!
to denying your involvement.
What’s so special about them? Is it ok to have a candidate that does well with white males but poorly with black females, for example? Or are you making the same mistake that “Party Democrats” often make in taking for granted the minority and female vote?
You wind up with Scott Brown, Charlie Baker…..
The Democrats have plenty to offer white suburban males but I’ll admit expressing this runs into difficulties.
Is swingy white suburban males? 5 percent, tops?
Sure that’s big enough that if you could somehow grab ALL of them and not lose any other group, it matters. But you’ll never grab all of them, and trying loses outreach to other groups.