Great map by 538 showing where the polling shifts have occurred.
… Texas… wow.
I’m only still worried because of Brexit. Wasn’t Brexit at 10% or less odds of passing?
I guess UK polling is less reliable for some reason. They also had David Cameron and Ed Milliband tied going into election night, I think.
Except JohnT that that is not a map based on polling shifts.
It’s interesting and it likely actually does represent what’s going on to no small degree but it is not based on observed polling shifts.
Maholoth, nope. Betting markets had given remain a firm favorite position (not 90% or anywhere close) but the polling had it as a toss up.
That is a great map, thanks for that. What the hell is wrong with the people in the Berkshires?
Naw, Brexit was leading in the polls, albeit within the MOE for many of them. A model that gave Brexit 10% chance of winning the day the vote was taken was made of pure fantasy.
Yeah, flubbed that up, didn’t I? Appreciate the clarification.
JohnT, note that that map is just showing shifts, not current position. So yes, Texas is getting bluer… but it’s still pretty solidly red, just not quite as solid as it used to be.
Mahaloth, the polling on Brexit, at least shortly before the vote, actually had it as more likely than not to pass. The big problem was pundits ignoring the polls, or using old polls before the public opinion shifted.
It’s really nothing like Brexit. A referendum is just different from an election for one thing. The polling was actually pretty tight on Brexit compared to current Presidential polling. Compare the HuffPo descriptions:
I wonder what’s going on with Maine? That’s an awful lot of red up there. Has Maine been struggling economically?
As far as Texas goes, I live in one of the blue counties on that map. I have noticed that a lot of the Republicans I know are not going to vote for Trump, and I could easily see Nueces county going for Clinton.
But really it is very interesting.
No, I understood what it was saying, I misunderstood how the data was derived.
Except it’s not actually showing shifts either, is it? It’s one of the most confusing maps I’ve ever looked at. What it seems to be showing (and I admit I haven’t had enough caffeine yet this morning) is this:
Minorities and college-educated whites are polling for Clinton. Let’s say if you are in that group you “should” be a Clinton voter. By the same token, non-college-educated whites “should” vote for Trump. Based on that, how many people who voted for Romney in 2012 “should” vote for Clinton this year? How many people who voted for Obama “should” vote for Trump this time? If in a county there are more Romney voters who “should” vote for Clinton than Obama voters who “should” vote for Trump, then Silver colors the county blue, and if there are more should-be Trumpers who voted for Obama than should-be Clinton supporters who voted Romney, then he colors it red.
Which is fine, but even if you follow all that, what does it mean? Damned if I know. I think it’s trying to show potential shifts, but my brain doesn’t handle that sort of second-derivative thinking very well, and most of those “potential” shifts will have already occurred and be baked into the current polling map. So what are we supposed to get out of it?
Northern Maine … the 2nd district … may as well be southern Mississippi for all the red in it. Mostly found on necks.
Ok, I think I found a better way to think about that map. It’s not about Clinton and Trump. It’s about the 2012 election, and it shows where Romney and Obama did better than the demographics would suggest (based on this year’s trends). Places where Romney did well among minorities and educated whites (like Texas and Miami-Dade, with lots of conservative Hispanics) are blue. Places where Obama did well among uneducated whites (like Minnesota and northern Maine) are red.
As the country gets more polarized along race and education lines, Silver is “predicting” (though it’s really post-dicting) what shifts will take place. The more that map reflects actual shifts, the more Silver can claim to have “explained” changes between the 2012 election and the current one.
Since Arizona’s actually looking close, I was wondering just how long ago I’d been saying the Dems ought to be putting some effort into that state. Since last September, at least:
*This was in a thread about which blue/purple states the GOP could turn red.
I’m not gonna pat myself on the back for mentioning NC, where 2008 clearly wasn’t going to be a one-off, the way Obama’s 2008 win in Indiana has been. But a year or more ago, few people were talking about Arizona as a place where the Dems could make inroads.
The NYT has an interesting, if wonky, article about that LA Times poll. As an example of how its unorthodox choices are affecting it, it talks about a 19 year old black male in Illinois who is a Trump supporter. Because of how the poll weighs micro-groups (it has its own category for males 18-21) and the weighing of his race, he has 30x the influence of the average voter in the poll and, each time the panel contacts him, Trump’s share of the black vote in the poll surges into the double digits.
Mahaloth, the polling on Brexit, at least shortly before the vote, actually had it as more likely than not to pass. The big problem was pundits ignoring the polls, or using old polls before the public opinion shifted.
[/QUOTE]
Thanks. Why did the Cameron/Milliband election look so close? I do remember polls indicating too close to call. Cameron killed him.
Multi-way elections in the Westminster system are more chaotic and difficult to predict; if you look at attempts in the UK, Canada, or other similar countries to do what Sam Wang and Nate Silver do, you’ll find there has been far less success. I’d assign this to two obvious problems:
- In the USA there’s basically two parties, with third parties rarely playing a major role. In the UK general election there were four parties vying for significant numbers of votes; the fourth place party, the Liberal Democrats, got 8% of the vote, which is a substantial number.
In such a system a small number of votes can go a long way or can accomplish little. Cameron’s Conservatives got 37% of the vote. That was more than expected, but not a huge amount more. But in a multi-party first-past-the-post system, 33% is a minority government and 38 is a substantial majority. The Tories beat LAbour by just 6.5%, not a slaughter by any means, but won more than 25% more seats.
In the 2015 Canadian general election the phenomenon was even more pronounced; the Liberal Party got 39.5% of the vote, which translated to 54% of the seats. The Conservatives got 32 percent of the vote, which netted them only 29% of the seats, and the NDP’s 20% of the vote got them just 12% of the seats. (If you simply reorder the names of the parties Conservative, NDP, Liberal, but keep the numbers the same, you get the results of the previous election almost bang on.)
- You also have the weird phenomenon of parties that have heavy support in a specific region and none elsewhere - in the case of the UK, the Scottish National Party, and in the case of Canada, the Bloc Quebecois. This causes bizarre mismatches between vote percentages as a national share and seat totals. In the 2015 UK election, UKIP got 12.5% of the vote and for all that effort won… one seat in Parliament. The SNP got 8.6% of the vote and got 50 seats. The Ulster Unionist Party, which exists only in Northern Ireland, got 1/30th as many votes as UKIP and won two seats. This makes the vote percentage-to-seat count weirder.
While this sort of thing is technically possible in a Presidential election, it generally doesn’t happen.
- Westminster-style elections are broken into hundreds of districts and winning districts is what wins you the election. No pollster has the resources to run a number of polls in every district, and so local movement in a district (which might can go undetected and you always have some shocks on Election Night.
By comparison, it’s relatively easy, and indeed is worth the while, to run a number of well-sampled polls on any state even remotely important, and there’s only 50 states, versus 650 districts in the UK or 338 in Canada. (In Canada we call them ridings, but I’m just using generic terms.) The smallest state is many times more populous than any UK of Canadian electoral district; the UK and Canada both have just one district each with 100,000 voters, at least as of 2015. The smallest STATE has something like 400,000 (Wyoming). To use an example, North Carolina, a swing state, has been polled by legitimate pollsters ten times in the last month and will be polled at least 12-15 more times. By November 8 we’ll have a rather clear picture of where North Carolina sits.
RickJay, nice job making the American electoral system seem reasonable. :eek:
It looks like 538’s Polls-Plus is rapidly converging with Polls-Only and Now-Cast.
System is entirely reasonable, we aren’t.