There is a difference between a qualitative approach and a quantitative approach. For a qualitative judgement on who would win between Tweedledum and TweedleDee, head-to-head polls even one year out can give us a very strong signal. For absolute accuracy on the final percentage vote totals they are much poorer. They get the sign right (who wins) but get the value wrong.

538 is obviously interested in absolute accuracy and getting the final margin wrong by 10% would be considered awful. But knowing a year in advance that if TweedleDee faces TweedleDum in the general election, then TweedleDee will win is very useful information – regardless the final margin of victory. I.e., the head-to-head polls are not complete BS *unless* you’re only interested in the absolute accuracy of the final vote total percentages.

zebra: “Desertphile 16, I am not a big fan of exit/entry polling either, with or without conspiratorial possibilities.”

I am not convinced that “conspiracy ideation” is involved in my head, but I am willing to concede I am wrong about the year 2000 elections and the balloting fraud I think occurred. Everyone I know has suggested I’m paranoid, about everything.

However, the USA Senate held a hearing on the issue, and witnesses testified under oath (after being granted immunity from prosecution) that they tampered with electronic balloting machines while employed at Diebold Electronics Inc. and at Electronic Voting Machines Inc.— there are even videos of some of the hearing sessions, available on YouTube. They explained how they tampered with the machines, how many machines they tampered with, who paid them to do it, where, and why.

Other witnesses testified, under immunity, that they located the most balloting machines in “Republican Party” precincts, and restricted machines in “Democrat Party” precincts. Hundreds of machines that were scheduled to be moved to “Democrat Party” precincts stayed in their warehouses.

In Northern New Mexico (and elsewhere) thousands of voters who looked Mexican and/or spoke Spanish had their ballots “challenged” by Republican Party “monitors” which means they were never even looked at to see who the voters cast their ballots for— over 23,000 ballots in New Mexico were “challenged” and never counted.

In Nevada and Iowa many black people said they were called on the telephone and told they would be arrested if they showed up to vote; a hell of a lot of black people in Nevada said they were called a day or two before the election and told to go to the wrong polling locations, far from where they lived.

Regardless of the fact I’m a raving schizophrenic paranoid, all this is public record.

When the exit polls showed that millions of people who said they voted for Al Gore didn’t show up in the official ballot counts, the Republican Party’s leader in the Senate said the Democrats had some how tampered with the exit pool data.

Look, I realize people have a hard time admitting they were wrong. But your characterization of head-to-head polls at this time of year was wrong.

As for the correlations of the head-to-head polls in Jennings and Wlezien; again, I can only surmise you didn’t read it. Go look at figure 7 on page 35 “Adjusted R-Squareds for Presidential and Parliamentary Elections Taken Separately.”

Do I need to spoon-feed you on how to read a graph? The r-squared is clearly 0.7 at 250 days prior to the election.

You got one thing right, this *is* like dealing with climate change deniers. Unable to read the science, unable to understand it when they do, and make laughable claims as a result. What you got wrong is that you’re the one on the denial side.

And I notice your complete silence on the head-to-heads one year out correctly getting the winner 12 out of the past 14 Presidential elections. Complete BS is what I call *your* claim.

Thanks for the info. I think the problem with Greg’s model is that it doesn’t account for what the campaigns may do in real time– it is a static picture.

So, as campaigns make tactical/strategic decisions about committing resources and particular approaches to particular demographics, we may see variations one way or the other for individual states.

Your last two posts are incomprehensible.

Sorry, but you just seem to be stringing together “statistics-like” words with no understanding.

Here’s what my 538 reference says (among other agreements with my take):

“Wait. Shrug off polls until just before primaries, or until after the conventions for the general election. General election polls are far more accurate on the eve of an election, and the candidate who leads after the major party conventions is likely to win.”

Even the paper you refer to, which is not US-specific and includes parliamentary and other systems, as well as caveats about changes in polling practice over time, shows a .7 correlation only beginning at 150 days out, which is approximately around the time of the conventions for the US.

Maybe you should take greater care in reading, and reading graphs?

In South Carolina, according to my model, if Clinton’s strategy holds, she will win 29 delegates, and Sanders will win 24 delegates. If the Sanders strategy pertains, they will tie, or possibly, Clinton will win one more delegate than Sanders

The first of the results are in.

The AP is reporting in South Carolina Clinton picked up 39 delegates, Sanders 14.

The model uses statewide demographic data, or, for South Carolina, about 35% black. But according to NBC exit polls, the voters in the primary were about 60% black. So, at least in states with large non-white populations (which includes several Super Tuesday states), the model over-estimates Sanders support.

Also, the model assumes a that Clinton would get about 70% of the black vote, while the NBC exit polling shows closer to 85%.

Which, if this holds for other southern states, shows Clinton in a much stronger position than the model suggests.

“Young voters: At this rate, all future candidates might as well give up on trying to build a campaign around the support of young people. Despite tens of thousands of people showing up to Sanders rallies, and young people favoring him overwhelmingly so far, they just aren’t showing up to vote like he needs them to. Fewer than 1 in 6 voters in South Carolina were under the age of the 30. And it follows a pattern of young people just not giving Sanders the turnout he needs. In fact, young voters were less of the electorate there than in any of the first three states.”
https://www.washingtonpost.com/news/the-fix/wp/2016/02/27/winners-and-losers-from-the-south-carolina-democratic-primary/

“The problem — which also happens to be the ongoing, perpetual problem for candidates who bank on the youth vote — is that older voters usually over-perform…

We’ve noted repeatedly that young people simply don’t turn out to vote as much as older voters…

Hillary Clinton won Nevada in large part because of her overwhelming support from black voters. She won Iowa because older voters turned out more than they did in 2008 and because younger voters turned out less. What’s more, there’s not much evidence that Sanders’s younger voters are more motivated to turn out for him than are Clinton’s older ones.

This is why campaigns that need younger voters in order to win often don’t. Younger voters — who move more often, work weirder hours and aren’t in the habit of voting — simply don’t vote as much.”
https://www.washingtonpost.com/news/the-fix/wp/2016/02/22/younger-votes-didnt-come-through-for-bernie-sanders-because-they-almost-never-do/

Another factor that needs to be taken into account is that Sanders may not do well in caucuses that are limited to registered Democrats:

“In an interview last week, Ben Tulchin, a pollster with the Sanders campaign, did not seem optimistic about his candidate’s chances in closed caucuses. He said that there was opportunity to organize, but that such caucuses were difficult for independents, who would have to change their party registration to participate, and young voters — two sources of strength for Mr. Sanders. “They’re not tailor-made for us,” he said. “It’s still structurally challenging for us to do well.” He was more optimistic about open caucuses, in states like Nebraska and Alaska. Whether Mr. Tulchin’s thoughts reflect the campaign’s views or represent an effort to lower expectations is hard to say.
What’s clear is that modest Sanders wins in Massachusetts, Minnesota, Colorado and Oklahoma would not indicate that he had much of a chance to overtake Mrs. Clinton’s delegate lead.”
http://www.nytimes.com/2016/02/28/upshot/hillary-clintons-winning-numbers-in-south-carolina-suggest-sweep-in-south.html?_r=0