Sure, if Nate gets 30 or so states wrong and a bunch of Senate and House seats wrong he'll take a huge hit in credibility.

And what happened to me? People made fun of me.

So the argument is that as long as Nate isn't calling it 100.0% for Obama, he can't be wrong.[/quote

Ray. Can we start with the assumption that you understand the basics of predictive modelling?

If it was 95% and Romney won, would people here concede that Silver blew it? What are we talking about here? Silver can't be graded? Then what good is he?

What value 538 provides, for the most part, comes from his writing and analysis. He's really good at explaining some complex methodological and statistical questions. I assume he'll retain those capacities even in the event of a polling disaster.

Nate does another step of aggregation, which is reasonably useful for readers, but doesn't actually produce anything that wasn't in the Pollster numbers to begin with.

If it was 95% and Romney won, would people here concede that Silver blew it?

One Of Us.

I believe Rasmussen does provide the full breakdown for its subscribers, but I'm not sure.

Have you actually seen the people that are employed? Dick Morris, David Gergen, John Zogby, and that's just the very tip of a gigantic iceberg. Nate's competition for prognostication is a whole lot of people that couldn't predict a coin flip if they got to make two guesses.

So the argument is that as long as Nate isn't calling it 100.0% for Obama, he can't be wrong.

Isn't most of the difference between RCP and Nate on national polls the difference between:
1) excluding Rand entirely and including it with some downgrade?
2) then averaging all equally vs rebalancing them based on past effects?

I think my bosses cringe anytime I'm doing a radio spot or making a blog comment to the effect that the goal of a projection system is to be slightly less horribly inaccurate than everyone else.

Mitt's off to a good start - he doubled down on wanting to dismantle FEMA this morning.

Not this morning. Romney's FEMA comments were during a primary debate.

Polls are controversial today because the presidential race is tight. If either candidate had a 10-point lead, few would care how different polls decide who’s a “likely” voter or how various population segments are weighted. These details usually don’t matter much. But when the margin is a few points, they can determine which candidate is the front-runner and has “momentum.”

More menacing is whether cellphones and shrinking popular participation subvert polls’ accuracy, because samples are less representative. The answer seems to be: “not yet.” Pew did a study comparing responses to basic questions (are you: U.S. citizen, homeowner, married?) from its surveys to larger government surveys. On most questions, responses were identical or close: 37 percent of respondents in both had children in the house; 75 percent in both were registered to vote. There were a few troubling discrepancies — but no fundamental break.

Less reassuring is telephone polling’s steep and rising costs, which could cause cash-strapped media organizations to balk. Contacting cellphones is expensive, because numbers must be dialed by hand. By contrast, computers can automatically dial landline numbers, making it easier to reach live people. (Congress prohibited this for cellphones to protect people from paying for unsolicited incoming calls.) A typical survey costs Pew from $60,000 to $100,000, says Keeter. That would cover renting tens of thousands of landline and cellphone numbers to produce 1,500 interviews of about 20 minutes each.

The solution seems obvious: switch to the Internet. But technically, that’s hard. Internet users may not be a representative sample of the U.S. population. Does the person behind that e-mail live in the United States? Permanent panels of respondents may act differently from randomly contacted people. Experiments are under way. Meanwhile, pollsters are stretched between a past that’s growing untenable and a future that doesn’t yet exist.

Well, if Obama doesn't make any major screwup and if a bipartisan group of governors and local officials wind up praising FEMA's reaction to Sandy, that comment of Romney's might wind up demonstrating the point that it's not the government per se that's at fault when they don't do their job, it's the fault of the people in charge of the government. Which is exactly the point that sane people have been making for the past 80 years. The response of "Brownie" & Co. to Katrina is hardly an argument for dismantling FEMA, but it might have been a good argument for dismantling Brownie.

So, to recap, PPP polls are worthless and biased because they provide full demographic crosstabs in their poll PDFs instead of in their press releases, which makes it too hard for you to find. And Rasmussen polls are totally fine and trustworthy because you think they might provide some of this information to subscribers.

Since RCP is emphatically not just an averager but is instead a subjective decision about what to include and what not to include, a subjective decision in keeping with the founders' explicit goals of increasing conservative coverage, then will its supporters here renounce it and refuse to visit it if Romney in fact does not win the popular vote?

If someone says that, and then asks if I'll stop visiting Nate, then I can deal with that question. But someone who won't deal with RCP's own subjectivity and then make demands upon other people is just concern trolling.

The 538 model uses the state-by-state polling as evidence of the national race, and the state-by-state polling is consistent with a national race where Obama leads by about three points.

Nonsense. Rasmussen, almost alone, weights poll results based on its survey of party ID, which it releases on a daily or near-daily basis all year, every year.

PPP didn't disclose the sample in the text of the blog post, press release, or tweets it posted in hyping the alleged Obama lead. Instead, people had to dig into the PDF to find this key information. The topline poll results are essentially meaningless without the context provided by the sample breakdown.

So what were the demographic breakdowns of Rasmussen's latest Ohio poll? As a reminder, you said PPP's latest Ohio poll was biased and unreliable because they provided this information in a PDF instead of in their press release:

PPP didn't disclose the sample in the text of the blog post, press release, or tweets it posted in hyping the alleged Obama lead. Instead, people had to dig into the PDF to find this key information. The topline poll results are essentially meaningless without the context provided by the sample breakdown.

You have frequently referred to Rasmussen's Ohio polling instead.

I have? Until today, Rasmussen had Obama in the lead in Ohio for months, so I kind of doubt I've been pimping Rasmussen's Ohio results here.

Well, the RCP average then, which includes Rasmussen.

Seems strange to dismiss a poll as "meaningless" because of its party ID results, then repeatedly tout other polls/averages of which you don't even know the party ID results.

There is some suggestion that the BLS Jobs Report may be delayed, which may raise eyebrows since the surveys are complete and one would think enough people could be designated as "essential" to wrap it up, especially given that the immediate DC area seems to be doing OK.

It's not like UE is going over 9

Thanks, but I didn't bother checking either, until the fifth or sixth time the stat was cited. Sloth is my go-to deadly sin.

Ohio General Election Results + Results of presidential elections.

Obviously there are multiple ways to turn and present data. But you were talking about how GOP candidates have traditionally overperformed in Ohio relative to the national popular vote. Addressing that premise, what's more germane than directly comparing how the candidates performed in Ohio to how they performed in the national popular vote?

This isn't how I'd previously seen it presented around the internet and it's less compelling than the percentage comparison, but it still suggests the GOP tends to outperform in Ohio.

... I.e., the straight comparison of percentages is more compelling than the margin-of-victory comparison in the chart I linked.

since Ohio historically goes +2.0 for the Republican relative to the national vote.

Over every election for which we have data, Ohio has never gone less for the GOP candidate than the GOP candidate secured in the national vote, and the average is +2.0 for the GOP.

when we also know that Ohio goes, on average, +2.0 for the GOP candidate relative to the national vote, that's precisely the type of anomaly that Nate's model is supposed to explain rather than expand

Now that Gonfalon Bubble has debunked that myth, and you yourself have admitted that comparing the Ohio numbers to the national vote are the "more compelling" way to analyze the numbers, do you think that Nate's percentage (a) in Ohio and (b) for the general are accurate?

For the record, in terms of your repeated reliance on the 2-point myth, here is but a sampling:

It's not a myth that the GOP overperforms in Ohio relative to the national result. The GOP overperforms by less in a straight percentage-to-percentage comparison, which is how the +2.0 thing has been conflated around the internet recently,

Nate has Obama as a 3:1 favorite for reelection despite a horrible economy, high unemployment, sub-50 approval ratings, horrendous right-track/wrong-track numbers, Romney leading in the RCP, and the fact that incumbent presidents generally either outperform their first election or lose — all, apparently, on the basis of Obama leading in Ohio by 2 points. The idea that a 2-point Obama lead in Ohio polling not only trumps all of the preceding negatives but makes Obama a 3:1 favorite seems both incredibly optimistic and entirely unexplained.

Is that not how you were presenting it in 4901, 4917, and 4966 (and other places)?

So now it is your position that because the GOP outperforms the national *margin,* in Ohio, Nate's model must be wrong? Maybe that is right, but that wasn't your criticism before. Your criticism before was that the GOP outperforms the national vote by 2%. That assertion is incorrect.

Do you think Nate's model doesn't account for all the factors listed above? Do you have any evidence that it does not? Aren't all of those things baked into the polling - Nate's primary data points? If you'd argue that they aren't, how so? As MCoA points out, Nate has written extensively about his methodology, and is hardly running a black box. There are numerous posts in this thread that touch on his methodology. Yet you keep asserting variants of your statement above - that his current projection is "entirely unexplained." This assertion has no basis in fact. His prediction is in line with the other predictions I've seen - a close popular vote with an Obama victory in the EC.

If you claim to be regressing to the mean, but your model projects something that has never happened before, then you aren't regressing to the mean.

He's a favorite because the opposition has proved itself so unpalatable that despite the significant advantages you cite, they are losing anyway.

Nate shows Ohio's polling average as Obama +2.8, an adjusted polling average of Obama +2.4, "State fundamentals" as Romney +1.1, and an overall both now-cast and final projection of Obama +2.2. Moving from +2.8 to +2.2 on the basis of Ohio having pro-Romney fundamentals would seem to be a fairly textbook example of regression to the mean.

Nate has Obama as a 3:1 favorite for reelection despite a horrible economy, high unemployment, sub-50 approval ratings, horrendous right-track/wrong-track numbers, Romney leading in the RCP, and the fact that incumbent presidents generally either outperform their first election or lose — all, apparently, on the basis of Obama leading in Ohio by 2 points.

Most of those (all but "Romney leading in the RCP") are things that Nate's model explicitly does not look at. If poll-aggregating can lead to accurate predictions, it has to be by looking only at the polls. If you want to construct a prediction system that relies on economic trends, feel free. But I don't think it's fair to criticize Nate Silver for not doing that. Things like "incumbent presidents generally either outperform their first election or lose" are outside the scope of the project.

The most significant change to our presidential forecast model this year is that it contains an economic index, which is used to guide forecasts along with the polls.

In fact, as you may have seen since we began our short daily summaries of the model’s output, new economic data often has just as much influence over the forecast as the latest poll from Ohio or Florida. ...

If you want to construct a prediction system that relies on economic trends, feel free.

Such as the already discussed (read: mocked) the University of Colorado forecaster that does that and against all odds has Romney winning Minnesota and New Mexico.

I don't see any mention of it on his methodology page, but Nate has repeatedly mentioned that economic data is factored into his model.

Well, the first half, anyway; I still think "incumbent presidents generally either outperform their first election or lose" is too vague to plug into a model.

In any case, Nate Silver doesn't need to offer an explanation, and doesn't pretend to have a compelling explanation. He isn't a pundit pushing a theory about who will win. He's an analyst presenting a model that aggregates state-by-state poll numbers.

Or Nate's.

Unlike the University of Colorado's model, Nate Silver's model is neither built on economic trends nor predicts Romney to win NM or MN. But yeah, other than that they are exactly the same.

Your mocking reply to the quote in #5570 suggested you were unaware that Nate's model indeed partly "relies on economic trends," and that you had perhaps missed the link in #5569. Either way, not a big deal.

I apologize for assuming that you understand there is a difference between "includes" and "based on."

Do you really think it will end then? There will be the discussions of why the polls/538/other aggregators did/didn't work.

Next Tues Spring. can't come soon enough. I don't care who wins at this point, all this crap about pollspolitics needs to end.

Next Tues. can't come soon enough. I don't care who wins at this point, all this crap about polls needs to end.

Do you really think it will end then? There will be the discussions of why the polls/538/other aggregators did/didn't work.

Just to piss off Sam, earlier this afternoon ABC News moved Pennsylvania and Montana ...

That article is insane.

Just to piss off Sam, earlier this afternoon ABC News moved Pennsylvania and Montana from "safe Obama" to "lean Obama."