The polls are measuring different things, but by what measure do you think national polls use a larger sample size? I mean they sample more people, but they are also sampling out of a larger population. Tell me the MOE and I’ll tell you which is more accurate.
It’s increasingly looking like Romney absolutely must win Ohio. This is his only path.
It takes a hit if he thinks there’s a 60% chance of Obama winning even if he’s trailing. That’s a fairly bold prediction, basically saying the polls are wrong.
Obama leads in the state polls. Romney leads in the national popular vote poll. You must somehow resolve that contradiction. It does not hurt someone’s credibility if they offer a reasoned justification for resolving the contradiction in Obama’s favor.
Unless you think it is irrational to give the state polling more weight than the national polling, your argument has no merit.
I’m not saying that Silver’s credibility would be hurt if Romney won today, but let’s say the polls in the coming days show Romney leading in Ohio. Would that not put Romney at better than a 50% chance of victory?
Based on Nate’s methodology, it would depend on how many polls, and how large a lead. If all the polls for the next 5 days average to Romney with a +2 lead in OH, then it’s very likely that it will move above 50% (I believe) in Nate’s prediction.
Holding everything else equal, then yes, I think that’s right.
But so what? You are conjecturing that in such a circumstance the 538 model would not show Romney breaking 50%?
That’s what I’d expect as well. As long as SOME set of polls shows Obama with an advantage, Nate can credibly put Obama in the lead even though the aggregate of national polls says he’s trailing. But if Ohio shows Romney with a lead, there’s no longer any rationale for favoring Obama.
But it’s not just some polls in OH, it’s most of them. You can look at the right side of his website- a lot more polls show Obama ahead in OH, and the average is significantly in Obama’s favor.
But Nate Silver is not taking a position that is contrary to the polls, you are. The five most recent polls of Ohio range from a tie to Obama +5. His estimate of the odds are fundamentally based on the polls, and the other factors he allows to influence his prediction are losing weight in his model as each day goes by.
This post shows you have no idea what his methodology is. Here, he would only show Obama in the lead if the statistical model arrived at that conclusion. Seriously, he’s not the hack you’re making him out to be.
Exactly. He’s designed his model to be right. Full stop. And his model runs 10’s of thousands of simulations every day. He just posts the results of those simulations, and expounds on things in his blog.
If his model sucks and is terribly wrong, that’s very bad for Nate Silver. He just wants to be accurate.
This is kind of funny. What the Hell do you guys think could possibly happen between now and election day that would put Romney in the lead in Ohio?
Many of my friends have already voted.
Will Mourdock’s idiocy make a difference for Romney?: Obama pressures Romney on support for candidate who said rape from pregnancy is God’s will – The Mercury News
Anyone see a problem with that headline?
Ha! It took me a minute. That’s…unfortunate.
If Romney wins the national vote by 2 - 3 percentage points and carries Independents by 8+ points, he’s going to win the election. Point blank.
What I actually find odd about 538 is that Nate has Obama receiving 50.2% of the vote. How many polls out have Obama hitting the 50% mark? I can find one. Otoh, there are three different polls right now which show Romney at the 50% mark. It’d be interesting to note where the 50.2% comes from.
A modified Electoral College map for 2012: http://25.media.tumblr.com/tumblr_mcgkt8fluq1r55d2io1_500.jpg
The first link I followed currently has three in Obama’s favor. Fivethirtyeight also builds their own estimate of what the national numbers would be from the state polling, which is generally more accurate. In addition, not all national polls have the same level of accuracy, so if you are aiming to be as accurate as possible, you will have to weight them differently. Mr. Silver has done quite a bit of justification for how he weights his pollsters. You may disagree, but he is really pretty rigorous, and explains his reasons clearly.
P.S.: I linked to fivethirtyeight’s Senate methodology earlier, not the Presidential one. It still shows how he’s not basing his models on what he wants or feels, but what he thinks.