Why Obama is Toast

[silverscreenselect]

This is a long article, but this guy is either a genius or someone who won't want to get out of bed tomorrow:

http://seanmalstrom.wordpress.com/2008/11/03/toast/

I think there is a lot to what he says, especially about how the candidates are contesting the various states.

[mrkelley23]

I would be much more inclined to give him some credence if he could put a coherent English sentence together. I made it through about five paragraphs before I gave up. He may be right about some things, but you won't be able to prove it by me.

[mrkelley23]

This is a long article, but this guy is either a genius or someone who won't want to get out of bed tomorrow:

http://seanmalstrom.wordpress.com/2008/11/03/toast/

I think there is a lot to what he says, especially about how the candidates are contesting the various states.

Looking like Mr. Malstrom may want to take some Ambien tonight.

[ToLiveIsToFly]

I realize this is all hindsight now, so who knows? But I like to think I would have said most of the same stuff before the polls closed yesterday as I'm saying now. I haven't been on the internet hardly at all since Friday, since I was asked to fly to Colorado on a few hours' notice and I've spent the last 4 days organizing precincts to get out the vote, which means coordinating a group of volunteers to walk around neighborhoods and doing the same thing myself. I think I walked >40 miles between Saturday and yesterday, and my feet hurt.

I don't think I can really fault the writer for most of this stuff. There's a long tradition of talking yourself into thinking your guy can win. Some people did this for Kerry in 2004. A LOT of people did this for Bush in 2000.

But I have to take issue with his singling out Nate Silver at fivethirtyeight.com. He's saying the numbers at the site are biased because Nate has a partisan lean, and he implies that Nate was somehow trying to hide it. Nothing could be farther from the truth - he discloses again and again that he's partisan. All of his commentaries are partisan and there's no trying to hide it. And yet his focus on the numbers was an attempt to keep them accurate, not optimistic, from day one. Which is hardly surprising since he comes from the world of baseball statistics analysis, where the tradition has long been to keep the numbers reflecting reality despite having a clear rooting interest in one team.

[silverscreenselect]

But I have to take issue with his singling out Nate Silver at fivethirtyeight.com. He's saying the numbers at the site are biased because Nate has a partisan lean, and he implies that Nate was somehow trying to hide it. Nothing could be farther from the truth - he discloses again and again that he's partisan. All of his commentaries are partisan and there's no trying to hide it. And yet his focus on the numbers was an attempt to keep them accurate, not optimistic, from day one. Which is hardly surprising since he comes from the world of baseball statistics analysis, where the tradition has long been to keep the numbers reflecting reality despite having a clear rooting interest in one team.

If you look at the raw vote totals, McCain came a lot closer than virtually every final poll, which was in the 6-12 percent range. That didn't translate too well at the state level, but it does show that the polls were off by a significant amount which could have made a difference in predicting the winner if the overall race had been closer. From a statistical standpoint, picking someone to win by 20 and have him win by 10 is just as big an error as picking him to win by 5 when he loses by 5.

So what a lot of people were saying was correct. Due to a number of factors, McCain's totals were higher than what the polls were saying (I think a lot of that was due to more McCain voters refusing to be polled or concealing their intentions). It's just that the final margin was too much for this error to change the final electoral college results.

I do believe that the pollsters want to be right, but I also believe they are human and it takes a lot of guts to stick by your guns when you are saying one thing and 20 other people are saying something else. It's easier to question your own methodology and adjust to bring it in line with everyone else.

Further, every election is different so that what works for pollsters now may not be accurate in four more years when the demographics change even more.

[SportsFan68]

I realize this is all hindsight now, so who knows? But I like to think I would have said most of the same stuff before the polls closed yesterday as I'm saying now. I haven't been on the internet hardly at all since Friday, since I was asked to fly to Colorado on a few hours' notice and I've spent the last 4 days organizing precincts to get out the vote, which means coordinating a group of volunteers to walk around neighborhoods and doing the same thing myself. I think I walked >40 miles between Saturday and yesterday, and my feet hurt.

Thanks, Fly!

[ToLiveIsToFly]

But I have to take issue with his singling out Nate Silver at fivethirtyeight.com. He's saying the numbers at the site are biased because Nate has a partisan lean, and he implies that Nate was somehow trying to hide it. Nothing could be farther from the truth - he discloses again and again that he's partisan. All of his commentaries are partisan and there's no trying to hide it. And yet his focus on the numbers was an attempt to keep them accurate, not optimistic, from day one. Which is hardly surprising since he comes from the world of baseball statistics analysis, where the tradition has long been to keep the numbers reflecting reality despite having a clear rooting interest in one team.

If you look at the raw vote totals, McCain came a lot closer than virtually every final poll, which was in the 6-12 percent range. That didn't translate too well at the state level, but it does show that the polls were off by a significant amount which could have made a difference in predicting the winner if the overall race had been closer. From a statistical standpoint, picking someone to win by 20 and have him win by 10 is just as big an error as picking him to win by 5 when he loses by 5.

So what a lot of people were saying was correct. Due to a number of factors, McCain's totals were higher than what the polls were saying (I think a lot of that was due to more McCain voters refusing to be polled or concealing their intentions). It's just that the final margin was too much for this error to change the final electoral college results.

I do believe that the pollsters want to be right, but I also believe they are human and it takes a lot of guts to stick by your guns when you are saying one thing and 20 other people are saying something else. It's easier to question your own methodology and adjust to bring it in line with everyone else.

Further, every election is different so that what works for pollsters now may not be accurate in four more years when the demographics change even more.

You're giving me a lot of reasons why the pollsters were wrong. But what I said was that I didn't understand why the article singled out Nate Silver and fivethirtyeight.com. He's not a pollster - he uses the data from the pollsters to make predictions.

I just got home from the airport. So I haven't had a chance to look at a lot of the state-by-state numbers. I just looked on cnn.com and cspan.com and didn't see full national totals with all candidates concluded. So for now I'm going to take fivethirtyeight at its word that the final percentages for the national vote total were 52.4% Obama and 46.3% McCain. Please correct me if that's wrong.

Fivethirtyeight's final prediction: 52.3% Obama, 46.2% McCain. Pretty inaccurate. And those state predictions he made? Out of the 48 states that have been called so far, he only got 47 right. Sure, he also predicted Missouri for McCain and North Carolina for Obama, and that looks like the way it's going to turn out, but there's no denying he got Indiana wrong - he predicted it would go to McCain. His numbers were obviously in the tank for Obama.

[Weyoun]

But I have to take issue with his singling out Nate Silver at fivethirtyeight.com. He's saying the numbers at the site are biased because Nate has a partisan lean, and he implies that Nate was somehow trying to hide it. Nothing could be farther from the truth - he discloses again and again that he's partisan. All of his commentaries are partisan and there's no trying to hide it. And yet his focus on the numbers was an attempt to keep them accurate, not optimistic, from day one. Which is hardly surprising since he comes from the world of baseball statistics analysis, where the tradition has long been to keep the numbers reflecting reality despite having a clear rooting interest in one team.

If you look at the raw vote totals, McCain came a lot closer than virtually every final poll, which was in the 6-12 percent range. That didn't translate too well at the state level, but it does show that the polls were off by a significant amount which could have made a difference in predicting the winner if the overall race had been closer. From a statistical standpoint, picking someone to win by 20 and have him win by 10 is just as big an error as picking him to win by 5 when he loses by 5.

So what a lot of people were saying was correct. Due to a number of factors, McCain's totals were higher than what the polls were saying (I think a lot of that was due to more McCain voters refusing to be polled or concealing their intentions). It's just that the final margin was too much for this error to change the final electoral college results.

I do believe that the pollsters want to be right, but I also believe they are human and it takes a lot of guts to stick by your guns when you are saying one thing and 20 other people are saying something else. It's easier to question your own methodology and adjust to bring it in line with everyone else.

Further, every election is different so that what works for pollsters now may not be accurate in four more years when the demographics change even more.

You're giving me a lot of reasons why the pollsters were wrong. But what I said was that I didn't understand why the article singled out Nate Silver and fivethirtyeight.com. He's not a pollster - he uses the data from the pollsters to make predictions.

I just got home from the airport. So I haven't had a chance to look at a lot of the state-by-state numbers. I just looked on cnn.com and cspan.com and didn't see full national totals with all candidates concluded. So for now I'm going to take fivethirtyeight at its word that the final percentages for the national vote total were 52.4% Obama and 46.3% McCain. Please correct me if that's wrong.

Fivethirtyeight's final prediction: 52.3% Obama, 46.2% McCain. Pretty inaccurate. And those state predictions he made? Out of the 48 states that have been called so far, he only got 47 right. Sure, he also predicted Missouri for McCain and North Carolina for Obama, and that looks like the way it's going to turn out, but there's no denying he got Indiana wrong - he predicted it would go to McCain. His numbers were obviously in the tank for Obama.

Silver was obviously pretty close. But I think the original article was correct in stating that, if somehow McCain took PA, Obama would still almost certainly win. I'd argue if PA was that much of a surprise, at a minimum we should cast aside any attempt at predicting the results.

[silverscreenselect]

Silver was obviously pretty close. But I think the original article was correct in stating that, if somehow McCain took PA, Obama would still almost certainly win. I'd argue if PA was that much of a surprise, at a minimum we should cast aside any attempt at predicting the results.

Each election has a different dynamic and the pollsters have to change their models to predict that dynamic. There were two main sets of polling trends this year. About half the pollsters wound up with Obama in the 6-8 range, which is fairly accurate. The others had him in the 10-12 range, which means they either overweighted some factors in his favor or underweighted McCain's. Again, in this election, it's no problem but in a close election, you get a Dewey Beats Truman scenario.

One of the problems that both sides have in dealing with anecdotal evidence is that you usually hear only a part of it. A number of Obama supporters claim that people came up to them and said "I'm really supporting Obama but I don't want my friends to know" while McCain supporters said the same thing ("I'm really supporting McCain"). Both are probably telling the truth, because I sure am not going up to an Obama supporter on the street and telling him that I'm a lifelong Democrat voting for McCain. You see favorable trends and ignore unfavorable ones, which is how a lot of people were convinced Kerry would win and others equally convinced McCain would win.

Pollsters are not infallible and when they are wrong, as in NH this year, the results are spectacularly bad. But the best ones are mathematicians, not politicians, and they view mistakes only as learning opportunities to refine their models.