Political Roundtable Cosmic String of Cataclysm - Part V

[fishercob]

I think Silver is somewhat under attack because of his blog's affiliation with the New York Times. Lots of Republicans hate the New York Times.

Back in '08 -- when I first happened upon him -- he was just 538.blogspot.com or something along such lines. I was simply drawn to him because I had never seen consistent projections based on polls of the individual states. And since we still have this silly Electoral College, those polls on the evening news that said "Obama and McCain locked in dead heat" weren't very meaningful to me.

He was doing something new and different -- or something new and different to me. Maybe others were doing similar analysis and I just hadn't come upon them because I don't live and breathe politics. I'm no statistican. But the more I learned of his track record the more confidence I developed in his projections. That seems reasonable, no? If he was wrong a lot of the time, he'd be less interesting. So Tuesday is a pretty big night for old Nate.

But if he's wrong it's definitely because he's gay.

[Nivek]

So, again...Silver's analysis can't be so easily dismissed with he's biased or gay. He's been right an awful lot in the past two election cycles. That doesn't mean he should be taken as gospel. But one of these commentators trashing him might want to come up with an analysis of Silver's model (which is published in full detail) that explains the error of his ways AND explains how he's been right so consistently despite these problems with his model.

Maybe I don't know enough lefties, as I don't see him being embraced as gospel because he supports Obama as much as for the openness and accessibility of his process when compared to most polling and prognostication processes I see. That's why I started reading him, because he gave so much information about the process. And, as Nivek noted, his track record is brief but broad. I assume any partisan would struggle with bias, but even if Obama wins, if it turns out much closer than Silver predicted, I would expect him to view that as a shortcoming to be addressed. His candidate winning and his process being accurate as two separate things.

Worth mentioning the difference between Silver's projected outcomes and the probability his system has for a specific outcome. In the presidential race, his system gives 79% odds for Obama winning. His popular vote projection is very close, though -- 50.5% to 48.6%. This is akin to in-game win probabilities posted by Brian Burke at Advanced NFL Stats. Burke's program predicts the probability of a given outcome based on the game state -- score, time on the clock, field position, etc. But there's still game to be played. A couple times this season, the Redskins have had in-game probabilities of winning that exceeded 90% and still managed to lose.

Of course, a 90% probability of something happening means there's a 10% probability that it won't happen.

[nate33]

The efforts from the right to trash Nate Silver are about as funny as the quasi-religious belief in his prognostications from the left. They guy has a track record of just one presidential election. He is 1 for 1. Call it 2 for 2 if you want to consider the 2010 congressional elections to be the same thing as a presidential election.

To expand on Jim's point, if you count the presidential race and the congressional races as 1, then yeah, he's 2-2. Except, the presidential race is really 50 state contests. And you'd have to ignore how accurate he was in predicting primary outcomes. In terms of the presidential race, his predictions were correct everywhere except Indiana. As Jim pointed out, he also predicted the outcome for every senate race that year.

In 2010, Silver's model predicted the correct winner in 34 of the 37 senate races. Silver predicted a net gain of 7 Senate seats for the GOP that year -- they got 6.

Also in 2010, he predicted the correct winner in 36 out of 37 gubernatorial races.

So...the sample size is a bit larger than 2.

Yale University economist Ray Fair designed a statistical model that has successfully predicted the winner in 7 of the last 8 elections. (The only one he missed was 1992 when Perot through a monkey wrench into the process.) Fair predicts Romney will win a close election.

And he might be right again. It's definitely going to be a close election.

So, again...Silver's analysis can't be so easily dismissed. He's been right an awful lot in the past two election cycles. That doesn't mean he should be taken as gospel. But one of these commentators trashing him might want to come up with an analysis of Silver's model (which is published in full detail) that explains the error of his ways AND explains how he's been right so consistently despite these problems with his model.

I'm not dismissing Silver's analysis. I'm just trying to put it into context. There are lots of people out there with a longer track record of accuracy. I don't understand how Silver's predictions suddenly became gospel. I readily cede that his short track record has been pretty accurate. But that's all it is. A short track record.

And let's not overstate his track record. Getting 49 out of 50 states right sound great, but let's face it, probably 45 of so of them were no-brainers where one state had a clear lead in the polls. It was really more like being 4 out of 5 on the close states. Likewise, predicting 34 out of 37 Senatorial campaigns in 2010 may no be all that impressive either. How many of those campaigns were within 2% or so in the polls? Again, he was probably more like 7 out of 10 on close races, and another 27 for 27 on the no-brainers.

[Nivek]

Umm, maybe you just missed the part where I wrote, "That doesn't mean he should be taken as gospel."

[montestewart]

Agreed that his track record is only significant in the close races. Much like picking the winner vs. picking against the spread, most of them are pretty easy. I just like the viewable process a lot, and came upon him in much the same way as fishercob, before he was affiliated with a big media outlet (although his leanings were still pretty clear). I thought he had arrived at election analysis via sports analysis (I can't recall now exactly why, other than the person who turned me to the site was a SABR-type).

[Nivek]

I came across him in a similar way. I'd been reading his baseball stuff off and on, along with some of the other sabrmetricians out there. The games are really different, but you never know when they'll come up with an idea applicable to hoops. Then I started reading his politics stuff, and...the rest.

[nate33]

Umm, maybe you just missed the part where I wrote, "That doesn't mean he should be taken as gospel."

Sorry, Kevin. I wasn't referring to you when I talked about people taking his word as gospel. Mostly, I'm referring to about 100 of my liberal Facebook friends who use Nate Silver as a virtual guarantee that Obama will win the election.

[Nivek]

Yeah, there are PLENTY of folks who need a lesson on what "probabilities" mean. 79% probability does not mean it's going to happen. It means it's PROBABLY going to happen.

[nate33]

The biggest gripe conservatives have with the polling is that it seems that most polls are taking samplings that are overweighted with Democrats. Take the recent CBS/NYTimes/Quinnipiac of Ohio, for example. The poll has Obama at 50% and Romney at 45%. Yet, the Independents side with Romney by a 6% margin. How is this possible? It's possible because the poll assumes an electorate with a +8% Party affiliation of Democrats (D37, R29, I30). What is the likelihood of this happening? In 2008, with an all-time high Democrat turnout behind a once in a generation historical presidential campaign in conjunction with a wave of state Republican party scandals, the electorate was just +8% Democrat. Do they really think they're going to get the same pro-Democrat turnout this year? After all, in 2004, Ohio had a +4% Republican affiliation.

[Zonkerbl]

Yeah, but you can use regression analysis to adjust those polls. I think that's the point.

[Severn Hoos]

The biggest gripe conservatives have with the polling is that it seems that most polls are taking samplings that are overweighted with Democrats. Take the recent CBS/NYTimes/Quinnipiac of Ohio, for example. The poll has Obama at 50% and Romney at 45%. Yet, the Independents side with Romney by a 6% margin. How is this possible? It's possible because the poll assumes an electorate with a +8% Party affiliation of Democrats (D37, R29, I30). What is the likelihood of this happening? In 2008, with an all-time high Democrat turnout behind a once in a generation historical presidential campaign in conjunction with a wave of state Republican party scandals, the electorate was just +8% Democrat. Do they really think they're going to get the same pro-Democrat turnout this year? After all, in 2004, Ohio had a +4% Republican affiliation.

nate, I agree that the electorate distributions look really skewed - although that's what the Left complains about Rasmussen as well. But let's suppose that the real split is more like 48-47 than 50-45. I get that the pollster might "want" to show an Obama lead to try to drive up enthusiasm and start the narrative of inevitability. But I wonder how that will play out in real life. Will Obama supporters be enthusiastic because their guy is up, or will they be complacent, thinking it's in the bag because Nate Silver says so? That sword would seem to have two edges....

And FWIW, anyone willing to cop to cursing the Electoral College after 2000? If there's another split (Romney wins Popular vote, Obama wins EC), will you still think the Electoral College is an insidious anachronism that must be replaced at all cost?

[montestewart]

nate, I agree that the electorate distributions look really skewed - although that's what the Left complains about Rasmussen as well. But let's suppose that the real split is more like 48-47 than 50-45. I get that the pollster might "want" to show an Obama lead to try to drive up enthusiasm and start the narrative of inevitability. But I wonder how that will play out in real life. Will Obama supporters be enthusiastic because their guy is up, or will they be complacent, thinking it's in the bag because Nate Silver says so? That sword would seem to have two edges....

And FWIW, anyone willing to cop to cursing the Electoral College after 2000? If there's another split (Romney wins Popular vote, Obama wins EC), will you still think the Electoral College is an insidious anachronism that must be replaced at all cost?

After the 2000 election, there didn't seem to be wide support within the Democratic Party leadership to eliminate the electoral college system, and if the reverse happens this year, I wonder whether the Republican Party would follow suit? I think both parties like being able to concentrate their "tactics" in a few key places, but eliminating that ability might result in more honest elections. Or maybe it would just result in widespread disinformation in vastly more expensive campaigns.

I remember the days when Virginia was clearly Republican and DC and Maryland (after Agnew) were clearly Democrat, and nobody bothered spending much money around here for the presidential election. But now there's a swing state in the hood. Woo hoo! Somebody around here matters!

[Nivek]

Saw a headline a few days ago about a Democrat proposal to award 29 electoral votes to the candidate that won the popular vote. Didn't read the article, so I couldn't say how they came up with 29.

[fishercob]

And FWIW, anyone willing to cop to cursing the Electoral College after 2000? If there's another split (Romney wins Popular vote, Obama wins EC), will you still think the Electoral College is an insidious anachronism that must be replaced at all cost?

I didn't say it needed to be replaced at all cost. I have said that it seems stupid and outdated. And when I said that, I wasn't thinking about Bush/Gore, which is ancient history for my feeble, young (36) mind.

But I have yet to see any sort of rebuttal as to why we in fact need the Electoral College. Who is it serving? If Romney wins the popular vote, he should be President. Anything other than that says that people who live in one place are more important than others, which strikes me as stupid, wrong, and immoral.

[Nivek]

My 8th grader surprised me the other day when he said he supported having the Electoral College. He said without it there'd be no reason for presidential candidates to pay any attention to states with smaller populations. The campaigns would be focused almost exclusively on big media centers. I'm paraphrasing, of course.

[montestewart]

My 8th grader surprised me the other day when he said he supported having the Electoral College. He said without it there'd be no reason for presidential candidates to pay any attention to states with smaller populations. The campaigns would be focused almost exclusively on big media centers. I'm paraphrasing, of course.

Similar to part of the the logic behind senate representation, and not a bad argument, but under the current system, heavily populated states that clearly lean one way or the other are also frequently ignored. Maybe swing states should be the center of presidential campaigns, since a closely divided electorate might demand a better discussion than a much more partisan electorate.

[popper]

And FWIW, anyone willing to cop to cursing the Electoral College after 2000? If there's another split (Romney wins Popular vote, Obama wins EC), will you still think the Electoral College is an insidious anachronism that must be replaced at all cost?

I didn't say it needed to be replaced at all cost. I have said that it seems stupid and outdated. And when I said that, I wasn't thinking about Bush/Gore, which is ancient history for my feeble, young (36) mind.

But I have yet to see any sort of rebuttal as to why we in fact need the Electoral College. Who is it serving? If Romney wins the popular vote, he should be President. Anything other than that says that people who live in one place are more important than others, which strikes me as stupid, wrong, and immoral.

It is serving the populations in the smaller states. Without the Electoral College our Republic, as we know it, would be drastically different. If I'm not mistaken, the Electoral College was the price extracted by the smaller states to join the union (I'm no historian but that's the way I understand it).

[Spence]

Yeah, there are PLENTY of folks who need a lesson on what "probabilities" mean. 79% probability does not mean it's going to happen. It means it's PROBABLY going to happen.

If I told you there was a 21% chance of you getting hit by a bus tomorrow, you'd probably think that's an alarmingly high number and do your best to avoid buses the next day. Assuming you believed my original assertion, of course.

[Spence]

This arguing about polls is so silly. Believe whatever you like, the truth will be revealed on Nov 6. I read all the polls and consider some more reliable than others, but I never whine about the results of a poll. What's the point? Unless...

Anyone planning to complain that his or her favored candidate lost the election due to "biased" polls should be locked in a closet with that thing on Donald Trump's head.

[Spence]

The electoral college doesn't ensure small states get lots of attention from national candidates, it ensures marginal constituencies receive a disproportionate amount of attention from national candidates. [Notice a lot of Obama or Romney trips to Rhode Island, Idaho, Montana, etc? Of course not. New Hampshire, which is closely divided, gets lots of attention from politicos. Vermont, right next door, which is not closely divided, receives no attention from politicos.] Whether you think it is good for marginal constituencies to receive a disproportionate amount of attention from national campaigns is a matter of opinion, but we should be clear about the actual impact of the electoral college on national campaigning.

I suspect federal land use rights and global warming would receive a lot more attention from national campaigns if candidates campaigned in Wyoming and California. Again, whether you think that is a good thing or not is a matter of opinion.