I've linked to electoral-vote.com often, since it is a site which breaks down polls state by state and collects them into an overall picture of how the important numbers may shake out in the election. A similar site, with better analysis, is Five Thirty Eight, a New York Times blog run by Nate Silver, which uses some complicated math formulas to forecast the probabilities of each candidate winning. (538 is the total number of electoral votes available from all 50 states.)
[...]
This season, Five Thirty Eight has generally skewed slightly better for Republicans than Electoral Vote has. Today, for instance, EV predicts a victory of 328-221, with 19 approximate ties. Five Thirty Eight, however, predicts 308-229. The difference seems to exist because the latter also factors in the various probabilities that any state could flip the other way, while the former only considers who's ahead right now.
However, another interesting number that is provided by Five Thirty Eight is the overall probability that either candidate will win everything. That number has drifted up and down, and right now it shows that Obama has a 76% chance of winning.
Now, assuming you are hoping for an Obama victory, that's pretty good, but not overwhelming. Personally, I'd feel much better about a 90% chance, or a 95% chance. If I thought there was really a 24% chance of Romney winning, I would never have made the blunt prediction that "Obama is going to beat Romney, and it's not going to be very close."
While I respect Nate Silver's analysis, I've never been entirely convinced that his predictive model factors in all the relevant information that goes beyond raw numbers. There are a lot of intangibles here, which seem apparent to me based on my instincts as a gamer. It has to do with the momentum of winning and losing.
Mitt Romney has been very far behind in the electoral college since he became the de facto candidate in April. (That was when Rick Santorum quit the race, notwithstanding the continued participation of Ron Paul, who is never ever going to win a chance to represent the Republican base.) Where matters stand now, Obama can lose several swing states and still win comfortably; Romney has to win most of them to have a shot. In the last few weeks, Obama's lead on Nate's graph has been widening, at a time when it should be shrinking.
The major opportunities for Romney to turn this around have come and gone -- the Vice Presidential nomination and the convention were the big ones, and neither of those gained him any traction. All that he has left are the debates, and I don't think anybody seriously thinks that Romney will outperform Obama so much that he'll move a very large majority of undecided voters, or get committed Obama voters to switch teams.
Of course, there's still the issue of Romney outspending Obama on advertising. The Supreme Court ruling in Citizens United v. Federal Election Commission meant that corporations are now allowed to spend unlimited money in an opaque manner on the candidate of their choice, where previously there were spending limits on individual contributors. Since Mitt "Corporations are people, my friend" Romney is generally preferred by corporations for opposing regulations and business taxes, it was easy to predict that this would greatly benefit the Romney campaign. This prediction panned out because in my estimation, Romney has outspent Obama by a factor of about 6:1.
But the thing is... that already happened. Romney has been the beneficiary of PAC money for the last several months, and that was during the whole time when he was losing badly and never caught up. It's not that I don't think money matters at all in a presidential election; but there may simply be diminishing returns as the media becomes completely saturated with campaign ads. This may be even more true with each election, as people increasingly prefer to entertain themselves with web surfing, video games, and commercial-free Tivo recordings and mostly can't be targeted by the traditional television and newspaper ads. And now, with less than two months to go, Romney has been mysteriously giving up on buying ads in many swing states.
But here's my main intangible reason why Nate Silver's formula might be much too generous in giving Romney a 24% chance to win. Romney is currently perceived as losing, and perception still has six weeks to snowball.
I've written before that I like to look at politics through the lens of competitive gaming. I am a big Starcraft II fan, and a lot of my philosophy of gaming has been influenced by David Sirlin's outstanding writing in Playing To Win. (I'm not kidding... every gamer should read this.) So I read a lot about meta-strategy in games, and I'm frequently reminded of something I read once about judging how a match is going. I'd like to make a Starcraft analogy, but I am aware that this is a pretty narrow niche for people, so I'm going to cast it in a more accessible metaphor about chess.
Very often a game is won or lost based on very small advantages that are achieved early in the game. Suppose one player loses a pawn without gaining any positional advantage in return. An amateur chess player might look at the board and conclude that the player who took the pawn is ahead by a very small amount and it is a close match. But having that advantage can open up a lot of new opportunities: for instance, the attacking player can now set up situations where he confronts one unprotected pawn with two of his own, and therefore makes it easy to increase his lead. When you're ahead, even a fair trade of pieces can increase your lead. When you're behind, your losses provide a bad feedback loop, and also sometimes causes psychological mistakes because the player gets desperate.
So even a seemingly small material lead, in the hands of a good player, can be the source of a very big advantage that isn't apparent if your only method of judging is adding up all they points on the board. From my point of view, it looks like everything Romney does is putting him farther behind, and I don't think Nate's analysis can account for that kind of momentum.
Mitt Romney has been very far behind in the electoral college since he became the de facto candidate in April. (That was when Rick Santorum quit the race, notwithstanding the continued participation of Ron Paul, who is never ever going to win a chance to represent the Republican base.) Where matters stand now, Obama can lose several swing states and still win comfortably; Romney has to win most of them to have a shot. In the last few weeks, Obama's lead on Nate's graph has been widening, at a time when it should be shrinking.
The major opportunities for Romney to turn this around have come and gone -- the Vice Presidential nomination and the convention were the big ones, and neither of those gained him any traction. All that he has left are the debates, and I don't think anybody seriously thinks that Romney will outperform Obama so much that he'll move a very large majority of undecided voters, or get committed Obama voters to switch teams.
Of course, there's still the issue of Romney outspending Obama on advertising. The Supreme Court ruling in Citizens United v. Federal Election Commission meant that corporations are now allowed to spend unlimited money in an opaque manner on the candidate of their choice, where previously there were spending limits on individual contributors. Since Mitt "Corporations are people, my friend" Romney is generally preferred by corporations for opposing regulations and business taxes, it was easy to predict that this would greatly benefit the Romney campaign. This prediction panned out because in my estimation, Romney has outspent Obama by a factor of about 6:1.
But the thing is... that already happened. Romney has been the beneficiary of PAC money for the last several months, and that was during the whole time when he was losing badly and never caught up. It's not that I don't think money matters at all in a presidential election; but there may simply be diminishing returns as the media becomes completely saturated with campaign ads. This may be even more true with each election, as people increasingly prefer to entertain themselves with web surfing, video games, and commercial-free Tivo recordings and mostly can't be targeted by the traditional television and newspaper ads. And now, with less than two months to go, Romney has been mysteriously giving up on buying ads in many swing states.
But here's my main intangible reason why Nate Silver's formula might be much too generous in giving Romney a 24% chance to win. Romney is currently perceived as losing, and perception still has six weeks to snowball.
I've written before that I like to look at politics through the lens of competitive gaming. I am a big Starcraft II fan, and a lot of my philosophy of gaming has been influenced by David Sirlin's outstanding writing in Playing To Win. (I'm not kidding... every gamer should read this.) So I read a lot about meta-strategy in games, and I'm frequently reminded of something I read once about judging how a match is going. I'd like to make a Starcraft analogy, but I am aware that this is a pretty narrow niche for people, so I'm going to cast it in a more accessible metaphor about chess.
Very often a game is won or lost based on very small advantages that are achieved early in the game. Suppose one player loses a pawn without gaining any positional advantage in return. An amateur chess player might look at the board and conclude that the player who took the pawn is ahead by a very small amount and it is a close match. But having that advantage can open up a lot of new opportunities: for instance, the attacking player can now set up situations where he confronts one unprotected pawn with two of his own, and therefore makes it easy to increase his lead. When you're ahead, even a fair trade of pieces can increase your lead. When you're behind, your losses provide a bad feedback loop, and also sometimes causes psychological mistakes because the player gets desperate.
So even a seemingly small material lead, in the hands of a good player, can be the source of a very big advantage that isn't apparent if your only method of judging is adding up all they points on the board. From my point of view, it looks like everything Romney does is putting him farther behind, and I don't think Nate's analysis can account for that kind of momentum.
I tend to agree with your overall conclusion, but it seems like you're bashing 538 little for not behaving in a certain way when it isn't supposed to be working that way.
ReplyDeleteIt's like asking why we don't have a crockoduck.
More to the point, 75% is freaking huge. Consider that there is a good 45-48% of the population that is going to vote Republican regardless, and a shift of 1% of the voting population is huge, and build in some uncertainty based on polling methodologies and the fact that we're weeks away and well, if it isn't impossible to get into the mid to high 90s, it's really hard.
Also, if you look at the now cast, which tries to reflect the probabilities if the election were held right now, Obama stands at 94.8% which I think reflects your thinking.
Also, I used to live in Massachusetts and it was equally clear to me that should Romney be the nominee, Obama was going to win fairly easily, and if not, Obama was going to win even more easily.
The fact that the Republicans are trying to lose just amuses me.
I'm not bashing 538 -- they're a great resource that provides valuable information. I'm just saying that the fact that 538 forecasts a 75% chance to win based on pure poll data, doesn't directly translate into a genuine 75% chance that that's how it will shake out in the end.
ReplyDeleteI do enjoy reading your views on this, but I don't think the game analogy works quite that well. While, say, in chess every little step counts and influences next options, in the case of election you're dealing with people who might be deciding much less logically and more emotionally. Therefore, something that should be one of those small advantages that ultimately prove crucial to the result may well diminish in people's minds in comparison to more recent or prominent things and lose its decisive power. Simply put, the game never forgets, while people do. Of course it will probably never happen across the board, but it's enough to make this "small advantages law" less valid in election than in a game.
ReplyDeleteMidori,
ReplyDeleteRereading it, I think maybe I didn't bring back the analogy back to the point in the way I intended to. I mentioned at first that Mitt is being perceived as a loser. I think voters love to jump on bandwagons. When a candidate is perceived a winning, he gets treated more favorably in the press. When he's perceived as vulnerable, a narrative condenses around him that is hard to shake off.
For instance, Al Gore in 2000 was the unfortunate recipient of the narrative that he made up fake stories to make himself look good. He correctly took create for his role in the legislation which created the internet, and some key technological leaders (i.e. Vinton Cerf) supported that claim; yet this false phrasing that he "invented" the internet stuck. At one point, an obvious joke about being raised on a song about unions was pulled out without the crowd reaction and played as if he seriously meant that his mom sang him a song which (a) is not really for kids, and (b) didn't exist when he was a kid. The more he got saddled with this meme, the more the press went out of their way to look for examples.
A similar thing happened with John Kerry and "flip-flopping" once it became clear that he couldn't effectively clarify his stance on the war. He really took the high road when confronted with opposition like Swift Boaters, refusing to respond directly to the attacks, and I think it cost him what could have been an easy win. His storyline was that he was kind of a European sounding wuss, and the fact that he wouldn't fight back against direct political attacks fed into that story.
In Romney's case, the fact that people see him as losing is translating into other Republicans bad mouthing him publicly, quitting the campaign or trying to distance themselves from him. With friends like that, who needs official Obama attack ads?
Russell,
ReplyDeleteYou're basically saying that it's hard to turn things around in a campaign if you have such a disadvantage, and I agree with you. In a game, however, the rules are more strictly defined and determine the options in a precisely quantifiable way (unless indeed you somehow pull a Kobayashi:).
A political campaign is different in this aspect. It's much more difficult to predict with certainty which narratives will stick, for how long are they going to be around, how influential will they turn out to be and, most importantly, whether something else might come along that will overshadow them.
That's pretty much what I meant when I said that this principle is less valid in elections. I didn't say it was completely invalid, because I agree it does describe a certain general trend. I just wanted to point out that there are limitations to that analogy, especially when we're talking about predictions.
If you haven't already checked out today's 538, you should do so.
ReplyDeleteIt will make you giggle.
http://fivethirtyeight.blogs.nytimes.com/2012/09/27/sept-26-could-2012-be-like-2008/