Wednesday, November 7, 2012

Election outcome and prediction markets


http://www.intrade.com

Sophisticated prediction analysis:

http://money.cnn.com/2012/11/07/news/companies/nate-silver-election/index.html


http://www.ritholtz.com/blog/2012/11/nate-silver-and-the-lessons-of-2012/

http://fivethirtyeight.blogs.nytimes.com/2012/11/01/oct-31-obamas-electoral-college-firewall-holding-in-polls/


Oct. 31: Obama’s Electoral College ‘Firewall’ Holding in Polls

On Oct. 11, this blog posed the question of whether President Obama’s “firewall” in battleground states was all that it was cracked up to be.
At that point, Mr. Obama still technically held the lead in the FiveThirtyEight forecast in enough states to give him 270 electoral votes. But Colorado, Florida and Virginia had turned red in our map, meaning that our forecast suggested that Mitt Romney had better-than-even odds of winning them. Iowa was just on the verge of doing so. And Mr. Obama’s lead was down to just a percentage point or so in Ohio, which would have collapsed his firewall at its foundation.
Theories that the decline in Mr. Obama’s polls that followed the first presidential debate in Denver would somehow skip the swing states were not looking good — as dubious as the idea that tornadoes “skip” houses.
Instead, at that point, Mr. Obama’s position in the FiveThirtyEight forecast had declined for seven consecutive days. If he stopped the bleeding there, he might still be the Electoral College favorite, albeit a narrow one. But it wasn’t clear where the bottom was.
It turned out, however, that the worst was almost over for him. Mr. Obama had one more terrible day in the polls, on Friday, Oct. 12, when Mr. Romney’s chances of winning the Electoral College rose to almost 40 percent in the forecast. But that was when Mr. Romney’s momentum stopped.
Since then, Mr. Obama’s standing has rebounded slightly. His position in the national polls has stabilized; although the national polls continue to tell a different story about the race than the state polls do; it can no longer be said that they have Mr. Obama behind. (More about that in a moment.)
Meanwhile, Mr. Obama continues to hold the lead in the vast majority of polls in Iowa, Nevada, Ohio and Wisconsin, the states that represent his path of least resistance toward winning the Electoral College. This was particularly apparent on Wednesday, a day when there were a remarkable number of polls, 27, released in the battleground states.
There were 12 polls published on Wednesday among Iowa, Nevada, Ohio and Wisconsin. Mr. Obama held the lead in 11 of the 12 surveys; the exception was a survey by the University of Iowa, which had Mr. Obama down by about one point there, but also had a very small sample size (about 300 likely voters). On average, Mr. Obama led in the polls of these states by 3.9 percentage points.
None of this ought to have been surprising, exactly, if you have been attentive to the polls rather than the pundits. It was a pretty good day of surveys for Mr. Obama but not a great one: for the most part, the polls were coming in close to FiveThirtyEight forecasts in each state, give or take a modest outlier here and there.
Rather, the polls in these states confirmed what we already knew: that Mr. Obama remains the favorite in the Electoral College.
Mr. Obama is not a sure thing, by any means. It is a close race. His chances of holding onto his Electoral College lead and converting it into another term are equivalent to the chances of an N.F.L. team winning when it leads by a field goal with three minutes left to play in the fourth quarter. There are plenty of things that could go wrong, and sometimes they will.
But it turns out that an N.F.L. team that leads by a field goal with three minutes left to go winds up winning the game 79 percent of the time. Those were Mr. Obama’s chances in the FiveThirtyEight forecast as of Wednesday: 79 percent.
Not coincidentally, these are also about Mr. Obama’s chances of winning Ohio, according to the forecast.
Regular readers will have seen the chart below once or twice before. It sorts the competitive states in order of Mr. Obama’s current projected margin of victory or defeat in each one, keeping a running tally of the number of electoral votes that Mr. Obama is accumulating.
Ohio remains the tipping-point state in the forecast, the one that puts him over the top to 270 electoral votes. There, Mr. Obama leads by 2.6 percentage points, which should convert to a victory about 80 percent of the time given the historical accuracy of polls at this late stage of the race.
Mr. Romney’s chances of winning the Electoral College without Ohio — a prospect we had defended as being plausible before — are looking more tenuous based on the most recent polling.
If Mr. Obama wins Ohio, and all the states above it on the chart, he’d have 281 electoral votes, meaning that he has 11 to spare. That means he could shed New Hampshire from his list, along with either Iowa or Nevada (although not both).
Of these two states, Nevada appears to be the slightly safer one for Mr. Obama; there, Mr. Obama leads by 3.5 percentage points in the forecast, as opposed to 2.9 percentage points in Iowa. The polling has also been somewhat more consistent in Nevada than in Iowa, another factor that the forecast considers in evaluating the probability of an upset.
One fortunate aspect of these two particular states, from Mr. Obama’s view, is that they are not very similar to one another demographically.
Iowa is quite rural. Nevada occupies a huge geographical territory, but its population is very urban, mostly living in Las Vegas and its suburbs.
Iowa is overwhelmingly white, and has a lot of moderate and middle-income, but highly educated, voters. Nevada certainly has an independent streak, but winning there usually depends more upon building a 50 percent coalition among diverse groups and then turning it out to vote. Iowa has a pretty good economy, all things considered; Nevada’s is still terrible.
Since Mr. Obama only needs to carry one of these states, it helps him that they form a diverse portfolio. If Mr. Obama’s turnout operation is strong, then Nevada should be one of the states where he benefits the most. If, instead, Mr. Obama has little “ground game” advantage, but he holds his own among independent and undecided voters, perhaps persuading them that the economy has improved enough to merit his re-election, Iowa may fall for him.
Mr. Romney could also circumvent his need to win Ohio by carrying Wisconsin, but that is looking tough for him. In Ohio, Mr. Romney is behind by two or three percentage points, on average, in the polls. In Wisconsin, Mr. Romney’s better polls have him down by two or three points, while his worst ones have him six to eight points instead. There’s still enough upside to winning Wisconsin that Mr. Romney should not give up on it, in my view, but his chances are down to 12 percent in the forecast, and most of those cases involve outcomes where he has already won Ohio anyway.
The more debatable cases are Pennsylvania, Michigan and Minnesota. Mr. Romney is the clear underdog in each one. But his campaign has so much money that it probably doesn’t hurt Mr. Romney much to spend a little bit of it there to maximize whatever residual chances he might have in case the polls are wrong. (Arguably, it was a poor strategic decision for Mr. Romney to make a half-hearted effort to compete in these states.)
Still, for Mr. Romney to win Michigan, Minnesota, or Pennsylvania, the polls would have to be much further off than they are in Ohio.
It doesn’t help Mr. Romney, either, that all of these states are in the same part of the country as Iowa, Ohio, and Wisconsin, meaning that they are unlikely to leapfrog them and become the tipping-point state on Tuesday. If, hypothetically, Mr. Romney’s polling were a bit better in culturally and geographically disparate states, like Oregon, New Jersey or New Mexico, they might represent better targets.
If Mr. Obama were to lose Ohio (but hold the other states), the tipping-point would then become Colorado. There, Mr. Obama holds a much more tenuous lead, about one percentage point in our forecast, which converts to about a 60 percent chance of winning. But at least it’s a lead rather than a deficit, whereas Mr. Romney’s non-Ohio paths would require him to win states where he is now three or four percentage points behind.
Mr. Obama also remains about a 60 percent favorite in Virginia. Another option would be Florida, although it is a resource-intensive state and we give him about a 40 percent chance of winning there.
While state polls dominated the news on Wednesday, there were also a handful of national polls out, even as others have been suspended in the wake of Hurricane Sandy.
On average, Mr. Obama led by just over one percentage point in these national polls, although it is an odd distribution, with two polls showing him up by four or five points, several showing a tied race, and one (the Rasmussen Reports tracking poll) putting him two points down.
The FiveThirtyEight model calculates a national poll average, using a more sophisticated method than the simple average I’ve taken in the chart. (The model doesn’t “forget” about the Gallup poll, for example, just because it has been suspended for a couple of days.)
We don’t usually print this number, because it would sow confusion: our estimate of the national popular vote, which we do publish, instead represents a combination of national polls and the implied standing of the candidates based on state polls.
But, for what it’s worth, our national poll average shows Mr. Obama up by about half a percentage point right now. This is within the range of other Web sites: Real Clear Politics has an exactly tied race in its national poll average; HuffPost Pollster has Mr. Obama down by three-tenths of a point; Talking Points Memo has Mr. Obama ahead by about one percentage point.
Again: we don’t take the average of the national polls to be tantamount to a forecast of the national popular vote, since state polls, if considered carefully, can provide considerable information about the national race as well.
Suppose, however, that Mr. Obama were to tie Mr. Romney in the popular vote on Tuesday. The way that the forecast model works, this would require subtracting some from Mr. Obama in each state in order for the arithmetic to add up.
Even under these conditions, Mr. Obama would still be a favorite in the forecast. In fact, he’d be about a 70 percent favorite to win the Electoral College conditional upon the national popular vote being tied, according to our simulations.
A tie in the national popular vote is a tolerable condition for Mr. Obama, in other words. His position is robust enough in states like Ohio that he has some slack. With a lead of about 2.5 percentage points in the tipping-point states, Mr. Obama could underperform his state polls by a point or two and still win.
Conversely, Mr. Romney has few chances to win unless the state polls are systematically wrong.
I don’t mean for this to sound dismissive; the polling error could quite easily be correlated across the different states, and the national polls are one reason to be suspicious of the state polls.
But we’re at the point now where Mr. Obama may be a modest favorite even if the national polls are right. Two weeks ago, when Mr. Obama appeared to trail Mr. Romney by a point or so in the national polls, that would not have been the case.
Is it possible that Mr. Obama has benefited, politically, from his handling of Hurricane Sandy? He has gotten high marks for it so far, according to the tracking poll run by The Washington Post and ABC News.
Our database contains roughly a dozen polls that conducted the bulk of their interviews on Tuesday or Wednesday, after Hurricane Sandy became the dominant news story. Most of these are state polls, and most were conducted in states that were isolated from the major effects of the storm.
Our analysis of the trend lines in the polls suggest that they have been a somewhat above-average group for Mr. Obama, perhaps suggesting a percentage point or so of improvement for him.
The model is not yet pricing in very much of this into its forecast, as trends like that can occur fairly easily because of statistical noise. But if the storm has a discernible effect in the polls, it seems more likely to help Mr. Obama than to hurt him based on what we’ve seen so far.
This is something to monitor as more national polls come back online. I think describing the race as a “toss-up” reflects an uninformed interpretation of the evidence, but there is surely room to debate how much of a favorite Mr. Obama is. However, Mr. Romney is not in a position to tolerate any movement in Mr. Obama’s favor given how close we are to the finish line.

No comments:

Post a Comment

Economic news - CNNMoney.com