Here's another take on this: the probability, as determined by logistic regression of our latest simulation run, of the candidate winning the electoral vote based on his share of the popular vote.
So a 1-point popular vote win translates to about a 75 percent chance of winning the Electoral College, a 2-point win about a 90 percent chance, a 3-point win a 97 percent chance, and a 4-point win a 99 percent chance.
This is, of course, assuming that my simulation model is getting this approximately right. Note that I've lumped together McCain and Obama's numbers here, making the distribution exactly symmetrical. I don't know to what extent these numbers would hold with two different candidates and with a different set of states in play, but I'd bet it's pretty close to the mark.
Thursday, July 10, 2008
Popular Vote v Electoral Vote, Part II
-- Nate Silver at 7:31 AM
Labels: electoral math, meta, popular vote
38 comments
Was that curve significantly different the last time you ran a simulation? can you superimpose the different results?
Nice work, Nate!
By the way, Glenn Greenwald -- apparently really peeved because Al Giordano won't join his FISA crusade -- accused Giordano of lying when Al mentioned the illegal surveillance and wiretapping that's been done for years in connection with the "Drug War". Giordano, however, was not and was able to prove it: http://narcosphere.narconews.com/thefield/salons-glenn-greenwald-writes-the-field#comments
Since you've been reamed out by people who are angry at you for not joining Greenwald's crusade, I thought this might cheer you up a bit. :-)
Nate,
Score another one for the model. This curve conforms to the conventional wisdom that a 4% win of the PV will always (well, almost always) result in an EV victory.
Add my applause, Nate. After the last two elections talk of still another election where one candidate wins the popular vote and loses the EC continues to crop up. In fact, of course, that outcome depends heavily on a very, very close popular vote election.
While it's theoretically possible for a candidate to pile up huge majorities in some states and narrowly lose others, the demographics and partisan distribution of the electorate makes it a more a "theoretical" than a real possibility.
Thanks for putting some quantitative muscle into that truism.
Thanks for a great site. As a numbers guy, this is heaven.
Excellent!
This is probably a weird idea, but I am wondering if you have looked at the effect of Swift Boat type incursions on past campaigns. I am tolerably convinced that Obama will be stonewalled by Swift Boat type ads of the type Director posted in another comment area here.
I doubt that such ads will be met with sufficient force to mute their effect -- unless the Will.i.ams (sic) of the world get to work dealing with McCain's vulnerabilities which include compulsive gambling and temper proclivities.
What was the effect of these incursions when Dukakis and Kerry were done in by them?
What would be the likely effect of a similar frontal attack on Obama?
Sounds like polling from ND today from Rasmussen!
Looks like you got your wish, Nate.
-Mark
Nitpick: the "loses OH, wins election" and "wins FL, loses OH" rows have 3,173 instances of Obama losing Ohio; the "wins VA/CO, loses OH" rows have 3,172 instances. (I think there might be a tie in the 3,173rd instance.)
This is all great stuff and I'm sure the simulation runs will become required practice for anyone working with predictions in the future. Nitpick: there's some weirdness in the colors of the states in the left-hand sidebar. You have FL and MO at lower win percentages for BO than e.g. NC, yet NC is red and FL and MO are purple. And MN only gets a purple for 90% BO, while IN gets a solid red for 52% JM.
Thanks for all the great work.
--UncleMonty
According to the numbers on the left, Obama's popular vote margin is +3%, so why is his winning percent not around +90%?
The state colors are based on 2004 results.
So to bring in your other line of work, this is a Jamesian "pythagoras" curve with exponent 50. That is
win prob = (votes for)^50/[ (votes for)^50 + (votes against)^50 ]
The 50 comes from the gain multiplier. A candidate with 0.505 of the vote, so 0.005 above even, ends up with a win percentage of 0.75, or 0.25 above even. That's a gain of 0.25/0.005=50.
Given that we have 50 states, it's kinda easy to remember.
Stephen C. Rose said "This is probably a weird idea, but I am wondering if you have looked at the effect of Swift Boat type incursions on past campaigns."
Stephen, for the effect on Dukakis in 1988, you can check page 10 of this paper, which clearly demonstrates the devastating effect of the Willie Horton ad, and Dukakis's failure to respond with a clear message.
As far as 2004, the effect of the "Swift boating" and Kerry's failure to effectively place his message are less clear, as Kerry never had a clear lead, and ran a generally lackluster campaign.
However, the lesson is clear. A candidate cannot let such attacks pass without a strong counter-response. It would appear that Obama has such a counter-response effort prepared. As a result, I am not convinced that 527 groups will be as effective in this campaign.
The use of 527 groups for attacks in this campaign may follow the old saw about "fighting the last war."
Nice model, I'm looking forward to see how it plays out.
I think it's interesting on the "Super Tracker" that as time goes by, you see the numbers tightening in their distribution. At the beginning of the year, you have such a wide distribution that gradually tightens up around March and April, then further into May and June. By the end of last month, the numbers are relatively clustered. Someone please remind me exactly what that's charting - is it the chances of either candidate of winning by "X" spread? In what, the popular?
Sorry, relatively new to the site, although I've been avidly watching the state-by-states and trying to make sense of everything I see.
I have said this before: Obama has such huge leads in NY, CA, IL that when you look at a national poll with close numbers (today’s Rasmussen had Obama down to 3%, yesterday’s Gallup was at 2%) it means that his electoral strength elsewhere must be weak. He is wasting votes in those big liberal states that would be better served by dispersing them elsewhere.
By contrast, McCain'sd support is not concentrated geographically as much as Obama's. Even the South (so many of you think) is not so solid for him. So should he begin to build even a modest lead in the polls, a landslide for him is more likely.
I think the current news cycle is taking its toll on Obama and I expect McCain to pass him in the Gallup and Rasmussen surveys by Sunday.
The flip flopping has destabilized his base. His comments about kids learning Spanish have not even made it to the masses, but they will. And given the hectoring and scolding way he delivered them and said he was “embarrassed” that American tourists cannot go to Europe and speak the local language I fully expect him to be further tarred with the “elitist” brush. Add to that the Iranian missile crisis (argues for keeping troops handy and next door and for drilling for our own oil) and you have the makings of a McCain surge. The only thing in the news cycle that helps him was what Jesse Jackson said about him. Now that was a gift!
Pete Kent posted:
"By contrast, McCain'sd support is not concentrated geographically as much as Obama's. Even the South (so many of you think) is not so solid for him. So should he begin to build even a modest lead in the polls, a landslide for him is more likely."
Keeping in mind that, with your comment about Obama's supposed identity as the "Antichrist" being "persuasive," you have no credibility, I wonder if you'd nevertheless humor me by giving a statistical explanation of your apparent wishful thinking above.
Rasmussen has Obama leading only 11 point in hie home state of Illinois in today's released poll.
No way, I'd bet my hose in Eureka, that Obama is at least 25 ahead in his home state of IL.
Really, something is wrong with Rasmussen polls......
> According to the numbers on the
> left, Obama's popular vote margin
> is +3%, so why is his winning
> percent not around +90%?
Because +3% is the popular vote margin today, but its not guaranteed to hold through election day. If you did a "probability of winning if the election were held today" (which Nate recently did), you should get close to 90%. Nate's model however attempts to do the near impossible, and that is to assign probabilities