One of the fun things we can do with our model is to change the election date to anything we like and see how it might impact the results. So what would the model project if the election were scheduled for tomorrow?
Notice how the red states turn redder and the blue states turn bluer. This is something you'll see happen very gradually over the course of the next four months. As we near the election, smaller leads will become more meaningful, and more states will be taken out of play.
Overall, our model thinks that Obama would have an 88.6 percent chance of winning an election held tomorrow. The most competitive states would be as follows:
1. Virginia. Obama 58% to win, McCain 42%
2. Florida. McCain 62% to win, Obama 38%
3. Indiana. McCain 63% to win, Obama 37%
4. Nevada. McCain 63% to win, Obama 37%
5. Missouri. McCain 65% to win, Obama 35%
6. North Dakota. McCain 79% to win, Obama 21%
7. North Carolina. McCain 81% to win, Obama 19%
8. Colorado. Obama 83% to win, McCain 17%
9. Louisiana. McCain 85% to win, Obama 15%
10. Ohio. Obama 88% to win, McCain 12%
11. New Mexico. Obama 88% to win, McCain 12%
12. South Dakota. McCain 89% to win, Obama 11%
13. Michigan. Obama 90% to win, McCain 10%
14. West Virginia. McCain 90% to win, Obama 10%
15. Montana. McCain 91% to win, Obama 9%
16. Alaska. McCain 92% to win, Obama 8%
17. South Carolina. McCain 94% to win, Obama 6%
All other states are at 95% or better for the leading candidate.
McCain's electoral prospects, unsurprisingly, would be very much tied to his chances of pulling out an upset in Ohio or Michigan. One thing that's going on here, by the way, is that the amount of polling data in a particular state is a factor. For example, our model thinks that McCain would win by 4.4 points in North Dakota in an election held tomorrow, but still assigns Obama a 21 percent chance of winning there because the polling data is so sparse that we can't quite be certain what's going on. In Ohio, meanwhile, the model projects an Obama win of 4.2 points -- less than McCain's projected margin in North Dakota -- but gives him an 88 percent chance of winning because that small lead is backed up by much more polling volume.
And there is still an outside chance that something totally wacky could happen, with the polling data off in the same direction in many states. As recently as 2000, there was about a 3-point miss in the final national polling averages ... Bush was expected to win that election by about 2.5 points but Gore actually won the popular vote by a half-point. In 1996, Bill Clinton was 12 points ahead in the final polling averages but won by "only" 8.5 points ... in 1980, Ronald Reagan led by just 3 points in Gallup's last poll but actually won by almost 10 points. Sometimes, there is movement that comes too late for the polls to detect, or everyone is off in the same direction with their turnout assumptions.
Saturday, June 28, 2008
If the Election WERE Held Today...
-- Nate Silver at 8:01 PM
Labels: history, scenario testing
30 comments
Interesting analysis and reminder of the difficulty in projecting turnout and closing margins of error. Though this is a bit off-topic for projections, you've probably run into different exit poll statistics...How have these fared in comparison to the final polls? And are you taking into account last-minute happenings that tended to sway votes?
nate,
this is why your charts would be so much more fun with live sliders and other doo-dads for all sorts of different variables...
it would be fun to turn on and off different polling firms, or click to normalize for party id, or lets say, choose party id based on latest registration data, etc...
at the least i'd like to be able to sort poll details by all the different columns..... i'd like to be able to make up my own scenarios to analyze, etc...
of course your skinner like ways have caused bad addiction so far... who knows what the doo dads would do...
(soapbox)
as for election day discrepancies, i have to think about all the vote fraud that has been widely reported, including caging lists, leaving punch card machines full of chad, moving voting locations to confuse voters, and all sorts of other stuff. that stuff adds up. our only way to deal with it is to be like fish and have plenty of eggs. so everybody try to get 10 of your low info friends to register to vote. elections matter.
(/soapbox)
Let me jump on the bandwagon here and back up the guy above me that suggested some interactive features for the map. I'm sure there have to be some ultra-tech savvy folks that would be willing to chip in. (I'm not one of them - though if you need 538 salsa, I'm the guy to talk to.)
538 salsa,
???
My goodness, he's still working on the technical stuff, I mean the mathematical stuff, and you guys want bells and whistles on the site itself?
OK, I'd like them too, but we have to be reasonable. This is still essentially a one-man operation; he's had some design help but doesn't have a programmer. If you want a site with a lot of "flash" but without the math, there are plenty of those.
Let's give it time -- and I'm sure Nate would be happy to have volunteers who have good technical design skills. BTW/ the programming of all the options that you want would take a lot of time.
An interactive map where u can see how all the states would change as a slider moved back and forth changing the margin of victory would be a great idea.
I think Nate has got a new project.
Nate
you rock !
this is a great site
Nate, thanks. I figured the chance of winning today was much higher than your fall projections.
The turnout error you mention sounds like a systematic error. If anything, I think it would go in Obama's direction. The primary polls underestimated Obama's support, I think by 3-4% on average, presumably because more Obama supporters turned out than were assumed by the pollsters. There is also going to be a lot of voter registration between now and November.
Perhaps some of this error is already included in your "polls tighten" effect.
btw, why does your Electoral Vote Distribution have a peak near 0?
-eyesopen
One reason for the extra large Reagan margin compared to the projections in 1980 was Carter's (typically) self-centered decision to concede early. It likely cost the candidate whose campaign I was managing the election (as well as screwing a lot of other Dems in the West.) One of a number of things he has done I have trouble forgiving him for.
I still have to quote the late great Tim Russert
OHIO
OHIO
OHIO
I think whoever wins Ohio wins the election. That being said because of the economy I think Obama wins Ohio with ease. Got I'm gonna miss Mr. Russert's face on election night.
"Carter's (typically) self-centered decision"
Excuse me? Jimmy Carter was one of the most caring, empathetic, and honest presidents we ever had. Just look at all the humanitarian work he's done since.
Hi Nate. Out of curiosity, is your consistent use of the word "data" as a singular, rather than plural, noun intentional? I'm not trying to be nitpicky, as I think the singular usage is widespread enough so as not to be prescriptively wrong. I was always taught, including in stats classes, to treat it as a plural, but I notice a lot of people doing the opposite.
I almost feel more comfortable with Kerry+IA+CO+NV than Kerry+OH.
Ohio has disappointed me too many times.
"Bush was expected to win that election by about 2.5 points but Gore actually won the popular vote by a half-point. In 1996, Bill Clinton was 12 points ahead in the final polling averages but won by "only" 8.5 points"
These two should not be compared. 2000 was close, 1996 was not. We don't know if the candidate leading in the polls will actually benefit, or not, from a large lead. There may be less enthusiasm on his own side, or less enthusiasm on the other side. That depends on a lot of factors, and 3.5 points could very well be within that enthusiasm gap.
Does anyone have any polls or data showing the effect of Hillary Clinton as potential veep on the presidential race? I've asked for this information in other comments, but have only gotten assertions that she hurts the ticket. I don't know until I see data.
One argument against her is she mobilizes Republicans. But she also mobilizies Democrats, and we know the data show there aren't as many Republicans as there used to be. So is it possible she is a net plus for Obama among party identifiers?
And then I have no idea if Obama-Clinton would be better among indepdents than Obama is polling now. Anyone have data on this?
The only recent poll I've seen shows Obama-Clinton helps Obama somewhat in Michigan. Don't know about other states.
See poll results here:
http://primebuzz.kcstar.com/?q=node/12076
Dear Ty, read this and then tell me what your agendum is?
http://www.askoxford.com/asktheexperts/faq/aboutgrammar/data
Some points:
1. The "election today" map actually shows Obama 306 - McCain 232.
2. Given an Obama lead of 36 EVs, he could lose EITHER Michigan OR Ohio, but not both. This is not generally considered a shocker.
3. Winning both would make it virtually impossible for McCain even if he won Virginia (likely in the end, Virginia hasn't voted for a Democratic President since 1964).
4. I still think what we're really looking at is Kerry + NH + IA + CO + NM = 273. That's Obama's