Reader Posts

« previous | TPM CAFÉ READER POSTS HOME | next »

Probabilities of Electability Update

Here I posted results of a Monte Carlo simulation I had done of the election, using polling data from Votemaster Andrew Tanenbaum's electoral-vote.com. My analysis then showed Barack Obama would win more frequently in a head-to-head matchup with John McCain than Hillary Clinton would, even though if you simply totalled electoral votes by who was leading the state, Clinton had a larger margin over McCain than Obama did.

The votemaster has updated with some newer polls, and so I've rerun my analysis. Again I've done 10,000 trials for each of a Hillary-McCain and Obama McCain matchup. Now Hillary wins against McCain little more often than Obama does:
Obama wins 45.25%, averages 269.8 EV
McCain wins 49.04%, averages 268.2 EV
Electoral tie  5.71%

Clinton wins 50.35%, averages 269.1 EV
McCain wins 46.19%, averages 268.9 EV
Electoral tie 3.46%

The headline summary of these results has changed from before: where earlier Obama won more often, now Clinton does. But the big picture really isn't that different. McCain is now one point ahead of Obama in Pennsylvania, whereas before it was in Obama's column. Hillary had previously led McCain by just 1 in Minnesota, but now she has a 5 point lead.

Interestingly, even though McCain wins more often than Obama, Obama has a higher average electoral vote total. This is largely because of Texas - in the modeling, Obama wins Texas about a third of the time, and swinging its 34 electoral votes to Obama pulls his average above McCain's. In fact, in my simulation Obama's average electoral votes are slightly higher than Clinton's, but for all practical purposes they're tied, both with McCain and each other.

My model is a simple Monte Carlo simulation: For each trial I add a random variable with a to the spread between the candidates in each state, assuming each state is an independent trial. Then I assign the state's electoral votes to the winner, and compute the overall winner of that trial.

This model is quite sensitive to the actual numbers reported by the poll: if you have a 1 point lead in the state, in the model you'll win that state about 2/3 of the time, with a 2 point lead, it you win about 85% of the time, and with a lead of 5% or more you win well over 90% of the time. Hence when Obama went from an 8 point lead over McCain (in the April 09 poll) to a 1 point deficit, the state swung from definitely Obama to probable McCain. My gut says Obama would still be a strong favorite to win Pennsylvania in November. But also, I'd say McCain is still far more likely to win Texas than the 2/3 the model projects from the current one point spread.

The big picture remains essentially the same: based on the latest per-state polling, either candidate would be in a tough fight with McCain. At present, Clinton is closer to her ceiling than Obama is, but as of today's data, she's a little more likely to win.


Comments (27)

Minor correction to the above: with a 5 point lead in my model, you win the state well over 99% of the time, not 90%. So basically a 5% or more lead is a "guaranteed" win in the state.

Rather than just giving the average electoral vote total, it'd be better to report the percentage of wins. After all, when it comes to the electoral college, "an inch is as good as a mile". I.e., losing by 35 points vs. losing by 1 point has the same outcome.

I reported both:


Obama wins 45.25%, averages 269.8 EV
McCain wins 49.04%, averages 268.2 EV
Electoral tie 5.71%

Clinton wins 50.35%, averages 269.1 EV
McCain wins 46.19%, averages 268.9 EV
Electoral tie 3.46%

It's essentially a dead-heat with both candidates right now, but where Obama was over 50% based on the data I used earlier in the week, now it's Hillary who's slightly over 50%.

My script outputs how often each candidate won a given state in the simulation, but that's a lot of data to type into a TPM blog posting box. Plus, I don't think they support tags, so the formatting wouldn't be too pretty...

Yeah, massive reading comprehension failure. Mea culpa.

I'm losing heart on this election, given that McAttack is not wandering in Mondalian wilderness. He should be laughed out of the race, not nearly tied.

Does this mean too many Americans are really stupid? Bamboozled? Propagandized? Does it indict the media?

Or does it mean too many Americans are just mean SOBs that would rather be broke than help their neighbors?

Something is terribly wrong.

It means that we haven't really begun to engage McCain yet. I really expect things to change significantly shortly after we end the Democratic primary race.

I'm with Ben here. McCain probably has better name recognition than Obama, and he has fewer negatives (so far) than Hillary. People still view him as a "moderate" Republican, which, especially now that he wants to make the Bush tax cuts permanent, he no longer is (if he ever really was one).

The charge that McCain is running for Bush's third term will stick, and his 100 years in Iraq quote will hurt him, especially if Obama is the nominee.

If I were betting, I'd still say the Democrats have at least a 2/3 chance to win in November.

But, as Ben said, the Democrats haven't really engaged him yet. I think this is pretty much McCain's ceiling, unless we find magic ponies that bring a sudden, glorious, and victorious peace to Iraq...

It means, just this once, just this frakkin once, Democrats are going to have to stop doing the Chicken Little dance every goddam time the weather changes, toughen up and soldier on.

And McCain hasn't been vetted yet. He's been damned by some republicans, but most don't know him or how scary he is.

And when he puts Condi Rice on the ticket he will be an even more formidable opponent. Clinton will have a tough time against that ticket. Obama will suddenly have no chance. Rice is a devout fundamentalist Christian, a woman, an African American who is aligned with McCain on the occupation. She will seriously cut into Obama's support from black women or Clinton's support from blacks in general. She will energize the fundamentalist Christian base. IMHO she is the greatest threat to either democratic candidate right now.

Dick Morris wrote a book 2-3 years ago predicting that the 2008 election would be between Clinton and Rice, and that Rice would win.

That would be an interesting gambit for McCain, and I think it would be a big plus if he's running against Clinton. Many Obama supporters, especially African-Americans, are likely to be disillusioned under most plausible scenarios by which Clinton could become the nominee. Picking Rice would help McCain make serious inroads among African-Americans.

Against Obama, it would be less effective, as it more clearly binds McCain to Bush and the Iraq war. I seriously doubt Rice would cut into Obama's African-American support, and it's not like being pro-war is going to be especially appealing to most women. I'd say Obama would beat a McCain/Rice ticket rather easily. Clinton would be in trouble, though.

But if it's Obama against McCain/Rice, where does the racist and sexist vote turn? That's a problem we've never had to face before! Could someone convince David Duke to run as a third party candidate?

Well, the sexist vote could turn to Obama, as long as they're not also racist (which, of course, is often the case). Unless, Obama also picks a female VP candidate.

Btw, you seem to neglect the possibility of Rice picking up the female voters who would be disillusioned by Obama "stealing" the vote from Clinton (since we all know she deserves it).

You're right - I probably am underestimating that possibility.

But Obama's campaign hasn't tried to make an issue of Hillary's sex, nor is there a widespread perception that he has. Whether you believe Clinton has tried to tar Obama as a "black" candidate or not, certainly some Obama supporters believe she has, and those would be more receptive to a McCain/Rice ticket.

So despite the exit polls from Pennsylvania, which suggested Hillary supporters would be more likely to support McCain than Obama supporters, I still think Obama supporters are more likely to switch or stay home if Clinton is the nominee.

Heh, I'd love to see McCain put Condoleeza Rice on the ballot. If he isn't having enough trouble energizing his right-wing fundamentalist base, imagine a 71-y.o. white man trying to win as a Republican with a scary-looking, never-married black woman first in line to the throne -- sorry, Presidency. What if, God forbid, McCain were to die in office? A black woman President? I'd love to see the racists stay home with the Christian fundamentalists in November. Maybe they can play pinochle, or something.

I'm not worried that black women would flock to a Republican ticket with a black woman VP. Clinton supporters' race-baiting aside, black voters don't vote based on race or gender identity -- >90% of African-Americans who vote have voted for an all-white Democratic ticket every election for a couple of generations now. That's not because black voters love all-white tickets -- it's because black voters (quite sensibly) prefer the Democrats on policy grounds, even if the Dems rarely trouble to earn our loyalty.

Besides, I've never heard a black woman say anything nice about Rice. Well, one woman did wonder out loud to me, "How can she bear to socialize with Cheney and Rumsfeld and all those right-wing white guys who probably play golf in country clubs where black women aren't allowed? I know she has to work with them, but does she hang out with them socially?"

"How can she bear to socialize with Cheney and Rumsfeld and all those right-wing white guys who probably play golf in country clubs where black women aren't allowed? I know she has to work with them, but does she hang out with them socially?"

Condi didn't hang out with Cheney or Rumsfeld; she hung out with W. watching football on Sunday afternoons!

She also played the piano and sang at Sunday services in the WH. Minister's daughter? I think she'll turn the religious right out even heavier than Bush did. She's real. I think in their hearts they knew he was a little phony.

And don't forget, if she gets the first question in a debate, her opponent will never get to talk. Nobody can run the clock out like Condi!

*bows to your comment*

I don't know about Rice, since an Iraq-centered competence and credibility attack on her would be formidable and overlap with one of McCain's biggest liabilities -- but if Obama is the nominee it might be a smooth Machiavellian move for McCain both to pick a woman for running-mate AND declare (as Peggy Noonan has recommended) that he'd only serve for one term.

Hard to think of who, other than Rice, that might be. The popular and young R Governor of Alaska?

But however clever his choice may be, I don't see McCain ever being President.

Hi Billy.

The popularity of the Iraq War makes Condi a natural. Plus, your assertion earlier today that Obama would energize the GOP base (apparently even more than would Clinton) seemed to me to imply that religious conservative Caucasians (their base) would line up in droves to vote against the scary liberal African-American. [Also, the GOP base is war weary. That's where most Obama Republicans come from -- and note the special election in MS where the Dem won, based largely on GOP war-weariness.]

It remains to be seen how fired up that GOP base is about a periodically moderate Republican and an African-American woman. If Obama sets their blood pumping, does the thought of putting a black woman one heartbeat from the Presidency make them feel that much better?

I just don't think the magical antiblack voter argument works only against Democrats, unless its designedly really just an argument against Obama.

Yes, and arguably she is also guilty of war crimes. While I can't think of a ticket I'd like to see more than McCain/Lieberman, it should be no less fun to point out that John McCain has taken on a VP who authorized torture.

Obama's been taking serious and simultaneous flak from both HRC and McC for at least a month; he'll be fine in the GE.

He'll do only a little better than Dukakis

Or about the same as Kerry. What's the difference?

Great job, Fossberry. Actually, I can't quite follow all the model-speak, but the trends are faskinatin'.

I hope you'll keep us updated, and I'll keep recommending when you do.

As to the main post, the exercise will become more interesting when it less closely resembles a series of coin flips.

In a series of trials resulting in nearly 50 percent victory percentages by all participants in all trial runs, or in a series of trials where the electoral vote average all around is almost exactly 269 (half the total), all this shows is that the data do not have enough internal variation to permit meaningful insight into what happens.

There is no statistical significance apparent within the exercise.

But it does remind us that Democrats win the tie scenario, because if the election goes to the House, Democrats control 26 house delegations to 21 for the GOP, with three tied. Unless we give back that advantage, 269 will be fine.

I'd say the exercise becomes more interesting (and accurate) as the state polling becomes more frequent, at least in states that may actually be in play.

What is its significance? It demonstrates the following points about the current polling data:
1. Both Democratic candidates are "electable".
2. Hillary isn't as strong as she appears if you simply add up the EV of states she currently is leading McCain, because she's barely leading in far more states than he is.
3. Obama is a little stronger than he appears when you add up EVs, because he's close to McCain in more states.
4. McCain currently is running neck-and-neck against either Democrat.

Post a Comment

Inside Cafe



Cafe Features


October 6-10

Book Cover

October 13-17

Book Cover

October 20-24

Book Cover

November 17-21>

Book Cover

December 1-5

Book Cover





Book Club Archive



Masthead

Editor-in-Chief
Josh Marshall

Site Editor
Lila Shapiro

Intern
Claire Wilcox



Subscribe to TPMCafe's feed.
Subscribe to TPMCafe's reader blog feed.

Advertise Liberally
Share
Close Social Web Email

"To" Email Address

Your Name

Your Email Address