« previous | TPM CAFÉ READER POSTS HOME | next »
Updated Electability Snapshot
For the past two weeks I've been running Monte Carlo simulations of the general election using polling data from Votemaster Andrew Tanenbaum's www.electoral-vote.com.
When I first did this simulation, both Clinton and Obama were quite evenly matched with McCain, although Obama was doing a little better. Last week both Democrats pulled significantly ahead of McCain, this time with Clinton running stronger than Obama.
The results this week (again I've done 10,000 trials for each election):
Obama wins 53.2%, averages 272.5 EV
McCain wins 46.2%, averages 265.5 EV
Electoral tie 0.6%
Clinton wins 87.0%, average 278.1 EV
McCain wins 10.4%, averages 259.9 EV
Electoral tie 2.6%
Now Clinton still has a big lead over McCain, and while Obama also leads, his lead is much smaller.
The big change from last week is a new poll in Texas, which now gives McCain a 5 point lead over Obama, rising from a 1 point lead in the prior poll. McCain now also leads in Wisconsin by 4, compared to an Obama 4 point lead last week, and McCain also widened leads in North Carolina and New Hampshire. Obama's big lead in New York grew even more, and he closed a little in Missouri, although McCain still has a significant lead there.
McCain went from trailing Clinton by 1 in Missouri last week to a two point lead this week, which explains her slipping a little compared to last week. Clinton actually improved her margins in most new polls, but in all those states the leads are still at or beyond the margin of error, so it had negligible impact on the simulations.
This is a snapshot based on the most recent polls in each state, not a prediction of what I'd expect in November. I'm using the standard 4% margin of error as the only variability in the model, assuming the recent poll reflects the true voter sentiment difference between the candidates. That is, for state in each trial, I'll generate a Gaussian random variable with sigma of 2.0, and then add that to the current margin between the candidates in that state. I then sum electoral votes across all states and count how often each candidate wins, and I'm reporting both the winning percentage for each candidate and their average electoral vote total.
Recently, Clinton has done better in this simulation than Obama. Both Obama and Clinton are clearly electable, as the current data suggest either would be at least a mild favorite against McCain. So while based on current polling Clinton may match up a little better against McCain, the claim that Obama can't win in November is simply not true. Obama not only can win, but he's favored to do so in each of the weeks that I've crunched the numbers.
Comments (4)
I started doing this because I like the www.electoral-vote.com a lot, but I noticed a few weeks ago that while Hillary's electoral vote totals were higher than Obama's, Obama was close behind McCain in many more states. The simple process of awarding the votes to the current leader ignored the size of a lead, which is something my simulation considers.
I'm actually an Obama supporter: I voted for him and have donated money to his campaign (the first political donation I've ever made). I also believe, despite these numbers, that Obama would match up better against McCain this fall than Clinton would - this simulation is a snapshot of where we are now, not where I think we would be after a general election campaign. But I also agree that Clinton would be a formidable general election candidate also, and I can understand why she, and many of her supporters may think she would be stronger.
May 10, 2008 7:47 PM | Reply | Permalink
The polls regarding Hillary against McCain have no basis in reality. The ones with mcCain and Obama do.
There will be no real bombshells of implications if Obama is nominated---it is what is expected.
But if Hillary is nominated---after losing in the yada yada yada---it would be such a shock to the democrats, that there is no way to compute how how it would affect a McCain Clinton poll.
One matchup is with what is expected---the other is a matchup which would occur after the political equivalent of a nuclear explosion.
May 10, 2008 9:04 PM | Reply | Permalink
Hey Forsberry, this is a very good write up. That said, I will have to disagree with your simulations based on percentage of votes. I don't think Clinton will have a 78 point advantage over McCain. And, more realistically, if something happens and she ends up being the nominee and taking 88% of the populat vote, then there's no way the electoral vote totals will be that close...she will sweep McCain like Reagan did in 1984.
May 10, 2008 10:18 PM | Reply | Permalink
Thanks! (shameless self-promotion): If you like this post, please click the recommend button!
Point of clarification: my simulation doesn't say Hillary would win 78% of the popular vote; it means that she had more than 269 electoral votes in 78% of the trials that I ran. So if the current poll differences in each state stayed the same in November, Clinton would have a little better than a 3 in 4 chance of winning the election, and Obama would win over half the time.
Obviously a lot can change between now and November, and indeed a lot has changed for Clinton in the short time I've been doing this: two weeks ago, McCain beat Clinton over 50% of the time, but now she would win almost 80% of the time in a head-to-head matchup. Their electoral totals are close, so swinging even one or two states can greatly affect how often you win in a trial.
These numbers aren't a prediction, they're a snapshot in time.
But I do think they're a good summary of the most recent polling in each state, better than simply counting electoral votes based solely on who is leading in a given state. Even though McCain currently is leading against Obama in states with 273 electoral votes, Obama wins most of the simulation trials. This is because McCain has far more states with very small leads than Obama does, and in the simulations Obama wins some of those states often enough to push him above 269 electoral votes over half the time.
May 11, 2008 3:53 AM | Reply | Permalink
Post a Comment