Predicting Electoral Vote Distribution

One interesting use of Tradesports data is to derive things from their state-by-state predictions. Geekmedia has done this with a page that automatedly derives a projection of total electoral votes. They chart the chances on this map:
Presidential Election 2004 Tradesports Price Map (with transparency)

While the volume on the state contracts is not high, so the prices may not be accurate, this is still an interesting exercise. But it would be nice to also answer the question "What is the distribution of electoral votes that Bush may get?" Tradesports has contracts for Bush > 200, Bush > 250, Bush > 300, and so forth. So I wrote a script which takes the Tradesports state data and performs a Monte-Carlo simulation. I'll spare you the full results, but here are the predictions for totals which have Tradesports contracts.

Bush EVs Model Says TS Price
> 200 95.9% 92.5
> 250 67.8% 76.5
> 300 13.2% 49.5
> 350 0.3% 17.0
> 400 0% 6.6
> 450 0% 4.0

The main differences are in the high end. Essentially, my distribution has thin tails, and the market distribution has fat ones. There is an obvious explanation for this: correlation between the state-by-state outcomes. My model assumes that each state votes independently. But of course, part of the 43% chance that Bush will win Arkansas is the 5% chance that we get some great news from Iraq. If this happens, other states will also shift towards Bush.

Unlike primaries, election voting is pretty much simultaneous (exit polls change this a little), hence the election morning state-by-state prices can be treated as independent. But the September prices cannot, because state chances are likely to move somewhat in tandem. To estimate the strength of this effect, we'd want historical state-by-state prices through many election seasons. Then we could calculate the correlation in state price movement. A quicker but less accurate way to do this would be to just look at the correlation in Tradesports price changes during the current season. I have not done this.

Despite the caveats, I have trouble believing the tails are as fat as indicated. My model is surely wrong about +300, but it would take a huge simultaneous movement to make +350,+400, or +450 happen. It seems unlikely to me that there are 17%, 6.6%, and 4.0% chances of this happening.


Update: Many of you are likely visiting this website from google. Welcome! Take a look around. We are a blog focused on philosophy, politics, economics, science, law, and more. Take a look around, especially at the "Popular Entries" on the right sidebar. Share this

Nice article, Patri. THing

Nice article, Patri. THing is, though, that the betting at http://www.electoral-vote.com, while swinging back and forth with changing poll data from day to day, is trending out to be Bush: 335, Kerry: 192 on Nov 2nd. These are solid trends that are about two months in building, so unless something really bad happens to Bush, like he pulls a bin Laden out of his hat, but some bureaurat leaks that he's been in detention for months now, or Zarquawi executes a whole bunch of hostages the night before the election, Bush is gonna break 300 pretty easily, I think.

The problem here is the Dole Effect: Dole lost in 96 because people didn't know what he was for, or what they were for, only what they were against, and that Dole was a pretty grumpy guy. People don't go to the polls that often out of spite. Thing is that even the people who are voting for Kerry generally can't stand him as a person. Why would they really want to trade one president they can't stand for another they can't stand?

Mike - this doesn't make any

Mike - this doesn't make any sense. When I go to http://www.electoral-vote.com (which was part of my research before betting on Tradesports), I see:

a) no betting
b) Bush 273, Kerry 255, 10 undecided.

Were you talking about some different site? Or are you talking to me from a parallel universe :razz:? Or maybe someone is spoofing my net connection to make me bet poorly...

Did you paste the wrong URL?

No, if you look at the link

No, if you look at the link on electoral-vote.com that shows the final projected electoral vote (as opposed to the main display, which is the vote if the election were held today), you'd see a rather significant difference because the site takes all the polls over the past months and plots least mean squares trend lines to project the future results on Nov 2nd, which is now about 336 to 192 in favor of Bush.

Of course, the polls from state to state tend to be rather noisy, but most are showing distinctive trends (and of course states like California seem to be on another planet entirely) that have not been changing significantly in the past week, except change in favor of Bush. The site does caution not to count on these projections until we at least reach October, and of course, Bush is free to screw things up, or have things screwed up by some event outside his control, but it looks pretty favorable for Bush.

At this point, the betting for the Nov 2nd results, according to electoral-vote.com, should be Bush by a landslide of electoral votes, over 300 pretty easily. Whether he holds onto it depends on what the October Surprise ends up being this time...

Oh, and no, there is no

Oh, and no, there is no betting on electoral-vote.com. I would tend to trust futures markets more if they had more participants than a few hundred self-selected policy geeks. The market you are betting in is not representative of the population as a whole, and IMHO is letting its own California biases get the best of it.

Okay, looking in depth at

Okay, looking in depth at Tradesports exposes to me the weakness: it doesn't rate absolutely safe, non-battleground states as being 100 or 0 sure bets. C'mon, if Bush is polling 70% in Texas when the rest of the country has him barely ahead of Kerry in the popular vote, the guy is gonna get Texas, so stop rating that state as a 96, okay? It's a freakin 99-100.

All that Tradesports is doing, really, is translating the percent scale into a bell shaped log scale, for finer grained rating of states outside a 50-50 result.

Moreover, it isn't factoring in the impact of 3rd party votes leaching support: anti-war republicans and independents, anti-Kerry liberals, they could all spoil things one way or the other. The Greens are huge in California. They might drain enough votes from Kerry to cost him a state that may view him as nothing but Bush-lite with better enunciation. Badnarik has more chance of taking New Mexico from Kerry by leaching anti-war independents and Republicans who can't stomach Kerry's east coast elitism.

Ah, I hadn't seen that

Ah, I hadn't seen that projection. I'm pretty skeptical of it. If you look at some of the close states, the data is very noisy and the trend lines don't look very trustworthy.

Oh, and no, there is no betting on electoral-vote.com. I would tend to trust futures markets more if they had more participants than a few hundred self-selected policy geeks. The market you are betting in is not representative of the population as a whole, and IMHO is letting its own California biases get the best of it.

What are you talking about? Tradesports is based in Ireland, and I doubt very many of the bettors are from California.

As for self-selected policy geeks, Tradesports, as its name implies, mainly deals with sports betting. Other than the presidential election, the most popular contracts are generally about sporting events. The betting on the presidential election has been quite heavy, and I very much doubt that its restricted to a few hundred people.

Okay, looking in depth at

Okay, looking in depth at Tradesports exposes to me the weakness: it doesnâ??t rate absolutely safe, non-battleground states as being 100 or 0 sure bets.

The fact that you see Texas trading at 96 is not an indication that Tradesports is wrong. There are several issues. First, that just indicates that the most recent trade was 96. But if everyone agrees that Texas is a lock, there is probably no or almost no trading on Texas. That 96 might have been one idiot taking a chance, or it might have been placed a year ago when there was more uncertainty.

Second, there are margin issues that keep Tradesports prices from being pushed to the edge. If a contract is trading at 97 and should be at 99, you have to lock up $97 in your non-interest-bearing Tradesports account to make those $2. If it takes more than a couple months, its not worth it.

Tradesports prices should only be viewed as meaningful when the odds are not close to 0 or 1, and when there is substantial betting volume.

Moreover, it isnâ??t factoring in the impact of 3rd party votes leaching support

How can you possibly know this? The Tradesports prices are the aggregations of a large number of opinions. Yet they are a single number. Why do you think you know what the traders have and haven't taken into account?

If you are convinced that Tradesports bettors are not taking things into account, why not put in some money and start betting? I've put my money behind my opinions.

Let me just put it this way:

Let me just put it this way: I will be EXTREMELY surprised if Bush gets less than 290 electoral votes, and I put even money on 300 or better..

Mikes current odds:

290 for Bush: 80%
300: 50%
310: 40%
320: 30%

If Tradesports is very widely used, it suffers from two things: a) its bettors are focused on sports betting and not political betting, ergo their instincts are off, and b) forget the California Effect, look at www.us-election.com, which registers worldwide sentiment toward the election. If Tradesports bettors have a large fraction of non-American bettors, the information sources that those people are all focusing on are their own national, government-cacked-up, anti-American anti-Bush propaganda mills. Their anti-US bias is creeping into the mix, because they are not basing their opinions solely on the media that the American Voter sees. That 67 bid on the Bush win I saw yesterday should be even higher. I'd say one should equate every 1% spread in the projected popular vote with 5 points on the betting. A 67 bet for Bush is assuming that he is only going to get 53% of the popular vote.

If you want to make money in this election, bet on Bush on Tradesports. I'm not a betting man, nor do I have the free funds these days to do so (those are focused on the FSP, which you might call a surer or longer bet, depending on your outlook), but if I did, I'd be betting on Bush. I would say that a more realistic bet on Bush should be in the order of 75 or so, maybe more.

While you may be right that

While you may be right that Tradesports is wrong, it seems a bit strange to me to jump to such vitriolic conclusions about why. It could just as easiy be American Kerry fans keeping the price down.

And how can you know that a 1% spread in the popular vote => 5 opinion in the betting? That may be true, I don't know. But I don't think you know either. You are assuming it, and then taking it as a sign against Tradesports that its prices disagree with your assumptions. Not a particularly fair evaluation.

The winner of the election (which is what Tradesports 67 price is about) is really not about the popular vote. Its about who wins the battleground states. There is still plenty of time for events to happen which shift popular sentiment, and it doesn't take much of a shift to affect those key states.

Its easy enough to come up with ideas why markets might be wrong on any particular issue. But they have an excellent track record at prediction.

:idea::kiss::oops::roll::wall

:idea::kiss::oops::roll::wall::wink::eek::end::lol::mad::sad::no::beatnik::behead::gossip::juggle::stupid::sweat::???::bomb::cool::cry2::dead::deal::evil:

I agree with your projection

I agree with your projection of electoral votes. I reached mine differently. I added and or subtracted a percentage to offset the media bias. and came up with 335-370 electoral votes for Bush. So I agree with your assesment..

Let me just put it this way:

Let me just put it this way: I will be EXTREMELY surprised if Bush gets less than 290 electoral votes, and I put even money on 300 or better..

Mike, I respectfully request that you hang up your spurs.