Here’s my prediction for election night with audio commentary.
Note that all times are Pacific Time — add three hours to convert to Eastern!
The first six and half minutes are a quick rundown of the methodology behind the prediction. Feel free to skip that if you want to get to the fun part.
As I mentioned yesterday, USC will be posting video of yesterday’s panel discussion. I’ll update this post with a link to that as soon as I get it.
Update: I forgot to give them credit in the presentation, but poll closing times come from the wonderful web site The Green Papers. I’ve assumed that the networks will not call the results for any state until all polling has closed in that state. Some news articles (e.g. this morning’s Reuters article) suggest the we might start to get exit polls/vote counts earlier than my timeline in some states. The confusion comes from the states where polling closes at different times in the Eastern and Western parts of the state.
16 users commented in " Election Prediction Presentation "
Follow-up comment rss or Leave a TrackbackExcellent presentation. I disagree on only one state, Florida, however, I didn’t take into account extended polling time when predicting when enough states will be called for a winner to be determined.
Florida really looks as though it’ll be incredibly close again. I actually think it will take longer to call than the model predicts, and it’s anyone’s guess who’ll win it.
Mainly I hope that it doesn’t prove to be decisive one way or the other…
It’s not every day that a guy goes so far as not only calling states but calling the timeline. A very cool presentation that political journalists around the country will hope is correct — not necessarily because of who wins, but when. Not having to pull an all-nighter for coverage would be nice.
I’m not sure on Florida either, but as Bruce says, those last few are anyone’s guess.
Very nicely done. Good to see my home state of Montana keeping everyone on the edge of their seats until the very end. Can’t wait to see which way those 3 votes will go!
Great presentation. But I highly doubt the mathematical accuracy of your time-line. You’re basing your results off a single past election, and applying that data to the current election.
The variance on each of the graphs you showed is extraordinary, and you might be using variables that are statistically insignificant (fallout from the fact that there aren’t enough data points and history to gather which variables are significant).
Not to say this isn’t interesting, but readers should take this assessment with more than a grain of salt.
[...] «w00t» Tags: No Tags [...]
Brilliantly done, now I won’t even have to watch!
I’m not sure I agree about the last week swing towards republicans, based purely on the 2004 results. Perhaps if it was based on a number of historical elections it would be a more solid argument. I think that the last week swing might be more representative of personalities (of the candidates) in the last week and the momentum of the campaigns.
Otherwise, I thought the presentation was excellent, and I’ll be comparing it to the night itself.
S.J.,
You make some good points. I’m combining two models that both have fairly high variance in the individual results, so some states are likely to fall quite a long way from my prediction. I wouldn’t be making wagers (or dinner plans!) based on, say, an expectation that Virginia will be called for Obama by 8pm.
I think it is useful in highlighting the states that are going to be critical on the night for each candidate — specifically, Pennsylvania and Virgina — and also in giving a good general sense of how the evening is likely to proceed.
One thing in defense of the model: the swings between the later polls and the final result tend to be quite small, so the states that I highlight as “too close to call” for long periods are, I think, almost certainly going to be so. For example, I have John McCain winning Missouri by about 1%. The actual result will almost certainly be somewhere between +3% for McCain and +1% for Obama, and unless the networks change their behavior dramatically from 2004, they’ll be very cautious about calling the result in a race that close.
One other thing… there are about 30 states that are likely to be called almost immediately. That means the projection is pretty stable as a whole, even if the call times for a small number of states are highly variable. So, I think we can be pretty sure everyone will be feeling nervous around 6pm, for example!
[...] The Crunch – Election Prediction Presentation [...]
Bruce – very thoughtful and innovative analysis. You are a mathematical mastermind!
Election night: Make sure you get home early…
My old friend and co-émigré Bruce Nash has put together a great presentation about how tomorrow’s election will unfold, in terms of when the TV networks will call each of the states. From Bruce’s model, it looks like it’ll be……
Great prediction/presentation!
I agree with most of it, except couple of tweaks:
Obama will win in Florida and Montana.
Although it’s quite possible that McCain will win in North Carolina.
[...] if you’re really looking for predictions, stop by the site of Bruce Nash at “The Crunch.” He has predicted not only the outcome of the race, but also when networks will make their [...]
[...] his election-night preview, outlined in a video presentation on his blog, Nash started with CNN’s 2004 presidential-election coverage. He found that states won by [...]
[...] his election-night preview, outlined in a video presentation on his blog, Nash started with CNN’s 2004 presidential-election coverage. He found that states won by [...]
[...] his election-night preview, outlined in a video presentation on his blog, Nash started with CNN’s 2004 presidential-election coverage. He found that states won by [...]
Leave A Reply