Analysis: Election Night Prediction
November 1st, 2012
Update: See the end of the post for an update based on the final set of polls. Also, check out my election night live blog, where I'm analyzing results as they come in and the performance of the model.
In a break from our usual programming, here is my election night prediction for 2012.
Rather than predicting the result in each state, my prediction focuses on something slightly different: when each state will be called by CNN on election night. This is useful information, I think, for political junkies who are curious if and when they'll feel comfortable enough with the results to go to bed on Tuesday (or the early hours of Wednesday!). Since we're in for a close election, it also might tell us something about how likely it is we'll end election day without a clear winner, and if that will set us up for a long legal battle over who will emerge triumphant.
Click the video below to see the presentation, and see the full article below the fold for links to download the spreadsheets on which model is based.
To try out your own prediction, you can download the election night projection spreadsheet here.
The yellow fields in the spreadsheet can be changed to see how differing margins in each state might affect the time that CNN will call the state. You can also change the model itself by varying parameters at the bottom of the sheet (use with caution!). Finally, you can sort on Column O in the spreadsheet to reorder the results in the order that they will be called by CNN.
The spreadsheet containing the data on the margins in states in 2004 and 2008, and the times at which CNN called each state are in this spreadsheet. I didn't have time to mention it in the video, but in the spirit of full disclosure, there are a few outliers that I excluded from the model because they threw off the projection in one way or another (in particular, Missouri's long wait for a result in 2008 and the tendency for Mississippi and Louisiana to be called after long waits even when the margin between the candidates is quite large).
Finally, my election night prediction for 2008 can be found over at the old Crunch blog.
As I mention in the presentation, I'd love to get feedback and comments. As ever, my email address is firstname.lastname@example.org.
Election Day Update: I have updated the projection based on 538's final set of forecasts. Assuming Mr. Silver's projections and my model turn out to be accurate, we should get a final call as "early" as 2am, Eastern Time. I'll be back with more tomorrow, but for now here's the spreadsheet with the final projection based on 538's numbers. For Romney fans, I also added the most plausible path to a Romney victory (again, more on that tomorrow).
Election Day Update 2: Mr. Silver keeps on tweaking his forecast, so I've updated the forecast again based on his (final, maybe, please 538) 10.10am ET projection. The model now projects a final call in the race at 2.23am Eastern.
Election Day Update 3: I'll be live blogging election night and offering analysis of results as they come in starting as the first results come in at 7pm Eastern, 4pm Pacific.
Filed under: Analysis