RNC bounce smaller than DNC bounce!

Some of you may be familiar with my meta-analysis of state polls. This is a calculation in which I use the last three available polls in close states to calculate the probability of a Kerry win there. I then go through every possible combination of states to calculate the overall probability of winning the electoral college. A number of election prognosticators follow a similar procedure. My current result is Kerry 279 EV, Bush 259 EV, which is much more encouraging than national opinion polls.

Following the history over time reveals something interesting:

One interesting aspect of my calculation is that even though it is a little slower to react (because of the averaging), it does a very good job of smoothing over all the variation that can drive a poll-reader crazy. This is because at any given moment I am using over sixty polls.

A graph over time is interesting. You can see that big events had clear effects in terms of electoral votes. Such events included the F9/11 release, John Edwards, and the Democratic convention. This caught my eye: in terms of electoral votes, the Democratic convention produced a bounce of 30 EV, but the Republican convention only produced a bounce of 10 EV!

This is a very different picture from what national polls tell us. I am not sure why this is, but a plausible explanation is that Kerry and his convention appealed to battleground states, while Bush and his convention rallied the Republican base.

Sam Wang
Princeton University


Display:


Correction (none / 0)

I wrote in haste. Many poll-followers (such as Jerome Armstrong and Chris Bowers, of course! see the main page) look at the Electoral College. Mine is different because it uses multiple polls to gauge certainty of a state win.
by mindgeek on Sat Sep 11, 2004 at 12:23:11 AM EST

Re: Correction (none / 0)

I fixed your graph, very interesting analysis.
by Jerome Armstrong on Sat Sep 11, 2004 at 02:09:04 AM EST
[ Parent ]

Ohio (none / 0)

Yesterday, you had Kerry with a 64% chance to get Ohio.  (I think it was 64%-60something).  I'm not sure how you got that, since there hasn't been a poll where Kerry was leading in Ohio in three weeks or so, to my knowledge (although there were two ties-both Columbus Dispatch mail-in polls).  Now you have him with 15%, which I still think is too generous.

by Geotpf on Mon Sep 13, 2004 at 01:23:27 PM EST


You are not logged in.

In order to post a comment, you must be logged in. If you have a member account, please log in to comment.

If not, you can make an account right here. It's quick and free.