Monday, November 03, 2008

Prognosticating

I'm not going to post any election predictions because, while I don't believe in the Bradley effect, I am a bit wary of the bandwagon effect.

But here is something to mess with your minds.

The polls have fluctuated a bit, but Obama has been ahead nationally by 2 to 10 points and is, by poll average, ahead in enough states to win comfortably in the Electoral College.

But what if there is a systematic bias in favor of the Democrats (as there were in the exit polls in 2004) and a late surge for McCain. Libertarian blogger Vox Day (who is not voting for McCain) thinks you have to adjust the polls by 5% in favor of the GOP. I certainly wouldn't think there is any rule to that effect (I don't think that pollsters generally overstate the Democrat vote by 5%), but, if the poll results are affected by assumptions regarding turnout and party identification that don't hold up, it could happen.

Blogger Baseball Savant, tries that adjustment, using 538.com's projections. McCain wins with 274 electoral votes. If you use the RCP averages, you can get McCain to 271, but only if you ignore a half point in Colorado.

Of course, for this to happen, the error would have to be systematic. There would have to be enough polls with a methodology skewed in the same direction to have significantly pushed the averages to the left. (Note: I am not talking about the statistical margin of error here. That's another thing altogether.)

Testing this theory would require more work than I have the time or inclination to do. For example, how do the adjustments for each poll work (I suspect that all of that info isn't even publicly available) and is there any poll that shows a McCain win or a collection of states that exceeds 270 within the statistical margin of error or with a McCain lead? Fox/Rasmussen is close but not quite there. IBD/TIPP is within the margin of error nationally, but I am unaware that it does state polling.

So, you'd have to have a systematic Obama error in almost all of the polls and a surge to McCain.

This may be unlikely, but it's not impossible. We probably won't know more based on leaked exit polling because those are likely to overstate Obama's vote as they overstated Kerry's.

11 comments:

Anonymous said...

During the 2006 WI Gov's race most polls had Doyle up 4 to 6%. Yet he won by 10. These same pollsters have Obama up 8 to 12%.

In 2006 only the always left UW Poll and the Wisconsin Public Radio poll had Doyle in double digits.

If the polls lean left we didn't see much of it in 2006.

Anonymous said...

McCain's belief in the Surge will be his downfall.

Anonymous said...

Agree with your overall analysis, that a McCain victory is possible. In order for McCain to win, he will have to win the undecideds in overwhelming fashion, along with a little poll bias. Anything less will mean an Obama win.

Jay Bullock said...

I like pollster.com; I used their numbers in laying out McCain's map late last week. I spotted McCain 6 points of poll bias, and Obama came out ahead in electoral votes, and slightly ahead in the national vote. The numbers were not much different this morning.

Rick Esenberg said...

Using Pollster's numbers, six points of poll bias (and either surge or undecideds breaking for McCain) would leave you with a toss-up. Applying the rule to the decimal point would result in a 278-267 Obama win.

Anonymous said...

anon 4:16
What a dumb comment. The surge worked...even O admits that.

Anonymous said...

@Anon 5:11
Reference was to Rick's last gasp hope for a Surge in McCain's numbers as the campaign wraps up. Perhaps you noticed he used the term in his post vis-a-vis the election? Hello?

Anonymous said...

anon 12:40

oops. my bad. sorry.

tom paine said...

Want to make it an early night for the election results?

If McCain does not win PA, VA, and NC it's a done deal for Obama and you can go to bed early.

If he does win all three it might take a tad longer.

tom paine said...

Rick, you obviously are a bright guy but it seems that your lawyerly training and your bias get in your way sometimes. Consider how difficult you made it to predict this election and your cockamamie theory on how it might work in McCain's favor. Here are just some of the words and terms you used:

"...systematic bias in favor of the Democrats...a late surge...Libertarian...identification that don't hold up...Blogger Baseball Savant...538.com's systematic error...methodology skewed...significantly pushed the averages...statistical margin...IBD/TIPP..." to name a few.

Rick, Rick, Rick...you should have been able to figure out that if Obama was able to defeat a skilled political machine like the Clintons that going against 72 year John McCain and his 91% record of voting with President Bush and McCain's choice of Gov. Hockey Mom Moose Hunter (but who can see Russia and lives atop oil) was destined to flame out.

Poor Sen. McCain probably felt like a one-legged man in an ass kicking contest!

tom paine said...
This comment has been removed by the author.