Wednesday, November 5, 2008

PPP's 'Bias'

PPP is a Democratic company, no doubt. We do private polling for Democratic candidates and we were rooting hard for Obama. But that never had any impact on our polling results. I saw people on various GOP blogs always say they were going to add five points to the Republicans for any poll that we released, and as I said a couple of months ago, that's fine with me but if you let that shape your expectations you're going to be sorely disappointed on election day.

So how much of a lean did our polls have? Based on the current numbers we overestimated Obama by a whooping one eighth of a point across our 16 final week polls. And I expect Obama's margins to increase in Oregon and Colorado as more returns come in from Multnomah and Boulder Counties respectively, so our bias may go down to zero or even a one eighth of a point McCain bias if the Democratic lead increases by a couple points in each of those states in the final analysis.

Here's the data:

State

Final PPP Poll

Current Results

PPP Bias

Nevada

Obama +4

Obama +12

R+8

Pennsylvania

Obama +8

Obama +11

R+3

Michigan

Obama +13

Obama +16

R+3

Ohio

Obama +2

Obama +4

R+2

Oregon

Obama +15

Obama +16

R+1

North Carolina

Obama +1

Obama +1

None

Indiana

Obama +1

Obama +1

None

Florida

Obama +2

Obama +2

None

Missouri

Tie

Tie

None

West Virginia

McCain +13

McCain +13

None

Virginia

Obama +6

Obama +5

D+1

New Mexico

Obama +17

Obama +15

D+2

Georgia

McCain +2

McCain +5

D+3

Colorado

Obama +10

Obama +7

D+3

Montana

Obama +1

McCain +3

D+4

Minnesota

Obama +16

Obama +10

D+6

Average

Obama +5.0625

Obama +4.9375

D+.125

11 comments:

Unknown said...

Congratulations...
You guys were the best pollster in the 2008 on state polling... Unbelievable. You polled so many state and only got one wrong which was within the margin of error... And most of the states nailed the exact margin...

mrbossman said...

Excellent job. I wonder if people will pay attention to the fact that your polling showed no "house effect" whereas Rasmussen's state polling show a 3 point Republican lean.

jr1886 said...

NBC news just confirmed Obama is the winner in NC. Congrats PPP for being the best pollster this election cycle. You guys beat Rasmussen handily and I still like SurveyUSA though they could have predicted NC too if they didn't peg AA turnout at 20%...PPP got the turnout right, it was 22%.

Tom,
Why doesn't reacleearpolitics.com include your polls in its average???

Anonymous said...

I believe Obama won FLA by 3 not 2, so that would be a +1 Rep bias in your final FLA poll.

Anonymous said...

Tom, why do you think that so many pollsters (including you) had NV a lot closer than it really was?

RS said...

Rasmussen looks to have nailed the popular vote. But I didn't want to say "yay" because Ras could be right for the wrong reasons...
So I was comparing the exit polls with your internals for NC - it didn't look that good (e.g. exits showed a more typical advantage among women for Obama while you said almost-equal). Then I remembered the warnings on exit polls...

I wish there were better ways of looking at pollster performance than just the toplines. Perhaps a comparison of all state polls against the final tallies? You can be "right for the wrong reasons" on some of them, but not for all of them, surely.

So in that sense, is there a better way of looking at your topline data? Say, a regression of actual vs PPP-finals, rather than the average? Simple Excel regression on this table says slope 0.96, r2 0.84 - so the final margins were within 4%, and very well correlated.

Or something else?

Still - it looks like you did a great job. Congratulations!

Anonymous said...

To rs: I think rasmussen may be celebrating too soon with its pop vote call. Looks like all the remaining votes outstanding (literally 2% of nationwide vote is still outstanding) come from Dem heavy precints (Boulder and Adams county in CO, Portland and Eugene in OR, etc.,). So I suspect that Obama's final pop vote total will be somewhere between 52.6 to 52.7% and McCain's to be either 46.0 or 46.1%, which means using rounding 53% Obama to 46% McCain. Looks like CNN nailed it this time not Rasmussen.

Valdivia said...

Thanks Tom on the night of the election I kept saying to anyone who would listen that you were very much on the money. Great job!

Anonymous said...

Congratulations Tom,

You guys have been the best pollster overall since just before the Indiana/NC primaries.

Unknown said...

Yes CNN nailed the popular vote margin, not Rasmussen, they were close but, like usual Ras had their R lean, especially in their state polls...

Anonymous said...

Great job, Tom. Your polls were always fun to look forward to and you guys always had nice big samples!

Looking foward to your 2010 and 2012 polling!

=)

-Josh

 
Web Statistics