Saturday, October 18, 2008

Chickens and Eggs

There was an interesting article in the Wall Street Journal today about polls and party id weighting.

We are not weighting for party id this cycle. Although I don't think there are a lot of voters whose party id fluctuates by the week, month, or year I do think there are enough- probably no larger than 10% of the population- whose id is malleable enough that we're not comfortable pegging our polls to any specific figures for party.

We have found an increase in party id for Democrats in almost every single state this month as the race has moved toward Barack Obama:

State

Pre-Convention Margin

Pre-Convention Party ID

Current Margin

Current Party ID

Colorado

Obama +4

R+1

Obama +10

D+4

Florida

McCain +3

D+3

Obama +3

D+6

Michigan

Obama +3

D+3

Obama +10

D+7

Missouri

McCain +10

R+1

Obama +2

D+6

North Carolina

McCain +3

D+15

Obama +3

D+13

Ohio

Tied

D+7

Obama +6

D+10

Virginia

Obama +2

D+6

Obama +8

D+5

Average

McCain +1

D+4.6

Obama +6

D+7.3


On average Obama is doing seven points better than he was in August, but we're also finding almost a three point gain in Democratic party id. The two exceptions to the rule are Virginia and North Carolina. These are both southern states where a lot of self identified Democrats have tended to vote Republican for President. A great deal of Obama's forward movement in those two places over the last couple months has been the product of significantly increasing the share of votes he gets from folks in his own party.

Take those two states out of the analysis and here's your new chart:

State

Pre-Convention Margin

Pre-Convention Party ID

Current Margin

Current Party ID

Colorado

Obama +4

R+1

Obama +10

D+4

Florida

McCain +3

D+3

Obama +3

D+6

Michigan

Obama +3

D+3

Obama +10

D+7

Missouri

McCain +10

R+1

Obama +2

D+6

Ohio

Tied

D+7

Obama +6

D+10

Average

McCain +1.2

D+2.2

Obama +6.2

D+6.6


On average Obama's standing has improved by 7.4% and there's been almost as large an increase in Democratic party id, 4.4%

So here's the big question. Are more people identifying as Democrats because more people are voting for Obama? Or has party id remained flat over the last two months, in contrast to what we've found in our polls, and Obama is just doing better because we're over sampling Democrats?

The latter is what a lot of Republican polling pundits would like you to believe. I think the former is true. I think as voters move toward one candidate or another, they're also likely to move toward identifying with that candidate's party. Again, I don't think it's a huge number of voters shifting their party preferences, but I do think it's enough that weighting to predetermined party figures could keep your polls from reflecting real movement in the electorate.

I guess we'll find out who's right on November 5th.

2 comments:

Andrew said...

I wonder if some of this Party ID change depends on the response rates to pollster calls. Perhaps members of one party tend to ignore calls from unknown numbers or wierd names on their Caller ID - but these people may very well come out to vote.

Alan Reifman said...

I maintain a website devoted exclusively to sample weighting by Party ID.

http://courses.ttu.edu/hdfs3390-reifman/weighting.htm

 
Web Statistics