Tuesday, May 27, 2008

Self Identification of Republican Voters

I think we're going to have a lot of folks voting Republican but not calling themselves Republicans in North Carolina this year.

In our most recent survey, voters identifying themselves as Independents overwhelmingly went for the Republican candidates- 48-31 for John McCain over Hillary Clinton, 48-33 for McCain over Barack Obama, 54-34 for Elizabeth Dole over Kay Hagan, 50-36 for Pat McCrory over Bev Perdue, and the Republican candidate leads among independents in every other race as well with the exception of Roy Cooper, Elaine Marshall, and Beth Wood.

What's the explanation? I think there are a lot of Republican voters who are pretty disgusted with their party right now. They're going to keep on voting Republican because they certainly don't think the alternative is any better, but they're not calling themselves Republicans.

This is one of the issues with weighting for party. We might have an idea of what the party distribution of voters in the state should be, but how does that account for folks who would have called themselves Republicans four years ago and are continuing to vote that way but not to identify themselves that way?

Pew found earlier this year that there was a 13 point Democratic edge in party id in North Carolina. We've been roughly weighting our NC polls to that party distribution. SurveyUSA, which doesn't weight for party, found an even more Democratic electorate in the state in their last poll- 49 to 33 to 15. If we had not weighted for party in our tracking poll two weeks ago, we would have shown a 17% advantage for Democrats in party identification.

I still haven't made a final decision but because of the uncertainty with party self identification in this most unusual election cycle I am leaning more toward not weighting our general election polls for party the rest of the year.

1 comment:

Anonymous said...


Remember that party id and party registration are two very different things. You can have registered Democrats that if asked identify themselves as Republicans and vice versa.

Since you poll using a voter file you have the party registration of each respondent at your fingertips. You probably should anticipate approximately the same turnout among each party identification on a percentage basis as you saw in 2004. (Those numbers just like African-american turnout you discussed last week are available from the SBOE.)

Since you already know how those polled are registered, the interesting thing is to see what they are going to do.

If you are interested in how they are identifying themselves, you could ask them whether they identify more with the Republicans, Democrats or neither party. You could also give them a generic State Legislative ballot question to determine which party they lean towards. (Since most folks have no clue who their legislator is they're pretty likely to vote the party they currently favor.) Or you could ask them if they are Democrat, Republican or Unaffiliated and see how it matches up with the voter file.

In any event, you are negating one of the major advantages of polling off a voter list if you are trusting the respondents to tell you if they are Republican, Democrat or Unaffiliated when you already have that info.

I'm sure there are, as you point out, many registered Repubicans that are disgusted with the Republican Party and intend to vote Republican that may no longer consider themselves Republicans and tell pollsters they are independent. And you are right, they would really screw up your sample of unaffiliated voters.

Finally, it should to some extent reduce the need for weighting your poll which as you've pointed out on numerous occasions can introduce the error that causes a poll to be off in the first place.


Web Statistics