Sunday, January 20, 2008

South Carolina Polls: how we stacked up

By my count, nine companies did polls in South Carolina the last week before the Republican primary.

Six out of nine projected a John McCain victory in their final poll. The three that didn't were American Research Group, Insider Advantage, and Rasmussen.

No company got the John McCain/Mike Huckabee/Fred Thompson/Mitt Romney lineup correct but I think that's because Thompson probably passed Romney as late as election day when Romney failed to give the state much attention after winning Michigan.

The two companies that came closest to getting the race right were Mason Dixon on behalf of McClatchy and Survey USA. The former uses live interviewers, the latter uses the IVR technology we do.

Survey USA had McCain winning by four and Romney just a point up on Thompson while Mason Dixon had McCain winning by two and Romney up a couple on Thompson. The final margin was 3.3%.

Clemson, Fox, and us all showed McCain with a 7-8 point lead. We did a more accurate job on the Romney/Thompson dynamic than they did- we had Romney up a point on Thompson in our last poll while they had Romney up by three or four. In Clemson's defense their poll was done over the course of a week so some of their data was older than ours.

The ninth company, Zogby, got the McCain/Huckabee closer than we did but was off on the Romney/Thompson dynamic, showing Romney up four.

I think if we do a Democratic poll this week we'll probably do it on Thursday instead of Wednesday- I imagine we would have been closer on the McCain/Huckabee margin doing it a night later.

Still I'd say we were there with Zogby right behind Mason Dixon and Survey USA for accuracy out of the nine companies doing South Carolina, so not too bad. And the fact that both us and Survey USA did a pretty good job is a positive point for the technology we use- I guess not too many 14 year old males answered our surveys and claimed to be senior citizen women as folks like to claim happens when pollsters use IVR.

No comments:

Web Statistics