Wednesday, November 4, 2009

Good election for IVR

As Taegan Goddard has already noted, last night was a good one for IVR polling.

In New Jersey the pollster.com trend showed Jon Corzine leading Chris Christie 42-39 on live interviewer polls. On automated polls Christie led Corzine 44-41. There was never an automated poll showing Corzine in the lead for the entirety of the New Jersey race. On the final round of polls we showed Christie at 47%, Rasmussen had him at 46%, and SurveyUSA had him at 45%. No live interview poll showed Christie higher than 43% and one even had him as low as 36%.

Similar story in Virginia, although not to the same degree. The pollster.com trend for IVR polls had Bob McDonnell up 55-42 to a 53-42 lead with live interviewers. I think it needs to be noted particularly though that SurveyUSA (18 point McDonnell win), PPP (14 point McDonnell win), and Rasmussen (13 point McDonnell win) all came closer to the final outcome than the Washington Post poll, which had McDonnell winning by just 11. And if they want to make the argument that's because their final poll was released a week before the election, PPP had McDonnell up 15 and SurveyUSA had him up by 17 in two polls conducted with roughly the same field period.

The Post wrote a remarkably haughty column in May about why they won't cover IVR polls. Now I know the real reason for their refusal- they just don't want their readers to know our polling is more accurate than theirs. It will be interesting to see if anyone chooses to reconsider their poll reporting policies in the days ahead.

9 comments:

Rational J said...

Do you call cell-phone only households? Could that be the reason for the + R advantage you found in your polls?

Anonymous said...

Careful.

You totally botched NY-23 and were off badly in terms of measuring the President's approval rating in NJ (you had 47%), exits showed it to be a much healthier 57%.

Looks to me like you just weighted better in NJ than some others.

Also means we should take your and -- and Rasmussen's -- coverage of the President's approval rating with a heavy dose of salt until you figure out what you're doing wrong.

Tom Jensen said...

Assuming that the exit poll was correct on Obama's approval and we weren't when we got the horse race correct is a leap. I don't think we're doing anything 'wrong' in measuring his approval, just sounds like you're a poll cherry picker.

Russ said...

No need to cherry pick. It's kind of a no-brainer that I'd trust the final wave of exit polls over any survey done before the election.

Oh and I made a mistake, it appears. Your final poll for NJ had the President's approval at 45%. A lot of the rest of your internals were flaky as well.

Sorry, but that's an issue that should be looked at. If you don't feel that way well, oh well. Not much I can do about it. If all you care about is the final result, then why release the other sample questions/all internals?

Tom Jensen said...

Russ,

I don't think you understand poll weighting. The entire poll gets weighted, not just the horse race. In our last poll Christie was up 47-41 with a group of voters giving Obama a 45% approval rating. If we had found a 57% approval rating for Obama the Corzine/Christie numbers probably would have been reversed.

Russ said...

So then how do you explain the discrepancy between the vote number and the approval numbers?

Christie beat Corzine among a group of voters who have a 45% approval of the President.

Or:

Christie beat Corzine among a group of voters who have a 57% approval of the President.

I think that's fairly significant and should be looked into.

Tom Jensen said...

I'm sure it's some sort of methodological difference between exit polling and IVR. Ask Blumenthal to figure it out.

Mike Mokrzycki said...

Not to pick a fight but it's a little funny to see Tom allege cherry picking when he omitted any mention of the NY-23 "Hoffman leads big" debacle. Then again, because I like my cherries with context, that was an awfully volatile situation. And Tom also neglected to mention PPP was the only pollster of any stripe to have Yes at all ahead on the gay-marriage prop in Maine.

Mike Mokrzycki said...

re Obama job approval in NJ, it's hard not to come to the conclusion that PPP was an outlier, whatever the reason. Per data collected at Pollster.com, no other poll in the past month+, IVR or otherwise, had Obama at lower than 52% approve, compared to 45% in the last PPP. I don't buy that only a lower Obama job approval # could explain a Corzine defeat -- with Corzine's own approval as low as it was and the reasons for that, he was beyond Obama's help.

 
Web Statistics