Tuesday, September 12, 2006

Casey Lead Down in New Zogby Poll

Among the things we ignored on yesterday's anniversary . . . .

A Zogby online poll taken over the holiday weekend shows Santorum's support virtually unchanged from a month ago. However, Casey's numbers drop from 51% to 47%. The MOE is 3.3, potentially putting this race at a virtual dead heat -- according to Zogby, at least. (Statisticians will also tell you that, aside from the dead heat possiblity, the MOE could well mean that the "real" numbers are 50-40.)

I am guessing that there is some skewing in here due to the holiday weekend and (perhaps) increased involvement of Republicans in the Zogby survey. Crosstabs are not available, but Zogby has taken the unusual (for them) step of using spot telephone checks to verify information provided by the online poll takers. According to the Zogby special methodology report issued with its news release, they attempted to contact about 2% of the online respondents.

I am on their online interactive polling list and was invited to participate in this poll. I did not because, well, it was the holiday weekend and I just didn't make the time. But, it is unlikely that I would have given them a correct phone number (if they asked for it) when I registered. It was even more unlikely that I would have been around to answer the phone over the Labor Day weekend (unless they paged me at Yankee Stadium). I suspect that I am not unusual in that regard.

As to increased Republican involvement, I have no evidence, only anecdotal reports of solicitation of conservatives to participate in electronic polls. This should not affect the polling if Zogby correctly adjusts the results to properly balance the results to reflect the Commonwealth electorate at large. But, since the WSJ/Zogby polls don't give the essential data, we have no way of knowing.

As a general matter, I don't have any less faith in the accuracy of interactive online polls. Many of the allegations about the efficacy of online polling were also asserted against telephone polls. However, I am always troubled when the crosstabs are not provided, as here. It is also curious that Zogby took the additional step of issuing a special methodology report with the press release. It suggests to me that there is an issue with the polls which they are trying to address.

These results don't trouble me terribly. Santorum has not made any meaningful progress in his numbers since the start of the campaign. Although he has moved from polling in the upper 30s to the low 40s, he can't win with 43%.

Not to put too fine a point on it, if Santorum were the challenger, I'd say he was in good shape. But he is a multi-term incumbent. He is tied at the ideological hip with the President, who is given very low marks in Pennsylvania. His personal favorable/unfavorable has been the worst, or near the worst, of all US Senators for over a year. A huge chunk of Pennsylvanians see this country as headed in the wrong direction -- which is the very direction in which Rick Santorum has been going. Finally, Santorum has eeked out an entire 1% increase in his polling numbers after reportedly blowing several millions of dollars on an intensive early advertising buy -- wiping out his cash advantage for the rest of the run in the process. Mabel, keep that butter handy.

That said, Casey had best soon figure out why he isn't wiping the floor with Santorum.

2 comments:

Dale said...

Actually, a statistician would probably tell you that there is a 95% probability that the actual scores fall within the specific ranges [54.3-47.7% for Casey; 50.3-43.7% for Santorum]. It is therefore quite likely that Casey's real score is 50%, but very unlikely that Santorum's was as low as 40%. It is possible that Santorum is tied with or ahead of Casey, but it is far more likely that Casey enjoys a small lead.

Santorum's position has improved significantly over that past few months.

Casey isn't running away with this thing because he's a terrible campaigner whose moderation does not fire up the Democratic base and in an off-year election that's the name of the game. The winning strategy is to suppress the moderate and independent votes and to fire up your base. Casey's strategy is more appropriate to a presidential election year in which there will be a large turnout of independents and moderates.

A Big Fat Slob said...

Yeah, I've been told that polling MOEs bear little relationship to the stuff actual statisticians do.

In any event . . . it is still hard to see where Santorum's position has "improved significantly". In last month's Zogby, he pulled in 42%, this month, 43%. In January's Zogby he was at 41%.

Taking the last five polls, Casey's average lead is 11%. I don't see how Casey fails to win by less than ten points.

I am willing to bet that Santorum's internals show double digits -- else he would not have run out that horridly negative ad about Casey contributors (which it is all too easy to hold up to a mirror).