PA: Is Clinton-Obama more like Rendell-Fisher? More like Specter-Hoeffel? Or more like Bush-Kerry?

SurveyUSA Breaking News - 04/14/08 10:19 PM

SurveyUSA will release a poll in the Pennsylvania Primary tomorrow, Tuesday, 04/15/08. There is disagreement among pollsters in Pennsylvania. The variation is captured in this graphic created by Charles Franklin, U Wisconsin, and posted to the Pollster.com website.

If you click on the graphic, you will be taken to the Pollster.com website, for the complete context. Graphic does not (yet) include SurveyUSA datapoint from 04/15/08.

PA Poll Volatility as graphed by Charles Franklin at pollster dot com

Within recent days, polls have shown Obama leading by 2, trailing by 20, or somewhere in between.

SurveyUSA is reminded of two recent elections in which pollsters, including SurveyUSA, struggled  in Pennsylvania. In the 2002 Governor’s election, between Democrat Ed Rendell and Republican Mike Fisher, each of the 5 pollsters who worked the race (including SurveyUSA) were too Democratic. Each overstated the Democrat’s margin of victory and, more precisely, each understated the Republican’s strength. Here is SurveyUSA’s scorecard on the contest:

(click on the graphic until legible)

PA 2002 Governor Election Final Polls

Unlike, however, the 2008 Pennsylvania presidential primary, where the pollsters are divergent, the pollsters in 2002 were consistent. Just wrong.

In 2004, pollsters in Pennsylvania were also wrong in the U.S. Senate contest between incumbent Republican Arlen Specter and Democrat Joe Hoeffel. Except in Specter-Hoeffel, each of 8 competing pollsters overstated the Republican’s margin and, more precisely, understated the Democrat’s strength. (Meaning: the pollsters were off in the 2004 Pennsylvania U.S. Senate election in the exact opposite direction as they were off in the 2002 Pennsylvania Governor’s election.)

(click on the graphic until legible)

PA 2004 US Senate Final Polls

The polling is not consistently wrong in PA. The scorecard below from the PA 2004 presidential general election showed a tight clustering, with most of the 13 competing pollsters inside of the margin of sampling error.

(click on the graphic until legible)

PA 2004 POTUS final polls

The question for the next 7 days is whether the 2008 Obama-Clinton primary is more like Kerry-Bush (where the pollsters were close to perfect), more like Specter-Hoeffel (where the pollsters were too far to the right), or more like Rendell-Casey (where the pollsters were too far to the left).

DCCyclone says:

I think it’s looking at it the wrong way to say in the Senate and Governor’s races examined that the pollsters got it wrong in opposite ideological directions. Rather, I think a stronger way to look at it that also reveals consistency in the polling “mistakes” is that the pollsters underestimated the losing candidates’ vote shares. A reasonable explanation for this is that each party in PA has a reliable base vote of just over 40%, and that just 15-20% of voters are persuadable to vote either way. Polling may “detect” a larger share of persuadable voters than 15-20%, but really anything more than 20% are really leaners who “fall back” to one party or the other.

If you start with a presumption that a major party nominee, barring scandal, will almost always top 40% in a high-profile statewide race that is considered at least somewhat competitive, you’re going to be right almost every time.

Garry Ikalina says:

please release them ASAP, i’m excited!

bonncaruso says:

can’t wait to see your results…

Copyright 2014 SurveyUSA®, Clifton NJ, all rights reserved. Terms & Conditions.