Saturday, November 17, 2012

Joel Benenson: Gallup Poll Needs to Retool

     Benenson was Obama's lead pollster and he certainly hits the mark on Gallup.

     "President Obama’s lead pollster, Joel Benenson, took issue with the wild variations in the Gallup polling data – and the various public surveys that were widely off the mark – in a post-election interview with POLITICO."

     "Benenson, whose forecasting model was critical to the campaign, went through the major issues with the public polls that showed wild variations throughout the final weeks of a remarkably stable race."

     “I think it’s long overdue for an organization with a name as well-known as Gallup to recognize what the demographics of the American electorate actually are and figure out why their model has continued to skew too old, too white and less likely to be college educated than the nation’s voters,” Benenson said.

      The question facing Gallup is a double whammy: not just about the inaccuracy of its polling during this cycle, but also its name is supposed to denote a higher level of quality and prestige. Last week it offered some pretty feeble rationalizations. It argued that its last poll showed Obama losing by 1 among likely voters and the actual results had him losing by 2-it's actually moved closer to 3 for the record-and so they were off by 3 points which within the margin for error.

      However, that's as noted above, a pretty feeble justification. For one thing they only looked at the last polls, which, as Nate Silver often argues is misleading by itself anyway as outlying polls like Gallup was the last month usually adjust their numbers to bring them more in line with consensus.

      For more on this please see here

      In fact, if you look at the last 3 weeks, on average, Gallup deviated by a much wider margin from the result often showed Romney up as much as 7. Also, closeness to the marign is one gauge. Another is whether or not they picked the right candidate. Clearly they didn't. Their polls over the last few weeks all had Romney leading.

      In response to Beneson they have some new explanations. They think that exit polls are overdone

      "Jeff Jones, of Gallup, dismissed the idea that the exit polling is Gospel."
       “We don't assume to know what the electorate is before it happens,” Jones said. “We take our sample [and] we weight it to ... known targets that are rock solid.”

        "He added, “People basically accept the exit poll as the gospel. The problem is it's the exit poll and those exit polls ... and i think it's pretty well known especially in the case of 2004 that the exit poll has issues.”

         Yet, the Obama team relied on exit polls and their results were much more accurate than Gallup's:

        "However, exit data has been a guide for the Obama campaign in terms of how the proportion of the white vote has declined since 1996, and how other demographic changes have taken place."

        Gallup at this point sounds too defiant for a poll that has been until now considered the gold standard of polls and yet has turned in one of the election's weakest performances. It's insistence on learning nothing from its performance does not bode well.



No comments:

Post a Comment