Public Policy Polling scrubs one of their control questions after it showed polls skewed to liberals

PPP is running a great scam in 2011 by trying to establish a narrative that the Republican gains just five months ago have crumbled after only two month in office. Many of their polls are gimmicks according to Nate Silver at NY Times, like the Charlie Sheen versus Sarah Palin poll or the one about the MS segregationalists. I first started covering this trend a few weeks ago.

While PPP is associated with Daily Kos, they’re accuracy in 2010 elections won over my doubts. However, in a very new development, they are also teaming up with the SEIU and given their narrative making, I’m not sure PPP should really be getting the unbiased press that they’ve enjoyed over the last few months.

Despite their liberal bent, news organizations have been trying to create cover and argue that they are far left but accurate.

I’m glad to see Professor William Jacobson not buying it. Legal Insurrection has been at the forefront of questioning the accuracy of some of the polling. A few weeks ago, I pointed out that whenever PPP does a poll on a governor that was elected last November, they seem to handicap the newly elected governors by a) oversampling those that voted for the candidate that lost not more than a few months ago b) oversampling a group of nonactive voters.

I’ve kept an eye on the trend, for two reasons:

First, the group of respondents who claimed that they had not voted or don’t remember voting were a big share of the Republican decline:

Looking at the PPP crosstabs, we learn that Scott is not viewed well by those that voted for Democrat Sink. Anyone surprised? There seems to be some defection in those that voted for him in comparison to Sink, but the big population driving up Scott’s negatives are people who identified that they didn’t vote or don’t remember voting.

I suppose that this could translate into Scott being incredibly unpopular, but to be more accurate, he’s incredibly unpopular with those that don’t vote or are so disengaged they don’t remember voting.

Second: their population of respondents who didn’t vote in 2010 or don’t remember voting had been growing from poll to poll. Their OH poll on 3/15/11 had a sample size of those that didn’t vote or don’t remember of 5%. Their MI poll on 3/22/11 had an increased sample size of 8%. By the FL poll on 3/29/11, that sample size was up to 11%, over double the OH poll just two weeks earlier. Quite frankly, this looks deliberate given the fact that this population was having a major effect on how the polls showed shifting sentiments away from Republicans. I wanted to see if PPP’s sampling would continue to trend upward, remain in the low double digits or trended randomly in the long-run.

Both criticisms of handicapping Democrat November losers and oversampling those that don’t vote were based on a single control question asking respondents how they voted for governor last year. That question is strangely absent from recent polling ever since I pointed it out.

On 4/6/11, PPP released a GA poll discussing a November rematch. No surprise, the Democrat would win the rematch. Only, what happened to the one question missing? How did you vote last November?

Professor Jacobson is onto scam number two, which has to do with Party identification:

Here is the sample breakdown in the recent PPP poll, 38% Dem, 34% Rep, 28% Ind:

This doesn’t come close to reality. Here are Gallup’s most recent party affiliation numbers, 31% Dem, 29% Rep, 38% Ind:

Gallup’s recent party identification analysis by state shows an even worse situation for Democrats, with the number of solidly Democratic states declining from 2008 (30) to 2010 (14).

The PPP poll which purports to show a shift away from Republicans uses a sample which overweights Democrats both relative to Republicans and relative to the total voting population.

Another trend I found in looking at PPP’s polls was that while there samples for the 2010 elections seemed completely out of synch, the voter split in the state for the 2008 election of Obama seemed to be pretty close to the actual state results. Thus, PPP’s polling is primed like it was 2008, when Democrats had 36% voter identification instead of current 31%.

David Cantanese at Politico picked up on a similar trend as he compared PPP to Republican polling. PPP confirmed with David Cantanes and admitted it, that their polling is ignoring the 2010 elections and looking back to 2008. The poll in question was a MI poll, where according to Gallup, Democrat identification is down 6.3% from 2008 numbers.

Can someone please give me a reason for why PPP is treated as non-partisan by the media?

Update:

I’d almost forgot how PPP had Coakley winning MA.

Update: I Own the World linked – thanks!