Which poll is closer to reality? Quinnipiac or Monmouth?
How two respected independent pollsters could have such differing results for Governor Chris Christie’s approval ratings has been the subject of quite a bit of chatter this week since Monmouth University released their poll indicating that Christie’s ratings were 11 points lower than reported by the Qunnipiac poll released last week. Quinnipiac reported Chrisite’s approval rating at 59%-36% while Monmouth said that 50% of registered voters approve of the job that Christie’s doing compared to 38% who do not.
There’s been enough buzz about the difference that Patrick Murray, Polling Director at Monmouth, posted a piece on his blog (cross posted on Politickernj), that took a swipe at Quinnipiac for framing their approval question in such a way that Christie’s numbers would be higher. Murray said that because Quinnipiac first asked if Christie would be a good selection as a Vice Presidential nominee, respondents were more likely to give him higher marks when asked to evaluate his job performance.
Quinnipiac, on the other hand plays around with the order in which they ask the governor’s job rating question. In 8 polls over the past year, they asked Gov. Christie’s job rating as the first question in 3 cases and the 3rd question in one case. For the remaining four polls, the governor’s rating question was slotted from #10 and #13 in their questionnaire.
When it was the first question, the governor’s positive job rating was only 44% to 47%. At the number 3 slot, it was 53%. At #10 or later in the interview, it ranged from 55% to 59%. It’s worth noting that the lower poll numbers came early last year, and were either closer to or even lower than other polls conducted at that time. Hmmm.
In the most recent Quinnipiac poll, one of the questions preceding Gov. Christie’s rating presented him as a potential nominee for Vice President. In other words, the survey framed the governor as a national figure before asking voters to rate his job performance. Could this be why his rating among Republican voters in particular shot up to an astronomical 92%?
Pollsters know that job approval ratings can be impacted by the context of a poll interview. That’s why most pollsters try to place these key trend questions in the same place in every questionnaire. This increases our confidence that any changes in a politician’s ratings are due to real shifts in opinion and not an artifact of questionnaire inconsistencies.
I’m willing to venture that first naming Chris Christie as Mitt Romney’s potential running mate before asking New Jerseyans to rate their governor might have had a wee bit to do with the two polls’ divergent trends.
Mickey Carroll, Director of the Quinnipac Polling Institute, is not interested in getting into a pissing match with Murray. “Patrick Murray is a very good pollster,” Carroll said three times in a seven minute phone interview with MMM. “Every poll is different, something could have happened in the week in between the two polls,” Carroll said, “we asked the question the same way.”
When told that Murray said that Quinnipiac framed the approval question by first asking a question about Christie being a potential VP, Carroll said, “that could make a difference, but I think we asked the approval question first. Didn’t we? Patrick Murray is a good pollster, a savvy analyst and a smart guy.”
Republican strategists, who would only speak on background, were quick to criticise Murray and side with Quinnipiac.
“The Quinnipiac poll from last week showing the Governor’s job approval at 59% is closer to reality,” said one strategist who cited internal GOP numbers, “The problem with the Monmouth University poll is that it samples, ADULTS, rather than registered voters, or better still, likely voters. It is cheaper and easier to poll adults, because there are a lot more of them and they are easier to qualify. It is harder, and more expensive, to find and poll a likely voter – especially a likely voter who votes in non-Presidential year elections.”
When told that Murray blamed the difference on how Quinnipiac framed the question, the same strategist said, “Patrick is out of his mind.”
In fairness to Murray, MMM verified that Quinnipiac and Monmouth both sample adults who then self identify as registered voters. However, on their website Quinnipiac says that they ask screening questions, plural, to determine who is a registered voter. Murray said Monmouth only asks one question to determine if a respondent is registered to vote or not. He said that 80% of his respondents tend to be voters. 78% of New Jersey adults are registered to vote.
In his blog post, Murray acknowledged that his Monmouth poll results are consistently more favorable to Democrats while Quinnipiac’s are consistently more favorable to Republicans. MMM asked Murray how that could happen consistently if both polls were using random computer generated phone numbers. “Is it how you weight the sample?” we asked. “That’s part of it,” Murray said, “the rest is that we (Monmouth) call a greater percentage of cell phones. Cell phone users tend to be younger and more Democratic.”
With that answer, that weighting his samples more heavily towards Democrats and cell phone users, Murray seemed to be confirming the Republican complaints.
“That’s fine, I’m the only one who consistently asks the trend question in the same place,” was Murray’s retort, “that’s polling 101.”
Another Republican strategist was more upset about how Murray wrote up his poll release that he was with the numbers. “Political sands are shifting?” asked the Republican. “It’s a margin of error shift! Murray sounds like he is writing press releases for the Democratic State Committee, not acting as an independent pollster from an esteemed New Jersey university.”
A third Republican scoffed at the notion of even taking Murray’s numbers seriously, pointing out how badly Monmouth Gannett polled the 2009 gubernatorial race between Chrisite, former Governor Jon Corzine and Chris Daggett. “Murray’s last poll in that election had Corzine winning by 2 points and Daggett getting 8% of the vote. Christie won by 5 points a few days later.”
MMM set out to find a Democratic strategist to weigh in on the difference between the two polls, but no one would talk to us. “Try Patrick Murray,” was the best answer we got.
Murray called shortly after this piece was posted to ask that if his 2009 gubernatorial results were going to be used against him, that his correct calling of the 2010 CD-6 congressional race also be mentioned.
During the 2010 congressional race, MMM analysed a Monmouth Poll that indicated Congressman Frank Pallone was leading Anna Little by 11%. MMM concluded, using Monmouth’s data, that Pallone’s lead should be 9%. Murray agreed, “your turnout assumptions are as good as mine,” he said. Things got funky when the Little campaign issued a press release announcing that Murray had revised his numbers based upon MMM’s analysis and that Pallone’s lead was then in single digits. Murray issued a release stating that his “official” numbers hadn’t changed. He issued a later poll that indicated Pallone’s lead was down to 7% and in the final days of the campaign said a Little victory “could well happen.”
When Pallone eventually won by 11%, Murray said he was right all along.
It was all great fun for MMM, except that Murray stopped taking our calls for a while.
In his call this afternoon, Murray reiterated that the primary difference between the recent Quinnipiac poll and his poll about Christie’s approval ratings, is that Quinnipiac changed their methodology by altering the order of the questions.
Regarding polling adults vs polling registered voters, Murray said he was not in the business of electing or reelecting any candidate. He said his job is to report on what New Jersey residents are thinking.
Murray also said that his “poltical sands are shifting” comment in the poll release was a reference to New Jersey Democrats who being more aggressive in how than they go after Christie than they have been in the past. It was not a reference to the public’s approval of Christie, he said.