Today, we were able to lay our eyes on some state polling of note (including one in Florida that will be music to the ears of fans of the president).
But, more notably, we also were graced with a pair of national polls (Pew and AP/GfK) that give Barack Obama a lead of either 3-4 points. On the same day, curiously, we also see the two daily tracking polls (Gallup and Rasmussen) clinging tenaciously to the notion that Mitt Romney is actually the leader, by a similarly modest margin.
What gives? I have some thoughts about that.
But, first, the numbers:
PRESIDENTIAL GENERAL ELECTION TRIAL HEATS:
NATIONAL (Associated Press/GfK): Obama d. Romney (47-44)
NATIONAL (Gallup Tracking): Romney d. Obama (47-45)
NATIONAL (Pew Research): Obama d. Romney (50-46)
NATIONAL (Rasmussen Tracking): Romney d. Obama (47-43)
FLORIDA (Quinnipiac): Obama d. Romney (46-42)
MICHIGAN (Mitchell Research): Obama d. Romney (47-46)
NEW HAMPSHIRE (Rasmussen): Obama d. Romney (48-43)
DOWNBALLOT POLLING:
FL-SEN (Quinnipiac): Sen. Bill Nelson (D) 43, Connie Mack IV (R) 39; Nelson 45, Mike McCalister (R) 34; Nelson 47, George LeMieux (R) 32; Nelson 47, Dave Weldon (R) 31
FL-SEN--R (Quinnipiac): Connie Mack IV 41, George LeMieux 8, Mike McCalister 5, Dave Weldon 3
A few thoughts, as always, await you just past the jump ...
One of the regular readers and commenters on the Polling Wrap, Davidsfr, has been keeping vigil for quite a while now on a particular statistic. That stat, relevant to today's topic, is chronicling how long it has been since a poll other than the Gallup or Rasmussen daily tracking polls has shown a lead for Mitt Romney. That streak is now at over four weeks and counting, for those scoring at home.
What could account for that disparity? Well, I suspect that the reason is actually fairly simple. The two pollsters issuing daily tracking polls also happen to have structural factors in their polls that would lead to results that would be incrementally more favorable for a Republican candidate. The issue with Gallup was drawn out in great detail by Mark Blumenthal over the past week or so. The issue with Rasmussen is that they not only weight their polls by party affiliations, but their weighting is way the hell off of what damned near every other pollster in the world is seeing:
Among other targets, Rasmussen Reports weights data by political party affiliation using a dynamic weighting process. While partisan affiliation is generally quite stable over time, there are a fair number of people who waver between allegiance to a particular party or independent status. Our baseline targets are established based upon separate survey interviews with a sample of adults nationwide completed during the preceding three months (a total of 45,000 interviews) and targets are updated monthly. Currently, the baseline targets for the adult population are 35.7% Republicans, 33.4% Democrats, and 30.8% unaffiliated. Likely voter samples typically show a slightly larger advantage for the Republicans.
Even having a likely voter sample that is R+3 is a stretch, given that there has not been a presidential election in the last two decades where the electorate (as measured by the exit poll data) has skewed three points in favor of the GOP. Indeed, over the last three cycles, the average has been roughly D+4.
Given those two bits of detail, it is rather shocking that Barack Obama ever leads in their national tracking poll. Furthermore, it dings the credibility of those numbers, not that you'll hear the traditional media scream about it (but more on this a bit downstream).
The bottom line—both of the daily tracking pollsters have structural elements to their polling that could clearly impact the results that come out of them. In the name of giving their readers and viewers a clear picture, media outlets that cover polling data should make note of that fact. Virtually none of them, of course, bother to do it.
In other polling news:
- The juxtaposition of yesterday's Bloomberg poll and today's Pew and AP/GfK polls could give any observer a serious case of whiplash. A pretty good authority on the subject is telling us all to calm down and not read too much into the polling. That authority is none other than the pollster for Barack Obama, Joel Benenson:
“The only thing that’s bouncing around are the public polls,” Benenson said. “The electorate doesn’t bounce around like that. It’s more static than the noise in all these polls. If you watch the electorate over time, they don’t jump up and down. This is a process.”
Benenson also dings the existence of horse-race coverage in the popular press, a pretty surprising stance for a political pollster to take.
- Elsewhere in the same article, WaPo writer Greg Sargent speaks to a point I made yesterday, even as I discounted the Bloomberg poll, about a particular media double-standard that seems to accompany yesterday's coverage:
As it happens, many political scientists agree with Benenson’s sentiments, and suggest focusing on polling averages instead. And there’s another point to be added, which is that the selective focusing on daily polls can exacerbate distortions about the race. When a recent Times poll found Obama’s approval down at 41 percent, many news organizations treated it seriously, even though it was obviously an outlier. Yesterday’s Bloomberg poll showing Obama way up was mostly treated as ... the outlier that it was. One drove the conversation; the other didn’t.
Precisely. Today's Rasmussen tracker puts Mitt Romney up by four points. If we accept the AP and Pew polls at face value (Obama +3 or 4 points), then Rasmussen is nearly as far off of the mark as Bloomberg's poll was yesterday. Yet a Google News search of "Bloomberg outlier" yielded 184 results. "Rasmussen outlier"? Zero.
- Taking a quick look at the downballot polls today, we get a textbook example of a classic conundrum in media poll analysis. The new Q poll in Florida puts Democratic Sen. Bill Nelson up over Republican Connie Mack IV by four points. That's several points behind where PPP had the race earlier this month, but it is also several points better than Quinnipiac had the race the last time they polled Florida. The media coverage thus far has painted this poll as evidence of the race tightening, but that may well be an unfair conclusion. It leads to a specific quandary in analyzing the poll. Which is more important: the temporal context of the poll (especially when polls may come weeks, if not months, apart), or the context of the pollster, given the existence of house effects? Today's AP poll provides another example. Looking at it from a temporal perspective, you'd say that's a decent poll for Obama. Obama +3 is on the upper end of most recent national polling. It's certainly better than the median, as well as the Pollster average right now (though that is adulterated a bit by the existence of those aforementioned daily trackers). However, if you look at it in comparison to the last AP/GfK poll, it is good news for Romney. The Obama lead is down from eight points during the last AP/GfK poll. So, which is the fairer analysis? It is a worthy subject for debate.