If it weren't for that critical component of modern political public opinion polling known as the "likely voter" screen, the banner polling headline today would be that, despite the Paul Ryan "bump" and a mammoth spending advantage, Mitt Romney still trailed Barack Obama in the latest CNN/Opinion Research poll of the presidential race by a gaudy 52-43 margin.
Alas, said "likely voter" screen does exist, so instead, the headlines (as they have been for most of this cycle) will warn of a coin flip race, with the president staked to a mere 49-47 lead over the Republican challenger. Even worse, it is inevitable that some in the chattering classes will talk of a "bounce" for Romney since the last poll, which showed Obama up 52-45.
Such talk of movement in Mitt's direction is analytical malpractice, of course, because the previous CNN/OR poll did not have said likely voter screen. In the only apples-to-apples comparison we have here (RV to RV), it is the president whose standing has incrementally improved.
With all this in mind, I close this week of the Wrap (after the fold, of course), with what will be a rerun for long-time readers: a mini-rant about how I am perhaps the only person in the polling analysis game who feels that likely voter screens tend to suck.
But, first, on to the numbers:
PRESIDENTIAL GENERAL ELECTION TRIAL HEATS:
NATIONAL (CNN/Opinion Research): Obama d. Romney (52-43 among RV; 49-47 among LV)
NATIONAL (Gallup Tracking): Obama tied with Romney (46-46)
NATIONAL (Rasmussen Tracking): Romney d. Obama (46-45)
CALIFORNIA (Tulchin Research and M4 Strategies for USC School of Education): Obama d. Romney (56-33)
COLORADO (Keating/OnSight--D): Obama d. Romney (48-44)
MISSOURI (Rasmussen): Obama d. Romney (47-46)
VIRGINIA (Rasmussen): Obama tied with Romney (47-47)
DOWNBALLOT POLLING:
CA-PROP 30 (Tulchin Research and M4 Strategies for USC School of Education): Yes 55, No 36
CA-PROP 38 (Tulchin Research and M4 Strategies for USC School of Education): Yes 40, No 48
NM-SEN (GBA Strategies for the Heinrich Campaign): Martin Heinrich (D) 51, Heather Wilson (R) 44
PA-SEN (Muhlenberg College/Morning Call): Sen. Bob Casey (D) 49, Tom Smith (R) 30
VA-SEN (Rasmussen): Tim Kaine (D) 45, George Allen (R) 45
A few thoughts, as always, await you just past the jump ...
Before I wade too deep into my rant, a caveat: In many cases, the gap between registered voter results and likely voter results, in any given poll that posts data for both, are quite small. As Nate Silver noted just the other day during the poll-a-palooza we waded through midweek, when he makes an adjustment of a RV sample to compensate for the lack of a likely voter screen, he puts his thumb on the scale on the order of a two-point swing to the GOP. Except in cliffhanger elections, that's hardly a decisive shift.
What's more, my own research on the matter, which was conducted in the fall of 2010 and used polling data from 2006 and 2008, found that 55.4 percent of the polls released in those cycles that featured both RV and LV results only differed by 0-2 points between their margins.
That, however, is not what happened today with CNN's new presidential survey. The gap was seven points, and it legitimately changed the entire narrative of the poll. Had CNN stuck with a registered voter polling sample, the story would've been that Paul Ryan has been no help to Mitt Romney, and that Todd Akin may well be a millstone around the party's neck, since Barack Obama has stretched out what was, earlier in the month, already a respectable lead in the high single-digits. Instead, we get another "OMG, what a toss-up" polling summary tossed at us.
At the risk of stating the obvious, there is a vast difference in perception between a two-point race, and a nine-point race.
Skeptics would note that the likely voter screen gives the political press something that they desperately want: a perilously close election. Romney's campaign has been pretty disastrous, all in all, so much so that Charlie Cook penned an entire column dedicated to the theme that Barack Obama should be getting boat-raced, and somehow is not.
With that kind of mindset in the press corps, the idea that Obama could actually defeat Romney, and do so by a fairly comfortable margin, is inconceivable to them. Switching to what seems to be an awfully restrictive LV screen undoubtedly helps.
"But registered voter polls are less accurate!" comes the inevitable cry from the right, as well as many "neutral" polling analysts. You hear it all the time, so much so that it is an article of faith in discussions of polling. And, on a base level, that makes sense: We know a large swath of registered voters don't show up on Election Day. However, that analysis I alluded to earlier from 2010 disputes that claim. Indeed, in those 56 elections where pollsters offered their RV and LV results, the likely voter screens only produced the more accurate result 37.5 percent of the time. It was the less restrictive "registered voter" results that came closer to the mark the majority of the time.
In the future, when both are made available, I will put both results into the Wrap, leaving you, the readers, to decide on your own which one represents a better snapshot of the state of play in the 2012 election cycle.
In other polling news ...
- Normally, when a campaign pooh-poohs an opponent's internal polling, but refuses to release their own, I take that as a sign that said internal poll is probably pretty close to the fairway. That is why this statement from the campaign of Colorado Democratic Rep. Ed Perlmutter caught my eye:
“That is not even close to what ours show,” said Perlmutter spokeswoman Leslie Oliver, who declined a request by The Colorado Statesman to release any of her campaign’s internal polling. “We never talk about them, it’s our standard policy,” she said. “Ed’s doing well, Ed’s got good support. It’s broad, it’s bipartisan.”
However, in fairness to Perlmutter, that actually does appear to be his standard policy. A check of my 2008 and 2010 polling databases (with well over 4,000 polls in them) did not turn up a single Democratic-sponsored poll in the mix. And that included a closely watched 2010 battle with Ryan Frazier, which Perlmutter eventually won with 53 percent of the vote.
- Those California polls, on the competing ballot tax propositions brewing in the Golden State, raise the spectre of a potentially problematic campaign scenario. With the ballot measure for wealthy attorney Molly Munger trailing 40-48, but the measure for Gov. Jerry Brown's tax proposal garnering a small majority of support, Munger may find herself in a position where she must not only bolster her proposal, but also attack Brown's. The reason is the state's method for dealing with competing measure on the same ballot. When two ballot measures deal with the same area of policy, and would be in conflict with one another, the measure that takes the most "yes" votes takes effect, in the event that both pass. Therefore, Munger does not only need to find someway to turn around an eight-point deficit in support, she needs to chip away at the support for Brown's measure. This is precisely the scenario that Brown supporters have feared, because they worry (with some justification) that this could lead to the failure of both ballot measures.
- Speaking of ballot measures, the Glengariff Group polling for local media in Michigan that showed Democrats Barack Obama and Debbie Stabenow leading at the top of the ticket also polled many of the pending ballot initiatives in the state. The news, for progressives, was mostly lousy. Polling shows a reversal of the state's controversial emergency manager law seems unlikely, and that the enactment of a supermajority to raise taxes seems quite likely. Not all the news is grim, however: An initiative codifying collective bargaining rights, however, did enjoy a 19-point edge, and a clean-energy initiative also enjoys a wide lead.