Today’s Marist/NBC poll still shows President Obama with the lead in Ohio, Florida and Virginia, but Mitt Romney has apparently gained considerable ground since the March poll:
In Florida and Virginia, Obama leads Romney by an identical four-point margin, 48 percent to 44 percent, among registered voters, including those who are undecided but leaning toward a particular candidate.
In Ohio, the president is ahead by six points, 48 percent to 42 percent.
In March NBC/Marist polls — conducted during the middle of the GOP primary season — Obama led Romney by 12 points in Ohio (50 percent to 38 percent), and by a whopping 17 points in Virginia (52 percent to 35 percent). In January, Obama was ahead of Romney by eight points (49 percent to 41 percent).
The poll shows Romney’s support is increasing, while Obama’s support still lags below 50 percent. Obviously, there has been some predictable consolidation behind Romney since the peak of the GOP primaries in March. But the fact that Obama can’t get above 50 percent is a sign that these states are going to be more competitive than Democrats expected.
Ed Morrissey also sees some problems in the poll’s party breakdowns:
In other words, Marist uses a model in Florida that assumes Democratic turnout will be six points higher than in 2008 and seven points higher than in 2010, while Republican turnout remains largely the same, which means that independents vanish. In Ohio, Marist assumes Democratic turnout will be only slightly lower than 2008, while Republican turnout craters below their nadir in 2008 and ignores the big GOP turnout in 2010. Only Virginia looks like a somewhat rational predictive model for the 2012 election, although they significantly oversample independents.
Because the poll also shows that Republican enthusiasm is actually higher than Democrat enthusiasm in Florida and Virginia, the actual connotation of the poll might actually be worse for Obama than it looks at the outset.
I'd be interested in some truly expert analysis on these polls. We conservatives are quick to dig around for the partisan breakdown of the polls, and it often seems that they are stacked against Republicans. But at some point I have to wonder – are they? Why would pollsters consistently do that? Pushing a candidate? OK – but what about "brand name"? nBeyond that are issues I don't understand. For example, are the numbers of Dems, Reps and Indies polled what matters? Do these pollsters crunch those numbers according to a formula based upon expected voter turnout models? In other words, just because, say, 36% of those polled were Democrats and 28% were Republicans, does that necessarily mean that those were the percentages used to determine the poll results? Couldn't they then take those numbers and come up with results that assume an electorate that will be 34% Dem and 32% Rep (or whatever)? This approach would make sense, since it is probably difficult to get the percentages right if you're relying upon people 1) answering the phone and 2) taking the time to do the survey. n I've spent time trying to do the math to figure this out with some polls but, unfortunately, the task is beyond my limited abilities. Maybe someone can fill me (us?) in.