When NBC/WSJ released a poll showing Obama up by 3 points nationally but by 8 in the swing states, it predictably led many to conclude that Obama has a larger lead in the swing states than he does nationally. But as acknowledged here and here, the evidence for a structural Obama advantage in the Electoral College is unpersuasive, at least at this early stage.
Just one day later, NBC released three polls conducted by Marist University showing a tight race in three critical battlegrounds: North Carolina, New Hampshire, and Michigan. This represents the third triplet of NBC/Marist swing state polls over the last month, covering nine of the twelve swing states included in the NBC/WSJ swing state subsample. After weighting the results by 2008 turnout, Obama leads in the nine polls by 3 percent, down from 6 percent in 2008, and similar to the 4 and 3 percent margins by which Obama led the May and June NBC/WSJ national polls.
The NBC/Marist polls are largely consistent with other state polls, although FiveThirtyEight has observed a modest pro-Democratic house effect in NBC/Marist's past polls. These nine NBC/Marist polls sampled nearly 10,000 respondents—forty times as many as the NBC/WSJ subsample.
The three NBC/Marist polls were buried by the health care announcement, but they provide a more compelling picture of the battlegrounds than the better publicized NBC/WSJ swing state sub-sample. There was nothing wrong with the decision to release the NBC/WSJ swing state sub-sample, but NBC should have known better than to trumpet such a provocative finding supported by such little evidence, especially since NBC possessed a far more compelling set of data contradicting the finding. A more responsible approach either would have downplayed the swing state sub-sample or given equal attention—at least—to the collective findings of the nine NBC/Marist polls.