NBC and the Wall Street Journal combine again to deliver a survey with a sample that manages to get the Democrat partisan affiliation advantage wider than that of the 2008 presidential election — and Barack Obama still can’t get above 46% in overall approval.  The poll gives the GOP a three-point advantage on the generic Congressional ballot despite the skewed sample, and the Tea Party shows its strength, as the WSJ reports and as Allahpundit pointed out last night.  However, the poll skew and methodology are worth a second look.

The data shows the partisan affiliation in the sample:

Strong Democrat ……………………… 20
Not very strong Democrat ………….. 15
Independent/lean Democrat……….. 8
Strictly Independent ………………….. 13
Independent/lean Republican……… 11
Not very strong Republican………… 12
Strong Republican……………………. 14
Other (VOL) ……………………………. 5
Not sure ……………………………….. 2

Either that breaks out to a 35/26/32 breakdown without leaners, or a 43/37/8 with leaners.  Either is an absurd reading on the present electorate.  Put aside, for the moment, the myriad surveys that show the partisan gap between Democrats and Republicans being between 1-3 points this year (Rasmussen’s latest was 1.2 points for Democrats).  The actual vote in 2008 gave Obama a win with a 7-point advantage in a race that had considerable Republican and independent crossover to the Democrats.

In order to believe that this survey is at all predictive, one would have to agree that Barack Obama and the Democrats have significantly broadened their base in the following two years.  Do we have any evidence of that at all?  Not exactly, as this poll itself shows.  Obama’s overall job approval is underwater at 46/49, and his rating on the economy continues to tank at 42/54.  A year ago, that latter figure was 50/42, and at the beginning of his term, it was 56/31.  The overall right/wrong direction question is now 32/59; a year ago it was 39/48, and it was 43/43 in April 2009.

Furthermore, all of those numbers come from a sample of adults, not registered voters.  And five weeks before a midterm election, the NBC/WSJ survey doesn’t bother to apply a likely-voter screen, when almost all other pollsters have done so.  That includes, by the way, the 56/32 split on whether Obama inherited the economic mess or is now mostly responsible for it.  What do likely voters think in this sample?  We’ll never know.

In other words, this poll is almost useless for any analysis of voter behavior.