Polls are commissioned by a number of media outlets...should I reflexively ignore polls commissioned by the sources I don't like? Or should I look at the internals...methodology, sample size, MOE, party affiliation of respondents...before passig judgement?
No poll gets it exactly right. Each organization's predictions will be off by a certain amount which we in turn can use as a metric for that organization's 'accuracy.' Because they'll all be off, it is helpful to smooth out the bumps in the data and do a 'metapoll' that pools all of the results together
rather than take the result of any one organization as sancrosanct. That's what our friends at RCP do on a regular basis.
1. RCP's data still indicate an Obama victory is in the works, and has indicated as much for over a year. I hate to be overly deterministic about this but the data is clear. RCP's 'No Toss-up States' map has had the same outcome -- with minor variations in electoral college vote distribution -- since before Romney was the nominee. This state of affairs hasn't changed. Romney's rise in the polls means either there has been regression toward the mean or his debate performance really did have an impact, but, in any case, neither has been enough to change the fundamentals.
2. If we were to take the results from a single organization as sancrosanct -- something I just said we shouldn't do -- we would presumably want to look at the accuracy of such organizations and carefully make a choice. And this in turn implies that we would avoid selecting an organization that consistently oversamples one of the parties and which had
virtually the least accurate results of all polling organizations during the last election cycle (i.e., Rasmussen).