"Romney’s Strong Debate Performance Erases Obama’s Lead" | Pew Research Center for the People and the Press

"Romney’s Strong Debate Performance Erases Obama’s Lead" | Pew Research Center

"Among likely voters, Romney holds a slight 49% to 45% edge over Obama. He trailed by eight points among likely voters last month."

12 points? I can't recall ever having seen such a dramatic turnaround in such a short time reported by a respected polling organization.

To add more confusion to the matter, Pew is a polling organization that the Republicans consider to be highly biased in favor of Democrats! Here's what the National Review, comparing final polls to actual election results, wrote this summer:

"Starting in 1992, EVERY Pew poll appears to lean to one direction — always towards the Democrat, and by an average of more than 5 percentage points."

The analysts at 538 tend to agree with that point in general if not in detail. They recently studied a dozen polling organizations and found that Pew has the largest pro-Democratic "house effect" - about three points.

Yes, Pew has tended to be biased in favor of Democrats. In this case, however, the massive shift from September to October is a sampling issue. I checked the complete data for Pew's September and October samples.

In September, their "likely voter" population included 31% Democrats and 38% Republicans - a 7-point difference in favor of Democrats, which obviously accounted for the essence an 8-point Obama edge.

In October, their "likely voter" population included 36% Democrats and 33% Republicans - a 3-point difference in favor of Republicans, which just as obviously accounted for the essence of a 4-point Romney edge.

In other words, 10 points of the apparent 12-point shift can be attributed to sampling variations. (Take out the margin of error and you can probably claim fairly that it is all attributable to sampling variations.)

Could the sample variation be not just a shift in the sample, but a shift in the population? Could it mean that significantly more people now self-identify as Republicans? That's possible, but that conclusion cannot be drawn from my knowledge base.