Much Ado about Polling

[Geee]

National Review:

Everywhere I went Monday, conservatives were spooked by a Gallup poll showing President Obama with a six-point lead over Mitt Romney. Then on Tuesday, a Washington Post/ABC News poll had the margin down to a single-point Obama edge. What is going on?

As Election Day draws closer, most major public-opinion surveys shift from interviewing registered voters to interviewing those whom they identify as “likely voters.” The transition from polling “registered voters” to predicting who will actually vote is a tricky one and can involve some turbulence and differences between pollsters. Furthermore, some organizations make the switch earlier than others — in the case of this week’s polls, Gallup’s was of registered voters, the WP/ABC one was of likely voters. And that could have made the difference.

As Nate Silver, the statistics guru of the New York Times, reports, “In the past six presidential election years, the shift to likely voter models has always helped the Republican candidate, but the difference has also always been small.” Pollster Scott Rasmussen argued during the 2010 elections that his polls, which always interview only those voters he determines are most likely to vote, are more accurate. This group, he says, is trending more conservative these days because opponents of Obama’s policies are highly motivated to vote. Pollster John Zogby says polls tend to oversample Democrats, especially blue-collar women, who often don’t vote. Thus, “the results may be skewed toward the Democrats.”