01 October 2008

How Much is Enough?

Disclaimer: This post poses early morning questions without providing answers. More coffee, anyone?

How effective is a sampling of 614 voters? That's going to be a very important question as we draw closer to election day. According to a new Wall Street Journal/NBC News poll, only 54% of respondents answered that they would "definitely" vote.

The Journal clarifies that 3/4 of the 614 are newly registered voters, and another 1/4 "lapsed" voters, or voters who did not vote in the last presidential election. This is significant because, the Journal found, 61% of new voters favor Obama. If taken as an accurate predictor of election day (and early voting) behavior, these numbers seem to hold the potential to determine the outcome of the election. It's no secret that Democrats are counting on a "groundswell" of new voters to make their voices heard in historic fashion this year. Registration of new Democratic voters dramatically outpaces new voter registration for Republicans, with an estimate of 2 million newly registered Dems since 2004 compared to a net loss of some 350K Republicans. But if only 54% of newly registered voters actually turn out, what might that spell for Obama?

Conventional wisdom, and the new WSJ/NBC News poll, reinforce the idea that economic troubles weigh foremost on voters' minds. That could certainly play as a motivating factor to get people to cast ballots this year. I'm still stuck on this number, 614, though. Without knowing anything about the science of polling, 614 strikes me as an incredibly shallow pool. With a margin of error of ±4.0%, the potential for unreliability tracks ahead of larger polls that allow for a margin of error ±2-3%. I'm no math whiz, so I can't tell you how that affects the data. But the bigger question seems to me to be why publicize such a thin survey? Solely for the "chatter class"? Or is there something substantial here that I'm missing? How do we assess whether 614 voters is representative enough to be credible?