A pet peeve of mine is the tendency of some political observers to dismiss adverse public opinion data as “meaningless” because it’s not reliably predictive. Recently I attended a political panel in which two speakers were harping on primary exit poll numbers about potential “white-working-class” defections to John McCain if Obama is the nominee, and also stressing the Jeremiah Wright saga as a huge general-election problem for Obama. I asked them how they squared this belief with general election polls showing (1) Obama generally running ahead of McCain; (2) Obama generally running even or close with McCain among white voters; and (3) little evidence that voters cared that much about Wright. Both speakers responded by saying that general election polls this far out from November were “meaningless.”
This point of view is even more prevelent when it comes to state-by-state general election polls. That’s why I was interested in Brendan Nyan’s recent post summarizing the research on state-by-state polls in 2004, which suggested they are not precise when it comes to predicting close states, but are otherwise pretty much spot-on in terms of broader results.
And that, folks, means they are not “meaningless.”