Thursday, October 30, 2008

Can We Trust the Polls?

Which polls do we believe in when there is so much variance in the results? The polls are volatile now because we really don't have a good metric for assessing the elusive category pollsters call "likely voters," a category especially difficult to define in a year with so many new voter registrations are from demographics - youth (71%) and blacks (25%) - who have not typically turned out to vote in significant numbers.

As think-thanks and universities seeking a public profile are jumping on the polling bandwagon, as are new PR companies and media outlets with their own agendas, we are better off over-weighting the results from the gold-standard of the polling establishment, Gallup.

In only one time in the history of polling has a candidate won the election after trailing in the Gallup poll in the week before election. The candidate was Ronald Reagan. The reason why Reagan soared in the final week of October in 1980 was because of his strong debate performance. Without a similar event occuring in 2008, a reversal and repudiation of the polls would be unprecedented and improbable.

Forecasting science is more science than art or ideology. If politicians didn't believe in polls they wouldn't be commissioning them. It is the job of pundits to spin the interpretation of polls, as it is the job of partisans to manipulate the nature of the sample size. But if done well, polls can tell us rather precisely what we think. And since 1/3 of the electorate will vote before Nov 4, these polls are indicating how early voters are voting.

No comments: