Polls and Predicting the Presidential Election

The November issue of National Geographic features the predicitions of political prognosticator Jim Campbell. Two of Campbell's factoids struck me as particularly interesting:
  • Four states -- Missouri, Nevada, Ohio, and Tennessee -- have voted for the winner in 13 of the last 14 elections. Since 1948 every candidate who carried at least three of them won.
  • Since Eisenhower's second election in 1956, candidates who carried the under-30 vote won in all but two (9/11) presidential contests.
Current polls show Bush leading in three of the four states mentioned above, which would make him a shoe-in to win, right? Not exactly. First, although Bush now leads in Missouri (50-45), Nevada (52-45) and Tennessee (60-38), Kerry has led in polls in each of these states during the last six months, even in Tennessee, which was a virtual tie earlier this month, and which Kerry led in July and August. Ohio has been back and forth, but Kerry now leads by 6 points. Whoever does a better job of getting out the vote will win these states.

Second, polls are unreliable. They don't take into account newly registered voters, or voters who didn't vote in the last presidential election. Further, they skew more heavily towards Republican respondents, and are more likely to exclude young voters, which leads to the second point: whoever wins the under-30 vote is likely to win. A recent Harvard study shows that Kerry leads among college students. I suspect this is true among older twenty-somethings, too.