Thoughts on Prediction 6

Commenter OBloodyHell thinks this week's prediction is too pessimistic about McCain's chances, citing the Strata blog and Friday's Wall Street Journal. Essentially both argue that poll results depend on the sample population, particularly the definition of "likely voters;" both show McCain does better among voters traditionally classed as "likely" (as opposed to the larger universe of voters who claim they are "likely" to vote). The WSJ highlights Gallup's decision to conduct separate surveys using the traditional and self-selected "likely" definitions (plus a third tied to "registered" voters). ZombieTime says all the big pollsters undercount conservatives, for various reasons.

For what it's worth, House Speaker Nancy Pelosi (D-Ca.) says Obama's a lock. The media agrees. Sunday's Gallup poll claims Obama "widened" his advantage among registered voters to 10 points. Rasmussen gives Obama a 6 point lead. And FiveThirtyEight.com predicts almost the same electoral vote distribution (344-194) as my post (349-189), though that site tends to favor Dems.

Yet, there are signs the race is far closer, such as the Real Clear Politics "Battleground poll":


source: RCP
Zogby's poll published Sunday shows Obama's lead slipping to 3 points:

"For the first time in the polling McCain is up above 45 percent. There is no question something has happened," Zogby said.

Gallup's latest sample among traditionally defined "likely" voters reflects the same 3 point spread:





source: Gallup Daily

Still, national polls are irrelevent, as are blue states (Mark Steyn calls the latter "red states with a couple of big blue cities"). Due to the Electoral college system, only battleground states matter. I've identified six McCain must take to win: Nevada, Colorado, Missouri, Ohio, Virginia and Florida. So Sunday, I checked local sources about those states. I heard optimism about McCain's chances in Nevada, Ohio and Florida. Questions about Missouri (which Bush won twice but where Obama now leads) provoked uncertain bewilderment (as did North Carolina, now a toss-up). But Democrats were confident, and Republicans non-committal, about Colorado and Virginia, where polls favor the Democrat.

So, finally, my point. The '08 election scenario is easy: McCain has to win at least the states Bush took in 2000 (where he won by one), or create new red states to counter blue trends (such as New Hampshire (4 votes)). Which means, as a minimum, that he needs the same level of conservative support that voted "R" eight years ago in "purple," i.e., battleground, states. Does McCain pass that test? Though he did in the glow of the Palin nomination, Gallup says conservatives are leaking in Red States and fleeing from purple:





source: Gallup

Simply put, Obama "excites"; McCain doesn't.

Conclusion: OBH, Strata and the Journal are persuasive about the centrality of the definition of likely voters. So I'll pay more attention to Gallup's "traditional" sample spread (which, unfortunately, neither RCP nor FiveThirtyEight track). But Gallup shows McCain still behind nationwide, and he would have to win each of my six key states (plus North Carolina) to prevail by four electoral votes. Forget poll bias; never mind conservative pundits--McCain can't sweep purple states without an energized conservative base.

Yes, the "elephants in the room" are vote fraud and the so-called "Bradley effect." (Biden's already on the latter bandwagon, though there's contrary anecdotes.) Which I'll address next Sunday.