Our frequent correspondent Dafydd ab Hugh has sent us his predictions for tomorrow night. Suffice it to say that he is optimistic; he has President Bush winning by 51% to 47% in the popular vote, and carrying 34 states with 328 electoral votes.
What I really want to quote, however, is Dafydd’s take on the polls, which are weirdly schizophrenic this year:
Typically, pollsters simply take the easy way out and look at the turnout in previous elections; but since elections are unique and often turn on intensely local issues, weighting errors turn out to be the single most common reason why a poll can badly fail to predict results.
In this case, it’s clear that one set of pollsters are assuming that demographics favoring Bush are going to be low, while others are assuming those same demographics are going to be high. That’s how you get polls clumping into one bunch predicting an easy Bush win and a second bunch predicting a Kerry win. In this case, you *cannot simply split the difference*, as RealClearPolitics does by taking the mean average; you must decide which turnout model is going to be correct, the one or the other.
Simply put, I have to choose which set of pollsters to disbelieve.
I believe that the half of the pollsters who are naively using the turnout model of the presidential elections of 1992, 1996, and 2000 are simply wrong; in all three of those elections, Republican vote was depressed by extreme disenchantment with the candidate, the Democrats had a better get out the vote (GOTV) plan, and there was staggeringly high black turnout (particularly in 2000)….
A much more accurate turnout model, I believe, is 2002, despite it being a congressional election: if anything, the GOP GOTV plan is even better than it was in 2002, and every poll has shown black turnout dropping back to the more normal turnout rates of the 1970s and 1980s — they simply are not as fired up by patrician liberal and Boston Brahmin John Kerry as they were by Bill Clinton (“America’s first black president,” as Toni Morrison put it) or Rantin’ Al Gore; and this time, there was no NAACP commercial telling them that Bush personally drove the pickup truck that dragged James Byrd to his death in Texas.
I also believe that this time, evangelicals will turn out hard for Bush, unlike in the last three elections — they sat home in 2000 because of the DUI, they were lukewarm about liberal Blob Dole, and they never liked Bush senior. But George W. Bush has given them both the faith-based initiative and a ban on partial-birth abortion, the first now on hold because of Congress and the second because of the courts. They’ll turn out.
Finally, Florida has a specific difference: unlike last time, there is a Senate candidate, Mel Martinez, who is a Cuban refugee; the entire Hispanic population of Florida, and in particular the Cuban population, will turn out in much larger numbers; since the latter is strongly Republican and even the former is further to the right than the Hispanics in, say, California or Arizona, this is a secret weapon that will definitely lift Bush above the poll average (Bush +1.1%) in that state.
Dafydd also agrees with me that John Thune will beat Tom Daschle in South Dakota.