Tuesday, November 04, 2008

Obligatory predictions post

Apologies to all for my extreme negligence as a blogger over the last month. I have all the usual excuses (too much coursework, too many activities...), but I still should be more conscientious about posting, especially given all the astonishingly loyal readers who, according to my site statistics, keep coming back day after day to find nothing.

Of course, I can't miss the opportunity to add my voice to the 10 zillion other bloggers who have already posted their electoral college predictions, and thus I feel compelled to post tonight. My best guess for the map tomorrow is this:



My justification? Well, for the most part this is a straightforward reading of current poll results -- I'm not trying to display any mysterious predictive genius. Subjective judgment is only necessary in the few states that lie close to the margin between Obama and McCain. Explanations follow:
  • Colorado, Virginia, and Nevada: By all indications, Obama is leading by at least 4 percent in all of these states, and it's hard to imagine McCain closing the gap, especially when he's explicitly written these states out of his strategy. Widespread early voting makes any sudden change close to impossible.
  • Pennsylvania: McCain's last-minute change in strategy shifted the focus to Pennsylvania, considered more tractable than most other states thanks to its lack of early voting. Unfortunately for McCain, Pennsylvania is solid Democratic territory, hovering at a few percentage points higher than Obama's national total. The demographic fundamentals here are simply too strong for a McCain surge to be plausible, barring some dramatic last-minute shift in the race nationwide.
  • Ohio and Florida: Here the picture becomes a little blurrier. FiveThirtyEight.com projects +3.3 for Obama in Ohio and +1.2 in Florida, both of which seem reasonable. If there is indeed any kind of "Bradley Effect," one might imagine it to strike in states like Ohio, with its downstate overlap with Appalachia, or Florida, which is in some ways still a very Southern state. Still, they both have nontrivial black populations (~15% in Florida and ~12.5% in Ohio), whose turnout has probably been underestimated by conventional likely voter models, and they both also have statewide in-person early voting, which makes any last-minute swing less likely. All considered, I think that Obama is definitely going to win Ohio, and almost certainly going to win Florida.
  • North Carolina: North Carolina projects on FiveThirtyEight.com as +0.6 for Obama, but I'm just as optimistic about its prospects as I am for Florida. First, North Carolina has a healthy early voting program, locking in a decent percent of the vote based on voter opinion over the last few weeks, which has slightly favored Obama. As with many other states, this also makes a last-minute shift away from Obama a little less likely. Second, and more importantly, North Carolina is almost 23% black, and polls' underestimation of black turnout by even a few percentage points may provide a crucial boost for Obama above his current numbers. Finally, when states are this close, a wide range of intangible factors come into play, and although some are negative for Obama (the "Bradley effect"), I suspect that the potentially enormous cell-phone gap is more significant.
  • Missouri: This is the closest state of all, projecting at +0.3 for McCain on FiveThirtyEight. It's hard to say anything definite here -- some of the factors favoring Obama elsewhere, like large black populations and statewide early voting, are missing -- but just as in North Carolina, I suspect that the cell gap outweighs any Bradley effect in Missouri. (Indeed, the same should happen nationwide...) My best guess is that Obama will edge out McCain by a few tenths of a percent here.
  • Indiana: It's still a bit of a mystery to me why Indiana appears so competitive. Bush won by over 20% there in 2004, and for many decades Indiana has been as deep-red a state as any in the nation. FiveThirtyEight.com's regression, which is a good summary of where the demographic and political fundamentals of a state suggest that it should fall, puts McCain up by 4.5% in Indiana. And yet the polls suggest that it's competitive! In the end, this is probably some kind of "neighbor effect," where Indiana's proximity to Illinois increases Obama's competitiveness. Given the surprising divergence of demographic fundamentals and the polls, though, I wouldn't be surprised if the polls have been -- for whatever reason -- mistakenly putting Obama a little higher here than he really should be. FiveThirtyEight and Pollster already project McCain to win by .5-1.1% anyway, and together with the concerns I've expressed I think this is enough to conclude that McCain will probably take Indiana.
  • Montana and North Dakota: The lack of polling in these states leaves them with an air of mystery, but the polling we do have suggests that McCain is probably up by 1-4% in each, and this is consistent with my intuition, which doubts that the political character of these states has changed enough for them to go Obama just yet, unless there's a much broader landslide nationwide than anticipated.
  • Georgia: I, for one, believe that the astonishing early vote totals and the aggressive voter registration and turnout campaigns in Georgia will make the race closer than a quick reading of polls would indicate. Still, at this point I think that McCain's margin is probably too large to overcome: FiveThirtyEight projects him at +4.5%, and every poll from the last few weeks has shown him ahead. Georgia may be tantalizingly close to turning blue, but it's not there yet.
  • Arizona: It would be great... but there's no way.
The final verdict, then: a margin for Obama of 364-174 in the Electoral College, plus 6.7-7.0% in the popular vote, where I think that Obama will slightly outperform the predictions of national polls thanks to a combination of the cell-sampling problem, a far superior ground operation, higher black turnout than anticipated, and a "Bradley effect" that is less significant than the preceding factors.

Essentially, my predictions just take the FiveThirtyEight state-by-state projections and shift them toward Obama a little to account for my belief that Obama will do slightly better across-the-board than polls predict at this point. (Yes, I know... you could have just skipped the long-winded and repetitive discussion, because this is the core of my analysis!)

We'll see what happens in a few hours.