Over the last week, an anonymous blogger who writes under the pseudonym Poblano did something bold on his blog, FiveThirtyEight.com. He posted predictions for the upcoming primaries based not on polling data, but on a statistical model driven mostly by demographic and past vote data. His model predicted a 17-point victory for Barack Obama in North Carolina and a 2-point edge for Hillary Rodham Clinton in Indiana.
Critics scoffed. Most of the public polls pointed to a close race in North Carolina. Looking back at Poblano's efforts in Pennsylvania, pollster Dick Bennett decried the models as "stepwise regression run amok." Slate's Mickey Kaus predicted failure for "a sophisticated model that ignores... what's been happening in the campaign. Like Rev. [Jeremiah] Wright."
But a funny thing happened. The model got it right.
Obama carried North Carolina by 14 percentage points (56 percent to 42 percent) and Clinton prevailed in Indiana by exactly the 2-point margin Poblano predicted (51 percent to 49 percent). Moreover, the predictions were more accurate than any of the pollsters' results, as indicated by the graphic below (modified from a chart created by Brian Schaffner of the Center for Congressional and Presidential Studies). The model was certainly closer to the final result than our Pollster.com trend estimates based on all of the available public polling data (we had Clinton leading by 4 points in Indiana and Obama up by 7 in North Carolina).
I also asked SurveyUSA's Jay Leve how the model would have ranked on the scorecards he created to rate pollster accuracy in Indiana and North Carolina. Had it been a poll, Poblano's model would have been the top ranking "pollster" on 13 of 16 accuracy benchmarks applied in the two states.
How could that be?
When a statistical model succeeds in this way, it is usually because the it manages to quantify reality in some important way. In this case, the fundamental insight captured by Poblano's model is the remarkable consistency of vote preference in the Obama-Clinton race among key demographic subgroups.
Over the course of the primary season, especially since early February, the preferences of Democratic primary voters have been mostly stable. As my colleague Ronald Brownstein put it last week, Clinton has consistently prevailed among a "beer track" coalition of blue-collar whites, Latinos and seniors. Obama consistently dominates Clinton among blacks and younger white voters, and he draws additional strength from a "wine track" coalition of independents and well-educated white voters. What varies from primary to primary is less about the shifting allegiances of voters within these groups and more about differences in the demographic composition of each state.
Consider some specific examples. The outcome of the North Carolina primary was wildly different than Ohio or Pennsylvania, yet non-college white voters favored Clinton by virtually identical margins in each state (+44 in Ohio, +41 in Pennsylvania, and +45 in North Carolina, according to the Edison/Mitofsky National Election Pool exit polls provided by NBC and ABC News). Clinton's margin was far narrower among college-educated whites in each state (+7 in Ohio, +10 in Pennsylvania and +7 in North Carolina). And Obama won near monolithic support from blacks in all three states (87 percent in Ohio, 90 percent in Pennsylvania, 91 percent in North Carolina).
The different overall outcomes owed mostly to the varying demographic composition of each state. Blacks and college-educated whites made up roughly two-thirds of the North Carolina electorate, but only about half of the voters in Ohio and Pennsylvania.
Poblano's model succeeded because it captured the demographic coalitions in evidence from previous primaries and applied them forward to Indiana and North Carolina.
All of which brings us to the underlying story of the Democratic presidential primaries. Since Super Tuesday, it has mostly been the story of what hasn't happened. Over the last three months or so, for better or worse, the underlying coalitions of support for Obama and Clinton have remained largely constant.
So therein lies the bad news and the worse news for the Obama and Clinton campaigns, respectively. The bad news for the Obama campaign is that evidence of an expansion of his coalition is weak.
The far worse news for the Clinton campaign is that evidence that she made "important progress" among white voters in Indiana and North Carolina is weaker still. After weeks of pummeling over the controversies involving the Wright's remarks and Obama's comments about the "bitterness" of rural voters, Obama's coalition remains intact and on track to produce a majority of pledged delegates. The "game-changer" that the Clinton campaign counted on to alter the thinking of unpledged superdelegates did not materialize.