++Addition++The equation I came up with was faulty. It was actually a bit less accurate, rather than a bit more accurate, than the polls were. It predicted Hillary winning by only 4.4 points, not 8 points. I'm sorry for the mess up.
Going on the assumption that 16% of the turnout in today's Pennsylvania Democratic primary will be black, 4% will be Hispanic, and the remaining 80% will be white*, I plugged the racial/ethnic demographic numbers into a linear equation built on regression data from non-caucus states that have already voted to come up with a prediction for today's results. The method's outcome favors Hillary by eight points, 54%-46%.
My intention isn't to try and upstage the pollsters or show that I have some superior insight into the workings of American electoral politics. The method may miss the mark. Even if I were to be so audacious, it'd be tough to distinguish my prediction from what the pollsters are saying. Real Clear Politics amalgamated average has Hillary ahead by a little more than six points.
Instead, I want to show how important it is to know who is voting. Or, looking at it from another angle, to take note of how relatively unimportant what the candidates say and do in the media, in speeches and debates, at public forums, and in other places actually is.
Obama's relationship with Wright, among other shady characters, has (finally) become public knowledge. Hillary's been ridiculed for the blatant embellishing of her trip to Bosnia while looking 'good' in the most recent debates. Obama has a huge cash advantage going forward. Those things are in addition to all the other perpetual campaign-related stuff that's been going on for the last several months.
And still the last-minute polls, with all that excitement and momentum-shifting factored in, predict virtually the same outcome that'd be expected if only the racial/ethnic characteristics of Pennsylvania's voters were known. Throw in age, income, geographic, religious, and family structure data into the mix and the two point gap between that prediction and what the polls show probably gets even smaller.
* The figures for other groups are too limited to be included, and anyway they'll only comprise a couple percent of the state's overall vote.