In reviewing some of the comments in the Oregon thread, I realized that there were a couple of obvious things I could be doing to further improve my turnout model:
(i) Adding a time trend -- turnout has been VERY high in the past several primaries,
(ii) While still using the percentage of the voting-age population as the explanatory variable, include the number of Kerry voters in 2004 as an independent variable.
This increased my turnout estimate in Oregon to about 660,000 voters. Obama also gains about a point in his statewide margin against Clinton because we're now showing turnout to be especially heavy in the Portland areas.
So our "official" projections are Obama by 13.8 points, and turnout of 661,470. Full results are below. Kentucky projection to be up later this morning.