[Update] This diary had some serious analytical flaws that I wrote in the heat of the moment. Its still important, but I didn't prove what I thought I had. [\Update]
According to page 3 of CNN's Exits as of 7:10, almost 20% of the Democratic primary voters plan to vote for John McCain in November [Update] over Obama [\Update].
Hillary's margin among them is 87-12. So three quarters of that 20% of the vote is 15%. That means 15% of whatever margin she racks up tonight is voters [Update] who very well might not vote for her. [/Update]
[Updatex2] This is where I really messed up. Since some of those Hillary voters might order their preferences Clinton-McCain-Obama we can't really distinguish between Clinton voters who really dislike Obama and strategic voters among those 20%.[\Update]
<Update>Analytically what we really need is a question that asks whether people would vote for McCain over any Democrat. Or we could use a question which asked voters whether they would be happy if Hillary Clinton won the Presidency, not just the nomination. Even a head to head between Hillary and McCain would be nice but its not there in the exits I see.
So take your expectations about how the distinction between strategic voters and Clinton-McCain-Obama voters breaks down and readjust the 15% I originally claimed accordingly.
Also note that this means when the press maw-maws about the fact that fifty percent of Hillary's voters aren't going to support Obama, that's not necessarily because they're Democrats taking their balls and going home, its because they always intended to vote Republican. If all the 18% of the electorate or so that really voted for Clinton despite the fact that they'll vote for McCain against Obama in November were voting strategically, then that is more than a third of her supporters, which means that 33% should be added on to the percentage because neither Democrat had a chance at those voters in November</Update>