As I showed here, in Indiana, there were about 110,000 pro-McCain voters in the primary, 75% of whom supported Hillary Clinton.

How might these voters have impacted the demographics of the exit polls?

Without the cross-tabs, we can't be sure, but it's reasonable to estimate that that nearly all of these pro-McCain voters were white. After all, in 2004, 99% of Bush's support in Indiana was white. Assuming they are all white, then:

• 83,653 white Clinton voters were actually McCain backers
• 27,544 white Obama voters were actually McCain backers

Also from the exit poll, we know that:

• 596,790 Clinton voters were white (93% of her total)
   
• 397,860 Obama voters were white (63% of his total)

By subtracting the McCain backers from each candidate we learn that:

• 513,138 white Clinton voters were sincere
• 370,316 white Obama voters were sincere

Thus, excluding McCain meddlers, Clinton won among white voters 58% to 42%. That 16-point margin is significant, but it's also 4 points smaller than her 20-point 60%-40% lead when McCain meddlers are included.

Also, the 16-point margin is almost half that of the 30-point margin in Ohio, suggesting that the whole narrative about Obama losing support among whites is completely false. In fact, Obama is making significant progress.

The same pattern holds true in North Carolina. Including McCain meddlers, Clinton won by 24 points among white voters, 61%-37%. After excluding McCain meddlers, she still leads, but her margin shrinks to 18 points, 58%-40% -- also much closer than in Ohio.

Of course, since these numbers don't take gender into account -- Clinton's lead among white voters is attributable to white women -- it's questionable how meaningful they are. But given the Clinton campaign's obsession with Clinton's seemingly magical ability to win among white voters, it's useful to know what the real numbers are. And as these numbers show, the gap is closer than most people think -- and it seems to be closing as the campaign goes on.