Posted by Jed Lewison on Wed May 7, 2008 at 5:15 AM Pacific

Methodology: McCain supporters in North Carolina and Indiana

This post explains how I arrived at the numbers in "McCain Supporters Deliver Indiana To Clinton."

::

For the first time in the 2008 primary season, the North Carolina and Indiana exit poll surveys asked voters who they would support in hypothetical general election matchups between McCain and Clinton as well as between McCain and Obama.

Because all voters were asked about both matchups, it is possible to determine how many voters plan to vote for John McCain in the general election even if the candidate they voted for in the primary wins the Democratic nomination.

Typically, the concern in a divided primary is whether candidate A's supporters would be willing to support candidate B if candidate B won, or vice versa. This question allows us to see if candidate A's "supporters" are actually supporters of candidate C.

For these voters, perhaps candidate A is a legitimate second choice to candidate C, or perhaps they think candidate B is a stronger candidate. In either case, however, since they plan on voting for candidate C in the general election, they really shouldn't vote in the primary, and if they do, their votes ought to be discounted.

The following two tables show the raw data from the North Carolina and Indiana exit polls. Final two sets of data (by % of all voters and by % of column) are derived from the first two (% total, by % of row).

The cells with bold red text are voters who intend to vote for McCain in the general election even if the candidate they supported in the primary wins the nomination.

To determine their impact on the general election final results, I subtracted from each candidate the relevant figure in the "by % of all voters" columns, and then recomputed the percentage split.