Some reporters have speculated about the impact of the "Limbaugh effect" -- partisan Republicans crossing over to vote fr Hillary Clinton solely to help weaken the Democrats against John McCain. The sieze of the effect is hard to measure. But there is one numerical measurement, first pointed out to me by the Pew Survey's Richard Auxier following the Pennsylvania primary, that gives some sense of it.

One exit poll question asks Indiana voters who they would support in a Clinton-McCain contest. 17% of them say McCain. Of those voters, 41% say they would vote for McCain over Clinton. In other words, these voters, 7% of the Indiana electorate, voted for Clinton in the primary but have no intention of supporting her in the fall.

Now, this isn't a precise measure of the "Limbaugh effect" -- no doubt there are some Republicans who backed Obama in the primary out of anti-Clinton sentiment, but plan to vote for McCain in November. But it is a good place to start when making a ballpark estimate. And it's a sizeable number -- 7% may wind up being as big as her margin of victory.

Update: On MSNBC, Keith Olbermann and Norah O'Donnell were discussing the Limbaugh effect. They showed -- I'm going from recall here -- that 11% of Indiana voters were Republicans, Clinton carried them 56-42%, but 63% thought Obama would be a stronger nominee against McCain. Somehow they decided that this proved the Limbaugh effect was negligible. I have no idea how they got that conclusion -- it doesn't prove anything at all -- but Olbermann has been looking for ways to belittle Limbaugh. I propose that political pundits be forced to take a course in high school statistics before going on the air.

--Jonathan Chait