Y'know, the funny thing about John Miller's NYT op-ed (see full commentary and links at Instapundit) on Libertarians swinging Senate races to Democrats (a topic I've blogged several times below) is that this year's exhibit, South Dakota Libertarian Kurt Evans, dropped out of the race and endorsed John Thune. He dropped out too late to be removed from the ballot, and the 3,000 votes he got nonetheless were more than the Thune-Johnson gap.
First, that means that it's possible that Evans actually swung some votes Thune's way, and Thune lost anyways.
Second, [political scientist's hat on:] the fact that Evans had dropped out and endorsed Thune almost certainly means that those 3,000 voters were disproportionately not Libertarian-Republican swing voters. All those voters probably swung to Thune. The remaining 3,000 were either hardcore libertarians, who would have stayed home or left the Senate race blank rather than vote for either major-party candidate; or Libertarian-Democrat swing voters who didn't find Johnson sufficiently pro-civil-liberties but who wouldn't have voted Republican in any event. Under usual circumstances I think that Libertarian candidates draw more otherwise-Republican votes than otherwise-Democratic votes (though not by nearly the margin that Greens draw otherwise-Democratic over otherwise-Republican votes). But these weren't usual circumstances; Evans had already endorsed Thune. It's therefore actually more likely that Evans' absence from the ballot would have increased Johnson's lead than that it would have decreased it.
See more from Radley Balko, Eugene Volokh, Clayton Cramer.