"While establishmentarians continue to push two-party conformity," reads the tag line on Matt Welch's piece at Reason today, "there remains little evidence that other parties are having any sort of "spoiler" effect."
I hope he's wrong.
It's usually hard to tell, because there's not universal and rigorous exit polling to find out which candidates third party voters would have supported if they hadn't had third party candidates to vote for. When that kind of polling is done, it often indicates effects opposite of those assumed (for example, the complaint that Libertarian candidates always take their votes "from" Republican candidates).
In 2000, Libertarian Harry Browne racked up enough votes in New Mexico that, if he did "take" most of his votes from George W. Bush, cost Bush the state. And if Bush had carried New Mexico, Florida (where Browne knocked down, IIRC, 12,000 votes or so) wouldn't have mattered.
In my opinion, Libertarian candidates should EMBRACE the "spoiler" role, whether it's plausibly true in any given case or not.
But our asses to "cover the spread" by enough to seemingly matter, then let the Republicans and Democrats scream about it. It makes them look like whiners, it makes Libertarians look more powerful than perhaps we are, and it forces "major party" candidates to think about what they might offer Libertarian voters to reduce that "spoiler effect."