He has this theory of "idea traps," in which bad ideas lead to bad outcomes which lead to populists with more bad ideas, etc. If libertarian ideas are as good as we say they are (think of how well they worked when they were tried in our Constitution), then maybe the economy would do so well that populist schemes would lose support.
This exaggerates my original claim. I've argued that economic growth improves ideas, which improves policy, which improves growth. But I've never said that the elasticity is so great that libertarian reforms are likely to take a slippery slope to Libertopia.
Still, Arnold is pointing in the right direction. The Idea Trap mechanism is part of the reason why I guessed that a libertarian takeover of academia would have such a large long-run effect on policy.