While I could be in favor of some caron tax (definately not a retail gas tax hike like Mankiw offers however), Friedman makes the point that I've made before that in times of war especially, increases in gas taxes are often not offset by decreases in other taxes.
Mankiw takes issue with his stance as a slippery-slope. Here is my response to Mankiw:
As I tried to get across in your "politics aside post," it seems to be completely biased that you as a new keynesian economist advocating normative policy can say, "let's focus on individual's likely behavior and call it good normative policy." The problem is that this completely ignores the likely (historically, esp. in times of war) behavior of governments when face with the idea of a large ambiguous-effect carbon tax. It is simply not a fair logic to ask people to put likely political action aside when making up ones mind about whether or not to normatively advocate a policy. This is of course why, when it comes to policy decisions which must necessarily extend past the blackboard, academic economists should offer multiple Postitivist solutions only (sets of possible solutions and probable outcomes) and give policy makers the assumptions and data behind those solutions.
Economists should butt-out of having a direct hand in NORMATIVE policy PRECISELY because most economists tend to narrowly focus only on what they want to see (gas tax potentially effecting consumer behvior) and they completely ignore all else (government political behavior, etc). Let economists have their opinions, but don't let them push their narrow minded normative agenda. Let them merely contribute via postiive economics.
So I say kudos to David Friedman for being one of the few academic economists who "gets it."
Also, I take issue with the characterization of his argument as being slippery slope. It seems to me that that is just code word used by people who don't like thinking in shades of gray. The fact is, reality isn't always pretty, and decisionmakers need to be aware of all the likely outcomes of a change in X. Will they anticipate all outcomes - no. But it's a heck of a lot better than focusing on only one effect that X has and ignoring all else.