Pigouvian taxes for obesity?

I have always been a supporter of Pigouvian taxes (to replace other kinds, not in addition to them). The basic idea is that you tax things that have negative externalities to compensate society for the extra costs. If a company emits greenhouse gases, you can either fine them directly for the amount of the cleanup (if you can calculate it) or indirectly as a carbon tax. If someone smokes, you can tax their cigarettes based on what you expect them to cost society in added health care costs (i.e. Medicare). The added benefits of this approach is that people are discouraged from doing these things, so you get less smoking and less emissions.

Some recent debates have got me thinking about how to apply the idea in reality. There has been a lot of debate lately on curbing obesity because of the $150 billion per year that obesity costs the US healthcare system. Can we use Pigouvian taxation? We would have two choices.

We can be proactive by taxing foods that the best science tells us lead to obesity. But this gets tricky for several reasons. First, is the science reliable enough to tell us? Nuts are very healthy in moderation, but not in excess. Should we tax those? Second, what about skinny people who enjoy a Big Mac and never get fat? Why should they have to pay taxes for externalities they never create? My biggest fear is that lobbyists would have a field day influencing Congress on which foods to tax (or not to). McDonalds v the pork industry v Coke v Hostess . . .

The second option is retrospective. We can charge higher insurance premiums for people who are obese and use the extra money to pay for the externalities that these individuals create. This is more direct and therefore more fair economically. But of course there would be a huge debate as to whether this is discrimination. We would have to have very accurate actuarial data on the increase in costs. The second problem with this approach is behavioral. It would not be nearly as effective in reducing obesity because it is unlikely someone would push away from the McDonalds counter because of the fear of future increases in insurance premiums. A tax right then and there would be more salient.

Despite all these challenges, I suspect that either one is better than using income taxes to raise the money that pays for obesity related costs. But because of the complexity, I have this nagging feeling that Congress would find a way to make it worse.