Both Matt Yglesias, Karl Smith, and Mark Thoma and James Kwak are pointing out the (previously mentioned) Romer and Romer research that shows a fairly low income elasticity (0.19) to increases in marginal tax rates during the inter-war period in the U.S. This points to a top of the Laffer curve (at least back then) of 80+%.
This basically agreeswith research from Saez, Slemrod, and Giertz that looks at later periods and finds a top of the Laffer curve in the 70+% region.
So if our goal is to raise revenue, a top marginal tax rate on the 0.1% of around 70% probably makes the most sense. In our convoluted federal tax regime, this would probably mean a top marginal tax rate of 50-60%.
Could there be an argument against doing this? Yes, of course. One is the libertarian argument that taxation is never ok and so should be kept at an absolute minimum. There are two main counter arguments for that. First, the very rich could not have become very rich without the investments that society has made in public infrastructure, education, peace, etc. The more you have benefited from those investments, the more you should pay. Second, inequality really seems to be a negative externality of the free-market system (see the Spirit Level for some data) that undermines society and democracy. In that case, higher marginal taxes serve as a Pigovian tax to reduce this externality.
The other argument against very high top marginal taxes is that it may reduce growth. The argument goes that higher taxes reduce investment in productive technology which in turn reduces growth. This is purely an empirical question and I’m not aware of any research that supports this conclusion. In fact, in the U.S. increased inequality over the last 30 years seems to have led to slightly lower growth. But this is an area where I need to know the literature a little better.