If you tax cigarettes, then cigarette tax revenue tends to rise. At the same time, the higher taxes may inspire people to quit smoking or to buy packs on the black market both of which tend to make revenue fall. There's some revenue-maximizing price point out there beyond which higher taxes produce lower revenue. And the same, one would think, applies to income taxes. But where is that number? Christina Romer and David Romer have a new paper looking at evidence from the 1920s and 1930s and find that the revenue-maximizing rate on the highest earners is extremely high—over eighty percent. Among the top 0.05 percent of the income distribution they find an elasticity of taxable income of 0.19 percent which implies "implies that tax revenues would be maximized with a tax rate of 84 percent; that is, you could raise taxes up to 84 percent before people’s reduced incentives to make money would compensate for the higher tax rates."
This follows similar empirical results from Piketty and Saez and suggests that the cutting edge argument for lower marginal tax rates on high income individuals is going to have to shift to a hypthesis that high marginal rates have a deleterious longer-term impact on labor supply.