First, from Steve Landsburg:
To understand Mitt Romney’s tax burden, you have to compare him to his doppelganger Timm Romney, who lives on a planet with no taxes. In the year (say) 2000, Mitt and Timm both earned (say) a million dollars. Timm invested his million dollars, saw it double over the past decade or so, and cashed out his investment this year, leaving him with two million dollars. Mitt, by contrast, paid 35% tax in 2000, leaving him with $650,000. He invested it, saw it double, and cashed out last year, paying 15% tax on the $650,000 capital gain. That leaves him $1,202,500, which is about 60% of what Timm’s got. In other words, the tax system costs Mitt almost 40% of his income.
Second, from John Cochrane:
If you made money in dollars, paid taxes, then went to Canada and got $1.20 Canadian, it would make no sense to say "you made 20 cents of income, we'll tax it." It makes no more sense to pay taxes again on money that is moved over time. We decry that Americans don't save enough, the Chinese, the trade deficit and so on. Well, if you want people to save more, stop taxing it.
Make it three. Here is Mankiw:
All of these calculations are static. They ignore the general-equilibrium effects that arise as the true burden of taxation is shifted by behavioral responses. In essence, these calculations are made under the implicit assumption that factors of production are supplied inelastically, so the tax stays where legislators put it. Of course, that assumption is implausible, especially in the long run. True general-equilibrium tax incidence is very hard, and as far as I know, reliable estimates on it are not readily available.