Anyway, here is an approximate reconstruction of my comments:
Assigning me to comment on this paper is a great example of the NTA at either its best or its worst. That’s for all of you to judge, depending on how much value you think I add.
Reflecting the NTA’s interdisciplinary aspirations, we have here a lawyer – that would be me – commenting on an economics paper that is full of math, including numerous propositions and proofs. In the words of Martin Short, playing a Hollywood agent in a movie called The Big Picture with respect to a stack of scripts that his client was naïvely hoping he had fully evaluated, “I read almost all of them almost all the way through.” But perhaps I can try to add value in a different way. In particular, I’d like to offer six quick comments in the 5 minutes that I have available today.
First, the paper is interesting and important because the question it examines – how optimal tax analysis might be affected if the incidence of a personal income tax may be shifted, such as through effects on pre-tax wage levels – clearly matters, yet has often been ignored. But it’s a shame that the paper is interesting and important! Life would certainly be easier if we could simply assume away any such effects. The paper is convincing, however, when it argues that we cannot necessarily do this.
Two areas in which the literature has wrestled with income tax incidence are the taxation of saving, and corporate income taxation. As to the first, if the overall level of saving doesn’t respond much to the taxation of saving – which some evidence, admittedly not entirely conclusive, suggests – then incidence-shifting would presumably be slight. As to the latter, the reason we need to worry about the incidence of the corporate tax is not that it’s paid by a legal entity – pretty clearly, that makes it a tax on the shareholders who own the residual – but rather that its being shifted is highly plausible, especially when national-level corporate taxes are operating within an integrated global economy.
Second, in a sense that the paper nicely explains, incidence-shifting with respect to high-end wages makes the tax on them what I will call “as if lower.” If pre-tax wages at the top of the income distribution rise in response to highly graduated rates, that partly reverses both the intended distributional effects and the tax’s dampening effect on labor supply. So in some respects it is as if the tax rate were lower than if there were no incidence-shifting through wage effects.
This causes me to entertain the intuition that perhaps the optimal rate might turn out to be higher, rather than lower, in the presence of wage effects, so as to get back to approximately the same place. But in the paper’s model it comes out the other way. This presumably is due to how things work out with respect to the lower 99 percent of the wage distribution. It would be nice for the paper to explain more fully what gives rise to its result, in which wage-shifting lowers, rather than raises, the optimal rate.
Third, a question of interest is how robust the paper’s finding would be to other inputs into the determination of optimal tax rates. Suppose, for example, that optimal high-end rates reflect the policymaker’s belief that extreme high-end income or wealth inequality imposes negative distributional externalities on other in the society. It seems plausible to me that the pattern of effect would be similar (lowering the optimal rate relative to the case of no wage-shifting). But given my prior comment, I suppose I shouldn’t assume that my intuitions about how the model works are correct.
Fourth, the paper assumes competitive high-end wage markets, while noting that a desirable extension might be to test the consequences of changing this assumption. Clearly, this could be important, given that there may in fact be non-competitive wage markets at the top of the distribution. However, one challenge in incorporating this extension is that we don’t necessarily understand how noncompetitive wage markets actually operate. Suppose initially that one is extracting the maximum available rent, and that changing the high-end tax rate won’t affect this. That’s a pretty simple case to evaluate (the optimal rate presumably should be higher than otherwise if the pre-tax wage is fixed), but it is only one possibility.
But suppose instead that, say, CEO salaries set by sweetheart boards of directors are more responsive than this. Might it conceivably go in either direction? One scenario is that the board restores the CEO’s after-tax position by raising the pre-tax wage. But for an opposite scenario, consider the claim (made, for example, by Thomas Piketty) that lowering high-end individual marginal tax rates led to higher pretax wages for CEOs, as it was now more worthwhile playing the entire over-compensation game. Whether one agrees with that claim or not, it’s a real world claim about how CEO wages respond to marginal tax rates that might actually come out predicting that the pretax wage is lower, not higher, if the tax rate on the CEO goes up. But my main point here is simply that we have to understand how particular noncompetitive wage markets work in order to be confident about how best to extend the paper’s analysis to such settings.
Fifth, the paper assumes away special sectoral taxes in response to noncompetitive wages in particular settings, noting that these would present both administrative and political economy challenges. But at least as a political economy matter, it’s unclear if setting high wages in particular settings where the markets aren’t functioning well would be harder or easier than raising high rates generally. For example, you might avoid needing to fight with the entire high-wage sector if only some of them are having their taxes raised. My point here, however, is just to note that political economy effects can be complicated, hard to model, and can go in various directions.
Finally, a quick point about the earned income tax credit or EITC. As the paper notes, if it bids down pretax wages at the bottom of the distribution, that would partly undo the intended distributional effects of the EITC. Depending on one’s reasons for having the EITC, this might not be all bad. For example, suppose that reduced pretax wages trigger expanded low-wage employment through their supply side effects, and that this would be desirable for other reasons (e.g., building the workforce affiliation, skills, and habits of low-wage potential workers, in ways those individuals have not anticipated). This might further the EITC’s aim of increasing low-wage employment. Then again, there might be lots of other ripple effects on wages in the low end of the distribution that one would want to think about. (For example, suppose it also drove down other low-end wages.)
Again, my point here is simply that the paper’s unavoidable embrace of new areas of complication points the way to yet more complications that we may want to try to add to the analysis.
In sum, this is a good paper that should stimulate extensions and further thought.