Two recent papers on top-earner taxation have made an important contribution to the policy debate on the topic, but it seems to me that we still have some way to go before we have an understanding of the phenomenon that is robust enough to use as a basis for policy.
The price-taker assumption can perhaps be interpreted as an identifying assumption. If demand is less-than-perfectly elastic, the elasticity of taxable income becomes a function of the parameters of both labour supply and labour demand. For the purposes of calculating the revenue-maximising rate, this doesn't really matter: the reduced form is all you need.
But if you're interested in the incidence of the top-end tax - and if you're talking about distributional issues, you probably are - then how the elasticity of taxable income unpacks into its labour supply and labour demand components matters a great deal. I've discussed this point several times (,,); Tyler Cowen and Richard Green have done so recently. The incidence of increased taxes on top earners remains an open question.
The even more recent paper by Piketty, Saez and Stantcheva (pdf, VOX version) makes some progress in unpacking the elasticity of taxable income. In their model, lower tax rates increase the incentives for top earners to bargain harder in order to extract larger rents. And since those top incomes are rents, there is no efficiency loss in taxing them, and the incidence of the tax is presumably borne entirely by the top earners.
But there's another identification problem. The rise in top incomes coincides with a general reduction in marginal top rates, which is consistent with the Piketty-Saez-Stantcheva narrative. But they also coincide with a general increase in firm values, which is consistent with a narrative in which executives are being rewarded for generating shareholder value.
It's hard to make reality-based policy when available data support both one conclusion and its opposite.