Paul Edelman has posted this draft on SSRN.  Here is the abstract:

With the noting of probable jurisdiction in Evenwel v Abbott it appears that the Supreme Court will finally make clear what “one person, one vote” is meant to accomplish. Is it supposed to equate representation between districts, or is it supposed to equate voting power between districts? Having side-stepped this issue for 50 years it is time for the Court to make clear what its doctrine is about.

One possibility raised by the plaintiffs in Evenwel is that, at least in the context of the Texas state senate districts, one achieve both ends simultaneously. That is, the plaintiffs claim that it is possible to draw the districts so as to have both equality of total population as well as equality of citizen voting age population in every district. To support this claim they provide an affidavit claiming it can be done in this case as well as an appeal to the increasing power of the districting software.

If the Court wishes to go down the road of accommodating both interests it may need to be reassured that the option is available not only in Texas but in any districting situation. And even if the Court does not require that districting plans consider the power of the vote, individual states themselves may wish to if they can do that as well as achieve equality of representation. That is what I provide in this paper. I will show that there are very good theoretical reasons to believe that it will always be possible to draw districts that will be simultaneously close in both total population and citizen voting age population (or, indeed, any other additional population that the Court desires.)

Some of this analysis, however, depends on how the Court chooses to assess the deviation in voting power. I argue that the choice of a threshold for total deviation in the voting age population is intimately related to one’s model of voting power. I show that the relationship between the deviation of voting power and the deviation of voting populations is linear in the variable that characterizes the voting model. Thus, the Court must choose, either explicitly or implicitly, a model of voting power should it decide to take the power of the vote into consideration. Even so, I show that the standard of 10% deviation in the voting population leads to a deviation of less than 10% in voting power over a broad range of models.