Chris Elmendorf and Doug Spencer have posted this draft on SSRN, and it is potentially relevant to the Shelby County Voting Rights Act case pending before the Supreme Court.  Here is the abstract:

The geography of voter discrimination—that is, the tendency of voters within different geographic units to discriminate against minority candidates on the basis of the candidate’s race—is of central importance to the adjudication of claims under Section 2 of the Voting Rights Act, and to the constitutionality of Section 5. To date, however, efforts to create fine-grained estimates of the geography of voter discrimination have been hindered by the modest sample sizes of national surveys. Using multilevel regression with post-stratification, we create estimates of anti-black stereotyping at the state and county level, and we show how these estimates can be used to evaluate (or redesign) the coverage formula for Section 5 of the Voting Rights Act, and to implement Section 2. Among our principal conclusions is that the coverage formula for Section 5 is remarkably well tailored to the geography of anti-black prejudice. If the Supreme Court invalidates the coverage formula in Shelby County v. Holder, Congress could re-enact a substantially similar formula on the basis of our findings. Note to readers: due to some delays in obtaining data, we are not able to include county-level results in this draft. We are posting this incomplete draft because of the timeliness of the state-level results for Shelby County and potential congressional responses.