Advocates for redistricting reform may have felt disappointed last week when the Supreme Court sent this year's big partisan gerrymandering case, Gill v. Whitford, back for a do-over in lower court. There's still hope for broad-based reform—but only with the help of data science.
The unanimous decision, written by Chief Justice John Roberts, held that statewide measures of harm were unsuitable as proof of gerrymandering because the plaintiffs had claimed a violation under the Equal Protection Clause of the 14th Amendment. Under that theory, the Court said, the plaintiffs should identify at least one victimized voter in each gerrymandered district. Because the decision was unanimous, a distinct possibility exists that plaintiffs will succeed when the case returns to the high court.
In the abstract, this requirement might not seem like a big deal, but it raises practical barriers to proving a gerrymander. The most obvious one is procedural: In the current case, the 99-seat Wisconsin Assembly will require dozens of new plaintiffs. Many state legislatures are even larger, with as many as 400 seats. Even worse, lawyers will have to prove that each of those plaintiffs could have been drawn into a more competitive district where they had a chance to elect a representative of their choice.
Such a demonstration requires data—and lots of it. In order to evaluate partisanship in maps of contested districts, an enormous amount of granular data is required. A state the size of Wisconsin has thousands of precincts which contain critical partisan election data. Precinct boundaries are not kept in a central repository, but maintained by counties, often only on paper. This enormous data requirement imposes a barrier to citizens and organizations who want to get involved in redistricting, not only in Wisconsin but in any other cases like it in the future.
Data-gathering is, to put it mildly, onerous. Here are examples from two gerrymandered states. In Trumbull County, Ohio, precinct maps are kept in hardcopy form on 11" x 17" sheets of paper. In the city of Hampton, Virginia, a single large-scale precinct map, suitable for framing, is available for $10. In both cases, the maps have to be requested in person. Then, for use by redistricting software, these maps still have to be converted into a computer-usable digital format.
These are extreme examples, but even straightforward cases require making phone calls to busy county clerks and making Freedom Of Information Act requests. We estimate that collecting and maintaining precinct data for one medium-sized state would cost tens of thousands of dollars. Covering the whole nation would likely cost millions.
This enormous data requirement imposes a barrier to citizens and organizations who want to get involved in redistricting. State governments, legislatures, and party organizations are in the best position to gather this information. But they also come under the greatest suspicion for partisan malfeasance. Conversely, independent reform organizations can potentially play a watchdog role, but are hamstrung by the considerable upfront financial burden of data collection.
One route to reform would be passing a law requiring a state to make precinct data available for public use. Another would be to combine the efforts of academic and reform organizations. Groups across the country, including Princeton University, MIT, Tufts University, the University of Florida, and the OpenElections project, have begun such efforts. Ideally, this data-gathering should be done in a coordinated manner to avoid duplication of effort.
Even if the Supreme Court eventually declines to limit partisan gerrymandering, all this data will still be useful. A promising front in the battle is state-level action. Already, Pennsylvania has redrawn the congressional map for 2018 based on its own Supreme Court’s reading of the state constitution. This action is immune from federal intervention, since it is not based on federal law. And other state constitutions, such as North Carolina, have provisions that resemble the Equal Protection Clause, as well as First Amendment-like provisions, which Justice Kagan has cited as a possible basis for reform.
The work of data scientists can empower citizen-led movements for new legislation where reform efforts are growing. In Michigan, the citizen group Voters Not Politicians has successfully gotten a measure onto the November ballot which would establish a redistricting commission. That commission will need input from communities of interest across the state.
In Virginia, which lacks an initiative process, the group One Virginia 2021 must persuade legislators, the beneficiaries of gerrymandering, to pass bills or a constitutional amendment. There, precinct-level data will help them make concrete demonstrations of what a fairer map would look like. The same data can help local communities get involved in a current racial gerrymandering case, Bethune-Hill v. Virginia Board of Elections, where the construction of alternative maps plays a key role.
Uses of this data are not restricted to drawing individual maps. The data can also be used in courts to identify outliers. Powerful algorithms can rapidly explore billions of possibilities and evaluate whether a proposed plan is fair or extreme. And the same data can be used not just to combat gerrymandering, but also to address other questions in which local communities have a stake, such as the drawing of school districts.
This year’s passion for fair districting can drive change, and data science can play a key role. Although the Supreme Court has imposed a large burden on reformers, responding properly can eventually make government more responsive to voters at all levels.