top of page

How to spot a gerrymandered district? Compare it to fair ones.

Excerpted from the Harvard Gazette. Kosuke Imai is a Faculty Affiliate of the Harvard Data Science Initiative and a member of the HDSI’s Causal Inference Working Group. Read the original article here.


New waves of statisticians, including a team at Harvard, have developed tools they think can help police the longstanding problem of gerrymandering of congressional and legislative districts in states by parties seeking to tip the scales for their candidates.


One method started at Harvard in 2020 by Kosuke Imai, Professor of Statistics and of Government, has been quickly making an impact. It’s been used by researchers, journalists, and election analysts, and has played significant roles in recent legal cases where legislators were forced to throw out gerrymandered maps.


Called “redist,” the tool creates a vast pool of alternate nonpartisan plans (upwards of 5,000-10,000) that can be compared to a map that’s being proposed of has already been enacted by local legislators or redistricting committees. This pool of nonpartisan baseline maps makes it possible to see whether the new map fairly represents the new shifts show in the Census, or is an outlier.


“What the algorithm does is that using the geography and distribution of different voters within the state, it shows what kind of partisan outcome we should expect,” said Imai. “But if we see something very different in comparison to this nonpartisan baseline, favoring one party under the enacted plan, that’s evidence that there are some other factors influencing when the plan was drawn.”


Imai developed redist with Cory McCartan, a Ph.D. candidate at the Graduate School of Arts and Sciences focusing on statistics. The pair found traditional methods for evaluating the fairness of redistricting plans weren’t working well, because they didn’t provide a neutral baseline to make objective comparisons. Fairness often became a subjective call, they said.


“For a long time, people have done gerrymandering and the question is ‘OK, how do I prove it?’” McCartan said. “It’s one thing to say, ‘Hey, I think that map looks unfair because the boundaries are super squiggly.’ But these things get litigated in court, so a judge has to clearly be able to decide: Is this fair or not?”


Redist has been used by plaintiffs in gerrymandering cases, including actions in Alabama, New York, Ohio, and South Carolina. In New York and Ohio, courts ultimately ordered districts to be revised based on the tool’s findings. In Alabama, the algorithm is also being used in a case brought before the U.S. Supreme Courtinvolving allegations of racial gerrymandering. Imai served as an expert witness for the plaintiffs, arguing that its new congressional map intentionally dilutes the Black vote. The case, which relies on protections in the Voting Rights Act, could eliminate one of the few remaining nationwide safeguards against rigged legislative maps.


Redist has become a major focus in Imai’s research group at Harvard called the Algorithm-Assisted Redistricting Methodology (ALARM) Project. The group recently launched the 50-State Redistricting Simulation Project and is using the software to evaluate the redistricting plans across the country by producing 5,000 alternative maps for each state.


“There’s always a question of how different ways of drawing boundaries can benefit some voters and harm others,” Imai said. “It’s important for social scientists to understand the nature of these types of political manipulations and address it.”

4 views0 comments

Recent Posts

See All

Looking back at two days of data science discovery and learning, brought to you by the Harvard Data Science Initiative Dear 2022 conference attendees and speakers, This year, we were delighted to host

bottom of page