The first time Kim Brace drew electoral district maps for the state of Illinois, more than 40 years ago, things moved slowly.
He and his colleagues hung maps of the state on the walls in the office of the speaker of the state house of representatives. Someone would climb a ladder, moving different blocks of people into different districts while another took notes below. In the evenings, they would go to the largest bank in Springfield and use a mainframe computer to generate a daily report. Over the course of the four-month legislative session, Brace was able to draw about 10 possibilities for electoral maps.
Ten years later, in the 1990 redistricting cycle, Brace, the president of Election Data Services, a redistricting consulting firm, was back at the drawing board. But this time, he and his colleagues didn’t have to draw on walls. They rigged up two personal computers – one couldn’t handle all the data they needed – with rudimentary mapping software. They drew about 100 potential maps.
By 2000, Brace was able to draw about 1,000 plans. In 2010, the last time he drew maps, he was able to produce 10,000 possible maps. “It lets you see and imagine different alternatives,” Brace said. “It gives me that capability of understanding the parameters and playing field that I’m playing in.”
It’s an exponential growth that reflects just how quickly the nature of redistricting – the decennial process of redrawing electoral districts – has changed. Every 10 years, when mapmakers sit down to draw district lines, they take on a God-like role, grouping tiny census blocks – the smallest unit of geography the Census Bureau defines – into different districts.
They’re looking not just at demographic information like age, sex, race and ethnicity and income level, but also at years of election results in presidential, gubernatorial, senatorial, US House and other races. In recent years, as American politics has become more polarized, it’s become easier to predict how voters will cast their ballots, political scientists say. Using that data, the mapmakers can precisely forecast how elections will turn out for years to come.
The ease with which mapmakers can move around pieces of the puzzle in creating a map now allows them to see more variety, tweak more and make their maps more and more precise.
In the coming weeks, new technology will play a huge role in helping Brace and other mapmakers carve up America’s 435 congressional districts in the US House and even more state and local districts. There will also be fewer guardrails in place than ever before; in 2019, the US supreme court said for the first time that there were no federal limits on how severely politicians could draw districts to give their party a political advantage, a practice called gerrymandering.
“What used to be a dark art is now a dark science,” said Michael Li, a redistricting expert at the Brennan Center for Justice. “Before, you weren’t sure about the data, but now you’re much more certain so you’re able to draw things in ways that can be more aggressive.”
Over the last decade, mathematicians and others have also begun to automate the map-drawing. New algorithms allow mapmakers to very quickly generate thousands of sample maps based on whatever criteria they input. They could immediately generate thousands of gerrymandered maps, for example, that give one party a significant advantage while also meeting other neutral redistricting criteria like keeping districts compact and meeting the requirements of the Voting Rights Act. The point isn’t necessarily to use a computer to draw a map, experts say, but to explore the possibilities of what’s possible.
“That’s a big deal. Sure, there were algorithms 10 years ago, but they were absolute stone age,” said Moon Duchin, a mathematician who leads the MGGG redistricting lab at Tufts University. “You just didn’t have, 10 years ago, good techniques for really seeing a lot of variety and now we do. And that’s a superpower you can use for good or evil.”
Nicholas Stephanopoulos, a law professor at Harvard who helped develop a tool to measure gerrymandering a few years ago, said that algorithms might help mapmakers explore possibilities that they might not have consi
dered on their own.
“An algorithm can help if you want to do a lot of things at once. If you want to do a maximal gerrymander and you want it to look pretty nice, and you want to respect county and municipality boundaries, then an algorithm can be helpful in identifying certain solutions that a human just might not stumble on to,” he said.
Even so, Stephanopoulos questioned how much sophisticated technology was needed to gerrymander.
“The power of technology for gerrymandering I think is somewhat overstated. Primarily because it’s so easy to gerrymander without the technology,” he said. “An ordinary human is perfectly able to design a very effective gerrymander virtually everywhere. It’s such an obvious technique.”
“I don’t know how to put this nicely – gerrymandering is not really rocket science,” added Samuel Wang, a Princeton professor who leads the Princeton Gerrymandering Project. “You can be reasonably clever and at the level of an excellent checkers player or a reasonably good board gamer and do a good job of drawing a map that confers partisan advantage.”
Duchin and other experts are working to make sure the algorithms are used for good. While the algorithms can be used to generate extreme maps, they can also be used to identify them by generating hundreds or even thousands of possible sample maps according to neutral criteria. Armed with those sample maps, experts say they will be able to more easily see when a map lawmakers are considering is more extreme than what’s expected.
“You can still build extreme maps, maybe even better than ever. But now we kind of have a method to kind of show that they’re extreme,” Duchin said.
Reformers have already seen how powerful these algorithms can be in fighting gerrymandering.
In 2017, Jowei Chen, a professor at the University of Michigan, used a computer algorithm to draw 1,000 theoretical maps for Pennsylvania’s 18 congressional seats. The algorithm built districts based on “non-partisan, traditional districting criteria”, like keeping county and municipal boundaries intact as well as equalizing populations and keeping districts compact.
Chen also told the algorithm to favor protecting incumbent members of congress. When he compared the 1,000 sample maps to the one Pennsylvania Republicans enacted in 2011, it was clear that the actual map in place was an extreme outlier, far more partisan than if lawmakers were trying to fulfill non-partisan criteria.
When the Pennsylvania supreme court struck down the maps in 2018, the majority pointed to Chen’s analysis as “perhaps the most compelling evidence” the map was so gerrymandered that it violated the state’s constitution.
The software most widely used for redistricting is called Maptitude, created and licensed by the Caliper Corporation, a Boston-based company with a few dozen employees that specializes in transportation software. Initially, the company didn’t think the process required specialized software, but it started hearing from redistricting consultants who wanted something that would help them draw political maps. Caliper rolled out its first version of Maptitude for redistricting in the 1990s. Today, a license can cost anywhere from $1,000 to $10,000, said Howard Slavin, the group’s president.
In 2010, Republicans took advantage of redistricting like they never had before. The party launched a concerted effort, called Project Redmap, to win control of state legislatures and then aggressively drew districts that entrenched Republican control.
“We were horrified with what some people had done with our software,” Slavin said. “We were software guys, math guys. We were making tools and stuff. And we weren’t invested in, you know, trying to make one side win against another or anything like that.
“Part of the problem is that when you make a tool you don’t get to control what it’s used for. You can use an axe to chop down a tree and I guess you can also use it as a murder weapon. We didn’t anticipate it,” he added.
After seeing what happened 10 years ago, Slavin began tweaking Maptitude to make it harder for mapmakers to get away with extreme gerrymanders. This time around, the software includes new metrics that will lave a kind of paper trail and make it easier to identify extreme gerrymanders.
“Now, with our software, it’s going to be pretty much impossible for anyone to hide a ge
rrymander,” Slavin said. (In previous years, however, mapmakers have gone to extreme lengths to conceal discussions from mapmaking from ever becoming public.)
While mapmaking has long been done in secret, there’s been an explosion of publicly available, high-quality tools that the public can use to draw districts for free online. Watchdog groups have also developed easy-to-use online systems that can quickly score maps to see just how gerrymandered they are.
Duchin, the Tufts mathematician, has developed software that allows ordinary citizens in places like Michigan and Wisconsin to their own sample districts to show lawmakers which parts of the state should be preserved.
“One of the big differences from 10 years ago and especially from 20 years ago is the leveling of the playing field, where anybody can have access to voting data and to scoring software that allows the evaluation of a map for fairness or unfairness,” said Wang, whose groups plans to publicly score maps as they are released. “That’s a big change in the positive direction in terms of pro-democracy and pro-disclosure.”
Those scoring tools, Wang said, will allow a vigilant public to identify gerrymanders that aren’t obvious to the naked eye and hold lawmakers accountable, Wang said.
“The fact that there’s just armies of nerds out there ready to look at these things, ready to … score things, that’s a real change from 10 years ago,” he said.