Incorporating “Geo Graphs” into Redistricting: How a New Computer Algorithm Could Stop Partisan Gerrymandering

The gerrymandering of congressional districts is a huge issue in North Carolina politics. The N.C. state legislature has exhibited a pattern of redrawing voting districts for their own political gain, sometimes subtly and at other times blatantly. In 2010 the census was taken and in 2011 maps that reflected North Carolina’s Congressional districts were redrawn. The census requires that these districts are redrawn into groups of substantially equal population size every 10 years for the purpose of voting for representatives. However, the North Carolina Supreme Court and then the U.S. Supreme Court eventually decided that during the 2011 redrawing, two districts in North Carolina, Districts 1 and 12, had been racially gerrymandered. These districts were ruled unconstitutional in the 2017 decision Cooper v. Harris.

When one political party gains control in a state, it will often order a redrawing of the maps in order to ensure that more districts will turn red or blue. For example, before the 2011 redrawing, North Carolina’s Congressional District 12 already took the odd shape of a snaking line cutting vertically through the state before the gerrymandered redistricting. During the redrawing the district was condensed even further with the purpose of “connect[ing] black communities without picking up white voters in between.” Racial gerrymandering is clearly unconstitutional and also illegal under the federal Voting Rights Act of 1965, but partisan gerrymandering is another issue. Courts have been hesitant to rule on cases in which one group or political party claims that the maps were specifically drawn to favor the other party or get more candidates into office. In the 1986 U.S. Supreme Court case of Davis v. Bandemer, the Court declared that it had the authority to rule on cases involving partisan gerrymandering, “but it declined to do so because it lacked a clear measure to indicate when this had occurred.”

The gerrymandering was blatantly political… with Representative David Lewis claiming, “I propose that we draw the maps to give a partisan advantage to 10 Republicans and 3 Democrats because I do not believe it’s possible to draw a map with 11 Republicans and 2 Democrats.”

It is true that there are many different criteria around which voting maps can be redrawn, but none of them necessarily guarantee a fair map. These factors include to what extent the districts are compact, foster competition in elections, represent a diversity of racial groups or ages, have a ratio of Democrats v. Republicans that is representative of the state, are continuous, etc. There are multiple web tools and apps that allow people to redraw districts to determine what different election outcomes would be, such as Dave Bradlee’s Redistricting site. The FiveThirtyEight blog used Bradlee’s web tool and the same census data in the same state to create multiple maps: one led to more Democratic seat victories than Republicans, one led to more Republican victories than Democrats, one was representative of the state’s actual political makeup, and one created very competitive or purple districts.

This exercise shows that it is quite possible to take census data showing that a state will vote 50% Democrat and ensure that more Republicans win Congressional seats than Democrats. In the 2012 North Carolina election, Republicans won nine out of thirteen available Congressional seats, even though 50.3% of votes went to Democratic candidates. However, there are some mathematicians and computer scientists working to make this process more transparent. Wendy Cho and her team and the University of Illinois Urbana-Champage came up with a new algorithm that would redraw voting districts in 2016. Their program takes the current or proposed voting district map for a region and compares it with “millions of hypothetical alternatives to determine whether the original map is a statistical outlier,” which would show that the map is drawn with bias. This process generates tons of data for each individual state being analyzed, since it incorporates the millions of alternative maps, and therefore requires a super computer to run. So while this algorithm is helpful in uncovering political bias, it is not the most affordable or accessible program for states to use.

Sheldon Jacobson and his colleagues at the University of Illinois have created a different algorithm to determine whether or not there is political bias in maps. They improve census blocks by adding “geo graphs” to maps. They analyze data in each census block, of which there are at least 24,000 in every state, and then use them to create congressional districts that are optimized for the different factors listed above (i.e. compactness, political competition, racial diversity, etc.). The purpose of Jacobson’s program is transparency. He wants people to be able to look at the way his geo graph maps come out, then compare them with the proposed maps to determine if there is political bias. His team claims that this algorithm is cheaper and more efficient because of the way it organizes data, and does not require a super computer to run.                

However, it is important to remember that these computer programs are not claiming to be automated cartographers. They are merely a tool to compare to human-drawn maps that will show what level of political bias exists. Currently there are no rules that require computer algorithms to be run before redistricting maps are finalized, although tools like this have been in use since the 1960’s. These redistricting tools may be helpful in Rucho v. Common Cause, a N.C. partisan gerrymandering case that is still making its way through the courts. The government watchdog group brought a suit against the N.C. state legislature for illegal partisan gerrymandering. The gerrymandering was blatantly political and Republicans had no qualms confessing this, with Representative David Lewis claiming “I propose that we draw the maps to give a partisan advantage to 10 Republicans and 3 Democrats because I do not believe it’s possible to draw a map with 11 Republicans and 2 Democrats.” The U.S. Supreme Court announced in early January 2019 that it will hear an appeal of this case.

Caroline Martin, 28 January 2019