As lawmakers move forward with drawing new maps, experts explain the tricks of the trade, how we can do better, and why gerrymandering is so hard to eradicate
State lawmakers voted to approve the new criteria  for redrawing congressional, state House and Senate districts last week.
Legislative redistricting starts with a county clustering process because of a provision in the state constitution that legislative districts should avoid splitting county lines. Congressional districts do not need to follow the county clustering rule.
Operating under the new criteria, legislators still have a lot of leeway, and it remains to be seen how the process plays out. However, years of litigation have prompted researchers to apply scientific rigor to the potentially contentious process. Here’s are some key facts you need to know:
County clustering: the first step
If we compare redistricting to the process of making and splitting a cake, census data give us the measurement of each ingredient.
According to the 2020 census, North Carolina’s population is 10,439,388. That is the number used for state House and Senate redistricting.
Dividing the total by the 120 state House seats and 50 Senate seats, we arrive at the number of people that should reside in each district. Ideally, each House district should have 86,995 residents and each Senate district 208,788. Districts can vary 5% above or below the ideal population.
Our ingredients are mixed, and the cake is baked. Now it’s time to divide it into legislative districts, which is like cutting it and giving a slice to groups of people.
But population imbalances make it impossible to draw districts — to cut pieces of cake — that are both equal in population and keep counties intact, as legally required.
A 2002 state Supreme Court case, Stephenson vs. Barlett , addressed this conflict by calling for “county clusters.” As geographer and mapping consultant Blake Esselstyn explained , we do this by finding groups of counties whose populations equal that of one or more districts.
Each county cluster gets a slice of the cake more or less proportional to the number of people. The more people in the group, the bigger the slice.
Once the state is sliced into county clusters, lines are drawn within them for each district.
Esselstyn  told Policy Watch that areas with the greatest population increases will affect the district maps the most, especially in the Mecklenburg County and the Triangle.
For example, Mecklenburg and Wake counties each have more than 1.11 million residents. With the new census numbers, they each could have 13 state House seats, up from the 12 and 11, respectively, under the previous census.
He emphasized that it remains to be seen which clustering will be chosen as the base map, and there remains a great deal of uncertainty as to where district lines will be drawn within those clusters.
Christopher Cooper, a political science professor at Western Carolina University, said that just because the legislature is following the recipe laid out in the Stephenson court case, that does not rule out gerrymandering.
“There are multiple optimal solutions in terms of which clusters [could form], and then within clusters, there is a whole lot of latitude,” Cooper said.
Considering the complexity of county clustering, Sen. Ben Clark, a Democrat representing Cumberland and Hoke counties, introduced a bill that would have delegated the task to the math department of Duke or the School of Government at UNC-Chapel Hill.
“Back in 2011, the ones that were used and had been presented as if they were the only options were generated… by a Republican map consultant,” Clark said. “That’s why I want an open and official process for actually creating or generating the total set of constitutionally compliant clusters for the House and for the Senate.”
However, the bill was never heard in the Senate Redistricting and Elections committee.
New report identifies optimal county clusters
Researchers found 16 possible sets of county groupings for the state Senate and eight for the House.
Duke University mathematician Gregory Herschlag is part of a group known as the Quantifying Gerrymandering Team  that’s headed by his colleague, Prof. Jonathan Mattingly and that’s been analyzing the challenges associated with redistricting in recent years. Herschlag said in an email that though some clusters remain unchanged from the 2016 groupings, many have significantly changed. This is consistent with the team’s previous finding that clusters are volatile, subject to population counts.
Some districts will likely need to “double bunk” their incumbents, as there are more incumbents than seats. Four Senate districts would contain two incumbents, and five House districts would double bunk two incumbents, the research shows.
Among them, Yadkin and Davie counties, where two Republican incumbents live — Reps. Lee Zachary and Julia Howard respectively — would have to consolidate in the same House district because neither county can combine with adjacent ones to form its own district.
Rep. Pricey Harrison, D-Guilford, said the way Guilford County is clustered with Alamance and Randolph counties in the Senate does not seem ideal. “We have a large portion of High Point African American voters … that are included in a Senate district in Randolph County, where obviously the differences are pretty stark in terms of interests of those different geographical areas.”
In the 2011 Senate mapping plan, Guilford had been grouped with Rockingham instead of Alamance and Randolph. Rockingham is home to Senate President Pro Tem Phil Berger, a Republican.
In a lawsuit  that successfully challenged the state House and Senate maps enacted in 2017, several experts independently analyzed the cluster and found evidence of extreme gerrymandering. However, the cluster itself was never changed.
The new county clustering study showed that Guilford and Rockingham should constitute their own cluster based on the 2020 census data.
Gerrymandering: easy enough to identify, but hard to root out
Experts at the Brennan Center for Justice  described the new redistricting cycle as “the most challenging ever” because of what they describe as a conservative backlash against the growing political power of communities of color.
In the past decade, the North Carolina legislature has engaged in litigation challenging its racial and partisan gerrymanders and the practice of diluting certain racial and political groups’ voting power.
This has resulted in the legislators’ promise to exclude partisan and racial data in the current cycle, though many argue  that legislators should study racial data in drawing district lines to ensure communities of color have the equal opportunity to select candidates of their choice.
Christopher Kenny, a Ph.D. candidate in government and redistricting researcher at Harvard University, said there are many indicators of the political landscape even when map drawers are not actively looking at election data.
For example, urban areas tend to vote Democratic. Accordingly, splitting up more urban areas “cracks” Democratic votes, Kenny said.
“If you’re allowed to look at certain racial groups, you might know that Black voters tend to vote more Democratic, urban Black voters tend to vote even more Democratic than non-urban Black voters.”
While it is often difficult to detect the use of partisan data during the redistricting process even with transparency measures in place, Kenny said, the main way to detect irregularity in maps is a redistricting simulation method.
To examine if a map was potentially rigged with partisan intent, researchers have found a way to compare it with a neutral baseline. Researchers ask computers to draw a collection of maps that only stick to a predetermined set of criteria without any biases. The collection is called an “ensemble.”
When layering partisan data on each map, researchers can see how many Republicans or Democrats are elected. Plotting the number in the collection, researchers get a distribution with a concentration of “normal” cases.
Researchers can then determine if the election outcome under the new map is an outlier.
Prof. Mattingly’s team was one of the first in the nation to implement the ensemble method in studying gerrymandering. Their work played a pivotal role in a series of court cases, first in Common Cause v. Rucho  challenging the congressional map in 2016, and later Common Cause v. Lewis , which resulted in the court’s order to throw out the state House and Senate plans enacted in 2017, followed by a redraw.
In a study  published in 2020, Mattingly’s team ran different gubernatorial, U.S. Senate, U.S. House, and state House election results on the 2016 congressional map. The team found the map consistently elected fewer Democrats than the remedial map redrawn in 2019, a plan proposed by independent judges, as well as the ensemble.
Herschlag said these maps make it more difficult for the disadvantaged party to gain seats. “That can really cause a lot of stability in the elections and make the elections very unresponsive to changes in how people vote,” Herschlag said.
There would be safe districts for the party in advantage, and the disadvantaged party would only hold more seats when their overall victory margin is disproportionately high.
This pattern exemplifies the classic method of gerrymandering — “cracking and packing,” in which the minority party’s votes are packed with a high concentration in certain districts they win, but in other districts, diluted and stretched thin.
Rep. Harrison said even with an explicit prohibition on the use of partisan data in map drawing, it is hard to avoid biases. “Legislators really know their districts well and know the partisan makeup of their precincts, so it’s sort of hard to unlearn that,” she said.
If you’re interested in knowing how a proposed map performs, use these map-rating tools developed by non-partisan groups, including the Princeton Gerrymandering Projec t and the Campaign Legal Center .