Gerrymandering and math in the era of state reform

Editor’s note: Hope Johnson is a data scientist at the Princeton Gerrymandering Project, where she is on a team developing OpenPrecincts, a database of precinct and electoral data to help citizens participate fully in redistricting. Hope graduated from Macalester College in 2019. The views expressed here are hers. 

Bug design courtesy of Lafayette College

We at the Princeton Gerrymandering Project (PGP) believe that extreme gerrymandering hampers democracy. In June of 2019, the Supreme Court failed to settle on a workable standard for partisan fairness in redistricting. In doing so, the court permitted politicians drawing election district maps to discriminate along party lines. To prevent this from happening, my team at the Princeton Gerrymandering Project are trying to bridge the gap between mathematics and the law to pursue the strongest route to reform: state-by-state.

States can take action to create fair, impartial, and transparent processes for redistricting. These actions include establishing an independent redistricting commission, creating voter initiatives, bringing court cases, and changing the state law. Given the variation in state politics and law, what works to create fair maps in one state may not be an effective strategy in the next. But in order for any strategy to prove effective, citizens must first be engaged around the issue of partisan gerrymandering! That’s where the data branch of PGP, OpenPrecincts, comes in.

As part of its broader mission to support fair districting using data, math, and law, my team established OpenPrecincts as a one-stop shop for any organization seeking data to achieve their representational goals.

Working closely with partner organizations, we collect data from many governmental sources. Then we validate the data, convert it to a usable format, integrate maps with voting and Census data, and make it freely available. Starting with highest-priority states, we are working towards creating a resource for all 50 states, the District of Columbia, and Puerto Rico.

Photo courtesy of Princeton Gerrymandering Project

In short, we aspire to make OpenPrecincts a key resource for citizen-driven redistricting reform. Public engagement sends a powerful message to legislators that citizens are paying attention to redistricting processes and outcomes. Data-sharing projects like the Public Mapping Project and  free software like Dave’s Redistricting and PlanScore help citizens draw their own maps. Using these tools, and integrating high quality data from OpenPrecincts, aids citizens in exposing uncompetitive and unfair district plans.

There are many possible ways to use granular, high-quality data to investigate gerrymandering. Perhaps the most basic way is to calculate simple summary statistics of electoral results over time within a given geographic unit. In areas where we worry about partisan gerrymandering, I sometimes notice that the vote share for one party hovers around 50% for a few years, and after redistricting the vote share in the same area shoots far above or below the 50% winning threshold. Although other factors might explain this pattern, it is a red flag for packing and/or cracking. Data visualization is an incredibly effective tool for spotlighting outliers like the one that I described above.

Calculating summary statistics is always my first step when exploring data, and there are formalized statistical tests to continue investigating. The widely-used student’s t-test, the difference between the mean and the median of the vote share for one party, and the divergence from proportional representation are all in the suite of tools to quantify partisan gerrymandering. The tests for partisan gerrymandering all come with their own advantages and disadvantages (as is the case for most statistical answers to political questions). At PGP, we always aim to contextualize quantitative results with local and historical information about voting.

Another, increasingly popular, method for quantifying gerrymandering uses sampling. Using Markov Chain Monte Carlo methodologies, various groups around the country have specialized in producing a vast number of possible districting plans and compare a contentious plan to the mass of sample plans. Comparing the outcome and competitiveness of the hypothetical plans and the actual districts is a great way of bringing mathematical rigor to questions about partisan gerrymandering.

In a working democracy, power lies with the people, but the practice of gerrymandering threatens that promise. Making use of statistical and mathematical tools, we can take a creative, and multi-faceted reform strategy to curb the practice of gerrymandering and make the redistricting process independent and nonpartisan. Ultimately there is a way to draw fair district lines in all fifty states, and math and data can help us take partisanship out of the process!