Team Members: Chuyan Millie Huang, Wenwei Vicky Liu, Kim-Anh-Nhi Nguyen
In Collaboration with the Metric Geometry and Gerrymandering Group (MGGG)
Gerrymandering is a practice intended to establish a political advantage for a particular party of group by manipulating district boundaries. Historically, different metrics of gerrymandering have been presented to courts to support rationales to claim illegal gerrymandering. There is not yet a universally agreed-upon metric for evaluating splitting of municipal units with districting plans. Commonly used metrics usually focus on either geographical compactness (number of cuts and splits) or partisan symmetry and vote efficiency (efficiency gap, mean-median, number of seats won by a certain Party).
The wide variety in rules applied to districting problems (even in the same state) means that any single measure of gerrymandering will be insufficient/exploitable. Therefore, one approach is to generate large ensembles of districting plans to explore the underlying political geography of a given states. Rather than focusing on any particular plan, the “ensemble approach” uses simulated data to study the universe of possible plans and conduct outlier analysis by comparing to large ensembles of other feasible plans.
However, computational redistricting is not a solved problem! Even though scholars have developed different statistical methods that courts can use to spot manipulative districting, there is no one coherent and consistent standard. Additionally, there is a great deal of interplay between the legal constraints and the metrics of interest and among the metrics themselves. Researchers have struggled to effectively represent the distributions of values within the high dimensional data and inspect the distributions under multiple metrics in an unbiased manner.
We partner with researchers at MIT CSAIL working in the Metric Geometry and Gerrymandering Group (MGGG) to design an interactive data visualization system to:
Choose a particular metric you are interested in and choose what you consider as a reasonable range of values using the sliding bar filtering function. Observe what this does to the distributions of other metrics!
To start with, please choose a state to analyze: Virgina or Pennsylvania
Number of cuts
Democratic Votes (in %) for the Most Democratic District
Mean-Median
Number of Democratic seats
Efficiency Gap (in %)
Click on a map to see its corresponding metrics compared to the global distributions of these metrics. How much of a outlier are they?
Map 1
Map 2
Map 3
Number of Cuts
Democratic Votes for the Most Democratic District (in %)
Mean-Median
Number of Democratic Seats
Efficiency Gap (in %)
This is why you should not compare metrics across states - because of differences in inherent geopolitical characteristics for each state.
Thank you to: