Visualizing Gerrymandering

6.894 Interactive Data Visualization Final Project

Visualizing Gerrymandering

Team Members: Chuyan Millie Huang, Wenwei Vicky Liu, Kim-Anh-Nhi Nguyen

In Collaboration with the Metric Geometry and Gerrymandering Group (MGGG) @ MIT CSAIL

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).

However, there is intricate interplay between the legal constraints and the measurements of interes, and among the measurements themselves. Researchers have struggled to understand the trade-offs between these intertwined metrics.

We partner with researchers from the Metric Geometry and Gerrymandering Group (MGGG) at MIT CSAIL to design an interactive data visualization system to:

Unveil some hidden interactions between these metrics;
Demystify some common false beliefs;
Propose a novel way of looking at gerrymandering for policy makers and the general public.

Here are some of the common metrics used to measure gerrymandering. Click on any metric to see their definition and common interpretation.

Now, 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!

Choose a metric to filter on

Number of cuts
Democratic vote shares for the most democratic district
Mean-Median
Number of Democratic seats
Efficiency gap

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 %)

Now, using the same tool on the right, let’s take a look at some common beliefs:

Gerrymandering should be flagged if Efficiency Gap is at 8 percent or more.
- Observe the Efficiency Gap distribution, across 100,000 plans, how much of the time does it fall within +/- 8%?
Higher number of cuts indicates gerrymandering.
- Filter for the Efficiency Gap to be within the +- 8% range, how does this affect the distribution of number of cuts?
Lower number of cuts would lead to “better” Efficiency Gap.
- Is this actually true?

After you have spent some time playing around with the tool above, check the section below.

Did you see?

Using Virginia as an example:

More than half of the possible values for Efficiency Gap fall out of the +/- 8% range!
Surprise! Filter on Efficiency Gap to be within +/- 8% uniformly changes the distribution of number of cuts (not in one direction!)
Lower number of cuts doesn’t not necessarily put Efficiency Gap into the +/- 8% range!

Hopefully, at this point, you are familiarized with the idea of “distribution” of metrics. Often times, political groups come to research scientists asking for fairness evaluation of a certain districting plan. This is not meaningful! We propose that an alternative viable evaluation method would be to compare any particular redistricting plan to the global distributions of all possible plans and the distributions of their corresponding metrics.

Click on a given map to see their corresponding metrics compared to the global distributions of these metrics. How much of a outlier are they?

You should be able to see that map 1 and 3’s metrics fall very close to the median/mode of the distributions, whereas map 2’s metrics seem like outliers from the distributions. This indicates a high likelihood that gerrymandering has taken place in map 2.

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 %)

We hope that by now, you feel more educated on:

Why some common beliefs are not necessarily true
Why you should not look at any particular plan or metric but look at distributions of metrics across possible plans
Why you should not compare metric statistics across different states

To discover more about gerrymandering, you are welcome to visit the Metric Geometry and Gerrymandering Group's webpage.

Interactive plan metric distributions

Below are three districting plans from Virginia that were generated using simulated data.
Can you tell if gerrymandering has taken place?

Lastly, we want you to understand that states differ drastically in terms of these metrics.
Look a look below to see how different the state of Pennsylvania is from Virginia!

Interactive plan metric distributions

Below are three districting plans from Virginia that were generated using simulated data. Can you tell if gerrymandering has taken place?

Lastly, we want you to understand that states differ drastically in terms of these metrics.Look a look below to see how different the state of Pennsylvania is from Virginia!

Below are three districting plans from Virginia that were generated using simulated data.
Can you tell if gerrymandering has taken place?

Lastly, we want you to understand that states differ drastically in terms of these metrics.
Look a look below to see how different the state of Pennsylvania is from Virginia!