rdabase

Redistricting analytics data & shared code

Installation

To get the data locally, clone the repository:

$ git clone https://github.com/rdatools/rdabase
$ cd rdabase

To run the scripts, install the dependencies:

pip install -r requirements.txt

To use the shared code in another project, install the package:

$ pip install rdabase

Data

The data are stored in the data directory by state. These pages describe each dataset:

• Data: Census and election data by precinct.
• Shapes: Shape properties by precinct.
• Graph: Precinct adjacency graph.

At present, data has been extracted for the 24 states with 7 or more congressional districts: AL, AZ, CA, CO, FL, GA, IL, IN, MD, MA, MI, MN, MO, NJ, NY, NC, OH, PA, SC, TN, TX, VA, WA, and WI. FL uses DRA's corrected precinct shapes (GeoJSON), with intersections removed, and converted to a shapefile, before simplifying them. In the future, we may extract data for other states.

Code

Some shared code and scripts are described here:

• Shared Code: Common code used in multiple applications.
• Scripts: Scripts to re-format the data for specific applications.

Sources

The data comes from the following sources:

• The total census population & VAP demographics data came from the 2020_census_XX-N.csv files in the DRA vtd_data GitHub repository, where XX is the state abbreviation and N is the suffix. We took the latest version of the data, which is the one with the highest N. DRA got the original data from the Census Bureau's Redistricting File (Public Law 94-171) Dataset or adjusted versions and American Community Survey (ACS) estimates.
• The election data comes from the 2020_election_XX-N.csv in the same repo. DRA got most of the election data from data partners, especially Voting and Election Science Team (VEST).
• The shapes are copies of tl_2020_FF_vtd20.zip from the Census Bureau's [Tiger Line Shapefiles]](https://www2.census.gov/geo/tiger/TIGER2020PL/LAYER/VTD/2020/), where FF is the state FIPS code, e.g., 37 for North Carolina.

Some things to be aware of:

• If it exists, we used the adjusted population data instead of the official 2020 census total population data.
• For Florida, the official VTDs from the Census Bureau were bad. We used DRA's corrected precinct shapes (GeoJSON), removed the intersections, and then converted it to a shapefile.
• We simplify the precinct shapes (see extract_shape_data.py) to approximate the simplification that DRA does, so compactness measurements align.

Testing

$ pytest