Ideation
On May 14, 2020, the House of Representatives passed a resolution to allow members of Congress to vote by proxy as a health safety measure at the beginning of the coronavirus pandemic. Voting by proxy meant that older or immunocompromised members of the House could designate someone to vote on the floor on their behalf. To do so, each member had to submit a photo to the House Clerk’s office with their designee’s name. Members could also change their proxy by submitting a new letter. The letters filed live on The House Clerk’s official website and can be sorted by when they were filed. However, the Clerk’s office made it so that the only way to see each designate was to click a PDF and read the hand-scribed letter. When the office began compiling these this was, I was working at C-SPAN as a digital producer and writing a newsletter on Congress. The intentionally opaque way of formatting this public collection made me wonder what hidden connections could be seen in a network analysis if all letters could be analyzed.
Data collection & process
There are currently more than 7K letters filed from May 2020 to December 2022, when proxy voting ended at the conclusion of the 117th Congress. I pulled the first 72 letters filed from May 20th-26th to get a snapshot of the first members to partake in proxy voting.
I created a node list with information about every member who requested proxy or acted as a proxy including their party, state, status (designator or designatee which is not a word but works for this purpose), and the committees they are assigned to. Since no Republican would not take part of proxy voting until July of 2020 — Florida Rep. Francis Rooney nominated Democratic Rep. Don Beyer to vote on his behalf — the connections are all between Democrats. I was interested in whether members were designating proxy’s based on their geographic location, shared committee or potentially timed served in Congress. Since this dataset was pulled by hand, I decided to focus on the first two potential connections.
Visual examples
This first example is what I could create in Gephi and export into Illustrator. I wasn’t able to see any of the edits I made to the preview in Gephi, so I made some guesses as what it would show and pulled it into Illustrator. I wanted to highlight the primary named of people who voted on behalf of multiple people. Every member highlighted voted on behalf of three or more people. The colors designate the state each person represents. By removing the names and the background, you can see that a majority of the connections are between members of California or Florida (below).
Fruchterman Reingold version of the graph above:
From here, I moved to Flourish to see what capability it had in visualizing this network outside of Gephi. The graphs below are all interactive. One major critique of Flourish is that it was not capable of doing any of the analysis for you, you had to have the edges and nodes list pre-created and ready for visualization.
The below graph is a directional network graph. If you hover over the individual points, it will show you the related network graph. But by having the individual points separate of some of their connections, these dots look like individual, unconnected points. You also can only edit the colors by by color palette, which leaves states with unclear coloring (why is Arizona blue?). Still, like in the state-based color network above, the amount of California (dark blue) connections are visibly clear. The Florida connections (pink) lose their prominence in this example.
Below is a radial graph of the connections. Collecting the states into groups makes sense to me but Flourish also doesn’t condense connections. Since the source is the person who requested a proxy and the target is the proxy, the source list is unique but the target list is not. In this graph, targets are repeated multiple times, but if you hover over them, you can see their multitude of connections.
Finally, I tried to crated a weighted radial graph and swapped the source and target list so this would be weighted by in-degree. This example does a better job of showing which members have more designations as proxy that the others.
Challenges & future updates
Pulling 7K results by hand and adding them to a spreadsheet would be a time consuming task, but it would give much greater depth to this analysis. Republicans were initially resistant to proxy voting, however, it is notable that the first Republican to vote by proxy did so by designating a Democrat to vote on their behalf. And, while Republicans voted by proxy less than Democrats, the connections they chose to lean on show the rare ties of bipartisanship where they exist. In a future iteration, I would like to finish the list for at least the full 116th Congress, which concluded at the end of 2020. The first year of a pandemic colliding with a presidential election year and a second impeachment of the then-president could be illusive for where connections exist.
Other challenges I faced were with Gephi itself. I was never able to see the preview view on my computer despite closing all other applications on my computer and giving it time itself to load. I ended setting some display options and then exporting the .svg into Illustrator to try to style and further explore the information. This also led me to using mostly Flourish to view the networks and do visualization for this data. Flourish, while a beautiful and mostly-free software, had it’s own limitations and challenges for presentation.