(Moving towards) a network analysis of U.S. Senators

In determining what exactly constitutes a network analysis, Miriam Posner’s website directed me to an older version of Scott Weingart’s website, with his introduction to networks available here. (Weingart received the 2011 Paul Fortier Prize, among other recognitions at the top of our field, and/but/so his website — a WordPress! — is a great resource across the board. Highly recommended.) He writes that, “[i]f you’re studying something with networks, odds are you’re doing so because you think the objects of your study are interdependent rather than independent. Representing information as a network implicitly suggests not only that connections matter, but that they are required to understand whatever’s going on.”

This week, I wanted to focus on connections or relationships with a topic that I believe reflects interdependency, so I took this third praxis assignment as an opportunity to explore the concentration of power that has long defined the upper levels of government in the United States. I was curious about how network analysis might make sense of what I view as a relatively closed circuit of people, predominantly white and predominantly men, whose relationships with each other may go back to college — or earlier, in the case of certain recent high-profile hearings — and serve to further consolidate their influence.

To explore this concept in more detail, I aimed to do a network analysis of the current 100 United States Senators to discover connections in their young adult lives through their time as an undergraduates. Many Senators attended one of a few law schools, but I was curious about finding other similarities even earlier in their higher education, whether a shared undergraduate institution or a similar set of experiences as an undergraduate at different institutions. It came to me that I could possibly use the tools for this week to visualize who — if anyone — would have been at the same school at the same time, for example. I ended up creating a dataset of each Senator’s name, college, graduation date, degree type, field of study, and additional information, including fraternity or sorority affiliation and if they served as student body president, were the first in their family to graduate college, graduated as a member of Phi Beta Kappa, or a few other categories that appeared in academic/professional overviews around the Internet. It’s not that I think being in a fraternity or sorority is an accomplishment (I did have a net positive experience with my own Greek life experience, but I’d hardly call it an achievement), but I wanted to include this element of undergraduate life as a potential link between Senators.

To consolidate all of this information in a Numbers file, I took Wikipedia at its word, and dug for details about several Senators as well. Some biographies described how so-and-so “graduated from [institution] in [year] with a [B.A./B.S./B.B.A.] in [field],” which was optimal for my purpose, but many biographies instead included partial information that left rows blank in my spreadsheet: reading that a Senator “holds a B.A. from [institution” or “graduated summa cum laude from [institution] in [year]” would send me to other sources, including alumni magazine profiles and commencement speaker information. For a few Senators, even this secondary step did not turn up the details I wanted, particularly with comparatively older Senators who graduated from college in the 1960s and 1970s whose fields of study remained hidden deeper than I could dig for this week.

At first, I included a column for “additional information” all together, separated with semicolons, but this catch-all approach had its limits. For one obvious thing, this approach didn’t allow me to sort  the data by any one consideration (if I wanted to see the Senators who graduated cum laude or higher, for example) that would illuminate a potential overlap in experience. To try and fix this, I broke the column up into a few not-catchy additional columns: “PBK?”, “other academic Greek,” “Greek social,” “first-gen,” “leadership,” and “athletics.” The column for “leadership” then became “class president?” and “valedictorian?”, making the table harder to navigate than before when these considerations were in one place. I soon realized another limit of this new approach: that some of this information might not be as accurate as I wanted it to be. Besides deliberately untrue information included in a biography, if a Senator graduated as their college or university’s valedictorian, it might not have made it into their Wikipedia biography (which anyone can edit) because so much else later in life eclipsed that one title or any other reason, or maybe the college or university did not even identify a valedictorian in the first place but the person still graduated at the top of their class.

The final consideration, which is fundamental in hindsight, is that I assumed that shared titles or organizations would lead to shared experiences that would lend themselves to network analysis. To be fair, there might be something about a group of people in the same Greek-letter organization at the same institution at the same time, or even across time, as the fascinating social experience of Homecoming illustrates, that ties them together, but it is hard to generalize from this feeling to a network. Do Doug Jones (who graduated from the University of Alabama, Class of 1976, with his B.S. in Political Science) and Michael Bennet (Wesleyan University, 1987, B.A., History) feel any kinship at all for their shared status as brothers in Beta Theta Pi, for example, and would it be fair to say that this kinship has affected their politics in any way, which is what my initial interest in this entire dataset seems to suggest? Or do John Cornyn (Trinity University, 1973) and Pat Roberts (Kansas State University, 1958) have some unspoken bond thanks to their B.A. in journalism?

I originally began pulling together this information to create a network analysis of what United States Senators might have shared early in their adult lives — institutions, honors, social organizations — but encountered fundamental problems with this very curiosity, not to mention the steep learning curve to putting the data into place. In working to express the data through Palladio and Gephi, I found that the platforms did not respond to my organizational approach or my questions, giving me a string of error messages and forcing me to return to the data over and over again to fill it in and rework its structure. I am going to try a few more rounds of editing my spreadsheet and exporting it to a csv file over the next few days, but also have considered that a different method entirely might give more insight.

Much of this process of compiling and adjusting the dataset was the challenge of figuring out how to organize the data as little as possible while aiming for accuracy and consistency. To paraphrase Micki Kaufman’s answer to questions last week about her method for working with a large quantity of documents processed with optical character recognition, I wanted to remember that we look for patterns where the data is cleanest, so the moment that you begin cleaning data, you begin influencing it, even subconsciously. This lesson only felt more important the more I messed with this data, even on a small, limited scale, and realized how much of my own interests and decisions affected any potential takeaways.

To return to Weingart’s post about network analysis, “Relationships (presumably) exist. Friendships, similarities, web links, authorships, and wires all fall into this category. Network analysis generally deals with one or a small handful of types of relationships, and then a multitude of examples of that type.” He uses the examples of authorship and collaboration as types or ways to describe relationships between types of nodes and introduces the distinction between asymmetric relationships, or directed edges that can be visualized with an arrow flowing one way, and symmetric relationships, or undirected edges that can be visualized with a line between nodes implying that the flow of the relationship is the same in both directions. For my purposes, I was only interested in finding the potential directed edges, the undergraduate-level features that current Senators have in common that could possibly indicate shared experiences and start the process of understanding various intangible “benefits of the doubt” that seem to hold real weight in political situations. Moving forward with trying to explore the concentrations of political power in the federal government, I think it might make sense to incorporate a greater sense of asymmetric relationships (who has clerked for who, for example, rather than who were classmates on the same level or who shared an experience “equally”), or else to work with nodes that offer less room for interpretation on my end.

1 thought on “(Moving towards) a network analysis of U.S. Senators

  1. Nancy Foasberg

    I wonder if what you’re looking at is a set of networks lying on top of each other and overlapping. It’s certainly fair to describe a fraternity as a network, or a university, and there’s definitely a relationship between them — but as to whether you can look at them both at once, that seems like it’d be difficult with these tools.
    I’m also really interested in how you want to catch the affective aspects of the network, and have no idea how that could be done. It reminds me of Johanna Drucker’s piece on visualization and trying to catch all the nuance in a visual format instead of reducing everything to math. And like Drucker’s projects, it seems both fascinating and really difficult.

Comments are closed.