See Design Process
Data Exploration
First Draft
Second Draft
Final Draft
Visualization Methodology Explainer
I chose to use bubbles to demonstrate the gender disparities in Congress because I felt that the contrast was better expressed through size than numerical comparisons.
I establish the use of color to denote political parties. I will continue to use this schema for the rest of the visuals, relying on the knowledge of the viewer (i.e. that they know Democrats' colors are blue and white and Republicans' colors are red and white).
I didn't use a bubble graph to visualize the distribution of members based on race because I wanted to emphasize the numerical contrasts between each column and because there were too many values. The initial drafts (see second draft above) looked too chaotic to properly compare the information.
I decided against using a logarithmic scale in this diagram, despite the extreme disparity between one category and the rest, because I wanted the viewer to know the exact number of House Representatives of each race. A logarithmic scale would have led to an increased cognitive burden that this graph did not need to use.
I chose to keep the scale between 0 and n-Members (n = number of members in the largest grouping) because having the scale between 0 and 435 concealed the values of the lower groupings. I decided that it was more important to clearly demonstrate the number of representatives in a particular group than contrast the number with the total number of representatives.
I decided against using a logarithmic scale in this diagram, despite the extreme disparity between one category and the rest, because I wanted the viewer to know the exact number of House Representatives of each race. A logarithmic scale would have led to an increased cognitive burden that this graph did not need to use.
I decided against using a logarithmic scale in this diagram, despite the extreme disparity between one category and the rest, because I wanted the viewer to know the exact number of House Representatives of each race. A logarithmic scale would have led to an increased cognitive burden that this graph did not need to use.
I chose to make a horizontal bar graph instead of a vertical one because it was easier to match the length of the bar to the number of representatives with four different categories.
I made sure to order each bar graph from most members of a group to least to make the contrast clear.
Despite the amount of information in this graph, I decided to keep the contrast between the parties to demonstrate that even members of different parties have similar backgrounds when they first enter politics. I thought that was an important aspect to communicate to viewers.
I also thought it was interesting in the smaller amounts contrasting the prior careers of Republicans and Democrats (i.e. a pastor vs. an activist).
I chose to display this data with a treemap because I wanted to display the quantiative-ratio data with size, which I couldn't accomplish with a bar chart. I also felt like the design choice was useful to distinguish between each of the idealogues in the same dimmension.
I originally wanted to display this information using a radar map, but I couldn't find an easy implementation of it using Tableau.
What Does the Average Member of the U.S. House of Representatives Look Like?
Most Average Member of Congress:
Eric Swallwell
Distribution of Gender in the 116th Congress
2018 was known as the 'Year of the Woman' by political pundits. I wanted to examine what the effect of the election was on the average member of congress.
One feature I noted was that Democrats are inching closer to gender parity between men and women while the Republicans seem to be heading in the opposite direction.
Distribution of Race in the 116th Congress
The 2018 Elections made headlines for being the most diverse congressional class ever. I wanted to see how much of a dent the election actually made in reducing racial disparities in Congress.
Despite the progress in multiracial candidates competing and winning that the 2018 Election brought, there is still a massive gap between White representatives and people of Color in Congress.
I constructed my 2016 dataset from scratch, but in future iterations of this research, I would like to investigate the racial makeup of Congress over time as a comparison exercise. I think it would be interesting to find percentage changes as time goes on, and then examine the relationship between multiracial candidates running and winning with the political atmosphere at the time. In other words, I'd like to compare if Democratic or Republican presidential administrations encourage POCs to run for office.
One trend noted in my visualization is the disparity between racial representation in the Republican Party. The party is almost exclusively white, with a few Hispanic and African American exceptions. One interesting finding was that there are an equal number of Native American politicians from both the Democratic and Republican parties.
Ivy League Graduates in the 116th Congress
Political pundits often comment that someone's Ivy League background makes them an ideal recruit for the House. I knew offhand that the majority of my congressional representatives from North Carolina and Illinois didn't go to an Ivy League, so I wanted to see if there was a trend that the majority of government officials went to Ivy Leagues.
There's also a stereotype of Democrats as being elitist and going to these elite schools, so I wanted to see if there were more Democrats who went to Ivies than Republicans.
Distribution of Military Service in the 116th Congress
I was reading a book on House election strategy and noted that several candidates were recruited by Democrats because they were veterans. It seemed that the party leaders in D.C. believe that Democrats who served in the military make better candidates against incumbent Republicans. What I wanted to see was if military Democrats represent more conservative districts or were first elected by beating an incumbent.
I was reading a book on House election strategy and noted that several candidates were recruited by Democrats because they were veterans. It seemed that the party leaders in D.C. believe that Democrats who served in the military make better candidates against incumbent Republicans. What I wanted to see was if military Democrats represent more conservative districts or were first elected by beating an incumbent.
Previous Family Member Being a Politician in the 116th Congress
The district I live in (IL-03) is represented by a congressman whose father was also the congressman before him, and he practically gave him the seat by running in the primary and then retiring. I was interested in seeing how many members of congress had family members represent their district (or a neighboring one). I was pleasantly surprised to find that it there were a very small number of hereditary politicians.
There wasn't a large discrepancy between Democrats and Republicans in hereditary positions (Democrats had maybe 5 more).
Method of Victory in their First Election
How a member first won their seat can be very informative. If they waged a successful primary against a tenured member, it can indicate that big shifts are happening in the party. If they won through redistricting, it means that demographic shifts favored that kind of candidate. If they beat an incumbent, it can represent shifts in favor of that party in the national political atmosphere.
Most members won their seats after its previous holder retired, pursued another office, passed away, or resigned due to scandal.
How a member first won their seat can be very informative. If they waged a successful primary against a tenured member, it can indicate that big shifts are happening in the party. If they won through redistricting, it means that demographic shifts favored that kind of candidate. If they beat an incumbent, it can represent shifts in favor of that party in the national political atmosphere.
Most members won their seats after its previous holder retired, pursued another office, passed away, or resigned due to scandal.
Distribution of Prior Careers in the 116th Congress
I really wanted to see what the most common prior career Representatives had before they entered Congress. To no surprise, more than a third were already involved in government as members of State Legislatures. It is apparently common for state lawmakers to make the transition to national government when a seat becomes available.
I was surprised at how many Democrats and Republicans shared origins. I expected the both parties were going to have lots of members who were state lawmakers, but I was interested that nearly identical numbers of Democrats and Republicans had also been attorneys, state executives, mayors, and businesspeople.
The only real noticeable differences came at the bottom rung of information (i.e. only had a count of 5 or less), where congresspeople who had previously been pastors were exclusively Republican while congresspeople who had previously been activists and professors were exclusively Democratic.
Distribution of Ideological Wings in the 116th Congress
Arguably the most subjective aspect of my dataset, the ideological parings of each House representative were a subjective choice based on three factors: 1) their membership in one of several ideological caucuses (such as the Blue Dog Coalition or the Freedom Caucus) - whose votes tracked with whatever leading figure, 2) their public affiliation with a particular candidate through endorsements, 3) their Lugar Center Bipartisan Score.
As you can see from this dataset - which because Democrats currently hold the majority in the House, it is biased towards left-leaning political figures - most members identify ideologically with former President Barack Obama. Most Republicans identify with former President George W. Bush.
This visualization is interesting because it displays the contrast between the actual majority ideology and minority ideologies that wield an outsized influence on media discourse. That leads many people to believe that a single mindset describes many members of the House, when in reality, they represent a variety of differing viewpoints.
Data Walkthrough & Analysis
For this project, I used a self-coded dataset on House Congressional Representatives in the 116th Congress, drawing on information from individual congressional websites, MIT's election data, GitHub repositories, and news articles.
Download The Dataset Here
This dataset contains information on each of the 435 representatives in Congress, including their party affiliation, how long they've been serving, their gender, ethnicity, prospects for 2020, prior job, whether or not they served in the military or went to an ivy league university, their age, their ideological equivalent (i.e. a President or presidential candidate), whether they've had a prior failed campaign before they became representatives, their first method of election, whether or not they had family members that were also politicians, whether they've been accused of illegal activity, their twitter account, their Cook 2020 Political Rating, and their 115th Congress Lugar Center Bipartisan Rank.
My Question
The field that I wanted to approach was congressional politics, and my question was what features does a typical member of the 116th Congress possess?
My Conclusions
My main conclusion from this experiment was how little predictive information can be gained from conventional wisdom from pundits on this subject. Several of the reasons I chose categories like whether or not the legislators had served in the military or went to an Ivy League school were because those were qualities described by pundits as ideal in candidates.
If my goal was to highlight features that were highly predictive of candidates’ future success in elections by finding commonalities, my research has thus far demonstrated that Ivy League education and military service aren’t effective predictors of electoral success.
To follow up on these results, I’d like to examine more community-based indicators. Details like if these candidates were born in the district they represent, they grew up there, they went to college there, number of years they lived in the districts etc. One reason I think I’ll find success in this area is because I noticed a correlation between Ivy League graduates and representatives from New England. It seems more likely that if you went to school near the area you represent, then you're more likely to win an election in that area.
On Tableau
Tableau helped a lot! It really makes you think about the type of variables that you're putting into it before you create a visualization, but it's extremely useful in rapid prototyping once you have an idea of what data you'd like to use. It's very versatile, and presents more than enough options to handle most tasks. However, I did mention that there were several graphs I wanted to create that I would have had to use some fancy workarounds in order to make them function on Tableau. The same graphs are comparatively easy to implement in JavaScript libraries like D3, so I wish Tableau had that same functionality.
First Draft of Project
Thought Process
Since my data had geographic points (i.e. all the states that each representative was from), I initially tried to demonstrate my results overlaid on a map. I thought that I might be able to create a chloropleth heat map for each district where I could indicate the variance between each category that I had mapped out. However, once I started mapping it became clear that I couldn't organize things by congressional district the exact way I wanted to, and I wanted to highlight multiple details about the person, not the district or the geographic area.
I also realized that a map wouldn't really help me with my question. I was looking to see what the most common features of each congressperson were across the board, and a map wouldn't cleanly demonstrate that.
Drawbacks
One of the main reasons why I couldn’t use a heat map was because my values weren’t quantitative, they were nominal. With a variety of labels, I couldn’t create any kind of range without assigning each of the labels a value (which I could have done, but I reasoned that the whole point of making the labels in the first place was to allow the implicit value of associating a congressperson with a particular political figure, prior career, etc. to be communicated). As a result, I thought that my data wasn’t well suited towards the heat map.
I also reasoned that my data (and the question that had prompted the data's creation) was more about the people themselves, not information about their Congressional District. I wanted visualizations that focused more on the people, not the geographic location.
Second Draft of Project
Thought Process
My goal for this next draft was to create a way of having all of my data on one graph. I knew that I didn't want to express that much information in my final draft unless it looked really good, so I wanted to experiment with Tableau and see what I could do. I had a line graph with each congressperson as a line, each quality as a column, and the rows as a measure of each of those qualities (kind of like the examples we've been working with in class lately) in mind.
Right away, it was clear that my nominal data wasn't well suited to this approach either - it was difficult to assign all the labels numerical values, and like I mentioned in the first graph, it didn't express the same kind of coded language that I wanted it to. I decided not to use that aggregation line graph because my values were either booleans (i.e. yes or no’s/true/false), and there wasn’t much range between 0 or 1 0 - or they were labels, the drawbacks of which I already discussed above.
Then I tried to do a scatterplot mapping all the qualities in a single plot, but I realized that in order to do positional information one member can’t have multiple qualities (i.e. it can’t occupy two different spaces in a 2D graph). If it were possible to display 9 dimensions on a single graph, that would be an interesting visual. However, going into anything larger than 3D wouldn’t be effective to convey my information.