No one selected (charts shows victory rate)
The U.S. Congress is a complex entity, with 100 senators and 435 members of the House of Representatives, or 535 members total. While political party is certainly a very important factor in terms of how an individual congressman will vote, it is not a purely determining factor as in other countries where parties often vote as blocks and legislators merely occupy a seat for their political party. Congressmen, especially in the Senate, often act individually, and taking a position contrary to the party's position is not unheard of (although Congress is becoming more polarized and crossing party lines is less common). Thus, the political structure of Congress is much more complex than simply which party holds the majority.
This project aims to visualize congressional voting patterns in terms of how similarly congressmen vote. If two congressmen tend to vote the same on bills or tend to cosponsor the same legislation, they may be considered "similar". When thinking about relationships between members of Congress in this way, we would like to visualize these relationships.
Currently, this project visualizes voting patterns in the U.S. Senate during the 114th Congress, from July 30th to November 30th, 2015.1 This period includes 50 Senate roll-call votes.
Above is a scatter plot and map useful for visualizing congressional similarity and relationships. The scatter plot shows each member of the congressional body as a point, with the vertical position denoting agreement2 with the selection, and horizontal position indicating political ideology3. The map encodes a state's congressional delegation's agreement4 with the selection as lightness (the lighter a state's color, the less its delegation agreed with the selection), and the political affiliation of the delegation as hue, with blue denoting Democrats, red Republicans, and yellow independents. In the case of a mixed delegation, the hue will appear redder or bluer (or yellower) depending on whether state Democrats or Republicans (or independents) agree more frequently with the selection.
When the selection is empty (the starting state), the visualizations show every member's "victory rate"5, or the rate at which a member votes with the winning side of a motion. Members can be added to the selection by clicking a point in the scatter plot or by entering their name in the text box. Clicking a state in the map will set the selection to that state's congressional delegation. The existing selection will be replaced with the new selection unless "Keep existing selection" is checked, in which case new members are added to the existing selection. Click the "Clear selection" button at any time to empty the current selection.
Currently, neither the Republicans nor the Democrats are able to dominate the Senate. This can be seen when the visualizations show each member's "victory rate". Both sides have victory rates that center around 50%. If the Senate were ideologically driven, we would see the data points centering around a trendline, with the slope depending on the dominant group (an upward slope indicating a conservative Senate, and a downward slope for a liberal Senate).
The state with the highest victory rate is Ohio. The Ohio delegation consists of both a Republican (Sen. Portman) and a Democrat (Sen. Brown), who is the Ranking Member of the Senate Banking, Housing, and Urban Affairs Committee. Both of these candidates are fairly ideological, so it is not too surprising that the Ohio delegation, overall, has a high victory rate; there is a good chance that either one of these members will vote with the winning coalition. Among Republican-only states, the Utah delegation has the highest victory rate, and among Democtrat-only states, Delaware has the highest victory rate.
There are a handful of Senators with very low victory rates, all Republican: Sen. Graham (R-SC), Sen. Rubio (R-FL), Sen. Vitter (R-LA), Sen. Paul (R-KY), and Sen. Cruz (R-TX). With the exception of Sen. Vitter (R-LA), all of these members are running for President, and they all have unusually high rates of not voting. Thus, their low "victory rate" is not due to frequently being members of losing coalitions, but because they are not voting on issues at all. (With the exception of Vitter (R-LA), these members are likely distracted by their Presidential campaigns.)
There are strong ideological battle lines in the Senate, with a wide gap between Democrats and Republicans (excluding the members with low victory rates, who vote so infrequently it appears as if they are not very ideologically driven). There are two Democrats, Sen. Donnelly (R-IN) and Sen. Manchin (D-WV), who are ideologically similar to moderate Republicans; both of these members are from states that are considered "red states". There are no liberal Republicans, though both independent members (Sen. King (I-ME) and Sen. Sanders (I-VT)) are liberals. In this period, Sen. Risch (R-ID) was the most conservative member of the Senate, and Sen. Wyden (D-OR) and Sen. Merkley (D-OR), both from Oregon, were the most liberal. Sen. Manchin (D-WV) was the most conservative Democrat, and Sen. Collins (R-ME) was the most liberal Republican (Sen. Graham (R-SC) not being considered because of the large number of times he did not vote).
When looking at agreements for various senators, we see that for senators close to the extremes, there is a linear relationship between agreement rate and ideology, which is not surprising (ideology is computed via dimensionality reduction, so agreement rate and ideology contain some of the same information). What is more interesting are members not at the extremes. We initially thought that we would see a low correlation for members not at the extreme, with "low correlation" meaning a cloud-like scatterplot. What we actually found was that more moderate members do have a relationship with the votes of other members: they are more likely to agree with other moderates and less likely to agree with extremists, so the scatterplot has a mountain-like shape for moderates. Try selecting members on the scatterplot from the left to the right (make sure the "keep existing selection" box is not checked) to see this for yourself.
Members from different political parties are quite distant in regard to political ideology, even in they are from the same state (West Virginia seems to be the only exception to this; the Democrat from West Virginia, Sen. Manchin, votes like a moderate Republican). This means that when two members with very different ideological positions agree, all the other members have high agreement with that coalition. This may be because such a coalition forms only for issues involving little partisan divide. Try adding the most extremist senators to the above selection to see an example.
When a state is selected, a user can use the shape of the scatterplot to obtain a sense of what the ideological leaning of that delegation is. Often, state delegations vote like a highly ideological member of their political party, though this is not always the case. Interestingly, states with delegations split between the two parties do not look like moderates; since these members are often ideologically distant from each other, if they are in agreement, many of the other members in the Senate are in agreement as well (in a sense, this delegation votes together only for nonpartisan issues, so the rest of the senate votes in agreement). West Virginia's delegation, though, seems to vote like a moderate Repubican.
One question we wanted to address with these visualizations was whether we see agreement depending strongly on political party or whether politics was largely regional. There are various ways to see this in the visualizations, like clicking a member and seeing what states members in high agreement with the selected member hail from. We do not see strong evidence of regional politics in this period; it appears that political party plays a stronger role in determining voting behavior than region (neglecting the fact that many regions tend to be dominated by one political party, like the South being dominated by the Republican Party and the Northeast and Pacific States being dominated by the Democratic Party). For example, Sen. Ayotte (R-NH) is just as likely to be in agreement with Sen. Collins (R-ME) as she is with Sen. Kirk (R-IL), and she has high agreement rates with members in the South but not with many Northern Democrats, even though she hails from that region.
We hope to expand this project in the future to include the House of Representatives and past Congresses. In the meantime, we hope you have found this visualization enlightening, informative, and entertaining.
Interested users can read the project journal, which describes the steps taken to create this visualization along with code documentation. The project's github repository is available here.
We developed our visualizations using the JavaScript library D3, by Mike Bostock. His scatter plot example served as the basis for the scatter plot visualization we developed. There were numerous other sources of code used in our visualizations. GitHub user mshafrir created a dictionary of full state names and abbreviations that we used for data manipulation. We obtained CSS code for creating the horizontal rules seen in this document here. We followed Scott Murray's demonstration for creating tooltips, using his code. For the member input textbox, we used the example described on the jQuery website. The fonts used in this document are from Google Fonts; the text and subheadings are Georg Duffner's EB Garamond font, and the title is Nicole Fally's Pinyon Script. Finally, we used the HTML5 Boilerplate template to begin construction of this webpage. Much of the data preprocessing was done using R.