CS E64 Final Project
May 22, 2009
Michael Tracey Zellmann
CS E64 Final Project
May 22, 2009
Each state gets a share of the 435 seats in the US Congress, based on their population in the census taken every ten years. The Constitution lets the states decide how to establish their voting districts. History shows that the incumbent party can manipulate those boundaries to its advantage. Gerrymandering is the term used and Wikipedia has a good description of the history with examples, if you are interested.
To counter this possiblity, several approaches have been proposed to objectively divide a state, so politicians can't influence the outcome. Often they use computer algorithms to create the districts. One example is the work by Brian Olson, who developed an application that seeks the best map for each state - such that "for all districts and all people. The best district map is the one where people have the lowest average distance to the center of their district." Below are illustrative results for California, Texas and Massachusetts. The current arragement is on the left, and the alternative is on the right.
I was intrigued, but wondered how a citizen might decide what this meant for them. For me - just an image wouldn't be enough. So, I conceived a visualization that would help you understand both plans, and see how they might affect you. Then you could make an informed decision about the merits, and possibly choose to influence your legislators to support an alternative.
Specifically, I wanted to address these questions.
To build an example, I applied this to Massachusetts. Typically, The current algorithms divide a state using census tracts, regardless of existing city and town boundaries. I didn't think that was right - two voters in the same city should have the same Representative. So, I used Biran Olson's solution as a starting point, and created a plan for Massachusetts that kept its 351 cities and towns intact. The measures of merit are district population equality and average distance from a city center to the district center, weighted by the population of each city. While the Constitution does not have any guidelines about district plans, US Supreme Court cases following the 1964 Voting Rights Act have found that districts that are bizarrely shaped, are not contiguous or have populations that vary by more than 1% may be examples of inappropriate plans. With a state population of 6,349,097, the target for each district is 634,910, with a range of 633,401 - 636,419. The current Massachusetts plan splits four cities - Boston, Fall River, Wayland and Hanson, across districts. The 'Current Districts' map outlines those cities in red.
Boston and Brookline share a bizzarre shape, and I wanted them to be together. Unfortunately, their combined population is 646,248, already over the high limit.
With this starting point, my plan misses the target, and without an algorithm to search for a better solution, I didn't improve my results. However, the districts are more compact. I believe I can meet the population range by implementing several algorithms in the future. To illustrate the results, the application starts with a panel showing maps of the two plans, along with a comparison panel showing bar graphs of the populations and average distances for each district in each plan. The "target" population range is highlighted.
You can interact with the application to answer the other questions, and get more information.
The radio buttons below the maps will show the election results for 2000-2008, for President, US Senator and US Congress. The district colors now show Democratic/Republican pluralities. As you move through the elections, it becomes evident that the alternative districts don't change any results. Everything stays blue. Massachusetts is strongly Democratic. I expect another state would show noticeable differences.
Clicking on a district in either map will drill down to show the detailed map of that single district. In this view, the election results are shown for each city.
Finally, to see what this means for any one of the cities, choose from the selection list. The current map and the alternative map will show the district for that city. Additionally, to the right of each map will be a data panel summarizing the demographic make-up of that district. A similar panel in the top-left corner shows the data for the entire state, and in the lower-left a panel shows the data for that specific city.
The top of the data panel shows election statistics. The 'R' bar counts voter registration. Red and blue denote Democratic and Republican registrations. The green counts un-enrolled, a major factor in Massachusetts, along with any third party members. The 'P', 'S' and 'C' bars show the counts for the Presidential, Senatorial and Congressional elections. For the elections, the green bar indicates blank ballots - or sometimes third party candidates, but that never totals much. The central vertical bar indicates 50%. Hence, if the blue bar extends to the right of the center, the Democratic majority is > 50%. You can quickly see that all Presidential results had a Democratic majority, and even stronger Democratic majorities for Senator and Congress. The age bars count population within each age bracket. The racial mix shows the balance between White, Hispanic, African, Asian and Other. The histograms show the number of cities at each level of income or home value, with the state-wide median in red.
I believe the visualization provides a way for citizens to answer the three questions for themselves. They can also explore the data more deeply in any way that matters to them. They can see the population and distance comparisons. They can see the election patterns for both plans. They can see what happens to a specific city, and gain insight into the population differences.
At present, the application runs as a desktop application, using a large screen - 1440 * 1030. I plan to release it as an applet with a smaller screen and a Java Swing Web-Start application for the full-screen version.
I plan to get a solution working for Massachusetts that meets the 1% population target, by separating Brookline and Boston, and implementing some of my own partitioning algorithms. After that, I will show the results to people in the Massachusetts government, who are planning for the 2010 census, as well as members of both political parties to get their feedback and try to incorporate their suggestions. After that, I will extend the application to handle other states, probably concentrating on California and Texas, where there are real issues and a lot more data. These developments will show up on my web site: http://www.nautilusone.biz, so check there if you are interested.
I will add another measure to compare the plans - the difference between the total of all district perimeters and the total perimeter of the state. Compact districts will have smaller perimeters.
I believe this visualization is a practical way to illustrate the results of different algorithms for partitioning each state, the major theme of my planned Master's Thesis.
Geographic information describing the boundaries of the cities and towns, as well as the current Congressional districts came from MassGIS, a service of the Commonwealth of Massachusetts.
Election statistics came from Public Document 43, for the years 200, 2002, 2004, 2006 and 2008, published by the Elections Division of The Commonwealth of Massachusetts, William F. Galvin, Secretary of State.
Census data on population, age brackets, racial mix, family income and home values was provided by the American Fact-Finder Service of the US Census Bureau.