The United States' history and politics have been shaped by how political boundaries and districts are drawn. Gerrymandering—the drawing of strange districts for political purposes—is considered business-as-usual. California is a particularly difficult state to divide into political units because it is so large, its population is now growing faster than the U.S. as a whole, and its citizens do not agree on many basic issues. The one thing everyone seems to agree on is that voters do not understand the redistricting process.

Redistricting is complex because there are many vague, conflicting guidelines and court cases dealing with what makes an ideal political district. Some locations such as Colorado have offered online mapping sites to their citizens to help them understand and comment on proposed districts. Most states offer Web sites with redistricting information, data, or maps (McDonald). However, citizens are usually left out of the redistricting process. My first inspiration for this project was a high school student in Berkeley, California, who drew his own city council district map (by hand, with help from the U.S. Census Web site), which was considered in the 2001 redistricting. Due to his knowledge of his city, he was able to pinpoint where the students (which could be considered a community of interest) lived. My second inspiration was the representative for Assembly District 5 (North Sacramento County/South Placer County), Roger Niello, whom I heard question some of the boundaries of his district.

There are many software packages to aid analysts in creating districts. Due to the increasing availability of GIS software and to today's powerful computers, many citizen and activist groups could afford to set up a sophisticated redistricting system. However, these GIS-based packages can also make it easier for political partisans to manipulate district boundaries for their particular purposes--i.e. to gerrymander more efficiently and less obviously through technology (Abramsky, 2003). After talking to someone who has used these specialized systems and some reading I have done, I began to suspect that these systems take a mechanistic cut-and-weigh (or nibble and count) approach to redrawing boundaries, based primarily on moving random census blocks rather than using technology to implement as many good government guidelines as possible.

I focused on populated places because Assembly District 5 includes many unincorporated areas that have defined boundaries. Some of these communities have considered or are considering incorporating as cities, which would make their boundaries even more important. Had I focused on a city-centered district such as the assembly district that covers most of Sacramento, I would have collected neighborhood and school-service area boundaries. My personal experience led me to believe that many Californians, especially those with children, would also look at the people living around a school as an important community of interest; I also found a quote about this topic on the Internet (Ashmore, 2001).

I started by researching the legal guidelines for redistricting. There are many conflicting and vague guidelines, but one that I was not aware of is that California is much more flexible than the federal government when balancing population among districts. This means districts can vary in population by as much as 10 percent, though population balance is still important; this makes it possible to adopt citizen-suggested district lines without major changes, while extremely strict population balancing probably means that expert redistricters, using specialized tools and many iterations, would have to draw the final lines.

Through the Internet and a fellow student, I found the official source of redistricting data in California, the Statewide Database (SWDB), a free and nonpartisan resource maintained by the University of California's Institute for Governmental Studies (IGS). The SWDB repackages the official U.S. Census data that must be used in redistricting, along with offering political boundary files and voter data. I decided to use the 2001 census data, despite the fact that its population data is now out of date, because similar data will be used in the next redistricting in 2011. I also looked for data that would provide the background layers that would orient a map reader, in particular, data that would help a user decide where a neighborhood or "community of interest", such as a student housing project or a defined neighborhood, is located and bounded. In the maps of my focus area, I utilized geographic layers covering street, waterway, census block, census tract, census populated places, school districts, highways, parks, institutions, and state and federal political districts, along with precinct and census tables obtained from the SWDB site. Some of this data came from ESRI (Redlands, CA) or from the U.S. Census site.

I set up an ArcMap project with these layers and clipped and projected it (UTM Zone 10) to fit my Northern California focus. The only additional data preparation I did before mapping was to merge the three types of school districts (elementary, secondary, unified) recognized by the U.S. Census into one layer. There were additional data layers I should have included, such as city council and county supervisor districts, but I did not find them on the Internet and I felt I had enough information for my analysis.

I used ArcMap to explore my area of interest, Assembly District 5, at various levels of detail and scale. I looked at the edges of the district in particular. GIS layer transparency settings made it easy to overlay and view various boundaries and district options. I produced maps comparing the 1991 and 2001 district lines (Figure 1) and comparing the current (2001) district with other boundaries, such as State Senate lines, school district lines, census divisions, and populated place boundaries.

In order to create an ideal district following the criterion of protecting existing boundaries, I started with the existing district shape as a guideline, despite its rather twisted and sprawling shape. I selected the district, then saved my selection as a new layer. I then used ArcMap Select by Location function to determine which, if any, school district boundaries were intact.

When I found that none were intact, I tried two techniques. First, I buffered the district by 1 or 3 miles to determine if simply expanding the existing district would bring the split districts fully back into the assembly district. Second, I modified my selection to choose school districts that had their center in the assembly district. I created a new layer from that selection. I repeated this process for populated places (in this district, this meant census-data places [CDPs], or unincorporated areas). The two selection shapes were combined through the ArcMap Union and Buffer tools in order to create one simple polygon representing my new ideal District 5 (Mydistrict in Figure 2). I also used the ArcMap Selected Features count view to create a results table comparing the different methods and selects (Figure 3).

Figure 1. Click on map for larger image.

After building my ideal district, I selected a small area that had not been included in the actual assembly district. I did this in order to determine how to do some elementary political analyses. I created a new layer of my selection and joined the census block features to a SWDB table that links Sacramento County census blocks to voter registration and census data. From this, I produced graphs of demographics and voter registration using ArcMap and Microsoft Excel (Figure 4).

My ideal district was larger than the existing district, which could mean it could be over the variance guidelines for population balance. This could easily be corrected by removing a small government unit such as a small school district, or by removing census blocks, then recalculating the number of residents.

Figure 2

My ideal district (Mydistrict in Figures 2 and 3) is not that different than the actual district, yet, because my decisions were driven by school district and populated place boundaries, it preserves 12 populated-place boundaries and six school district boundaries. It does not include an entire county because of the high population of Sacramento and Placer counties (which is much higher than the maximum population of an assembly district).

Figure 3

The demographics of parts of this primarily Republican district are distinctly different than those of the district as a whole. In a small selected area of the district within the Sacramento city limits, there was a concentration of women, Democrats, and independents (Declined to State in Figure 4).

Figure 4

I did not make full use of the data that is uniquely provided by the Statewide Database resource mainly because I did not focus on registration or demographic analysis. The hierarchical precinct structure of California makes it difficult to do voter analysis across elections and to compare voter registration to election results. The Statewide Database site explains this problem well, but its conversion tables that attempt to solve these problems are difficult to understand and use. Given more time and some help from staff, I would have perhaps grasped them. However, in general, tables that aggregate information by various fields limit the end user, and the utility and currency of the information in such tables depends on which questions the designer thinks people will want answered or which questions they have asked in the past. As I am familiar with both GIS and database concepts, I would like to find a direct source of registration and voting precinct geographies, along with detailed attribute tables, so I could do my own direct analyses.

My analysis of District 5 would have benefitted from a population density layer, more boundary layers such as county supervisor districts, and (since I assume this area is growing quickly) more up-to-date school and street information. The collection of higher quality information would have required contacting local agencies or even focusing on a much smaller area for study.

The interactive mapping features available through ArcMap produce useful maps for improving the redistricting process. The ArcIMS Web-based mapping service has some limitations for redistricting, but with enough development time, a Web-based service could let citizens become more involved in redistricting. In contrast, the number of paper maps that would have to be produced to offer this amount of information would be enormous, and the mapmakers could make too many assumptions about what is important to map users. If a variety of political, census, and background layers are provided, a knowledgable resident can do a better job at identifying communities of interest than a fully automated system would. In addition, GIS technology helps citizens gauge the quality of a suggested redistricting plan. People can explore excluded and included areas graphically, which requires almost no GIS experience, or they can use more sophisticated GIS selection and summarizing tools to locate and document problems. I, for example, found graphical evidence (Figures 1and 5) of something I had read—that the last redistricting of Assembly District 5 was a status quo, census tract-driven one. (DeBow, 2005)

Figure 5: AD5 with Census Tract Boundaries