ASLCS - The Journal - Computer Technology in the Redistricting Process

Computer Technology in the Redistricting Process

By:
Mary Galligan
Principal Analyst
Legislative Research Department
Kansas

Volume 5, Number 1 Summer 1999

This article is adapted from a presentation given at the 1998 Professional Development Seminar of the National Association of Legislative Information Technology in Baton Rouge, Louisiana on October 10, 1998. The presentation was developed in collaboration with Glenn Koepp, Assistant Secretary of the Louisiana Senate and Linda Meggers, Director of Reapportionment Services Unit for the Georgia Legislature.

Introduction

The first part of this article is a short course on the redistricting process. It is followed by a brief discussion of the computer support used by the Kansas Legislature in 1989 and 1992. The final section includes some of the main lessons learned from the two most recent rounds of redistricting. While this article describes redistricting by the Legislature that results in a bill that must pass both chambers and be signed by the Governor, many of the basics and the lessons are applicable to states that use other decision making formats.

The Kansas Legislative Research Department (KLRD) is a central non-partisan staff agency. KLRD supports both chambers and both parties in redistricting. The role of KLRD staff in redistricting traditionally has been to develop, under the direction of committees, subcommittees, or individual legislators, maps that depict potential plans for new districts and to prepare reports of population for each proposed plan so members can determine whether a district passes the first threshold of acceptability, that is, encompassing the ideal district population.

The Kansas Legislature drew new districts most recently in 1989 and 1992. Based on those experiences, it is safe to say that redistricting is not a computer project. It is the most fundamental policy-making exercise in which the Legislature engages because Legislative districts are the basis for representation and participation in all the decisions that come before the Legislature. The policy-making effort that results in new districts, like many undertaken by the Legislature, is facilitated by the use of computers.

The Short Course

Redistricting can be conceptualized as the process of accumulating bits of territory, for instance voting precincts, into districts that are as close to the same population as possible. In order to do that two categories of data are required:

1. population figures for the geographic units used as district building blocks; and
2. maps of those geographic units.
Maps are necessary because districts must be composed of contiguous building blocks and in most cases must be as compact as possible.

Conceptually, this is a very simple process. One adds precincts to a potential district until the ideal population is reached, draws a district line around those precincts, and starts accumulating another set of precincts to be the next district.

It is even simple technically. Redistricting does not require higher math skills. All one needs to know is:

• the ideal population for each district, which is the state population divided by the number of congressional seats or number of seats in the House or Senate, depending on the plan in question;

• the counts and locations of significant concentrations of racial or ethnic minority population;

• the permissible deviation for district populations—0 for Congress and + or - 5 percent for the legislature; and

• the populations and locations of precincts or other units of geography that comprise building blocks of districts.

Armed with that information, one can sit down with a calculator, a map, and a set of colored pencils and create a district plan.

The process becomes more complex when one recognizes that there are many possible combinations of precincts that can be grouped together to create a district that meets the ideal population requirement. For example, in Kansas in 1992 approximately 200 House district plans were created for consideration. The Legislature had almost as many Senate district plan proposals to consider and another 50-60 Congressional district plans.

But, despite that complexity, one can manually produce maps and reports for members to review. However, just because one can does not mean one should.

Why Use Computers?

The volume of data that must be manipulated and the short time frame for redistricting makes it a natural for computerization. Members want and need to be able to see not just an answer to the question: "How can these 3,200 (or 1,000 or 4,000) precincts be assembled into districts of equal population?" They need to see the best answer to that question. As with every decision made by legislatures, "best" is not a technical term, it is a description of the collective wisdom of a majority of the House, the Senate, and the Governor.

Redistricting is right next to developing a new tax plan or financing for K-12 education, as a process that readily lends itself to computer support. That is because redistricting requires skills and abilities that computers accomplish so much better than people.

• Redistricting by its very nature requires tedious rounds of recalculation after each addition or deletion of a precinct from a potential district. Even a simple spreadsheet does that much better than a committee staffer with a calculator.

• Accuracy is paramount. The legislature’s new districts will be challenged in court. Plan on it. Every analysis that was used by the legislature in making its decision to adopt a plan will be subject to scrutiny—from the summing of precinct populations to the calculation of the ideal district size, deviation from that ideal, percent racial minority population, and any statistics used to identify communities of interest.

• Speed is important. Computers do all the things they do well very fast. That means more time for the assessment and deliberation that lead to decisions. It also means that members can consider many more options than even a huge staff could possibly provide for them via calculator and colored pencils.

Being able to see all the possible permutations of the correct answer to the problem of developing the best district is where a Geographic Information System (GIS) is extremely useful. The most elementary description of a GIS is a computer program that ties together point data representing geography and attributes of those points. But in reality, redistricting does not fully employ even a stripped down GIS.

Many, if not most, legislatures used computer support of some sort for the last several rounds of redistricting. The sophistication of that support has increased over time, but the basic needs of the legislature for technological support have remained essentially the same. Few legislatures have employed artificial intelligence to develop the "perfect plan." For a variety of reasons, legislators have jealously guarded their roles as ultimate decision makers in redistricting. Especially in regard to artificial intelligence, just because you can do it, does not mean you should.

Complexity Is Not A Requirement — But May Be Inevitable

Computer support for redistricting may be very simple. In 1989, the Kansas Legislature contracted with a local firm to develop software that integrated a computer-aided design (CAD) program with a relational database to support redistricting. The polygons representing Kansas precincts were created in the CAD. Each of those 3,000+ polygons was associated with the corresponding population figure in the database. Each polygon was associated with one, and only one, population figure. Reports were generated from the database that showed the total population of each precinct and the total district population which was the aggregate of all the precincts comprising a district.

Legislators presented to committee staff lists of precincts they wanted in districts or drew districts on paper maps that showed precincts. The role of the single personal computer on which the CAD/database program existed was to develop a report of the population of the proposed districts and a plot showing the desired district.

The system price was a few thousand dollars. Two staff members were trained to use it and did so under the "other duties as assigned" provisions in their job descriptions.

So, if the Legislature can get by with only the most rudimentary abilities of computers, what is the big deal about redistricting?

The big deal is litigation. Every decision, every number, every report, every note, comment, and direction in and around the redistricting process can, and in all likelihood will, become part of a lawsuit. All of the planning and system building that precedes redistricting is preparation to defend the legislature’s work product in court.

By 1992, the Legislature’s understanding of the task at hand had matured considerably. It was lucky in 1989. That set of House districts was not challenged. Members knew, however, that their luck could not hold.

In 1992, the Legislature used six VAX 3100 workstations and a VAX 3400 server communicating over a dedicated network with a proprietary software system to manipulate 3,200 precincts composed of over 160,000 census blocks. Each of those blocks had associated with it:

• 20 attributes describing the population;
• 21 election result attributes for two election cycles;
• 3 voter registration attributes; and
• geographic attributes identifying it as a component of a tract, precinct, city, township, and county.

The Legislature decided to contract for software and support services specific to redistricting. Considerations that went into that decision included:

• Reluctance to hire the necessary development staff to tailor a GIS to the redistricting job. The Legislature did not at that time, and does not now, have computer application development staff. Most GIS programs available in the late 1980s required anything from tweaking to major development to make them efficient for redistricting. The group of legislators charged with planning for redistricting did not want to "staff up" for this effort, only to lay off all those employees after redistricting was over. Even with an in-house development staff, the same considerations apply, unless there are "excess" staff who can be taken off ongoing assignments for the next few years.

• A desire to have a redistricting system that was sufficiently easy to use that legislators could work virtually unassisted at a workstation. Creation of really usable, foolproof user interfaces adds significantly to development time.

• A perceived need for redistricting-specific technical expertise to support in-house staff assigned to the project. The threat of litigation contributed to this decision to the extent that one of the services the Legislature sought from its contractor was expert testimony in any court challenge of the plan adopted.

Preparing for the Next Round

The legislative leadership in Kansas appointed a six-member planning group, three senators and three house members, during the 1998 interim to begin the planning process. That group will remain in existence until 2002 when the Legislature will adopt new district plans.

Staff support for the Kansas planning group includes two researchers and two attorneys, all of whom do this work along with their other duties. Since Kansas does not have a central computer system development staff, KLRD provides most of the required technical expertise. This planning group will depend on staff to provide them with information for the planning process. Based on that information, they will decide whether to use in-house staff or contract for support. They also will dictate the functions of the redistricting computer system, for example, whether it will generate districts automatically and whether the system will be simple enough for legislators to use without staff assistance. They will decide what non-population attributes need to be included in the database. They also will decide the number of workstations at which plans may be developed and whether or how the public will be able to access the legislative system. They will decide where and when to hold public hearings. They will create the record that will become the defense against accusations of unfair districts. The technical and legal staff will lend crucial support to the decision-making and record-building effort.

All of the staff involved and all of the legislators who guide that staff must keep the potential for litigation foremost in their minds at all times. None of the players—the legislators, the technical staff, the legal staff—can adequately make preparations alone. Litigation preparation is a team effort.

Lessons

Kansas has learned many lessons from two rounds of computer assisted redistricting.

1. Redistricting will continue to be a computer-supported effort because of the tremendous amount of data that must be accurately analyzed in a very short period of time.
2. Redistricting applications do not need to be cutting edge. They do need to be absolutely reliable, accurate, and fast. They need to be secure to protect data integrity and to protect the confidentiality of each plan prior to its public presentation.
3. A "turn-key" redistricting system that meets legislators’ needs during the intense burst of redistricting activity may well be worth the money even if post-redistricting applications are limited.
4. There is never enough storage space for all the electronic and hard-copy maps and reports generated in this process. Notwithstanding the dearth of storage capacity, all the documents, logs, and notes that led to the Legislature’s decision need to be preserved for defense of the plan in court. Consult frequently with the legal staff regarding what needs to be preserved.
5. The Legislature must be able to quickly and accurately evaluate plans developed by individuals and groups outside the Legislature. That function must be included in any redistricting system. The availability of GIS software that can be used to create districts on desktop personal computers makes the ability to evaluate outside plans more important than ever before.
6. Development and implementation of redistricting computer support are unlike any other in the legislative environment. This job breaks the legislative rule that "there is never time to do it right, but there is always time to do it again." Even if you are in North Carolina, where redistricting has become a decade-long rather than once-a-decade activity, the redistricting computer support has to work perfectly the first time.
7. The combined effect of the timing of the release of census data and legislative sessions results in inadequate testing time for the programs with the real data. In Kansas’ case, the general timing will go like this: Population data will be received from the Census Bureau sometime in March 2001—while support staff are busy with their duties in the 2001 Session. After adjournment in May, legislators will begin holding public hearings related to redistricting. Members will begin developing congressional plans in July or August. Proposed plans will be considered by the Legislature beginning in January 2002. By May 2002, it is over. In November 2002, members will be elected from those new districts.
8. Staff and vendors are too inclined to say, "we can do that." If contractors are solicited, require a real-life demonstration of the software on your equipment at your site. Be sure the demonstration is conducted with prototype data provided by the Census Bureau to ensure that the software can actually handle the volume and structure of the data. Neither staff nor vendors will have time to rewrite code to handle different file layouts after the census data arrives.
9. Scope creep, the tendency for any technology application to grow during development, is a reality. Focus on the goal of developing or procuring a redistricting application that will do the job it is asked to do. Be content with building auxiliary applications after redistricting is done.
10. The small things count. The appearance of maps and reports is very important to members. Both maps and reports must be understandable. They must be readily reproducible and in large quantities. Do not overlook production of output that can go directly into your bill drafting system.
11. Redistricting probably will have an impact on other computerized operations and system performance, especially if a single network supports all legislative functions. Redistricting can generate a tremendous amount of network traffic because the files are very large. The network may need special preparations for increased traffic. It probably will not be acceptable to bring the bill processing system in the House to its knees every time the President of the Senate wants to review a plan.
12. A successful redistricting effort requires thoughtful planning. As long as the Legislature and the staff supporting redistricting go into the process with a firm grasp of the task at hand, solid planning, steadfast resistance to the "we will do it just because we can" approach, and unflappable senses of humor, redistricting and post-redistricting litigation can be successful.

For more information about ASLCS, write or call:

Joan Barilla
National Conference of State Legislatures
7700 East First Place
Denver, CO 80230
Phone: 303/856-1349
FAX: 303/364-7800
E-mail: joan.barilla@ncsl.org