This is Part 2 of a continuing series. All figures and references are numbered sequentially throughout the series. IQ lessons 1-4 appear in DM Review's October issue. (Part 1 of this series can also be found at
Improve Defective Election Processes
The 2000 U.S. presidential election exposed seriously flawed processes. This series will examine those processes and pertinent quality lessons:
- Voter Registration
- Ballot Design
- Election Day Voting Registration
- Voting Process
- Vote Recounting
- Absentee Voting
- Election Process Improvement
There is a way forward. Quick fix panacea? No! Permanent solutions with quick wins and sustainable improvement? Absolutely! There are proven quality process improvement methods that apply to information processes that can and must be applied to the election processes in order to achieve election reform that will be addressed in "Election Process Improvement."
We now review the various election processes and the pertinent information quality principles and techniques for improvement in the form of IQ (Information Quality) Lessons, describing quality principles, and QI (Quality Improvement) Recommendations, describing specific recommendations.
1. Voter Registration
Citizens cannot vote unless they are registered. Citizens must have easy access to a process to register. The voter-registration process must have integrity to assure persons presenting themselves are qualified to vote. Voter registration data must be easily available on election day or during early voting. Processes must be in place to capture updates to citizens' new addresses and voting jurisdictions to prevent denial of voting right and fraud. Currently, American residents have a mobility rate (change of address) of about 15 percent per year (with a high of 20 percent in 1984-85).15
IQ Lesson 5: Process integrity and information quality begin with clear definition of the data that must be known and integrity of the design of the databases to house that information. States must have well-defined voter registration databases with easy, yet secured, access by each other.
QI Recommendation: Design the voter registration database and process to capture "identity attributes" such as birth date, mother's maiden name, last four digits of Social Security number and driver's license number, digital signature and fraud-prevention information such as deceased and other attributes that disqualify one from the voting privilege. The process and database must also be designed to capture and maintain name changes as well as previous addresses of individuals with relative effective dates to handle the "information quality decay" of address changes and disqualification events.
QI Recommendation: Processes should be in place to capture current addresses easily. National Change of Address (NCOA) data from the U.S. Postal Service and motor vehicle registration should be utilized to capture new permanent addresses. Access to a database of deceased persons will help prevent voting fraud.
Registered voters should be able to update their name and address data easily. Enable updates through the Internet using secured services and proper identification verification to reduce problems on election day.
2. Ballot Design
The Palm Beach County butterfly ballot called attention to the importance of ballot design. With some 19,000 overvotes, this one county accounted for 17 percent of the 111,000 overvotes in 67 counties as well as one fifth (19.5 percent) of the Florida's total votes for Pat Buchanan, three times as many as the county with the next highest Buchanan vote (see Figure 2).
Figure 2: Buchanan Votes by County
This count represents a statistical outlier in a county where the demographics are not consistent in general with Buchanan's positions and where Reform party registration is a fraction of the total number of votes recorded for him. This vote-count anomaly can be explained by the design of the ballot in which there is overlap between the candidates listed and the selection punch holes. The butterfly ballot in Figure 3 illustrates the ambiguous location of selection punch holes, overlapping the candidate boxes in that ballot. The arrows pointing to the selection punch hole are small, weakly calling attention to the correct punch.
Figure 3: Palm Beach Butterfly Ballot. (Sun-Sentinel graphic/Daniel Niblock. http://cnews.tribune.com/news/image/0,1119,sunsentinel-nation- 82382,00.html
Figure 4 represents an improvement in presentation quality. The candidate boxes are separated from one another, center-justified so as to be closer to the action punch holes, with no ambiguity of relationship of punch hole to candidate. The color/shade separation and stronger pointing from candidate to selection box much more strongly associates the correct punch selection.
Figure 4: Improved Ballot Design
Palm Beach County was not alone in problems with ballot design. Gadsden County had a 12 percent spoilage rate with almost 2,000 double-punched ballots. The ballot had eight presidential candidates in one column and two in a second column that looked as if it could be a different election. Duval County had ballots spread over two pages, creating a 7 percent reject rate in predominantly white precincts. Compare this to a reject rate of less than 2 percent in the District of Columbia.16 This violates the "information chunking" principle in which all information about a given subject is contained together in a cohesive group.
Furthermore, there were numerous problems with the printed procedures for voting. The Palm Beach ballot instructed voters to vote for their candidate by punching the hole "to the right" of the candidate's name. However, the selection punch holes for all candidates on the right side of the butterfly, including Buchanan, were to the left of the candidates' names. Refer again to Figure 3. The Duval county sample ballot (with presidential candidates spread over two pages) that was inserted in the Florida Times-Union contained voting instructions that stated, "Vote all pages."17 This was corrected in the election day instructions to read, "Vote appropriate pages."
Instructions on both optical ballots and punch cards usually state, "Vote for Group" when referring to the President/ Vice-President pair. But where are these two ever referred to as a "group" in any typical context? This is confusing generally and is compounded by the often-accompanying text, "A vote for the candidate(s) [sometimes singular and sometimes plural] will actually be a vote for their electors."
IQ Lesson 6: There are three components to information quality: 1) Definition (you know the meaning of the data, e.g., "touching this button causes your vote to be recorded," or "party" denotes an established political group that supports specific principles and to which a candidate belongs); 2) Content (accurate value, e.g., the name of the party by the candidate's name is the correct party name); and 3) Presentation (information, such as instructions for voting, is presented in a way that is clear to the voter. When actions are required or information is requested, the information is presented in a way that leads the knowledge-worker to make a right action. Information presentation design is of utmost importance in any form for communicating information, requesting action or collecting information).
QI Recommendation: Design ballots with all candidates for one race chunked in a single space with unambiguous placement of candidate names with their selection mechanisms. After ballot design, write and test voting instructions and procedures against the ballot to assure consistency.
QI Recommendation: Review instruction wording for clarity. For example, the instructions "Vote for Group" should be replaced with a more clear, natural text, "Vote for one President/Vice President team by completely shading the oval next to their names."
QI Recommendation: Use multi-sensory design techniques. For example, include candidates' photographs as well as names for visual recognition. Electronic recording devices can use sound to indicate the selection of a candidate, to indicate when an incorrect choice is made and to indicate when the final recording of a vote has been made.
Again, we cannot "blame" the ballot designers or the people who signed off in approval of the ballot. A basic principle of accountability is we cannot hold someone accountable if they have not been trained or taught the principles.
IQ Lesson 7: Point 6 of information quality states, "Institute training on information quality."18 Deming asserts a universal truth that people cannot do a quality job if they do not know how.
Basic training in information quality principles must be provided for election officials, especially for ballot designers and procedure writers. This training can be standardized for general principles with modules that address the different voting systems. A key focus should be error-proofing techniques that help prevent common errors inherent in the different types of voting technologies.
Training must provide guidance in how to conduct a "usability test" of ballots for voter-friendliness by testing them in ways to simulate the experience voters will have in the voting booth.
As many as 178,857 voters in Florida alone may have been denied their right to have their intended votes counted in the presidential election.19 That many ballots were not counted by failure of the vote counting process to register a vote (undervotes) or were disqualified by having selected more than one candidate (overvotes). That represents as much as a three-percent "error" rate. In other words, the votes of as many as three out of every 100 voters who thought they voted for president were not counted. It is not possible to determine from an undervote with no perceivable mark whether the voter intended to vote for someone or not. National Election studies indicate approximately 0.73 percent of voters deliberately abstained from voting in the presidential races, and exit polls from Voter News Service captured sufficient data in 1992 to indicate 0.77 percent of voters abstained from voting in that presidential race.20
IQ Lesson 8: The absence of information can never be inferred to mean, "I chose to leave out information." There must be positive indication of a planned absence of information.
QI Recommendation: Implement a "no vote" or "abstain" option for the voter to select in order to assure that voter intent was to abstain from voting in a given race. This should be adopted especially for voting methods prone to high degrees of failure to sense a voter selection. This is the only way to assure that a "no vote" (undervote) was intentionally cast.
QI Recommendation: The term "residual vote" means "all uncounted votes, undervotes (whether the voter intended to abstain or not), spoiled votes and overvotes. Create a new term or redefine the meaning of "residual vote" to include only defective votes (unintentional undervotes, spoiled or overvotes). This is a more meaningful metric because it includes only defective votes for measuring voting process integrity. It requires the use of a means to identify a "no vote" in any election.
Editor's Note: Reference items #1- 14 appeared in Part 1 of this series in DM Review's October issue.
15. Jerding, Grant. "U.S. motility rate slipping," USA TODAY 19 January 2000, citing U.S. Census Bureau data.
16. Mintz, John and Dan Keating. "Fla. Ballot Spoilage Likelier For Blacks," The Washington Post 3 December 2000, p. A28.
17. "Samples, actual ballots differ in Duval," The Tennessean 2 December 2000, p. 7A.
18. English, Larry P. op. cit, p. 364.
19. Cauchon, Dennis. "USA TODAY, other newspapers examine overvote." USA TODAY, 11 May 2001, p. 2A.
20. "To Assure Pride and Confidence in the Electoral Process" report, The National Commission on Federal Election Reform. August, 2001, page 53. www.reform elections.org/data/reports/99_full_report.php.