This is Part 3 of a continuing series. All figures and references are numbered sequentially throughout the series.
3. Election Day Voting Registration
Many problems were cited where qualified voters were turned away at the polls because of inaccurate, incomplete or non-current voter registration lists. Types of failure here include errors created during voter registration process, no process to capture address change data, lack of access to the authoritative voter registration database or precinct voter lists created from non-current copies of voter registration data.
IQ Lesson 9: No matter how well- defined or how accurate and complete, data is nonquality unless it is available to processes that require it when it is needed. Processes should have access to the authoritative source of data or downloaded or printed data from that authoritative source on a timely basis.
QI Recommendation: Jurisdictions that had election-day registration problems should analyze causes of missing information. If printed lists are used, assure they are from the correct source and are made after all updates have been applied. If electronic files are used at the polling place, assure they are downloaded with current data. The best solution is to have secure access to the single authoritative voter registration database.
QI Recommendation: In no case should a properly registered voter be denied the right to vote. If a voter attests that he/she is eligible to vote, he/she should be allowed to vote with a conditional or provisional ballot. These provisional ballots should have appropriate identity attributes to allow next- day verification. If the provisional voter is found to be eligible, the vote should be included.
IQ Lesson 10: The maxim "haste makes waste" applies in election day processes. When voters have to fit voting into their workday schedules, they have a tendency to rush, increasing opportunity of error. Polls tend to have "rush hours" that likewise contribute to operational errors.
QI Recommendation: Make election day a national holiday. This reduces the pressure for both poll workers and voters.
If voter lists are used at polling stations, an additional scheduling technique can be used to "suggest" staggered times for voters based on the second letters of their last names, resulting in four or five relatively equal groups of voters spread across four or five time slots. For example, names with second letters of a-d might have a suggested voting time of 7:00 a.m. to 10:00 a.m. Persons with names Bach, Talbert, Edison and Udall would be encouraged, but not required, to go to the polls during that time to even out the demand at the polls.
QI Recommendation: Take advantage of normally wasted wait times in lines at the polling places by passing out voter instructions and sample ballots. Provide a videotape on continuous loop TV monitor(s) with "on-the-job" training that shows someone voting and validating a ballot. This should show types of problems the voter might encounter with the voting method used and what to do to prevent or correct those problems. Provide audio headsets for different languages.
QI Recommendation: Make expert poll workers available for people needing additional assistance and a sample voting booth to practice the voting process. This should be required for first-time voters.
4. Voting Process
The USA TODAY, The Miami Herald and Knight Ridder newspapers analyzed some 60,647 undervotes (ballots in which no candidate selection was sensed by the technology). Along with five other newspapers, they analyzed 111,261 overvotes (ballots in which more than one candidate was sensed by the vote-counting systems). They found that up to 18 percent (around 31,000) of the total 171,908 "defective" votes analyzed would have been able to be counted as legal votes, depending on which of four acceptance standards were used.21
Figure 5: The Winner, Using Four Standards of Judging
USA TODAY/The Miami Herald/Knight Ridder newspapers examined 60,647 punch-card ballots and optical-scan ballots in Florida that didn't register a vote for president by machine (undervotes) and 111,261 punch-card ballots and optical-scan ballots that registered more than one vote for president (overvotes). Of those 171,908 ballots, as many as 30,451 (18 percent) were credited to either George W. Bush or Al Gore. As shown in this Figure, Gore picked up 682 more overvote ballots than Bush (1,871 vs. 1,189) and 384 more optical-scan undervote ballots. The winner among punch-card undervote ballots depends on which of four counting standards is used.
Indeed, the U.S. Constitution provides the mandate for information quality in the election processes.
Most of these 171,908 voters, except the minority, who intentionally abstained from voting, were disenfranchised. Regardless of the cause, thousands of voters in Florida were disenfranchised as a result of broken processes that caused their votes to be invalidated and not counted. This count does not include all disenfranchised voters. Some were turned away at the polling locations because of errors or omissions in the voter lists; others did not make it to the polls because of work schedules or traffic. Some polling places had long lines and voters left. Others heard broadcasts of "election winners" and did not go to the polls.
Lack of standards creates chaos in the recount process. Depending on the standard applied (strict, 2-corner, Palm Beach or standard), there was the potential for four vote differences had all 67 Florida counties agreed on one of the aforementioned standards for the recount. George W. Bush wins using two of the standards, and Al Gore wins with the other two, as depicted in Figure 5. However, because each county has jurisdiction over its own "standards" for determining "voter intention" of defective votes, there is the possibility for many different potential outcome counts. With four different standards as analyzed and only 12 counties with punch-card technology (Florida had more than 24), there would be 412; in other words, 4 (different standards) times itself 12 times (number of counties); or more than 16 million different possible outcome results. This is only one of the problems with punch-card technology.
The problem is compounded by the fact that different voting processes have different inherent failure rates. This poses a problem for the "one person, one vote" principle. For jurisdictions with zero defects, this would be true; but for jurisdictions in which there was a 3 percent error rate, the one person, on average, has only 0.97 of a vote, according to Michael Grant, vice president of Tru-Vote International.
The Constitution of the United States of America warrants that "the citizens of each state shall be entitled to all privileges and immunities of citizens in the several states."22 However, the 2000 presidential election saw a six or sevenfold disparity between states with the lowest rate of residual or "defective" ballots (uncounted, undervote and overvote) and those with the highest. Maryland and Louisiana had the lowest rejected ballot rates (0.5 percent and 0.6 percent respectively), while Wyoming and Georgia had the highest with 3.6 percent and 3.5 percent rejected vote rates respectively.23 Do the citizens of Wyoming and Georgia have equal "privilege" in their right to vote and to have their votes recorded and counted accurately? Indeed, the U.S. Constitution provides the mandate for information quality in the election processes.
Technology itself is not the solution. Utah, with mostly punch-card equipment supplemented by paper ballots, produced a 1.8 percent defective vote rate, while New Mexico, with mostly electronic voting technology and optical technology for the remaining counties, produced a 2.8 percent defective vote rate (virtually equal to Florida's 2.9 percent rate), according to the Caltech/MIT study.
The real solution to election process reform is to apply quality principles to improve the processes to eliminate errors.
Figure 6: Voting Methods Used Across the U.S.
Each of the 3,140 counties sets up its own system for voting and recording votes.
IQ Lesson 11: Standards, including both process and data definition and technology standards, are required to produce consistent results. Currently there are six different basic technologies for voting systems, as exhibited in Figure 6. Defining the election process is a state's right charged to local jurisdictions. The more that processes and technologies can be standardized across the 3,140 county jurisdictions, the more consistent the results of the processes and the more easily a recount can be conducted successfully without turmoil.
QI Recommendation: States should create standards for technologies and process guidelines with the most error-free capabilities. States should have the same quality standards in every voting jurisdiction for state elections.
QI Recommendation: The Federal Election Commission should promote quality standards and guidelines for federal elections. The Federal Election Commission should provide training with such guidelines, standards and error-proofing techniques so that all jurisdictions have access to the best practices and quality improvements.
For 72 percent of the defective votes in Florida, voter intent was unable to be discerned with absolute certainty. However, statistical analysis of those votes and the companion votes makes a statistical likelihood of voter intent possible. Figure 7 lists the 20 most common overvote combinations. Anthony Salvanto, political scientist at The University at California-Irvine and specialist in analysis of voting patterns, examined voting patterns of 56,225 Florida overvotes. Findings:24
- Of the total 111,261 overvotes, Gore was marked on 84,197 while Bush was marked on 37,731. Of the 56,261 overvotes Salvanto examined:
- 83 percent of overvotes that included Gore, but not Bush, voted Democratic in the U.S. Senate race.
- 69 percent of overvotes that included Bush, but not Gore, voted Republican in the Senate race.
- 45 percent of voters who marked both Bush and Gore voted Republican in the Senate race while 42 percent voted Democratic, a nearly even split.
Figure 7: Florida Overvote Combinations
There were 2,036 possible overvote combinations. An analysis of overvote ballots showed the 20 most common combinations. The list is ranked by the number of ballots.
Some problems in punch-card ballots included misaligning the ballot on top of the voter booklet. On Miami- Dade County ballots, candidates were assigned numbered chads with No. 4 for Bush and No. 6 for Gore. Chad Nos. 5 and 7 were not assigned to any candidate. Nearly 2,000 ballots had punched chads not connected to any candidate. The No. 5 chad was punched cleanly on 676 ballots, while the No. 7 chad was punched on 991. These invalid punches can best be explained by the fact that if the punch card is laid atop the voter booklet, rather than sliding it into the voting machine, Bush lined up with the No. 5 chad and Gore with the No. 7 chad, a vote difference of 315 votes.25
5. Vote Recounting
The recount process has been required in the past because the processes of voting and vote counting themselves are flawed. Punch-card ballot recounts conducted with the same card readers frequently yield different counts because of the failure of the chads to be completely punched during voting and falling out between the first machine counting and the subsequent machine recount.
Manual recounts are required to attempt to decipher voters' intentions when the voting process fails to record a vote readable by the technology. Both punch-card and optical scanning technologies are subject to significant inadvertent undervote and overvote failures. Neither have the ability to prevent overvotes. Optical technology flaws exist in the form of failing to completely mark the selection space, marking with an "X" or check mark instead of shading the entire space or using a pen or pencil instead of the required marking device.
Electronic recording means have the capability to prevent voting for two candidates in one election and can be programmed to warn voters that they have not voted in a given election.
Manual recounts are dependent on clear definitions of standards for "interpreting" votes. The existence of four different standards for punch- card ballot recounts is a problem.
However, one problem with recounts in every voting method that is impossible to solve is to assure the vote, whether a cleanly punched chad or a completely marked optical vote, was for the candidate the voter intended.
IQ Lesson 12: A recount is basically an assessment process. It is an inspection of the ballots to assure that the official vote counting process recorded an accurate result. The best way to solve problems requiring recounts is to design quality into the process to minimize the errors that require inspection and interpretation.
Unfortunately, there is currently no process of assessing the accuracy of a ballot. Optical technology that is scanned while the voter is present allows such problems as a failure to detect a vote cast or multiple selections for a single race, but it may not allow the voter to confirm he/she marked the candidate of his/her choice.
QI Recommendation: As a long- term goal, replace voting methods that produce ambiguous voting results with voting systems that are designed with quality principles such as error-proofing techniques and audit verification capability. Even the use of optical technology with the one percent overvote rate in Florida would produce one million overvotes without other error proofing techniques applied. Additionally, many DREs (Direct Recording Electronic devices) that are designed similar to mechanical lever machines lack voter-friendly features. The Tru-Vote Validation and Verification Voting System is the best example of a voting system that implements quality principles with built-in voter quality assurance and audit verification to assure vote recording. Some of Tru-Vote's features are described in Section 7, Election Process Improvement.
In the short term, examine the nature of defective votes and define and implement vote- counting and recount standards and error-proofing improvements for error-prone voting processes until they can be replaced. See Section 7.
21. Cauchon, Dennis and Jim Drinkard, op. cit. p. 4A.
22. The Constitution of the United States of America, Article IV, Section 2.
23. Caltech/MIT Voting Technology Project, op. cit. p. 89.
24. Cauchon and Drinkard, "Florida voter errors cost Gore the election," p. 4A.
25. Cauchon, Dennis. "Recount winner depends on standard," USA TODAY 26 February 2001, p. 3A.