Go back to Voting Machine Webpage
http://www.itl.nist.gov/lab/specpubs/500-158.htm
Computer
Science
and Technology
--------------------------------------------
NBS Special
Publication 500-158
Accuracy,
Integrity, and Security in
Computerized
Vote-Tallying
Roy G. Saltman
Institute for Computer Sciences and Technology
National Bureau of Standards
Gaithersburg, MD 20899
Sponsored by:
John and Mary R. Markle Foundation
75 Rockefeller Plaza, Suite 1800
New York, NY 10019-6908
August 1988
U.S. DEPARTMENT OF COMMERCE
C. William Verity, Secretary
National Bureau of Standards
Ernest Ambler, Director
_____________________________________________________________________________________
4. SOME RECENT DIFFICULTIES IN COMPUTERIZED VOTE-TALLYING
The following are examples of difficulties that have occurred in the use of computerized vote-tallying, in the years 1980-1986. The purpose of providing these narratives is to demonstrate the types of difficulties that have been recently experienced. These reports provide support for recommendations concerning procedures that can be followed so that other election administrators may avoid similar difficulties. It is not the purpose of these reports to assess or assign responsibility or culpability for the difficulties experienced.
The examples were chosen for several reasons. Four of the situations (Carroll County, Maryland; Charleston, West Virginia; Elkhart County, Indiana; and Palm Beach County, Florida) had been identified in an article in the New York Times [6]. The situation in Dallas had been reported there in a separate article [36]. The remaining situations provide elucidation of particular types of problems and, in addition, reliable information about the situations was made available by authoritative or competent sources.
As in the 1975 vote-tallying report [1],in which examples of reported difficulties were given, the descriptions should be read with the following caution:
It is not intended that these descriptions be the definitive versions of what occurred. No absolute proof can be offered that the events occurred exactly as described. However, reported data have been supplied by named sources on the scene, and exact quotations by participants and observers are given when appropriate.
4.1 Carroll County, Maryland: November, 1984
Carroll is a county of about 100,000 population whose county seat, Westminster, is located about 30 miles northwest of the city of Baltimore. On November 8, two days after the Tuesday, November 6, 1984 general election, and in accordance with the rules of the Maryland State Administrative Board of Election Laws (SABEL), voted punch card ballots from two districts of Carroll County were taken to a neighboring county, Frederick, to be rerun on an independently-managed system. (Similarly, ballots from Frederick County were taken to Westminster to be rerun.) This rerun, in order to verify the original results, is necessary under Maryland regulations before the results may be certified.
It was clear from these reruns that one of the computers used was in error in determining the outcome of a contest between Wayne Cogswell and incumbent T. Edward Lippy, for Carroll County School Board. Manual counts of the votes on ballots from both Frederick and Carroll Counties showed that the Carroll County computer was the one that was incorrect. The initial but unofficial count, made public on the evening of the election, had incorrectly indicated that Cogswell was the winner.
An investigation, undertaken the next day (November 9) by Craig Jester, a county computer program contractor, demonstrated that a wrong utility computer program for reading the ballot cards had been used. After the correct utility program was installed, the results coincided with those obtained manually and with the Frederick County computer. The utility program, named COLBIN, had been previously written by Jester under contract to the county and had been successfully used in the May, 1984, primary election.
The purpose of the COLBIN utility program was to read the voted ballot cards in the "column binary" format used for voting, rather than in a simpler format. At the request of Carroll County data processing personnel and to reduce the price, Computer Election Systems, the vendor of the vote-tallying system, had supplied the system with an elementary utility program that could read cards only in the simpler format. With the simpler format, ballot cards would be required to have a maximum of one punch per column, not an acceptable situation for the Carroll County ballot. Carroll County contracted locally (with Pelorus, whose president was Craig Jester) for the COLBIN utility program.
On Saturday, November 10, the count was rerun (using the vote-tallying system including the COLBIN program). Members of the county Board of Elections and the County attorney were in attendance. The count indicated that Lippy was the winner. On Wednesday, November 14, eight days after the election, the Board of Elections certified the results, naming Lippy.
The cause of the error was reported in the Carroll Sun on Nov. 18, 1984 in an article by Steve Kelly [37]. A more complete explanation was provided in a letter, dated Nov. 26, 1984, from Thomas J. Van de Bussche, Administrator of the Data Processing Center of Carroll County, to Dr. Thomas Lewis, President of the Carroll County Election Board [38]. Mr. Van de Bussche's letter was included in a report submitted by Dr. Lewis on December 5, 1984 to Mrs. Marie Garber, Administrator of SABEL [39].
Mr. Van de Bussche admitted that, in testing an improved vote-tallying system provided by Computer Election Systems, he had inadvertently replaced the production version with a test version that did not include the COLBIN utility program. The logic and accuracy test of the vote-tallying system on Oct. 25, 1984, performed prior to the election in accordance with Maryland regulations, produced results consistent with the test ballots used. None of the test ballots had more than one punch in any column. Therefore, the test ballots did not reveal the error.
In the general election of November 6, 1984, the contest for the school board seat in question was listed in the same punch card columns as a home rule issue. The two contests were listed on different ballot pages of the "votomatic" ballot holder. The combination of votes for a school board candidate and a particular home rule position in the same column created a punch configuration that was not recognized as valid by the elementary utility program. As a result, some valid votes were not recorded in both the school board and home rule issue contests. Most of the votes not recorded (about 13,000) were for Lippy, because many Lippy voters chose the home rule position listed in the same card column. Votes not recorded on the home rule issue did not affect the ultimate outcome for that question. If the COLBIN utility program had been used, all votes on the contests would have been recognized as valid.
In summary, the incorrect announcement of the result of the school board contest on election night was due to mistakes by the Data Processing Center of Carroll County in using the wrong utility program and in using a perfunctory logic test that did not disclose the problem before the election. No factual evidence is available that contradicts the documentation submitted to Mrs. Garber by Dr. Lewis and Mr. Van de Bussche.
The incorrect announcement was not due to any error in the vote-tallying computer system supplied by the primary vendor, Computer Election Systems, nor any activity undertaken by its representatives. Nevertheless, on July 29, 1985, the New York Times, in referring to this particular situation, reported that "The vote counting program that has been challenged in .... Maryland was developed by Computer Election Systems of Berkeley, Calif." [6]
The error was discovered after the election but before certification because of a Maryland regulation that required recounting on an independently managed system. This specific regulation was based on a recommendation that "further confidence in the machine-counted results can be achieved if mandatory machine recounting of a percentage of the precincts for each race is carried out on a different, independently managed computing system than that used to produce the official count." [40]
On June 11, 1985, another recount of the school board race in question was carried out, using the Carroll County computer, again including the COLBIN program. This recount was undertaken at the request of the State court in which Mr. Cogswell, the defeated candidate, had filed a suit asking that the results of the election be re-examined. The recount verified the correctness of the election results certified on Nov. 14, 1984, although Mr. Van de Bussche has indicated that the recount results did not exactly match the count reported in the certification.
Mr. Van de Bussche has stated that the recount, carried out with all sides in attendance, was hurried and less than precise in that, with the permission of the court, card reader "checks" were ignored in the ballot-reading process. Usually, a card reader "check," indicative of a reading failure, would result in a decision to re-read the entire precinct of voted ballot cards. Instead, the card or cards causing the "check" remained unread and the reading process continued. The failure to re-read an entire precinct upon occurrence of a read "check" resulted in a small but random reduction of votes to both candidates, according to Mr. Van de Bussche. The differences were not significant enough to raise reasonable doubt as to the correctness of the certified results.
According to a July 11, 1985 story by Chris Guy in the Carroll County Times referring to the court-ordered recount, "...defeated candidate Wayne Cogswell had verification that use of an incorrect computer program caused a nearly 13,000-vote mistake in the unofficial totals released election night." [41]
4.2 Charleston, West Virginia: November, 1980
The following discussion was developed primarily from newspaper articles in the Charleston Daily Mail and Charleston Gazette.
Following the general election of November, 1980, three defeated candidates charged gross violations of election laws in Kanawha County, the county in which Charleston is located. Defeated U.S. Representative John Hutchinson first filed a complaint with the U.S. Department of Justice alleging that his civil rights had been violated. The other defeated candidates who joined Hutchinson in charging election law violations were former State Delegate Leonard Underwood and former Kanawha County Commissioner William Reese. Underwood, who was the first to initiate a suit, had filed for a recount after he was edged out in the election by Delegate John M. Wells. (Underwood was re-elected as a State Delegate in 1982.)
According to an article on June 2, 1981 in the Charleston Gazette [42], Darlene Kay Dotson, an employee in the office of the County Clerk, had stated in a deposition taken for Underwood's suit that the ballots from the election in question had been run through the computer on the day after the election to get "precinct-by-precinct reports." According to law, the ballots are to be secured for the official canvass, which was not done at that time. Two members of the County Commission, Robert Silverstein and Al Shepard, each possessed the only key to separate padlocks that were both needed to open the vault in which the ballots were kept, and both denied opening the vault or giving the keys to anyone else to do so.
According to the June 2, 1981 article, Carolyn Critchfield, also an employee of the office of the County Clerk, told Shepard that, to her knowledge, the ballots hadn't been out of the vault before the canvass. "Shepard said Mrs. Critchfield believed Ms. Dotson had events confused with the primary election when the ballots were run the day after the election," the article reported.
There was also concern about a test run of the computer's tabulation ability. Ms. Dotson had testified, according to the newspaper article, that neither County Clerk Margeret Miller, nor the county commissioners would sign the certification approving the test, and that she had to sign it.
Mr. Underwood had filed for a recount on December 2, 1980, and then, on December 16, he asked the circuit court to require that the election ballots be manually counted to compare them with the computer tabulation. In early February, 1981, a circuit judge denied Underwood's request, but on February 17, 1981, he filed an appeal with the state Supreme Court, asking for the manual count or a retabulation by the computer. However, following the circuit judge's denial of Underwood's request, the ballots had been destroyed by order of the County Clerk.
The state law in effect at that time stated that ballots shall be preserved for 60 days and "if there be no contest pending as to such election and their further preservation be not required by any order of a court, they shall be destroyed." According to Ray Dodson, lawyer for the County Commission, Mrs. Miller told him that she did not know that Mr. Underwood had filed an appeal in the state Supreme Court. Dodson said that there is some question as to whether Underwood's appeal constituted a true contest under the law.
On February 24, 1982, according to a February 25, 1982 article in the Charleston Daily Mail [43], County Clerk Margaret Miller was indicted by a special Kanawha County grand jury on six felony and nine misdemeanor charges. Three of the felony charges had to do with the removal of ballots from packages after the unofficial election night vote tally and before the County Commission's vote canvass that produces official totals. Two felony counts accused Mrs. Miller with allowing a vault containing the ballots to be opened between election night and the canvass. The other felony charge stated that she allowed the ballots from the general election to be destroyed while Underwood's suit was pending. In June, 1983, Mrs. Miller was found innocent of all charges. The jury had concluded that there was no "willful misconduct." [44]
However, in February, 1983, the three unsuccessful 1980 candidates filed a civil suit in Federal district court against Mrs. Miller and 15 other individuals. According to a newspaper article on February 5, 1983 [45], the suit alleged that the three were prevented from being elected because of a conspiracy on the part of the individuals named as defendants. In addition to Mrs. Miller, defendants included Mrs. Miller's husband Steven; former U.S. Representative Mick Staton, who defeated Hutchinson in 1980; Kanawha County Commissioners Henry Shores and Robert Silverstein; Kanawha County Prosecutor (and later mayor of Charleston) James Roark; several employees of the County Clerk's office including Darlene Dotson and Carolyn Critchfield; John Cavacini, Jr., a campaign worker in 1980 for Governor Jay Rockefeller; Computer Election Services, supplier of the electronic voting equipment used in 1980 in Kanawha County; several employees of that company; and Bernard H. Meadows, deputy clerk of the Boone County Commission.
Previous to the filing of the suit, former U.S. Representative John Hutchinson had been reported as charging that one of the defendants, John Cavacini, Jr. was in possession of final returns for the 1980 election two days after the Nov. 4, 1980 election and more than 30 days before the official canvass was completed on December 8, 1980. Hutchinson said, according to a June 5, 1982 article [46], that information in his possession
"...proves without any question, the machines didn't count at the canvass. I think the results were predetermined. It's conclusive in as much as the documentation shows after numerous absentee and challenged ballots were counted that the totals didn't change. The machine did not count the ballots."
Hutchinson also added that:
"There is a very great probability that the numbers were in the machine and the machine never counted anything."
In December, 1983, Hutchinson, Underwood, and Reese filed an amended lawsuit that specifically alleged that Kanawha County Clerk Margaret Miller and her staff rigged the vote-counting computers to predetermine the results, that ballots were mishandled, that some ballots were not opened until the canvass and then were counted to meet a predetermined result, that all ballots were not counted, that the vote canvass was falsified, and that the computing equipment was not properly tested prior to the election. The three plaintiffs also claimed that Boone County election officials tampered with the computer equipment before the election, that ballots were altered, and that an accurate ballot tally was not made. Irregularities in procuring the vote-counting equipment in both Kanawha and Boone Counties were also alleged [47].
In May, 1985, the suit was dismissed (charges against several defendants had been previously dropped), and the Charleston Daily Mail editorialized on May 3, 1985 as follows:
"[U.S. District Judge Charles] Haden sat through days of non-evidence on behalf of three plaintiffs -- former Congressman John Hutchinson, House of Delegates member Leonard Underwood and former county commission candidate William Reese -- all of whom lost election races in 1980 and have been mad about it now for more than four years.
"The plaintiffs were somehow allowed to bring this petty case to trial. Throughout, they failed to present a particle of proof to support their claims that 11 defendants conspired to deprive them of public office.
"People who run for election obviously like to win, but they must be prepared to lose. And barring solid evidence of fraud, they ought to accept their losses graciously without proceeding to harass either the individuals who ran against them or those who counted the votes." [48]
Appeals of the dismissal were similarly dismissed, and the U.S. Supreme Court announced on February 24, 1987 its refusal to hear the case.
Possibly influenced by Underwood's initial suit requesting a manual recount of the ballots, West Virginia amended its law on electronic voting systems in 1982. The revised subsection 3-4A-28 (4) is as follows:
During the canvass and any requested recount, at least five percent of the precincts shall be chosen at random and the ballot cards cast therein counted manually. The same random selection shall also be counted by the automatic tabulating equipment. If the variance between the random manual count and the automatic tabulating equipment count of the same random ballots is equal to or greater than one percent, then a manual recount of all ballot cards shall be required. In the course of any recount, if a candidate for an office shall so demand, or if the board of canvassers shall so elect to recount the votes cast for an office, the votes cast for that office in any precinct shall be recounted by manual count.
4.3 Dallas, Texas: April, 1985
In the election for Mayor of Dallas, held April 6, 1985, the incumbent Starke Taylor avoided a runoff by obtaining slightly more than the required 50% of the vote. There were three opponents to Taylor: Morehead, Goldblatt, and Daniel. Max Goldblatt, the leading opponent, requested a recount. A machine recount (including absentee ballots that were mixed in) was undertaken on April 11, 1985, by order of the District Court. The original results and the recount results are summarized below: [49]
|
Morehead |
Taylor |
Goldblatt |
Daniel |
Write-In |
Original |
2,318 |
38,998 |
35,081 |
621 |
10 |
Recount |
2,318 |
38,973 |
35,082 |
619 |
10 |
The machine recount showed that, overall, Taylor's votes decreased by 25, while Goldblatt's votes increased by just one, an insufficient change to cause a runoff election. Mr. Goldblatt did not raise a challenge to the accuracy of the recount at that time, and Mr. Taylor was confirmed as Mayor for the two-year term.
Taken as a whole, the recount results would appear to confirm the vote totals produced in the original count, but a precinct-by-precinct review [50] raises technical questions about the accuracy of the combined human and computer system that produced them. There were 250 precincts in this election, but only 89 precincts showed no change in votes for any of the four candidates in the contest for mayor. Taylor's votes changed in 100 of the 250 precincts, while Goldblatt's votes changed in 113 precincts.
In the recount, Taylor lost one vote in 45 precincts, two votes in each of ten precincts, three votes in each of four precincts, and five votes in each of two precincts. Taylor gained one vote in 25 precincts, two votes in each of nine precincts, three votes in each of three precincts, and five votes in each of two precincts. These changes sum to a loss of 25 votes.
Goldblatt lost one vote in 47 precincts, two votes in each of five precincts, three votes in each of two precincts, four votes in each of two precincts, and five votes in each of two precincts. Goldblatt gained one vote in 41 precincts, two votes in each of ten precincts, three votes in each of three precincts, and twelve votes in one precinct. These changes result in an overall gain of one vote.
In addition, the number of ballots counted changed in 109 of the 250 precincts, for a total loss of 27 ballots [51]. Fewer ballots were tallied in 59 precincts, while more ballots were tallied in 50 precincts. In precincts that lost ballots, one fewer was counted in 37 precincts, two fewer in each of eleven precincts, three fewer in each of seven precincts, four fewer in each of two precincts, five fewer in one precinct, and seven fewer in one precinct. In precincts that gained ballots counted, one more was counted in 33 precincts, two more in each of 13 precincts, three more in each of two precincts, and four more in each of two precincts.
For the recount, data on changes in precinct vote totals for the candidates were made available to the public. However, the number of ballots counted in the recount in each precinct was not included in the data released. The changes in ballots counted that are presented above are based on data collected by Ms. Terry Elkins, campaign manager for Max Goldblatt. Ms. Elkins was present at the recount, and copied down the number of ballots tallied for each precinct.
The causes of the changes in ballots and votes counted are not clear. However, changes occurred in such a manner as to yield both positive and negative values in approximately equal amounts, and the number of precincts in which changes occurred tended to decrease as the value of the change became more severe. This situation suggests a problem of system inaccuracy, rather than of deliberate bias. This inaccuracy could be of either human or mechanical origin, or both.
When numbers of votes change in a system using pre-scored punch cards, the problem is often ascribed to hanging chad, although failure of the card readers to read the cards accurately is a reasonable possibility. If the problem is said to be due to chad, the explanation is that a hanging chad may fall out, creating either a vote or an overvote; or, a hanging chad may be pressed back into place, usually eliminating a vote.
The cause of changes in ballots counted is more perplexing. In the system used in Dallas, ballots were hand-fed into the card readers one at a time, by both the voters and the recounters. Consequently, it is unlikely that, on several occasions, two cards were fed at one time and counted as one. However, the system used at the precincts by voters required each voter to handle his or her ballot twice. First, the voter fed the ballot into the precinct ballot computer. Assuming that the computer accepted the ballot, the voter was then supposed to retrieve the ballot from the output of the computer and drop it into a ballot box. In some cases, a voter may have put the ballot into the ballot box without first having fed the ballot through the computer, or conversely, the voter may have fed the ballot through the computer more than once. On the other hand, inaccurate card counting by the machines, due to card jams, is a possibility. Only a thorough audit of voter sign-in lists, compared with an accurate count of numbers of ballot cards delivered from the precincts to the election headquarters could provide a more confident basis for a statement of the cause of the identified ballot count discrepancies.
In September, 1986, Ms. Elkins concerns about the election were publicized. These concerns included matters other than the differences between the original count and the recount, as discussed above. Instead, a major aspect of the information provided by Ms. Elkins at that time concerned the total number of ballots cast for Mayor. She noted that the "Combined Canvass Report" produced by the Dallas County Election Department on the evening of April 6 stated on page 14 that there were 78,398 ballots cast, while the same document stated on page 27 that there were 80,208 ballots cast. Furthermore, the "Official Cumulative Report" (which serves as an overvote-undervote report), produced by the Election Administration on April 8, stated that there were 79,783 ballots cast. Ms. Elkins contended that documentation that she had gathered supported none of these numbers of total ballots cast.
Ms. Elkins also noted another apparent discrepancy in the results reported for a City Council seat. In that situation (District 7), the number of votes tallied was reported to be 10,365. This value exceeded the number of ballots reported to be cast, 9,679. In addition, in one precinct, the initial number of ballots cast, 263, was replaced later with another value, 515. These technical concerns of Ms. Elkins were supplemented by concerns that the results presented on the computer printouts were created independently of the actual totals of the voted ballots through a deliberate attempt to subvert the outcome. [52]
As a result of Ms. Elkins' complaints, it was reported that Attorney General Jim Mattox and Secretary of State Myra McDaniel began investigations into voting discrepancies. According to the Dallas Morning News of Sept. 23, 1986, "the probe centers on allegations that computerized voting equipment and computer programs used to tabulate state and local elections may have been tampered with to bring about `preprogrammed' results." [53]
In that same newspaper article, Ms. Elkins was quoted as saying that "the allegation is that the computer used to count the votes was given new instructions after it calculated that Max Goldblatt was leading Starke Taylor by 400 votes." Ms. Elkins has noted that the Dallas County computer had encountered difficulties shortly after 8 p.m. on election night, and that the candidate who was leading at 8 p.m., prior to the computer difficulties, was not leading when the computer reported again.
Ms. Conny McCormack, Dallas County Elections Administrator, admitted that the documentation for the April 6, 1985 election could appear contradictory. Her explanation was that the difficulty concerned the treatment of "split precincts," that is, those precincts bisected by the Dallas city boundary. There were 11 such split precincts. The value of 78,398 for ballots cast was produced by assuming zero ballots cast from these split precincts. The value of 80,208 for ballots cast was produced by adding the total ballots from the split precincts, including ballots cast outside of the city. The final value of 79,783 for ballots cast included only those ballots cast within the city of Dallas. Ms. McCormack contended that the recount generally confirmed the correctness of the originally reported outcome. [54]
Ms. McCormack's explanation of the problem of reporting split precincts was supported by the vendor of the vote-tallying system. In a memorandum on the subject, a vendor representative stated that there was a difference between the type of reports requested by Dallas County Elections Department for the PBCs (precinct ballot counters) and for the central computer. The central computer was used to accumulate totals reported by the PBCs. The coding for the central computer included provision for split precinct specification of ballots cast, but the coding for the PBCs did not allow for this. According to the vendor, "this extra statistical option was not requested by Dallas for that election." [55]
The vendor representative further stated, in the same memo, that:
"Since the [data] packs did not have this [split precinct] information to transmit, all of the precincts which were transmitted had a "zero ballots cast" for the districts. Again, total ballots cast and all candidate votes were present and correct.
"On the cumulative report, then, the votes received by the district candidates were much greater than the ballots cast figures for those districts. Although the explanation for this apparent anomaly is now clear, it clearly was a suspicious looking situation."
With regard to the change in ballot totals reported for one precinct, Ms. McCormack stated that this was due to a failure of a PBC data pack. This occurs about 2% to 4% of the time, she stated. The procedure when this happens is as follows:
"When [a failure of a PBC data pack] occurs, the actual ballot cards are counted at the central counting station. Such discrepancies from PBC tape to actual ballots cast is determined by examination of the specific Ballot and Seal Certificates. When there is a discrepancy between number of persons signing the signature roster at the precinct and the number of ballots cast according to the PBC tape, then the ballots from that precinct are counted centrally." [54]
A summary of the ballots cast in this election, as officially reported, is shown below, by district. It can be seen from the table that the largest component of ballots from split precincts occurs in District 7, where the increase from Column A to Column B is 1,661. However, 422 of these ballots were from outside the city, so that the total number of ballots cast in District 7 inside the city is 10,918. The latter number is shown in Column C, and it is appropriately higher than the total votes for candidates in District 7, which was 10,365. [56]
District |
(A)
Excluding
Split
Precincts |
(B)
Including All
Ballots From
Split Precincts |
(C)
Only Ballots
Cast in City |
1 |
9371 |
9395 |
9392 |
2 |
6314 |
6314 |
6314 |
3 |
16351 |
16382 |
16382 |
4 |
14529 |
14529 |
14529 |
5 |
13088 |
13088 |
13088 |
6 |
4118 |
4118 |
4118 |
7 |
9679 |
11340 |
10918 |
8 |
4948 |
5042 |
5042 |
Total |
78398 |
80208 |
79783 |
Certain technical problems raised by Ms. Elkins seem to have been given a credible explanation, on the basis of information made publicly available in late 1986. However, as Warner Croft stated in his testimony to the Texas House of Representatives Committee on Elections:
"There is an audit trail but there are holes in it. The audit trail should consist of everything from the ballots themselves to the console log being printed by the computer on election night. The present laws don't identify what the minimum requirements are, so that, with the absence of a minimum definition, it just does not exist. You go after these things, and the laws don't require that they be kept on file now, so they have been destroyed months ago. So you really couldn't tell if there was fact to these allegations are not. That has been one of our problems. Records aren't available; there are no auditable results." [22]
Ms. Elkins' charges that the results were "preprogrammed" independently of the actual votes cast were not put to rest in 1986. In March, 1987, the Texas Attorney-General's office asked the District Attorney of Dallas County to assist in reviewing the election complaints. The review concerned "the reliability of Dallas County's computerized election system and whether the equipment is vulnerable to fraud through subtle changes in computer programs." [57]
On October 14, 1987, the office of the District Attorney of Dallas County replied to the Texas Attorney General's Office with a letter [58] including the following:
"We have carefully considered each of the thirteen (13) "discrepancies" discussed in the report [submitted by your office], and ... each of the "discrepancies" has been explained to our satisfaction; and although we verified that a few coding errors were in fact made, we have concluded that they were the result of unintentional "human error." We find no evidence whatsoever to indicate any deliberate fraud in the 1985 election, nor do we find any credible evidence to indicate an attempt to manipulate the election or its outcome by anyone, be it candidate, election official, or vendor."
Some knowledgeable persons have found this statement puzzling, in view of Warner Croft's testimony that necessary documents constituting the audit trail had been previously destroyed.
In the April, 1987, mayoralty election, split precincts were eliminated by the Dallas County Election Administration. Precincts that had been bisected by the Dallas city boundary or City Council district boundary were divided into two or more separate precincts. Letters A and B, added to the precinct identifier, were used to distinguish the formerly combined precincts, as in 1193A and 1193B.
4.4 Elkhart County, Indiana: November, 1982 And November, 1986
4.4.1 November, 1982 General Election
Elkhart is a county of about 140,000 population on the northern border of Indiana, located about 90 miles east of Chicago. In lawsuits filed in state and Federal courts, several losing candidates alleged that election fraud occurred in the administration of the general election of November 2, 1982.
In this election, the computer facilities of a bank located in the county seat of Goshen were used for ballot-counting purposes. The office of the county clerk had limited computer expertise for carrying out its responsibilities for management of the ballot- counting process. Technical operations related to computer processing of the ballots were undertaken by bank employees and by an employee of the vendor of the vote-counting software. The bank's computer was capable of multiprogramming, and bank operations continued during ballot counting.
It is important to note at this point that punch card ballots in a general election in Indiana have "straight party" punch locations. A voter who votes one (and only one) of these punch locations automatically votes for all candidates of that party unless the voter specifically votes for a candidate of another party in a particular contest. A vote in a particular contest always overrides the straight party vote. This feature of Indiana ballots figures prominently in this situation.
Procedural and computer-related errors in the election affected at least three contests. The errors concerned votes for the Town Board of Wakarusa (a town within Elkhart County), votes for Districts 2 and 3 of the County Council, and votes for a State Representative contest.
The boundaries of the town of Wakarusa are wholly within Olive Township of Elkhart County. Some of the voters in this township live within Wakarusa, and some live outside the town. Only those residing within Wakarusa are entitled to vote for the Town Board. However, no arrangements were made to distinguish town residents from non-residents for purposes of casting ballots. All voters in the township were given the same ballot. The result of not separating ballots of residents and non-residents was that Town Board candidates received more votes than they were entitled to obtain. The extra votes were from voters residing outside the town, and these consisted primarily of straight party votes that were assigned by the vote-tallying computer program to Town Board candidates as well as to other candidates on the ballot. In addition, there may have been votes for individual Town Board candidates by non-resident voters who mistakenly voted for them.
The Wakarusa problem was not discovered until several days after the results were certified, when an election worker realized that the total number of votes for Town Board candidates was much higher than that number of votes which could have been cast by voters entitled to vote for Town Board. The problem was "solved" by an informal agreement approved by the Election Board under which all straight party votes were eliminated for Town Board candidates. This "solution" disenfranchised those voters within the town who lawfully cast straight party votes. The change overturned one outcome, but the loser was dissuaded from suing. According to Elkhart attorney David T. Stutsman, the agreement to change the outcome was in violation of Indiana law, and a legally correct solution would have been to hold a new Town Board election.
In the counting of votes for County Council, votes for candidates in Districts 2 and 3 were interchanged. This situation became apparent to one of the Election Board members when Higgins, the unopposed candidate in District 3, was initially shown to have a significant number of opposing votes. In addition, Barnes, the incumbent in District 2, was getting almost no votes, while Wiedenhoeft, Barnes' novice opponent, was winning a large majority. This unexpected condition was pointed out, and the interchange was discovered. The error was corrected by changes in control cards that were carried out by an employee of the software vendor. This latter individual had been sent from another state by the software vendor for the sole purpose of assisting in ballot-counting operations. He appeared for the first time on the afternoon of the election, and returned to his home state immediately following the completion of the computerized vote-tallying.
In the contest for State Representative, it become apparent late in the evening of November 2 that an incorrect punch position was being used to tally votes for one of the candidates, Philip T. Warner, in his race against Mike Puro. The software vendor's employee again changed some control cards, and a retabulation was undertaken. The retabulation resulted in a significant change in vote totals: an increase of 3,710 for Warner and a decrease of 53 votes for Puro. The reason that there was a decrease in Puro's totals was because of individual votes for Warner by voters who had also indicated a straight party punch for Puro's party. Before Warner's votes were properly tabulated, the straight party punches had added to Puro's total, but the specific votes for Warner on the same ballots negated these straight party votes.
It seems clear, from the Wakarusa situation, that the full implications of computer processing of ballots were not completely appreciated by the persons responsible for running the election: the Election Board and its employees. In addition, it seems clear, also, from the tabulating errors in the County Council and State Representative contests, that insufficient attention was paid by those responsible for running the election to examination of the "edit list" that identifies, for the computer, the assignment of punch locations to candidates. Adequate testing of the computer system prior to use appeared to be lacking, possibly to an extent inconsistent with Indiana law.
The alleged misfeasances have been described in lawsuit briefs filed by the losing candidates, and documentation on the charges has been obtained from David T. Stutsman, their attorney. A major source of problems in the election was the apparent failure to properly test the equipment prior to use. The losing candidates charged that no test of the automatic tabulating equipment was undertaken five days prior to the election, as required by the Indiana statute then in effect, and that only a superficial test was done at about 4 p.m. on election day.
According to Mr. Stutsman, this test on election day consisted of a count of just 13 test ballots each for only two precincts of the 63 precincts involved in the election. As a result of a lawsuit concerning this election, an alleged configuration of these 13 cards and the results that they generated on the computer have now come to light. According to this information, the test ballots should have generated three lawful votes for the Democratic senate candidate, but instead they generated four such votes. One test ballot that should not have produced a vote contained a straight Democratic punch, and individual punches for both Democratic and Republican senate candidates. A second test ballot that should not have produced a vote contained straight party punches for both the Democratic and Republican parties, that is, an uncountable straight party overvote. One of these test ballots probably supplied the incorrect fourth test vote. If this information about the test ballots and results is correct, the ballot counting process on November 2, 1982 should not have proceeded until the logical error was corrected. As stated in the Indiana statute I.C. 3-2-4-4(f) in force at the time:
"If any error is detected [in the test], the cause therefor shall be ascertained and corrected and an errorless count shall be made before the automatic tabulating equipment is approved."
Other errors in administration that have been alleged were that the changes in the control cards that were made were not accompanied by system tests, that the computer systems's clock was disabled (thereby preventing times of console actions from being reported), and that no precinct numbers were printed on the ballot cards.
Computer consultants hired by the losing candidates submitted statements including the following:
"The Edit List was not correct when the program was initiated.... Thorough tests would have caught these irregularities.
"Given the nature of the undocumented program correction ... it is impossible to know exactly how the program tallied the votes. The control log shows no test of accuracy after the [vendor's] representative modified the parameter cards.
"The lack of a clear audit trail and the lack of error reports, the convoluted code reports plagued with errors, and the unquestioned trust in an untested electronic process seems to be a major problem with this system comprised of hardware, software, and users.
"In the opinion of [the computer consultants,] it would be possible to modify program logic with the use of inserted control cards. It would be possible to change accumulated vote totals by reading in control cards at the appropriate time during the program execution. This would be unknown to the election officials or anyone but an experienced operator. This program uses alter verbs which allows program logic changes with the use of control cards inserted at execution time.
"A knowledgeable operator can change the program logic in execution with insertion of control cards ... many possibilities for change in program logic and vote tally exists." [59]
The losing candidates' case, brought before the U.S. District Court for the Northern District of Indiana, named the local board officials as defendants. It was alleged in the pleadings in that case that the computer system was not tested, that there was no error-free test of the system before the official count, that there was no actual count of the ballots and that the alleged count and certification of the vote count was fraudulent. The pleadings and briefs further stated that the control cards for the operation of the program were altered by the vendor representative during the counting, and that the acts by the election officials were willful, wanton, reckless and oppressive.
However, the court entered a summary judgment on Feb. 21, 1985 against the losing candidates because, in the court's opinion, there were no allegations of any "willful conduct which undermines the organic processes by which candidates are elected" (language of an important precedent, Hennings v. Grafton) [60]. The judgment of the lower court was affirmed by the U.S. Court of Appeals for the Seventh Circuit, which stated that "the appellants (i.e., the losing candidates) confuse fraud with what is at most willful neglect." [61] Mr. Stutsman notes that the plaintiffs were not allowed to present their evidence in the case. The lawsuit was dismissed as a matter of law, without a trial on the merits. Therefore, a jury was not allowed to hear the case.
A second Federal lawsuit, filed in the Southern District of Indiana, continues to be pending, with the vendor named by the New York Times as the principal defendant. This suit charges negligence, breach of contract, and strict liability, alleging that the computer system used in the election was sold in a defective condition and caused loss and damage to the losing candidates. In addition, it is alleged that the vote counting was a fraud and that the certification of the vote totals was false and fraudulent.
4.4.2 November, 1986 General Election
Following the 1986 general election, a State-mandated recount was undertaken that included ballots from Elkhart County. In this recount, directed by Dean David Link of the Notre Dame Law School, it was discovered that the computer program used to count ballots in Elkhart County was not counting correctly according to Indiana law. The problem occurred in the treatment of ballots that were overvoted in the "straight party" positions. According to Indiana law, these ballots should be voided for all partisan contests. Instead, the ballots had been counted in the contest being recounted. These ballots were eliminated in the recount.
4.5 Gwinnett County, Georgia: November, 1986
A recount undertaken in a State Senate race showed the loser in the first count winning the contest by 77 votes. The original tally had given challenger Steve Pate a winning margin of eight votes. However, the recount gave incumbent Donn Peevy 13,682 to Pate's 13,605. According to an article in the Atlanta Constitution, November 13, 1986, "the recount was the result of a computer hardware error ... affecting hundreds of uncounted votes ..." [62] The "computer hardware error" in question was believed to be a problem with the card readers that read the pre-scored punch card ballots.
Allegations raised in this election included charges of "hidden ballots." It appears that a box of ballots and documents received from one precinct was initially believed to consist of the complete set of voted ballots on top of other documents in the box. When the documents were removed from the box after the ballots on top had been counted, more ballots were found under the documents. In addition, it was "charged that ballots from two precincts did not arrive at the elections office until early on the morning of Nov. 5 [the day after the election] in the car of the elections board member," according to the article in the Constitution.
The original count was undertaken on a system consisting of two personal computers, each with a card reader, networked together to drive one printer. For the recount, it was agreed to separate the two computers and to count the ballots twice, once on each computer. Unfortunately, the recounts obtained from the two computers were slightly different, although not of a difference sufficiently large to overturn any contest.
The Georgia Tech Research Institute (GTRI) was asked by the office of the Secretary of State of Georgia to render any possible assistance. GTRI decided to review the situation and to determine the source the discrepancies between the two computer outputs. The investigatory team, headed by Dr. Britain Williams, hoped to separate the causes of the errors into at least two classes: those caused by handling of the ballots by voters and voting officials, and those caused by the hardware. According to Williams, "everything that could have been done to insure the accuracy of the recount was, in fact, done and the discrepancies observed are inherent in this type of system. A thorough analysis of these results will be conducted in an attempt to estimate the inherent error in using pre-scored ballots." [63]
A report has now been produced by Williams and an associate. [106] The report compares differences in the results in three counts: the general election, and the recounts on the two individual machines. The report states that:
"Errors were either tabs which were not yet dislocated from their pre-punched positions in the ballot, or stray tabs which filled other previously punched-out positions (+ and - errors)."
According to the report, errors were caused by such factors as handling procedures, the ballot puncher (which was of the "votomatic" type), the vote counter, the punched card's density, vote position on the ballot card, human error, and pure chance. The report concludes that:
"...there should be a system set up to make the voters and especially the volunteer workers aware of the effects procedural care has on the accuracy of tabulations (since this was where the main problem was discovered). Details like not putting large (or any) rubber-bands around the ballots would also be advisable (especially since custom ballot carriers are provided)."
4.6 Illinois - Statewide Testing Program
The Illinois State Board of Elections, Division of Voting Systems, under the direction of Michael L. Harty, (now Director of Elections, Maricopa County, Arizona) has undertaken tests of vote-counting computer systems. Between 1983 and 1987, the division conducted 48 tests of the automatic tabulating equipment and computer programs in 41 election jurisdictions. The tests have involved anywhere from 1,000 to 65,000 test ballots. The division found apparent computer program tabulation errors in 11 of the election jurisdictions tested. The division reports that most of these errors would not have been discovered without adequate testing. As a result of its testing experiences, the Division of Voting Systems has concluded in its Summary of Findings and Observations on State Board of Elections Computer Testing Program, revised February, 1987: [64]
"The testing of computer vote tabulation systems needs to be improved substantially. At a minimum, voting systems tests must be large; must test all voting positions; must test overvotes and undervotes; column binary punches; straight and split party votes; nonvoting position punches; and must test for every candidate in every ballot configuration in every precinct. Only by extensive testing of a computer vote tabulation system can we be reasonably assured that tabulation of the ballots will be entirely accurate".
The following descriptions of programming, program initialization, and hardware problems in local jurisdictions in Illinois are taken from the reference given above.
4.6.1 Programming and/or Program Initialization Errors
Whiteside County - 1986 General Primary Election: The system tabulated votes on ballots that contained invalid security codes (ballot style identifiers).
Morgan County - 1985 Consolidated Election: No straight party votes registered for the candidates of a party. However, this did not affect the individual candidate totals.
Peoria County - 1985 Consolidated Election: The computer program misassigned straight party punches for a candidate for township supervisor. The candidate received a tally from a straight party punch for the opposite party but failed to receive a tally from the straight party punch of his own party.
Sangamon County - 1985 Consolidated Election: The computer program would not accept ballots with proper ballot style identifiers. This error was not discovered by the test previously run by the local jurisdiction. In addition, ballots in precincts with more than one ballot style did not contain different style identifiers. Thus, it would not have been possible to separate the voted ballots of the different styles.
Logan County - 1985 Consolidated Primary Election: Tabulation errors occurred when precincts were split by ward boundaries. When the same punch position was assigned to different candidates in different wards in the same precinct, votes for one of the candidates were not tallied by the computer program.
Effingham County - 1984 General Election: A county-level office was not being tabulated in five precincts, though votes were assigned to the office.
Jackson County - 1984 General Election: A translation error between precinct returns and the summary report caused the summary report to fail to properly reflect the precinct sum totals for certain candidates.
LaSalle County - 1984 General Election: The straight party vote was not being tabulated for all candidates in a party. In addition, when an overvote occurred in the straight party column and also in an individual candidate's punches on the same ballot, the candidates involved actually lost a vote, i.e., had their vote totals reduced by a vote (instead of simply being denied a vote).
Grundy County - 1984 General Primary Election: Forty-seven percent of the precincts had one or more of the following types of errors: (1) the assignment of the wrong county board districts in the precincts, (2) the deletion of candidates in precincts, (3) the incorrect assignment of candidates to precincts, (4) assignment of only 2 vote for each vote cast, and (5) incorrect totals of precinct votes on the summary report for several candidates.
Rock Island County - 1984 General Primary Election: Two votes for each vote cast were being tabulated for a candidate. In addition, the "no" votes on a proposition were not being counted. Further, the summary report totals did not properly reflect precinct sum totals for several candidates.
4.6.2 Hardware and Punch Card Difficulties
City of Chicago - 1987 Consolidated General Election: The system test indicated an approximate 3% failure rate of program chips. The chips were improperly programmed or "burned." The malfunction would have been identified during the public test.
Boone County - 1987 Consolidated Primary Election: Due to substantial ballot quality defects, a system test could not be executed. New test ballots were ordered.
Pulaski County - 1986 General Primary Election: The principal disk that contained the vote-tallying program failed to operate for the system test. The duplicate (backup) disk was employed. The principal disk operated correctly for the public test. No reason for the problem was discovered.
Jackson County - 1984 General Election: Column binary punching appeared to cause severe tabulation delay. In addition, the card reader stopped occasionally during the tabulation of a precinct. When this condition occurred, the results already obtained had to be erased and all the ballots for the precinct had to be retabulated. The cause of this difficulty could not be immediately ascertained.
Will County - 1984 General Election: During the system test, the card reader was jammed twice by ballots. The ballots involved were almost completely destroyed in the process.
4.7 Maricopa County, Arizona: September, 1986
A clerical error that would have interchanged votes for the two major parties in this primary election was caught during testing. Pre-punches (ballot style identifiers) were incorrectly specified, and if the errors had not been caught, votes in the primary contests in each party would have been assigned to the other party during tallying. Poor communication between the county data processing department and the election administration contributed to the problem. A well-designed testing process caught the error, so that ballot counting during the actual election was not affected. [65]
4.8 Moline, Illinois: 1985 Consolidated Municipal and Township Election
The following report is primarily based on an article in Illinois Issues, November, 1985, that was republished in a newsletter of the National Center for Policy Alternatives. [66]
In this election on April 2, 1985, the failure of a card reader to read correctly caused a losing aldermanic candidate for Moline City Council to be put into office. The error was not rectified until about three months later.
The failure of the card reader was not initially apparent. It was only discovered after defeated Moline 3rd Ward aldermanic candidate Earl Wendt went to court to demand a recount. Wendt had lost by two votes, and the recount was ordered by Rock Island County Circuit Court. In the recount on May 14, Wendt picked up two votes, but his opponent Allen McCauley picked up 92 votes. The large change in McCauley's total demonstrated that a serious error in vote-tallying had occurred. The data showed that the error was in the failure of the system to tabulate many straight party votes for McCauley. Further investigation appeared to demonstrate that the problem was a slipping timing belt in one card reader.
As a result of the Wendt-McCauley recount, the party of losing 4th Ward aldermanic candidate Charles Reynolds also demanded a recount in a petition to the Moline City Council. Reynolds was of the same party as McCauley. However, the City Council denied this petition and two similar petitions from other candidates on the basis that the 30 days following the election allowed for such petitions had expired.
Following these denials, Dennis Faust, assistant State's Attorney submitted a petition of quo warranto to Chief Judge David DeDoncker of the 14th Judicial Circuit. The quo warranto petition, whose origin in Anglo-American law goes back to Edward I in 1275, is used to remove "any person [who] shall usurp, intrude into, or unlawfully hold or execute any office...", according to the Illinois Revised Statutes. Faust's theory was that if members of the Moline City Council hadn't received a majority of the votes, then they held office illegally.
On July 8, Judge DeDoncker agreed to a recount for the office of city clerk and for three aldermanic offices including that for the 4th Ward, despite the fact that the 30-day period had expired. He stated that there was no knowledge that a machine was not working until 40 or 50 days after the election. "We had legal American voters who voted and did not get their votes counted", DeDoncker said. The judge appointed representatives of the vendor of the vote-counting system to serve as officers of the court and ordered them to conduct a recount. A judge of a higher court refused to stay the recount, and it proceeded.
As a result of this recount, alderman Roy Lear picked up two votes, but his opponent Charles Reynolds added 135, making the final count 557 for Reynolds to 473 for Lear. Through a court order, Lear was removed from the city council. Reynolds was seated in his place. Lear did not appeal.
4.9 Oklahoma County, Oklahoma: November, 1986
In the general election of November, 1986, difficulties perceived by an independent group of local observers involved, among other items, the operability of the precinct-located, mark-sense computers, and the anomalous numbers of counted ballots that were reported. These observers, David Clampitt, Carolyn Burkes, and Sue Milton, submitted the data on which the discussion below is based.
The county signed a contract to purchase the mark-sense vote-counting equipment in the summer of 1984. The State of Oklahoma has no system of prior State approval of voting devices before such devices may be purchased by a county. However, no county may purchase a type of voting device before the State Board of Elections has implemented rules of operation for that type of device. Four counties in the State utilize mark-sense type voting devices. Oklahoma County is the only county of these four utilizing the particular model in question.
Engineering tests on the particular devices purchased by Oklahoma County were not carried out by the county until November and December, 1984, when the equipment was being delivered. A feature of the equipment noted in the tests was that a certain number of ballots were not being processed (were being treated as unreadable) by the machines. Approximately 1.5% of the ballots were so treated by each of the two different units being tested. According to the report, "non-processed [ballot] cards were not visibly different from those accepted and tabulated, nor were they always the same cards on successive runs." [67]
The vote-counting equipment was used in a special election at about the time of the test, and the report notes that:
"The frequency of non-processed ballots [5.22%] was observed in several precincts to be significantly higher than that observed during testing. This is attributed to improper insertion of the card by the voter, namely, failure to insert the card vertically and release it cleanly....We attribute this rate primarily to voter unfamiliarity with the system...."
The report recommended that:
"Although the [vendor's equipment] is judged to comply with all applicable statutes and regulations,....as [one of four] desirable product improvements, the vendor should be encouraged to reduce the frequency of random non-processing of acceptable ballots."
During the November 4, 1986 general election, the number of non-processed ballots was over 2% in a significant number of precincts. According to State rules, the county Board of Elections "has the authority" [68] (but is not required) to recount precincts with over 2% non-processed ballots. The county board has used its discretion in selecting particular precincts for reprocessing. Reprocessing, if done at all, is done on the county's central computer. Not all precincts with over 2% non-processed ballots were reprocessed in the November, 1986 general election.
Ballots that are not reprocessed and read by the central computer would never be counted at all, except under a court-ordered manual recount. There are no State or county rules which would permit non-processed ballots to be counted manually in order to have the manual results added to those results already obtained by machine.
Results of the 1986 general election in Oklahoma County show, in many cases, a lack of reconciliation of the number of voters signed in at each precinct with the number of ballots cast. Table 1 below demonstrates that fact.
Ballots Cast
Pct. |
A |
B |
C |
NP |
Total |
Voters Times 3 |
%
NP |
R |
1 |
1036 |
1047 |
1063 |
104 |
3250 |
3261 |
3.2 |
|
6 |
513 |
516 |
521 |
20 |
1570 |
1557 |
1.3 |
R |
10 |
389 |
396 |
400 |
48 |
1233 |
1221 |
3.9 |
R |
34 |
897 |
942 |
976 |
242 |
3057 |
3003 |
7.9 |
R |
39 |
1014 |
1094 |
1092 |
168 |
3368 |
3372 |
5.0 |
|
59 |
370 |
374 |
381 |
47 |
1172 |
1173 |
4.0 |
|
61 |
37 |
38 |
38 |
4 |
117 |
117 |
3.4 |
|
72 |
65 |
71 |
71 |
6 |
213 |
213 |
2.8 |
|
82 |
708 |
722 |
775 |
350 |
2555 |
2481 |
13.7 |
R |
98 |
380 |
398 |
406 |
46 |
1230 |
1284 |
3.7 |
|
106 |
546 |
514 |
590 |
201 |
1851 |
1857 |
10.9 |
|
117 |
22 |
26 |
24 |
7 |
79 |
78 |
8.9 |
R |
122 |
870 |
1002 |
987 |
350 |
3209 |
3126 |
10.9 |
R |
126 |
250 |
257 |
261 |
36 |
804 |
807 |
4.5 |
|
131 |
10 |
11 |
11 |
0 |
32 |
33 |
0.0 |
R |
135 |
156 |
284 |
340 |
350 |
1130 |
1047 |
31.0 |
R |
136 |
104 |
107 |
109 |
7 |
327 |
327 |
2.1 |
R |
154 |
164 |
186 |
187 |
46 |
583 |
570 |
7.9 |
R |
159 |
663 |
861 |
905 |
489 |
2918 |
2820 |
16.8 |
R |
161 |
904 |
948 |
950 |
110 |
2912 |
2889 |
3.8 |
R |
202 |
16 |
16 |
17 |
4 |
53 |
51 |
7.5 |
|
208 |
116 |
121 |
123 |
14 |
374 |
375 |
3.7 |
|
222 |
539 |
578 |
585 |
197 |
1899 |
1857 |
10.4 |
R |
256 |
249 |
258 |
266 |
29 |
802 |
807 |
3.6 |
|
268 |
682 |
697 |
696 |
51 |
2126 |
2154 |
2.4 |
|
270 |
953 |
1224 |
1319 |
703 |
4199 |
4062 |
16.7 |
R |
In this election, each voter was given three ballot cards, designated as A, B, and C. The number of processed A, B, and C ballot cards and the number of NP (non-processed) ballot cards are totalled in Table 1 for selected precincts. The total of the ballots tabulated (Total column) may be compared with three times the number of voters signed in at each precinct (Voters Times 3 column), as shown in the table. Values in these two columns in the table for the same precinct differ in most cases. According to a January 25, 1987 article in The Sunday Oklahoman entitled "Security of Elections Described," [69] "an examination of the ballot accounting forms used in last November's election in Oklahoma County revealed that a number of them are neither filled out nor signed." In addition, said the article, precinct voting materials are not logged in on a time sheet when received at the central location, except for the first and last precincts.
The "%NP" column in Table 1 indicates the percentage of non-processed ballots, over all ballots counted. The "R" column indicates whether, as a result of the non-processed ballots, the precinct was recounted on the county's central computer. Several precincts with significant percentages of non-processed ballots were not recounted, at the discretion of the County Election Board.
In most of the precincts identified in Table 1 that were recounted, the total number of ballots counted exceeds the three times the number of voters. From this situation, it may be inferred that the recounting of non-processed ballots (sometimes up to four times in an effort to get them to be read) was not properly accounted, and repetitive attempts at reading added incorrectly to the total number of ballots. The failure to match totals of ballots fed into the machines against voters signed in calls into question the reported individual candidate totals, as no proper cross-foot total can be obtained.
The failure of proper accounting naturally raises the issue of a manual recount. Although there was no request for a manual recount in this election, there have been such requests in other elections in Oklahoma County. It has been the position of the Oklahoma County Election Board that State law, until recently, did not allow manual recounting, since the Board's interpretation of State law is that the same system that was used for the original count must be used for recounting. However, this interpretation has been continually challenged in the courts and has been overturned. There have been manual recounts in Oklahoma County, over the protestations of the Election Board. In the most recent session of the State legislature, a law was passed allowing a recount petitioner to select either a manual recount or a machine recount, in those counties that use vote-tallying devices.
In addition, during the November 4, 1986, general election, over one-third of the 273 voting machines used in the election failed in some manner, some more than once. The most prevalent type of failure was that the machine simply stopped accepting ballots. [70]
Another anomaly of the reported results of this election in Oklahoma County is that the lower-level offices and obscure State questions on the ballot appeared to be as popular with voters as the office of governor. That is, the traditional "fall off" of voter interest in the lower-level offices and State questions did not occur, at least according to the reported results.
This condition is shown in Table 2 below. The votes for the offices of governor, State representative, and on State question #589 are given. The office of governor appeared on ballot card A with other State-wide and Federal offices. The State question appeared on ballot card B, and the office of State representative appeared on ballot card C.
Table
2
Precinct |
Votes for
Governor |
Votes for
St. Rep. |
Votes for
Question #589 |
1 |
1025 |
1039 |
982 |
6 |
509 |
515 |
484 |
10 |
379 |
398 |
359 |
34 |
881 |
941 |
914 |
39 |
989 |
1032 |
1038 |
59 |
362 |
367 |
343 |
61 |
37 |
38 |
37 |
72 |
64 |
69 |
69 |
82 |
696 |
Not reported |
665 |
98 |
376 |
389 |
351 |
106 |
540 |
581 |
474 |
117 |
21 |
23 |
25 |
122 |
855 |
962 |
918 |
126 |
248 |
255 |
228 |
131 |
10 |
11 |
10 |
135 |
156 |
323 |
258 |
136 |
103 |
106 |
104 |
154 |
162 |
180 |
168 |
159 |
652 |
878 |
808 |
161 |
885 |
927 |
899 |
202 |
16 |
17 |
16 |
208 |
115 |
117 |
117 |
222 |
529 |
557 |
536 |
256 |
246 |
260 |
243 |
268 |
664 |
683 |
631 |
270 |
926 |
1218 |
1091 |
In every case in which a vote was reported in the identified precincts, the office of State representative received more votes than the office of governor. In some cases, State question #589, an unimportant issue that would have been expected to receive little public interest, received more votes than the office of governor. Most independent election observers would have difficulty accepting this situation unless a very special condition of strong public concern could be demonstrated.
A commentary on this situation with regard to precinct #270, the last precinct listed in Table 2, was reported by The Sunday Oklahoman on January 25, 1987 in an article entitled "Uncertain Vote Count Puzzling to Analysts." [71] The article pointed out that in demographically comparable Precinct #459 in Tulsa County, the unusual reversal of drop-off did not occur. The race for governor in the comparable Tulsa County precinct received more votes than other races, as would be expected.
4.10 Palm Beach County, Florida: November, 1984
Following the November, 1984, general election, David Anderson, defeated candidate for Property Appraiser of Palm Beach County, sued to contest the election of his opponent Rebecca Walker. [72] Anderson asked that the Court order a hand recount of the ballots, or a hand recount of at least several precincts in that election. The issues on which Anderson sued included handling of the ballots, precinct procedures for signing in voters, ballot secrecy, counting of punch card ballots, and possible manipulation of the computer program. Anderson's initial complaint was dismissed in Palm Beach County Circuit Court on March 1, 1985 [73], but he filed an amended complaint. With regard to the computer-related points at issue, Anderson charged in his amended suit that:
"There were irregular counts from the computer on each total, each time it was run. The tabs caused by perforations punched in the voting process in the computer cards did not all tear loose, thereby seriously affecting the vote results. It is apparent that because of the type of equipment and method used, that it is impossible to accurately count any election." [74]
Clearly, there were problems of hanging chad ("tabs..[that] did not all tear loose") in this election. In a document filed in connection with this suit (Defendant Walker's First Request for Admission to Plaintiff) [75], the results of a computer recount of another contest in the same election were given. In the other contest, Owens versus Perry, Owens' original total of 137,994 increased by 279 votes in the recount, while Perry's original total of 137,817 increased by 214 votes. According to Ms. Jackie Winchester, Supervisor of Elections in Palm Beach County, these changes were due to chad that were not fully removed in the first count (thereby resulting in no vote) but which fell out before the recount (thereby adding to the original totals). However, no report of undervotes or overvotes is required by the State of Florida or by Palm Beach County, and therefore it is not possible to confidently deduce from the data that this reasonable supposition is correct.
In his amended complaint, Anderson also charged that:
"The election was run on machines that permit a means of changing the result on the ballots contrary to the votes cast by electors through an alter system in the commands in the computer program, or through the use of a single or several ballot cards specially marked with an alter code on them, which the Plaintiff has reason to believe was activated during the counting of the ballots." [76]
No documentary evidence in support of this claim was filed by Anderson, and in fact, in his initial suit that was dismissed, Anderson had alleged that the election "may have been" run on machines that should not qualify under state laws since "there may be" under the system a means of changing the result through an alter system in the commands in the computer program.
In a motion to dismiss, defendant Rebecca Walker noted that the Department of State of the State of Florida previously approved the voting system in use, and that there is no provision in the Florida election laws allowing for a hand recount of ballots cast through the use of an electronic voting system. [77] The Owens-Perry contest had been recounted by computer two days after the initial count because it was sufficiently close to meet the statutory recount test. The Anderson-Walker contest was insufficiently close to require a recount, but was actually recounted as a by-product of the Owens-Perry recount. The computer recount of the Anderson-Walker contest was unofficial, and the same result showing Walker winning certainly did not satisfy Anderson.
On March 28, 1985, according to Ms. Winchester, a hand count of ten precincts selected by Mr. Anderson was undertaken by the staff of the Supervisor of Elections. Mr. Anderson and his observers were in attendance. "Mr. Anderson and everyone else present agreed that were no substantive differences", reported Ms. Winchester. On September 10, 1985, Mr. Anderson's amended suit was "dismissed with prejudice" by Circuit Court Judge Richard I. Wennet. [78]
4.11 Salt Lake County, Utah: November, 1980
A last minute breakdown of one of Salt Lake County's two ballot reading computers caused a delay in production of the tally. No county totals were produced for two hours, and the final tally was produced at 5:39 a.m. the following morning. The situation was reported in an article in the Salt Lake Tribune on Nov. 6, 1980. [79]
Although spare parts were available, a decision was made during the count not to attempt to repair the machine, but to keep going with the one ballot computer that was working.
The situation might have been helped if a head start on ballot counting had been implemented, as was done with paper ballots. However, it was believed that this might have created confusion in the meshing of early-collected ballots with the later ones for the same precincts. In addition, no election workers were available to collect the early ballots. It would have cost extra money to hire additional workers, as others were occupied with regular jobs during the day.
It was noted that the complete 100% tally was available much earlier than such a tally would have been available if paper ballots had been used.
4.12 Stark County, Ohio: May, 1986
The following description is adapted from the account given in the July 21, 1986 issue of Election Administration Reports [80], Richard G. Smolka, Editor, with permission of the publisher.
Stark is a county of about 400,000 population whose county seat, Canton, is located about 60 miles south-southeast of Cleveland. An unprecedented court-ordered "audit" (hand recount) of a Stark County computer recount in a county commissioners primary contest again named as winner the candidate who had apparently won in the official results of the May 6, 1986 primary but lost in the computer recount. The "audit" revealed a computer program error that permitted over 100 invalid punchcard ballots to be counted in the recount.
At the end of the election-night count, Robert A. Capestrain held a 26-vote lead in the three-person contest to be Democratic nominee for county commissioner. A recount by computer on May 27 (held because of the closeness of the original tally) put Patty Miller ahead by 5 votes. For the computer recount, the computer program used to obtain the original results was not used. Instead, a special computer program was written, in order to count only the disputed contest and not the other contests on the ballot. The mystery, however, was why 165 additional votes had been tallied in the recount although the number of ballots read by the computer was the same.
The following table provides the votes for the three candidates in the computer tally of the primary, the computer recount, and the hand counted "audit":
Votes Counted for Stark County
Commissioner
Democratic Primary Election,
1986
Candidates |
Primary
May 8 |
Recount
May 27 |
Change |
Audit
July 8 |
Change |
Robert A. Capestrain |
12,967 |
13,018 |
+ 51 |
12,980 |
- 38 |
Donald E. Casar |
7,987 |
8,019 |
+ 32 |
7,989 |
- 30 |
Patty Miller |
12,941 |
13,023 |
+ 82 |
12,952 |
- 71 |
Totals |
33,895 |
34,060 |
+165 |
33,921 |
-139 |
The 165 additional votes in the recount were randomly distributed throughout the 481 precincts. Most precincts had no changes, and most of those with changes had a one-vote increase. All candidates gained votes. The names of the candidates were rotated by precinct in the ballot booklets in positions 98-100-102, and the extra votes were distributed among these numbers on the ballot cards. Each of the three positions received approximately the same number of additional votes.
Initially, there seemed to be no satisfactory explanation for the additional votes. Hanging chad was suspected as a possible cause. Fraud was much less likely because it would have required access to ballots from all affected precincts, working knowledge of the ballot rotations, plus sufficient time to locate and punch ballots which had not been voted for county commissioner.
Following the computer recount that indicated a reversal of the initial count, candidate Capestrain filed suit challenging the recount. Because of the unusual nature of the recount result and rumors of fraud, the candidates, attorneys, election board, and court agreed to "audit" procedures that would resolve any identifiable problems with "hanging chad" as well as ensure that the vote count would be complete and accurate. Most importantly, they all agreed that the audit would constitute a final resolution of the vote count dispute. Judge Harold E. DeHoff of the Stark County Court of Common Pleas included a provision in the agreement that all parties would waive any rights of appeal.
The audit included a manual count and a computer count. Forty two-person teams were assigned to manually count the ballots under specific rules. The court order also provided guidelines on removal of hanging chad and specified that only the two master commissioners, appointed by the court, could remove a suspected chad or hanging chad.
Before the start of the audit, Ohio Director of Elections Dorothy Woldorf and area manager Robert Braun of the vote-counting system vendor gave the counting teams both written and verbal instructions on procedures to be followed. The manual recount began at 9:00 a.m. and continued until completion at about 7:45 p.m.
After the first several precincts were manually counted, it became evident that the audit was producing totals more closely matching the original count rather than the recount. By 11:00 a.m., the recount program error had been uncovered. The error was due to the failure of the recount program to distinguish between Democratic, Republican, and unaffiliated ballots.
In the May 8 primary, voters were given Democratic, Republican, or unaffiliated ballots, depending on their party registration. The logic in the computer program and associated header cards that were used to tally the primary ballots was able to distinguish among the different types of ballots, even though all the ballots were tallied on the same computer equipment.
In the recount, all the ballots were again tallied together on the same equipment, but the logic of the recount program could not distinguish among the different ballot types. It was apparently believed by the author of the recount program that the assignment of unique ballot positions to each contest and candidate was sufficient to separate the ballots. However, some Republican and unaffiliated voters had "voted" (i.e., punched out chad) in a ballot position assigned to a candidate in the Democratic county commissioner contest. These ballots were not counted in the Democratic primary tally, but they were counted by mistake in the computer recount.
In the audit on July 8, the ballots were first separated by party before being given to the two-person teams. The separation was easily accomplished because the ballot types were distinguishable by color. Consequently, in the manual recount, "votes" by Republican and unaffiliated voters were not tallied.
During the audit, the master commissioners completed removal of chad on 28 ballot cards. Nine of these were identified as "hanging chad", and the others were termed "bulging chad". One commissioner said that it was obvious that the voter had detached the chad, but that it had been pressed back into position, probably when the cards were stacked. The removal of the chad by the commissioners had no effect on the outcome, but did increase the vote by a net of 26 over the original count.
4.13 Summary Of Problem Types
The problems in specific elections that have been described are categorized below in order to identify the most prevalent types and elucidate the specific difficulties.
4.13.1 Insufficient Pre-election Testing
Lack of sufficient pre-election testing appears to be a major source of operational difficulty. Problems in the following situations would have been avoided if significantly increased numbers of test ballots, using many different expected vote combinations, had been run through the machines and the results compared with expected answers:
* Carroll County, MD; Nov., 1984: A larger number of test ballots, using different combinations of voting possibilities, would have demonstrated the incorrect vote-counting due to the presence of the wrong utility program, and would have avoided the embarrassment and controversy that resulted.
* Elkhart County, IN; Nov., 1982: A larger number of test ballots should have brought to light the three separate coding errors: failure to distinguish Wakarusa voters from other voters in the same township, the reversed tallying in the County Council races, and the incorrect punch position used to tally votes in the State Representative contest. The fact that all three errors were identified after a significant number of votes had been actually tallied (and because the strange results raised suspicions) supports the concept that more test ballots would have brought the errors to light before the tallying began.
* Elkhart County, IN; Nov., 1986: A larger number of test ballots, testing the correctness of the logical rules, would have identified the incorrect logic of the straight party overvote implementation.
* Illinois Statewide Testing Program: The use of a larger number of test ballots would have made clear the logical errors in programming and coding that were identified. This need was recognized by the testers from the State Board. [64]
* Moline, Il; April, 1985: The implementation of pre-election testing might have brought to light the ballot-reader failure before the election, and therefore might have prevented the losing candidate from being certified as the winner.
* Stark County, OH; May, 1986: A more complete checkout using a large number of ballots would have identified the logical error that caused the program used in the recount to fail to distinguish between voters of different political parties.
* A situation not appearing in this category is Maricopa County, AZ. Adequate checkout procedures in that jurisdiction prevented the incorrect punch position assignment from being implemented in the election.
4.13.2 Failure to Implement An Adequate Audit Trail
* Dallas, TX; April, 1985: The failure to separately report numbers of ballots cast in each part of a split precinct produced ambiguous and suspect results.
* Elkhart County, IN; Nov., 1982: Changes in control cards, to compensate for errors that were discovered, were not documented; retest of the system, following the changes, was not done.
* Oklahoma County, OK; Nov., 1986: There were, in many precincts, significant differences between number of ballots tallied and number of voters reported as voting. In some cases the number of ballots tallied exceeded the number of reported voters. These differences made the results appear suspect.
4.13.3 Failure to Provide for a Partial Manual Recount
* Charleston, WV; Nov., 1980: The charges of conspiracy, and the expensive and time-consuming lawsuits, might have been avoided if the local laws had required, or allowed for, a partial manual recount without a court order. When the ballots were destroyed, the essential evidence to disprove an incorrect tally, of whatever cause, was made permanently unavailable.
* Dallas, TX; April, 1985: Although a recount was undertaken, it was a machine recount, using machines managed by the same organization. No manual recount, or recount managed by an independent organization, was done. Consequently, some suspicion remains.
* Moline, IL; April, 1985: A partial manual recount would have brought to light the incorrect results due to the ballot-reader failure, and would have prevented the losing candidate from being certified as the winner.
* Oklahoma County, OK; Nov., 1986: The unusual results in which lower-level contests received more votes than the contests on the top of the ballot could not be validated. Thus, a serious loss of public confidence could not be prevented.
* Palm Beach County, FL; Nov., 1984: The inability, under Florida law, for the defeated candidate to force a recount (because his race was insufficiently close), raised suspicions unnecessarily. A partial hand recount, as is done in California, might have prevented this situation from arising.
* A situation not appearing in this category, because an unofficial manual recount was taken in order to verify the different results obtained on an independently-managed system, is Carroll County, MD, November, 1984. The combined manual count and recount on an independently-managed system revealed the error in the original count.
4.13.4 Inadequate Ballots or Ballot Reader Operation
* Carroll County, MD; November, 1984: A recount, in which read "checks" were ignored, showed slightly different tallies than the certified results. Chad is also believed to be responsible for some of the difference.
* Dallas, TX; April, 1985: The recount taken a few days after the election showed changes in ballots cast in 109 of the 250 precincts, and changes in votes cast for at least one of the candidates in 161 of those precincts.
* Gwinnett County, GA; Nov., 1986: The differences between results obtained on different computers reading the same ballots indicates the limitations of pre-scored punch cards in their ability to provide reproducible results.
* Moline, IL; April, 1985: Apparently, a slipping timing belt caused the reader to read incorrectly. The failure to provide for a partial manual recount (see above) or for sufficient pre-election testing (see above) prevented identification of the problem in a timely manner.
* Oklahoma County, OK; Nov., 1986: The ballot readers could not read up to 11% of ballots cast in some precincts.
* Palm Beach County, FL; Nov., 1984: The changes in votes cast on successive tallies of the same punch card ballots, probably due to the presence of chad, reduced the confidence of the losing candidate in the validity of the reported outcome.
* Stark County, OH; May, 1986: The final (third) count of ballots verified the first count, excepting that the results were slightly different due to chad fallout or deliberate removal of hanging chad by the inspection boards.
4.13.5 Inadequate Security and Management Control
* Elkhart, IN, Nov., 1982: Vote-tallying was done in a borrowed facility with equipment and employees not under management control. Multiprogrammed computer operations independent of vote-tallying and not under management control were being carried out while vote-tallying proceeded. A vendor representative was permitted to operate the vote-tallying process, and to change control cards without adequate documentation, and without regard to requirements for re-testing.
4.13.6 Inadequate Contingency Planning
* Oklahoma County, OK; Nov., 1986: No administrative rules were available to allow the counting of the significant number of ballots that could not be counted by machine. Many voters were disenfranchised.
* Salt Lake County, UT; Nov., 1980: No backup was available when one of two computers failed during ballot counting. The result was a count much slower than expected.
4.13.7 Inadequate System Acceptance Procedures
* Oklahoma County, OK; Nov., 1986: The system procured was not adequately tested prior to acceptance. There was inadequate preparation to deal with the failures that later occurred.