Disclaimer: This commentary is a personal response to the Election 2000 report and does not represent either the Iowa Board of Examiners for Voting Machines and Electronic Voting Systems or the University of Iowa.Acknowledgement: I would like to thank Iowa Secretary of State Chet Culver for his interest and encouragement.
In December 2000 and January, 2001, the Iowa Secretary of State conducted a series of hearings around Iowa looking into the problems exposed by the very close general election of November 2000. The statewide margin between the top presidential candidates in this election was under half a percent (638,517 to 634,373), and as the recount process in Florida began, there was a very real possibility that there would be a call for recounts in Iowa and Wisconson.
The results of the hearings in Iowa is presented in a report, Iowa's Election 2000: Facts, Findings and Our Future. This is available on-line from the Secretary of State's web site,
http://www.sos.state.ia.us/pdfs/ElectionReport.pdf
This report focuses not on the facts surrounding the election in Iowa, although it does provide some valuable factual data, but rather, it focuses on people's perceptions of what ought to be done, as gathered at the public hearings. Basic factual information about this and other recent Iowa elections is available from the Secretary of State's web site, at
http://www.sos.state.ia.us/elections/election_results.html
In addition recommending immediate action, the Iowa's Election 2000 report asks that an Election Reform Task Force be convened to, at minimum, ask ...
I am no lawyer; I cannot therefore speak with any authority on what the decision requires. However, I have strong opinions on what ought to be required. I believe, strongly, that the Supreme Court was correct to identify non-uniformity in the counting of votes as a civil rights issue, and I believe that we are obligated to assure that every voter's ballot is counted.
At the same time, while I agree that use of identical voting technology across the state trivially assures this, I believe that we must resist this easy path for several reasons:
As I have repeatedly emphasized, no currently available technology that uses computers to count votes is adequately defended against corruption on the part of voting system vendor staff. Our current system of regulation under the "voluntary" FEC standards poses only a minor barrier to such corruption.
Until we have a regulatory system that is genuinely secure against such insider manipulation, we must not permit a monopoly vendor to emerge!
It is far easier to phase in new machines, updating the machines in a small county or in some of the precincts of a large county before committing to the massive purchasing involved in upgrading the machines in a metropolitan county.
If the minim "unit of purchase" for new voting machines is one state, only very large companies will be able to afford to compete in this market, and we will likely see the number of vendors shrink rapidly as successful vendors merge with each other or drop out of the market, with no new vendors entering the competition. This could lead rapidly to a national monopoly and all of the risks inherent in such a monopoly, even if the courts do not conclude that the Supreme Court required more than a statewide monopoly.
The argument against a hand recount rests on the premise that hand counting is inherently subjective and subject to bias, while machines are inherently unbiased. This was the argument that led to the widespread adoption of lever voting machines as replacements for hand counted paper ballots in the early 20th century, and it remains a powerful argument today.
The problem is, biased machines can be built, and if such machines incorporate computers, this bias can be made arbitrarily difficult to detect. There were indeed severe problems with the hand vote counting schemes in use prior to the adoption of lever machines. In regions dominated by entrenched and corrupt political machines, such vote counts were routinely manipulated in order to pre-determine the winner without any necessary reference to how the voters actually voted.
Unfortunately, the substitution of mechanism for hand labor did nothing to eliminate entrenched and corrupt political machines! Just as hand counting rules may be rigged in order to predetermine an election outcome, so may mechanisms be doctored. Furthermore, an exhaustive inspection and test of a lever voting machine is a very expensive proposition, involving checking thousands of moving parts for correct behavior. As a result, if it is ever done, it is done as a spot-check. If a machine is rigged to fix one election, the rigging may not be noticed during any kind of inspection or testing until years later!
Thorough testing of computerized voting systems is even more difficult, so much so as to be practically impossible. However, with lever machines, it is individual machines that are typically rigged. With computerized systems, the primary vulnerability is in the software used to program an entire family of machines. Therefore, regular independent checks on randomly selected machines can deter and detect corruption.
The most direct independent check on the function of a voting machine is a hand recount. If there is reason to believe that a machine is faulty or corrupt, we must have this option!
The current law allows candidates who were on the ballot or who received votes to request recounts. Clearly, these are interested parties and must retain the right to request a recount.
Current law does not allow recounts under certain circumstances where I believe they ought to be allowed: First, if the county commissioner of elections believes that a machine has malfunctioned, he must be able to initiate a recount! This happened in Johnson County in the recent general election, and a somewhat ad-hoc procedure was followed in order to recount the ballots in question. I want to emphasize that I believe that the right thing was done, despite the fact that current Iowa law does not appear to authorize such a procedure!
The second gap in the current law is a general provision for initiating an automatic recount in a race when the margin is small. Several states have automatic recount provisions based on simple margin thresholds such as one percent or half a percent. The recent general election would have forced a recount in Iowa if any of the threshold values I've mentioned had been in effect.
I favor a more specific criterion for triggering a recount. This rests on a somewhat more complex set of canvassing rules in which not only the votes are canvassed, but also the overvotes and undervotes. If the total number of votes, overvotes or undervotes does not match the total number of ballots counted, district wide, something is wrong and the initial canvass should be declared to be only tentative! If the margin of error is small compared to the margin of victory between the leading candidates, a winner can be safely declared, but if the opposite is true, recounting should begin immediately, and not just in selected precincts, but district wide.
The rules suggested above automatically include ballots known to have been lost or destroyed as part of the margin of error. In addition, I would like to see the number of as-yet unresolved challenged ballots included as part of this margin of error, and I would like to see the overvote included as part of the margin. If the total margin of error in the count is more than the margin of victory, recount should be automatic!
Overvotes on machine-counted paper ballots are almost always the result of problems, ranging from printing errors, problems with the ballot marking mechanism, smudged erasures, or flyspecks in the paper. I would prefer to see a law requiring that our counting machines have a high chance of counting such marginal marks as votes, rather than missing any deliberate marks made by the voter. This will create overvotes where a human reader would see no problems.
Under the above requirement that the counting machines count even faint marks as votes, the only undervotes that remain should be genuine undervotes where the voter indeed made no mark. The right of a voter to undervote will always exist on paper ballots of any kind, and in fact, for many races, undervotes will be very common. Therefore, I see no reason to consider machine counted undervotes as part of the margin of error in the vote count.
If the stated reason for the recount is the belief that a voting machine has malfunctioned or may be corrupt, a hand recount must be allowed! If a specific machine is believed to be faulty, a machine count with a different machine running the same software is an obvious alternative to a hand recount. If, on the other hand, the honesty of the software itself is in question, the obvious way to test the honesty of the machine is to do a hand recount.
If the recount is triggered by a close vote margin, independent of the details of how that margin is stated, I strongly urge that all overvoted and damaged ballots be subject to a hand recount. Machines will generally count flyspecks, smudges and similar marks as votes, while humans have a very easy time distinguishing these from deliberate votes. This does require that overvoted ballots be segregated from others, but modern voting machines routinely segregate ballots requiring hand examination, for example, write-in ballots, from the remainder, so this is not a difficult requirement.
One of the proposals for ammendment to Iowa's recount law suggests that the recount board could have a "true copy" or a "corrected" ballot made in order to allow machine counting of damaged and marginal ballots. This proposal has merit, so long as these copies or corrections are made with sufficient oversight and an appropriate audit trail is maintained.
The problem with the hand recounts in Florida arose, in significant part, from the prevalance of punched card technology in that state. On optical mark ballots, we face problems with flyspecks and smudges, something we have all learned to deal with since kindergarten. In contrast, in Florida, they faced problems with dimples, trapdoors and other arcania that are far from our commonplace experience. The result was a prolifieration of arbitrary sounding rules about what marginal punchings counted.
Another problem we face is that hand recounts are so rare that neither the county administrators, the parties nor the press are accustomed to the procedures involved. I believe that it is important to maintain a population of people who have experience with hand counting, so I strongly urge routine hand counting of some votes in every election.
Toward this end, I urge that, for example, one randomly selected race in one randomly selected precinct be hand recounted in every election. This hand count would serve as a check on the accuracy of the voting machine software in ddition to serving as practice for the large-scale hand counting that might be required if there was a contested election.
I am unsure of the intent of this question, although it does seem to open the doors to such ideas as instant runoff elections and similar extreme variations on the current vote counting schemes. If we exclude these, it seems that the candidate who receives the most votes should be declared to be the winner.
If we adopt a counting scheme that allows us to accumulate a number corresponding to the error in the count, however, we could declare a race to be a tie if, after recounting, the margin of victory of the leading candidate remains smaller than the error. I would strongly urge the adoption of such a rule.
I would hope that public is notified, sufficiently promptly that if the interested parties are paying attention, they would learn from this alone. The Draft Legislation included as an appendix to the Iowa's Election 2000 report suggests a reasonable approach, notifying all candidates by certified mail and attempting to contact all of them by phone. This does not address the case of legitimate write-in candidates; there is clearly no need to contact Mickey Mouse, Donald Duck and other facetious names that are frequently written in, but if a real person receives a significant number of write-in votes, it would seem reasonable to notify that person.
It is extremely important that the board contain an odd number of members so that it will not deadlock, and that it contain representatives of genuinely opposing candidates. In a pure two-party system, opposition is easy to guarantee. In a genuine multiparty system where two candidates are genuinely opposed to each other and stand out well ahead of the others, such as we appear to have today, representatives of those candidates plus one other should suffice to build a fair recount board.
The entire problem becomes far more complex in a genuine multiparty system where parties form transient coalitions against each other. In such a system, guaranteeing that adversaries serve opposite each other on the recount board can be very difficult!
A bond should only be required for a recount called by a candidate. I strongly urge that, if we do not trigger automatic recounts for close elections, then we should at least waive the requirement for such a bond on a close election, as the current law already allows.
Hand vote counters must be able to allow for ballots printing errors, flyspecks in the paper, smudges, erasures and accidental marks made by a voter (frequently called hesitation marks). Hand counters should be required to discount ballots containing deliberate mismarkings that might be used to disclose the voter's identity.
I assume that teams of two counters will examine each ballot, where the two represent opposing candidates or parties, and I assume that parties will be allowed to post observers who can closely observe the counters, and I assume that representatives of the press and public will be allowed to closely observe the count. In such circumstances, I expect that most of the counters will agree about most of the ballots, but there will remain some cases where the counters disagree.
Where counters disagree about the marking on a ballot, for example, if one counter says that a ballot contains a deliberate mark, while another says that the mark looks accidental, I strongly urge that that ballot be counted as "disputed" and set aside. If the number of disputed ballots exceeds the margin between the winning candidates in the recount, the disputed ballots and only the disputed ballots should be re-examined by either a larger panel or a more expert panel.
I have two suggestions for the constitution of this larger panel that differ significantly in their implications. One suggestion is that this larger panel be selected from among the electorate in the way a jury is selected, with challenge and questioning rules quite similar to those we use to build a jury for a criminal trial, and supervised by a judge in exactly the way the jury in a criminal proceeding is supervised. Since every voter is an expert at interpreting his own marks on a ballot, there should be no doubt about the competency of this jury to address the issue.
The alternative suggestion would put the disputed ballots before a panel that is highly trained, relative to the "run of the mill" vote counters. Most or all of the vote counters in a typical recount will be election judges -- precinct election workers. I would suggest that the expert panel be composed of, for example, county commissioners of elections, or even of judges who are well versed in election law.
I believe that either of these schemes can be made to work!
... Work in progress ...