comments and musings on voting systems elicited by
What Went Wrong in Ohio
The Conyers Report on the 2004 Presidential Election

This page:
Voting by machine
Punch-card systems
Computerized voting stations
Exit polls
The partisan divide
Really stupid ideas for improvement
My background and experience

Related pages:

Conyers report quotes

Ohio 2004 Presidential election results

analysis of the Kerry/Conally anomaly




index pages:

Voting by machine

Voting by hand-marked and hand-counted pieces of paper is still the most reliable method for ensuring each vote is counted as the voter intended. Except for accommodating disabilities. Except for accommodating voters with disabilities, the purpose of machinery is to make the counting more convenient (and cheaper).

In general, the problem with voting machines is they separate the counted representation of the vote from the physical representation controlled by the voters and examined by the officials. Optical-scan ballots do so less than most other systems, and are known to be nearly as reliable as hand counting.

top of page

Punch-card systems

In a punch-card system, the “machine” used by the voter consists of a ballot slot, a set of holes, and a stylus to be pushed into holes, allegedly punching corresponding positions in the ballot, along with a set of pages listing the candidates and indicating which slot to punch. As the pages are turned, different rows of holes are exposed for punching.

The ballot has only an arbitrary mapping to the races and candidates, easily skewed because the voter can’t see the punches or read what they mean. Failure to insert the ballot all the way can shift all the votes, as can using a machine that lists the candidates in the wrong order. Holes are not guaranteed to be punched fully, especially late on a busy day when the machine clogs up with chad (those little bits punched out – or not). Or early in the day, if someone has failed to clean the machines – possibly only for precincts where they prefer fewer counted votes.

These systems are so hard to use that trained election judges are generally supplemented by volunteers with minimal training. The 2004 Ohio vote was apparently affected by such volunteers, attempting to be helpful and not understanding the consequences.

In 2000, much of the fuss over hand-counting punch cards missed the point: these systems are incapable of producing more than a statistical approximation of the result. Punch cards counted multiple times cannot be expected to produce a consistent result, because of hanging or otherwise anomalous chad. In a close election, only a hand recount can yield a trustworthy result from punch-card ballots. Politicians and equipment company representatives claiming otherwise are lying. In fact, because the companies stand to lose business and election officials are faced with heightened expenses, they seem to have colluded in Ohio to fabricate an apparent consistency.

Case in point: the Palm Beach butterfly ballot

Both partisan interests and the desperately complacent have blamed Palm Beach voters for their difficulties with the illegal “butterfly ballot”; these attacks failed to appreciate the complexity of the problem. A picture of the butterfly ballot made the problem clear, but you’ll have to take my word for it; the Web link I used to have here has gone stale.

Certainly, it’s possible to figure out where to punch for each candidate. But these voters were used to Florida ballots, which by law are supposed to list all candidates in a given race in one column, with the Republican and Democratic party candidates in the top two slots. They had no reason to look for additional candidates on the back of the next sheet; the only ones they cared about were where they were already looking. Many punched the second hole because that’s where the second candidate’s vote goes.

Most of these then noticed the arrow and punched the third hole, assuming the second one would be disregarded because no arrow – on the page they were looking at – pointed to it. Had the ballot been legally designed, that would have been the case. No one wants to ask for a replacement ballot if they don’t see the need, especially if they’re hassling with a punch-card system, and practically no one was actually looking for any of the second-column options. (Those who were doing so avoided the errors, of course, along with anyone who intended to vote for the first candidate listed.)

The officials – Democrats, in fact – who allowed this design should have known it was confusing. They used it in 1996 and got over 15,000 overvotes. (They probably ignored this because it only takes votes from the second candidate listed, and Clinton alphabetically precedes Dole.) Four years later, they suffered the consequences of lazy complacency: their opponents launched a massive and racially biased vote-interference campaign, which a fair vote in Palm Beach would have prevented from succeeding. This time the crooked ballot altered enough Democratic votes to allow that campaign to succeed.

I’m entirely willing to blame the Palm Beach officials, but it’s contemptible to transfer blame to voters who were deprived of a fundamental right, when they were not primarily responsible for their errors.

top of page

Computerized voting stations

Computer interfaces offer the voter a great deal more confidence that their vote is cast correctly. Instant feedback is possible, and erroneous votes can be corrected without the inconvenience of reported a spoiled ballot, obtaining a replacement (if officials allow it), and starting over.

Much of this confidence may be misplaced. On many of these machines, there’s no physical record of the vote; a recount consists of re-reading the same reported numbers. In some Washington counties, that was the extent of the “full and fair” recount of the 2004 gubernatorial election. This may have affected the result and overturned a Republican victory.

Beyond the risks of ordinary error, these machines offer a wealth of opportunity for “Trojan Horse” voting interfaces, as demonstrated in Ohio. We programmers are entirely capable of changing a Yes to a No when you’re not looking, and not all of us can be trusted to refrain. Safeguarding the software and monitoring the system to ensure its integrity is not a simple matter.

I have two reactions to the Ohio reports of Trojan Horse voting interfaces.

Computerized voting interfaces do offer considerable benefit for disabled voters. A wider application to rigging elections might be prevented if they are restricted to supplementing more trustworthy systems, such as optical scanning. If only a fraction of the voters use the computerized interfaces, that would reduce the potential effect of, and therefore the motivation for, crooked programming. Cost savings might be possible as well.

top of page

Exit polls

I have disliked exit polling for a long time, because it has been used to “call” elections before the polls close, thereby affecting turnout. The problem is especially knotty for national elections, where time-zone differences have some states reporting results while polls are open in others.

However, both Florida 2000 and Ohio 2004 have revealed a second function of exit polling: it serves as a check on the reported results. In Florida, the polls correctly revealed that a plurality of voters had attempted to vote for Gore, as substantiated by subsequent review. The exit poll reflected the intention of the voters more accurately than the final official count. A similar result was available in Ohio, though the pollsters appear to have “revised” their results to match the official count after the fact.

In an era when top election officials simultaneously chair campaigns, as happened in both those states, any independent source of information is crucial.

top of page

The partisan divide

The Democratic and Republican parties each circulate horror stories about how the Other Guys are maneuvering to steal an election, usually by dishonest manipulation of voter registration or by intimidating voters. The major difference is in the chosen remedies: Republican solutions to the alleged problem have consistently tended to involve preventing people (usually selected by race, language, ethnicity, or vulnerable social class) from voting. Democrats have – so far as I have observed – concentrated their efforts on assisting legitimate voters with exercising their voting rights. Until that pattern changes, I will involve myself in only one party’s efforts.

Not all Republicans are tarred with the same brush. A Minnesota House committee chair recently allowed his committee to act on a change restricting challenger positions to state residents, partly because out-of-state challengers have a history of voter intimidation. The chair, a Republican, knew that his own party had just contracted for a large number of out-of-state challengers for 2006. He allowed the change anyway, because it was “the right thing to do.”

top of page

Really stupid ideas for improvement

Recent attention to election problems has led to a number of proposed improvements. Many of these have merit; some are just plain stupid, at least to anyone cognizant of Murphy’s Law. Among these are:

Abolishing the Electoral College

This proposal violates the fundamental rule of engineering: If it ain’t broke, don’t fix it. It was largely motivated by the fact that the loser in the 2000 election won a plurality of the popular vote.

But if the popular vote is counted directly, then any vote in any state can offset a contrary votes anywhere else. In other words, we’d have been looking at a nationwide recount, on terms worse than the Florida one.

The Electoral College functioned as intended. The popular vote was whisper-close, and the Electoral College was so close one state made the difference. One candidate picked up extra electoral votes corresponding to the Senators from the less populous states; the other picked up extra votes by winning large states with winner-take-all rules. Most proponents of change concentrate on one effect or the other, but actually they balanced out, and they can be expected to do so whenever the Presidential election is close.

That’s not to say the Electoral College functions perfectly. In 2004, one of ten Minnesota electors voted for the same person – his party’s Vice Presidential nominee – for both President and Vice President.

National voter identification

The proponents of this “solution” ignore the obvious problem: all you have to do to stop someone from voting is hide his voter ID, or take it away, or declare it out of date or suspicious or wrong, or just mail it late.

ID checking is, in fact, one of the most widespread and powerful methods of depriving voters of their rights.

Instant runoff voting (IRV)

Proponents of IRV seem to think it can correct all election abuses and cure warts. I think they’re wallowing in a private fantasy, and they haven’t bothered to study their own algorithm because they can think of a few examples where it would work. It’s like the classic “engineer’s proof” that all odd numbers are prime. *

* One’s a prime, three’s a prime, five’s a prime, seven’s a prime, nine’s experimental error....

IRV works by allowing voters to rank the candidates in order of precedence, and then successively discarding candidates who have the least first-place votes until someone achieves a majority. This makes it possible for someone to cast a sentimental vote for a hopeless minor candidate, with their vote automatically switching to a second choice when their first is discarded. The cases offered as examples by IRV proponents are simple ones that operate like a runoff election.

I assume, although they don’t always make it clear, that they would allow voters to completely exclude some candidates from their rating lists. If a vote requires rating all candidates, that creates a huge barrier to voting. It also allows a vote to be counted for a candidate the voter actually opposes, which is tantamount to election fraud by stealing votes.

With the assumption that rating lists can be incomplete, several flaws are apparent:

top of page

My background and experience

I have voted on a variety of systems, including punch cards. As a programmer/analyst with experience in data communications, I also have a technical appreciation of their capabilities, as well as an understanding of the nature of software.

I have observed a legally-mandated public test of voting machine configurations. (Employees were pleased to see two members of the public, and said they hadn’t ever had any attend before.) I have also attended a demonstration by a manufacturer of computerized voting stations, designed to assist voters with disabilities.

During the 2004 election, I volunteered for a Voter Protection project. As a party-appointed observer – the Minnesota title is “challenger” – I observed a full day of activity at one precinct. This has given me a much deeper understanding of the extensive, finely-tuned, and possibly fragile set of safeguards in our system.

Among the more important safeguards in Minnesota, I consider same-day registration vital, especially for a state where recent rosters have capriciously omitted longtime voters. It’s true that a couple of abuses have been identified (and prosecuted) in the area, involving two voters and three votes. The penalties are severe.

In just one precinct, I saw half a dozen instances of disappearing registrations. It seems clear that illegal voting is not occurring on anywhere near the same scale as arbitrary, inexplicable removals from the rolls.

As of last year, we also have a Voter’s Bill of Rights. In addition, safeguards include a very large set of procedures and audited physical records, all designed to protect against manipulation of systems with particular characteristics. Changes in the system may introduce new risks, not covered by the existing safeguards. Computerized voting equipment is just one of the threats.

top of page