Voting systems have come in for heightened scrutiny since the presidential election nearly 19 months ago. We've had charges of the nationwide undercounting of (presumably Democratic) minority and low-income votes. We've had panels of the obsessive discoursing about teensy cardboard convexities, or the lack thereof. Millions of dollars have been spent re-re-counting Florida ballots, to little effect.

In the process we became aware of the profusion of balloting methodologies, and we were confronted with the news that inaccuracy is inherent in any system. Here then is a summary of voting machine types, and a discussion of some of the problems that can occur with each of them.

Ballot methods presently used in the U.S.

• Votomatic-type punch card machines (used in counties with 32% of the U.S. population). Candidates' names, as well as issues to be resolved, are printed in a sheet or booklet that aligns with a set of holes in the panel of the device. The voter then makes selections by punching holes in a paper computer card. The card is pre-scored, so that the punch action can completely detach a little rectangle, the now famous "chad," leaving a hole to be found by a high-speed electronic reader.
  Answering a question you might well ask, is this the same as the old IBM punchcard (and before that, the Hollerith census card)? Yes, it is.
• Datavote punch card machines (4% of the population). The entire ballot - names, issues, propositions - is pre-printed on the card, and a stapler-like tool is used to punch the holes.
• Optiscan machines (27%) (same as used in SATs) which require the voter to use a special writing instrument (infra-red compatable ink or soft pencil) to fill in an oval or otherwise mark a candidate's name. These devices are used with central counting of the ballots or, in the case of 'precinct-counting,' the voter is permitted to insert the ballot into a test machine at the polling place, to verify correct voting.
• Mechanical lever machines (18%). These require the voter to set a small lever on the machine to select each candidate, and then pull a large handle to record the vote, incrementing a set of counters at the rear.

• Electronic systems (9%). Known as direct-recording electronic (DRE) devices, they allow the voter to select candidates by pushing buttons or touching spaces on a screen which displays the image of a ballot. The devices are programmed to prevent over-voting (and the consequent loss of your intended ballot), and can prompt for undervoting, for instance, by flashing a light above each office until a vote for that office has been recorded.
  There are touch-screen DREs and push-button DREs. A benefit of this technology is that there are accessory devices for blind voters, who thus don't need assistance (and the potential loss of the secrecy of the ballot).
  In the recent Charlottesville City Council election, the Hart Intercivic eSlate DRE device was tested.
• Paper ballots (1.4%). Simply marked with 'X'es and stuffed in a ballot box for hand counting later.

Note: percentages are for 1998, and show the employment of a particular system throughout an election district. About 8% of the population lives in counties or municipalities employing mixed systems, primarily in northern New England and Wisconsin. The source of the percentages is Election Data Services.

Possibilities for wrong counting

R. Doug Lewis, executive director of the Election Center (an international organization of election officials) has written about the ways that voters themselves can contribute to bad counts:
"Americans continue to amaze election officials with their creative ability to miscast votes. Give them a marking device with their paper ballot, and they'll take out their own pens instead - and the wet ink from a fountain pen may occlude the lens on the optical scanner. Or they'll use red ink and the infrared scanner won't detect it. Or they'll write notes, which a machine can't read. Give them an ATM-style touch screen, and they'll touch two candidates' boxes at once - and the screen will read neither, or both, or the box in-between."

But it's not just the voters.

Paper Ballots Opportunities for error start with ballot layout and instructions (all those little boxes), continue through hand tallying and interpretation of the markings, as well as the mistrust of a system that in the past has been open to all kinds of fiddling.

Votomatic Punch card machines: people's votes may not count if they incorrectly insert the card, or if it was improperly pre-scored; if there is confusion caused by the information in the printed booklet, or the arrangement of candidates on the page (the "Pat Buchanan in the middle" possibility); 'pregnant' and 'dimpled' and 'hanging' chads not completely punched; or even perfectly-punched chads not removed from the cards prior to counting; misalignment of the card in the machine. Since the card has no printed information, there's no way for the voter to verify the ballot by inspection after removing it from the device.

Datavote: There will be fewer punching errors, because the names of the candidates or the issues to be voted on are printed on the card. But this same benefit means a great deal more space is taken per item - this can easily result in multiple cards per voter, slowing the count considerably.

Lever machines: If the headings are misaligned, the voter may inadvertently select the wrong item. Bad vote counts can result if all the counters in the machines have not been properly reset to zero prior to use, or if the mechanical counters do not turn properly due to misalignment or excessive friction.

Another issue is cost - since lever machines are more expensive (about seven times the cost of a punch card machine, plus more expensive to maintain, transport and store) there are likely to be fewer of them in a jurisdiction, leading to longer voting lines and resultant voter disaffection - or not voting.

Optical scan: where special, infra-red-visible markers are required - if the voter uses a regular pen or pencil, the vote may not be counted. There is a 'fill-in' space - miss it, or circle it, or attempt to erase the mark, and the vote won't be counted.

Direct Recording Electronic Devices: Cost is an issue with DREs, with machines currently costing $3,000 to $6,000 each. However, their use can result in lower associated costs like election workers' time.

We're still waiting for final results of the 'user satisfaction' survey done by the city of Charlottesville in the recent council election. (Preliminary indications are that everything went quite smoothly).

A sensitive point that arises with DREs is that for each election the programming of the machines is done by the vendor, not local election officials. Also, since the counts of votes are entirely electronic, an audit trail of some kind must be created, raising the potential for problems through confirmation by alternative counting methods.

The use of this equipment clearly raises the prospect of employing the Internet for more voting, with the attractive notion of voting from home, office or cafe, or at municipal polls equipped with nothing more daunting than conventional computers. Issues of security and identity are similar to those now being addressed in other web-mediated applications.

Whose votes count?

At one point in the dust-up over the Florida counts, Al Gore claimed "the old and cheap, outdated machinery is usually found in areas with populations that are of lower income people, minorities and seniors on fixed incomes." But several studies indicate that, except in highly distinctive circumstances, this is not true. Because some counties (Los Angeles, Miami/Dade) have enormous populations of 'minority' voters, their selection of equipment skews all statistics.

On the other hand, according to the same studies, the simple fact is that the higher the representation in a district of poor and minority voters, the higher the error rates. According to a report prepared for the Committee on Government Reform of the U.S. House of Representatives, "On punch card machines, voters in low-income, high minority districts were 3.85 times as likely to have their votes thrown out than voters in affluent, low-minority districts." On precinct-counted machines, with the possibility of verification and re-casting one's vote, these people's ballots were still 2.2 times as likely to be thrown out.

The error rates are considerably lower using other technologies (although they remain higher for the disadvantaged), and the convenience and low cost offered by the punch card devices will increasingly be outweighed by considerations of fairness and the risk of future humiliation. (Dave Sagarin, May 29, 2002)

Note: Our thanks to Jim Heilman, election consultant and former Registrar of Albemarle County, and Sheri Iachetta, Registrar of the City of Charlottesville, for assistance in the preparation of this article.