The Engineering of Voting & Voting Machines, 1629 to Today

November 5, 2013

Credit: Amanda OliverOver the centuries, engineers have fielded a number of innovative voting machines to protect the integrity and anonymity of our electoral systems. From voice votes to brass balls and electrons, we’ve come a long way.

In the early days of American democracy, voting was a public affair. Voters would literally voice their selection at polling places, to be recorded and tallied by monitors. Using paper ballots wasn’t unknown, of course – the earliest North American election conducted by paper ballot was in 1629, in Salem – but voice voting was preferred for its simplicity and inherent fraud protection. Conducting elections in this way required no specialized and provided a clear audit trail, with accountability. You could buy a citizen’s vote, or suborn the monitor, but who voted for whom was clear to both the voter and community at large.

This feature is also a vulnerability, however. Voice voting makes intimidation trivial and the purchase of individual votes possible. To refine the election system, it was necessary to engineer a method for anonymous, secure voting.

The Australian Ballot

In the United States, paper ballots were initially preferred as an upgrade to the voice voting system, though these were more vulnerable to large-scale subversion than voice voting. “Stuffing” the ballot box – ie, inserting more than one vote per voter – was a common attack, as was manipulating the total after the fact by printing and adding ballots to corrupt the tally. The only patch to either vulnerability was to either require signed ballots or designate an election official to place each ballot in the box.

Credit: Wikimedia CommonsSigned ballots would only reestablish the flaws of older, voice vote systems. Nominating a single, key individual to place the ballots in their box would do much the same, while making large-scale fraud far easier to accomplish. (It’s cheaper to bribe one man than a plurality of voters.) What democracy needed was a system with accountability, anonymity, and impartiality.

The simplest solution came out of Australia, in 1858. As with most elegant answers to complex problems, the Australian Ballot is so obvious in retrospect that appreciating its initial novelty is difficult. Ballots, themselves, were standardized, printed, and distributed at polling places by the Australian government. Each voter was issued one ballot, with a blank box next to the name of available candidates. Voters would check a box, slip the ballot into a locked container, and that was that. The containers were then shipped to a central counting station, where they were unlocked and counted under supervision.


Against the background of fraud and intimidation which haunted the use of paper ballots, the Australian Ballot was indeed an innovation. Votes were anonymous, access to ballots controlled, and opportunities for fraud minimized. Within decades, this became the standard approach for paper ballot systems in a majority of Western democracies.

It didn’t take long for crooked political machines to shift focus, however. Instead of perpetuating voter fraud – which almost never happens, anymore – the battle shifted to challenging vote tallies on apparently spurious grounds. “The intent of the voter” could be obfuscated by quibbling over the precise location of an X’s center, or whether the angle on a check mark was sharp enough to count as a check mark. While the Australian ballot patched a number of systemic vulnerabilities to abuse, further standardization of inputs was needed to prevent our evolution towards a system of election by lawsuit and nitpicking.

The Chartist’s Brass Balls

In response to 19th century political corruption in the United Kingdom, the Chartists – a coalition of progressive Parliamentarians and tradesmen, some of whom came out of the first mass labor movements – published the 1838 People’s Charter. In it, they proposed to address issues of corruption, disenfranchisement, and voter intimidation through systemic electoral reform. (The Australian ballot, in fact, grew out of a compromise with Chartist elements in that nation.)

The Chartist solution to the problems of paper ballots was to update ancient practices with contemporary engineering.

Source: 1838 People's CharterIn cases where anonymous voting was to be preferred, or where paper ballots were prohibitively expensive, the historical solution was to use specific tokens to indicate a ‘yay’ or ‘nay’ vote on a given subject. This practice is almost literally as old as democracy itself, with the Greeks placing tokens in designated containers to indicate their selection. Secret fraternities and trade bodies would vote to accept members in much the same way, with white signifying ‘yes’ vote and a black, ‘no.’

This method remained in accessible public memory at least until the late 18th century, where it was at one time proposed as the method for selecting the American President. (Candidates would be proposed to the Senate, who would use a white-or-black ball system to anonymously vote their selection.)

In the Chartist’s proposal, each voter would be issued a brass ball upon arrival at a polling station. They would enter a screened area, with an input hopper with holes labeled for each candidate. The hopper would pass inputs though a mechanical tabulator – with its results screened from the public – while judges would see the ball pass out of the machine, confirming that a single vote had been cast, and pass it to the next voter.

While their machine addressed issues of intent, fraud, and intimidation, the weaknesses of the Chartist device are immediately apparent: elections are almost never for a single office, for one, and upgrading the machine for varying slates of candidates would entail a nearly complete rebuilt of the input hopper and ramp system. Decades later, an American mechanical engineer in New York refined the Chartist concept sufficiently that his machine, and its descendants, remained in production for almost a century.

The Meyers Automatic Booth Voting Machine

Credit: Wikimedia CommonsJacob H. Meyers designed the first mechanical lever-based voting machines, which entered the field in 1892. The Meyers Automatic Booth consisted of a mechanical tabulator, accepting inputs from a toggleable slate of candidates and open offices. A system of interlocks prevented voters from voting for more than one candidate in a given race, preventing user error, and the booth would open after the operation was complete, preventing ‘stuffing’.

Determining the intent of the voter was a simple binary, with no room for debate; either they voted, or they didn’t. Individual intimidation was impossible, as well, as the machine retained no information as to which voter cast which vote.

As you can imagine, a purely mechanical system with variable inputs and nested interlocks is non-trivially complex. At the time, the Meyers Automatic Booth was among the most complicated machines in production. This creates some difficulty in inspection and maintenance, naturally, but increased fraud protection drove the adoption of mechanical voting machines in all urban centers by 1930. By the time production ceased in 1982, lever-based mechanical voting machines manufactured by either AVM (Automatic Voting Machines) or Shoup were nearly ubiquitous.

These machines weren’t without weaknesses, as is to be expected with any sufficiently complex mechanism. Maintenance and inspection was incredibly expensive, their was no audit trail to detect under-counted or miscounted votes, and statistical measurement suggests weaknesses in the analog tabulator itself – there are more 99s reported in election results than chance would predict. The incredible expense of inspecting and repairing each machine before every election virtually guarantees that poorly functioning tabulators remain in the field, under-reporting votes by significant margins.

A replacement for mechanical, lever-based machines would have to offer the same level of fraud prevention and protect voter anonymity, while being more affordable to maintain and repair. Over the second half of the 20th century, engineers deployed a number of promising candidates.

From Looms to Voting Machines

Credit: Lars O.Storing mechanical instruction sets on punch cards began with mechanical looms in the 18th and early 19th century, with empty and intact squares corresponding to raising or lowering a bolus hook in the loom. The most commercially successful of these, the Jacquard loom, inspired engineers to develop methods for storing and retrieving information based on manipulating punch card patterns.

Institutional computing followed in less than a century later, with the census of 1890. Using reading and tabulating machinery developed by Herman Hollerith – founder of one-quarter of IBM – census takers punched holes in blank cards corresponding to questionnaire responses. This was the first large-scale use of machine-readable storage media, setting the stage for the first era of big data.

With the census as a proof of concept, it seems strange that another seventy years would pass before punch cards were proposed as a means of automating the tabulation of Australian-style ballots. Rather than check a box, voters could punch out a square on their card. Polling places would be much cheaper to set up and maintain, while mechanical tabulation could take place at single, central location, making maintenance and repair much easier.

Credit: Wikimedia CommonsMany areas of the country either still use these punch-card voting systems, or have only recently upgraded to more modern systems. (The 2000 Presidential election showcased a number of flaws with punch card voting machines – the ‘butterfly ballot’ being a direct descendant of the original Votomatic systems.)


Machines which could scan a ballot for visible marks, rather than punched-out holes, were first developed for use in the educational testing sector in the 50s and 60s. Engineering voting machines based on optical sensing began almost immediately, with entrants from Westinghouse, Cubic, and others. Financial chaos, mergers, and acquisitions left a few dedicated players standing, who provide many of the optical sensing voting machines in use today. Global Election Systems, Election Systems and Software, Sequoia Voting Systems, and Dominion Voting Systems are notable examples.

Electronic Voting Machines

Source: ProCon.orgThe first fully electronic voting machines debuted in the early 70s, with Thornber Election System’s Video Voter and the Votronic, by Election Systems and Software. Adoption of these machines was slow, at first – the Video Voter only saw use for four years in Illinois – but its descendants capture a greater share of the voting machines marketplace every year. In 2012, 39% of American voters used an electronic or computerized voting machine. During the last year in which Video Voter systems were used, 1980, only 1% of the American electorate used electronic voting machines.

These machines, in theory, offer enhanced security, reliability, and audit capabilities as compared with older systems. Implementation has yet to live up to this promise, however, as the majority of systems in use produce no auditable paper trail and are built around proprietary system software.

As of 2004, the electronic voting systems most commonly used in America were:

  • AccuVote – TS (Election Systems and Software, formerly Premiere Election Systems (Diebold)
  • AVC Advantage (Sequoia Voting Systems)
  • Microvote DRE, MV – 464 (Microvote Corporation)
  • Votronic and iVotronic (Election Systems and Software)
  • Shouptronic 1242 DRE (multiple vendors)

It is very much an open question whether or not the widespread adoption of these machines will finally deliver on the promise of a secure, fraud-resistant and reliable voting system. While the potential of the technology is clear – and evidence of real-world errors or fraud scarce – open source code and paper trail audit features would do a great deal to further secure these units against modern-day ballot stuffing.