The Efficiency of Elections... and All Processes

I am a citizen of three countries - US, Canada, Israel, although with my travel I should get a United Airlines passport as well - all inherited. It has given me insight into various election methods, not only how one is elected - proportional representation vs ridings/districts, parliamentary vs executive branch, government-financed vs private-financed, etc. - but also in how one actually casts a ballot. In the end, whatever your system (assuming it is open and democratic in nature), each citizen needs to make their selection known.

The United States, which voted 2.5 months ago, is most interesting, since each State has its own voting system; the US is, after all, a federation of independent States (hence, the "Federal" system and "Federal" government), including elections. In most elections I have seen in the US, there is some form of voting machine, in which the voter flips some switches to pick their vote, and then pulls a lever to cause a small cardboard paper to be punched correctly and then drop into a box. At the end of the day, each district counts the votes in its box, usually with representatives from each major party verifying each vote, and adds up the results.

More recently, several states have tried electronic voting. Although this is partially an efficiency play - counting hundreds of millions of votes by hand is terribly slow - for the most part, it is more of an effectiveness play. After the debacle of the "hanging chads" in Bush vs Gore in the 2000 elections, people felt that the physical card system was not so much inefficient as likely to lead to errors. Since electronic voting booths have no chads, no physical cards at all, it is impossible for them to have physical border errors. Put in other terms, the real world is analog, has lots of continuous states, and thus a chad can be punched through, still in place, a little perforated, and anything in between. Digital systems come down to a binary 1 or 0: either it is true or false.

Israel, which is voting today, has a very quaint slip and envelope system. The voter receives an envelope from the voting officers, and then goes behind a booth. In the booth there is a box with stacks of slips of paper. Each slip represents one party. The voter takes the slip for the selected party, puts it in the envelope, seals it, then leaves the booth and puts the sealed envelope into the voting box (looks a lot like a suggestion box) in full view of the voting officers, to prevent multiple envelopes.

Both physical systems seem very anachronistic, hardly something for the 21st Century. After all, I vote in shareholder elections, pay my taxes, and even register my kid for school in a digital fashion, often even from the comfort of my own home. Why do we use paper for one of the most important freedoms we have, the freedom to choose our own government?

As a result of this sentiment, there has been a lot of pressure from citizens to move towards greater electronic voting, even if only in polling stations (i.e. not online). It is important to understand what this *does* solve and what it does *not*. At heart, a voting system tries to perform the following:

1. Collect a single vote from each person - once a person has voted, they cannot vote again
2. Ensure the vote is secret - once a vote is cast, no one should be able to connect that vote to that person
3. Ensure the vote is accurate - if there are 400 actual votes for party A in a district and 200 for party B, then the total count should be 600, showing 400 for A and 200 for B
4. Resolve uncertainties - if a vote is unclear, then resolve it according to local rules

Both physical and digital systems can handle the first 3 of the above, and the digital system is more efficient, and slightly more accurate at the 1st and 3rd.

1. Single vote from each person - the physical system has each person go only to their district. After they register, their name is marked off on a physical list, so that if they come a second time, they are refused entry. A digital system can easily do the same, more efficiently, and with greater accuracy. After all, a voting official can accidentally mark off the name below mine, or misread it when checking if I have voted yet, whereas the digital system is less likely to do so. Note that digital systems also have margins of error, but they are far far smaller than human.
2. Ensure the vote is secret - the physical system does so either by having an machine that punches an anonymous card behind the booth, or by having an unmarked envelope placed into the box. The digital system can do so as well, but it is harder. It is not a challenge to have a single entry into a database that is digitally signed. But digital systems by their nature keep logs. A smart technician can match up the timestamps of vote entries with the timestamps of voters and get a pretty accurate picture of who voted for whom. There are technical solutions to scramble the results, but this is not the point. More importantly, the digital system does not have the perception of secrecy, which is at least as important as secrecy itself. If users do not believe their votes secret, even if they are, they will not trust the voting system as much. The need to believe in something's capabilities more than the actual capabilities is something inherently understood by every marketer, much less so by the engineers who build the system.
3. Ensure the vote is accurate - The physical system has actual humans count each piece of paper, usually with 2 or even 3 people verifying each vote. Nonetheless, the vote is subject to human errors and, even with multiple onlookers, fraud. The digital system is far more accurate and less likely to be subject to fraud. It is in this one area that the digital system outshines the physical system. However, it is important to note that both systems are subject to error. No system is 100% accurate - in elections, manufacturing or any process. It is simply a question of how large the error rate is. In general, automated systems have lower error rates than human systems, but not always.
4. Resolve uncertainties - here is the rub. Every system, however accurate or inaccurate, however secret or public, however precise or imprecise, will have areas of uncertainty, a physical or digital version of a hanging chad. Automated and digital systems do not eliminate uncertainty. In the physical world, the 3 or 4 voting officers can come to a consensus on what to do with an unclear vote, or, if they cannot agree, can agree to disregard the vote. In digital systems, there often is no voting officer. As such, the system itself might be making default decisions that humans would not have made, e.g. in the 2012 US Federal elections. Simply defaulting to one party or another, or discarding all uncertain votes, can have unintended consequences on the election, especially close ones. Further, rule #2, the secrecy requirement, makes following up to validate the uncertain vote impossible. Thus, unless and until the human voting officers can be inserted into the uncertainty resolution process, or the human decision process replicated for every possible uncertainty, from which technology is a very long way away, the digital machine will perform poorly compare to the physical.

Digital and automated systems do not solve every problem. Some, like accuracy and gating (1 vote for 1 person) it can perform markedly better than physical and human systems; others, like secrecy, it might perform similar to humans, but with a possibly high perception cost; and some it does materially worse. In all of the above, it will do it far faster and more efficiently, but as we know, automating a bad process does not make it into a good process; it makes the bad process fail faster.