As the hard-fought AV referendum draws to a close, James Heather reflects on both sides of the campaign. One of the most talked about points of difference is the alleged need for electronic counts under AV. While electronic counts are not used in Australia, and would not be needed here, it is still important to begin a wider debate on how much we value electoral accuracy.

Disagreements often bring out the worst in people. Many of the arguments on both sides of the AV debate have been dubious at best.

The No2AV campaign, largely driven by Conservatives who believe that their party would perform less well under AV, perpetuates the argument that AV encourages tactical voting. To anyone who understands how AV works, this is manifest nonsense: in almost all practical cases, one’s rational strategy under AV is to vote according to one’s true preferences, something that cannot be said of First Past The Post. But if nonsense helps the campaign, then let nonsense prevail.

The Yes to Fairer Votes campaign, on the other hand, heavily influenced by Liberal Democrats who believe that their party would gain more seats under AV, argues that we should vote Yes because the BNP want us to vote No. For a long time, the Yes web site had a large menacing-looking image of Nick Griffin on the front page, alongside the banner: “He’s voting No. How about you?”. There was no mention that Robin Tilbrook, the leader of the English Democrats and generally considered to stand rather further to the right than Griffin on the political spectrum, has made it clear that he will be voting Yes. Such selective propaganda techniques hardly make for coherent arguments.

There seems little common ground between the two campaigns. They disagree on whether AV is fairer than the current system; whether it means an end to safe seats; whether it is too confusing; whether it encourages tactical voting; and whether it produces unstable governments.

One of the key areas of debate concerns the cost of running an AV election. The No2AV campaign says that AV will require technology to assist in the counting, and puts the cost of this at £130million, based on counting equipment that it says is used in Australia. The Yes to Fairer Votes campaign describes these claims as “scare stories”, and says that machines will certainly not be needed.

The No2AV argument is beautifully misleading. Australia really does have vote-counting machines, as they claim, and they are an important part of the Australian electoral process – but they don’t use them for AV elections. They are used only for STV elections, which are an order of magnitude more complicated in terms of the tallying process.

But despite the smoke and mirrors, one point of consensus between the two sides does emerge: where elections are concerned, Technology Is Expensive. Whatever voting system we end up with, let’s not mess about with computers.

Are both sides right to defend pen and paper and manual counting? Or is there a place for technology in the democratic process?

Estimating the cost of either electronic counting or electronic voting is very difficult. In the case of e-counting, where paper ballots are scanned or manually entered after the polls close, scanning hand-written digits, as would be required for AV elections, is very much harder than detecting a single cross in a box. In Australian STV elections, this is done by either manual data entry or scanning and extensive manual checking. This is extremely costly.

E-voting options range from relatively cheap Direct Recording Equipment (DRE) interfaces, which simply log the vote and increment an internal counter, to more expensive and cryptographically subtle end-to-end verifiable voting systems like Prêt à Voter, which give much stronger security guarantees: no part of the system, including the voting terminal, ever learns the content of the vote; the voter is given a receipt that can be used later to check that the vote was not changed; but the receipt does not allow the voter to reveal the vote to a third party.

But even if we had a price list for all these systems, it would not help unless we also had an answer to the antecedent question: what is the value of the properties these systems provide?

It is certain that our current method of hand-counting does not give perfect accuracy: recounts always give different answers. Since AV tallying is essentially iterated First Past The Post tallying, a manual AV count must increase the error level. Suppose that roughly 1 in every 1,000 were miscounted, and that a new system could reduce this to 1 in 10,000. What is the cash value of such a system? Should we say that the error rates are already so small that they are inconsequential? Or should we consider that we disenfranchise those whose votes we miscount, and no price is too high for ensuring that votes are counted as accurately as possible?

End-to-end voting systems claim to prevent electoral fraud, by voters and officials. What level of fraud do we currently tolerate? Part of the problem, of course, is that no one knows: there are too many possible avenues of attack for undetected fraud. Suppose that a system could eliminate all ballot stuffing, coercion and count rigging. What price do we put on that? Do we believe that we are essentially running fraud-free elections already? Or is it worth paying the money to stop whatever fraud there may be? What, indeed, is the value of public trust in the electoral system?

One wonders how much the Florida debacle cost in the 2000 US Presidential election. The costs of the legal proceedings alone were vast; but are we even certain that the right result was eventually returned? What is the cost of running an election for the leader of the most powerful country in the world, and getting the wrong answer? How much is our democracy worth?

Please read our comments policy before commenting.