Back in the early years of this century, the UK was at the forefront of testing out e-voting for public elections. An extensive series of pilots were held and then … e-voting fell out of favour, because the pilots were not a success for a wide range of reasons. The issue still keeps on popping up, so having recently come across again what I wrote back in 2003 about those pilots, those lessons are worth restating. Here is what I wrote back in the summer of 2003. Luckily the last paragraph turned out to be wrong.
E-voting: triumph or disaster?
According to the e-voting industry, the e-pilots this May  were a triumph. “Everyone’s a winner” said the Athena Consortium, responsible for pilots in Swindon and Stroud. Stroud’s own press release talked of “E-voting success … The figures are a major success for the Council.”
Whilst one in five of the total votes cast in Stroud were cast electronically, overall turnout was down 7% from the last comparable elections. Letting people vote in a different way provided a choice which some took, but it didn’t result in more people voting overall.
It was a similar story elsewhere. Swindon saw its turnout fall by just over 1%. Across all of the e-voting pilots, turnout was unimpressive, being down slightly (0.4%) on the last comparable elections.
Given the failure of the e-pilots to boost turnout, it is not surprising that the Government and the Electoral Commission focused on other aspects of May’s pilots in their immediate reactions.
The Electoral Commission’s initial response has been to highlight the successes of all-postal pilots in raising turnout and to be largely quiet about the e-voting pilots.
The minister, Nick Raynsford, has been more forthcoming in praising e-voting pilots, though using the argument that they were a success because of the number of people who used them – regardless of the fact that often fewer people in total were voting – and glossing over the cost and security issues.
The more subtle version of the turnout argument, and the one which the government and Electoral Commission are increasingly using, is that e-voting and similar might not raise turnout now. But imagine in an even more wired world in 15 years time…
Well, maybe. But online banking hasn’t finished off phone banking. Phone banking hasn’t finished off postal banking. And none of them have finished off banking in person. The penny post was introduced the century before last and the postal service is still with us.
Even so, providing choice – a range of voting channels in the jargon – can’t be bad, can it? That depends on practicality.
In several of the pilots, very basic errors were made – for example in two of them there were no backup alternatives available for key pieces of IT equipment.
One of these was Sheffield, despite the city’s advantage of having also run an e-pilot last year as well. Indeed, the Sheffield pilot was littered with basic IT project management blunders. Two staff were made available by one of the contractors to maintain IT equipment on polling day – but they were not provided with any transport with which to get to the polling stations if a problem occurred!
One polling station never received its ISDN line for on-line checking of whether or not people had voted – so allowing people to vote twice on the day, once there and one at another polling station, without any checking taking place that would catch them.
It’s the future – it’s the back of an envelope
Presiding officers at polling stations were not all supplied with paper registers to use in case of IT failure – with the result that when an inevitable malfunction occurred in one polling station, staff had to improvise ways of trying to keep records of who was voting and, according to Richard Allan MP, the result was that they “literally used the ‘backs of envelopes’.”
Given these blunders, and others both in Sheffield and elsewhere, it is perhaps no great surprise that neither the Electoral Commission nor the Office of the Deputy Prime Minister (ODPM) have so far shown much interest in taking responsibility for the actual performance of the pilots on the ground.
But despite their experimental nature, the pilots are still about real elections – with real candidates, real voters and real decisions to be made. Is it really satisfactory for the public and candidates to have to go through an election which is being run so badly?
There are in theory two important quality controls on the e-voting pilots. The ODPM has drawn up a list of preferred suppliers for local authorities to work with and also has a say over the overall pilot programme each year. Yet these controls have manifestly failed – too many of the pilots in May were badly run.
It would be too easy to say that the lesson from the technical problems with May’s e-voting pilots is that for e-voting to work, we just have to get a bit better management. The pilots have rather exposed a much more widespread failure to manage IT projects successfully – across several councils and several contractors and with inadequate supervision from ODPM. These widespread flaws can’t be remedied by a brief set of exhaustions to get it right next time.
iMacs or brown envelopes?
There is no doubt that e-voting is rather sexier than the traditional ways of voting or even all-postal voting. A voting future which involves iMacs and websites is more eye-catching than one dependent on pencil stubs and brown envelopes. But that doesn’t mean it is any better, especially when the matter of security is added to the list of concerns.
The old IT joke is that the only secure computer is one which is switched off with all the cables pulled out. Very secure, but not very useful.
Despite the need for care, some of May’s e-pilots showed a naïve approach to security. One example was the Chorley pilot, where the statutory instrument passed by Parliament said that the ballot papers should have bar codes on the back which can be “read by an electronic scanner but not by eye.” Although it is an esoteric skill, reading barcodes by eye is possible – and no doubt there would be many training sessions at conference on how to read a bar code if this “security” system was to be widely used!
More generally, one of the most important requirements for an e-voting system the public can trust is to have a clear audit trail. To have confidence in the result, we not only need to know the result but also to be able to check how it was derived. It is rather like bank accounts – how many people would trust their bank if they were only told their balance at the end of the month, but never sent any statements or given information in-between?
However many of the e-voting systems piloted have very few, if any, audit trails that can be checked in practice and many of the involved have little interest in audit trails. For example, the Basingstoke returning officer stated that if there was a query over the number of votes the computers said were cast, he would only be willing to check the figures after he declared the official result and it became legally binding, only subject to very expensive legal action.
It would be like a bank telling a small business it was going to call in a loan and bankrupt it – but if you said it had got its figures wrong on the size of the loan it would only check them after you had been bankrupted rather than before.
Lib Dem agents save the day
Other major security problems at Basingstoke were only avoided by the intervention of the Liberal Democrat agent, Keith Watts. One concern with electronic machines is that someone may electronically stuff them with votes before polling begins. It is therefore important to be able to check that the machine is ‘zeroed’ before voting starts. Yet in Basingstoke the original plan was to let machines be used even if the machine refused to produce a zero print.
Similarly in Epping the Chief Executive was prepared to side-step the safeguards which normally come from the ability of agents and candidates to query a result before it is declared. In his case, he told a member of an ODPM study visit that he was prepared to declare results even without agents or candidates present.
In one ward in Epping, there were more spoilt votes than there were for one of the Labour candidates (just over 100 votes). Although the result was queried, and it was confirmed that there was roughly the same number of spoilt votes across each machine used in the ward, it is still very difficult to believe that there wasn’t a significant technical failure here, such as a collective configuration error.
It was nearly much worse. When Jon Whitehouse, one of the Lib Dem agents, went to a briefing session he was startled to hear that the machines were all set to be configured with a “none of the above” option on the ballot paper.
Never mind that the law does not allow this, nor that no-one had been consulted on this major change!
When Jon protested he got the machines changed, but it was only thanks to his last minute decision to attend the meeting (when another engagement was fortuitously cancelled) that the election did not go ahead with “none of the above” on the ballot paper by mistake.
As with many of the other e-pilots, the clear conclusion is that the pilot system is deeply flawed – letting far too many wrong or bad decisions through.
It is easy to slip into being too enthusiastic about e-voting – it sounds modern, talking about it attracts media interest (would journalists turn up to hear a minister talking about wanting a new envelope design for postal votes?) and there are plenty of companies with a sniff of big profits at a possible e-enabled general election making their case.
But what really have we seen so far? It’s expensive, it comes with extra security and fraud risks, many of the pilots have been embarrassingly basic errors and it barely raises turnout.
There are some benefits – e-voting systems can be designed to cope well with the needs of visually impaired people and some people certainly like e-voting. The downsides though are large – including the risk of technical failure wrecking an election, the security problems and the high costs so far.
Is the Government listening? Probably not so far – the ODPM’s press release on the pilot results only conceded, rather dismissively, “a few minor technical problems.”