28 June, 2004: Technological calamity update

[ Home page | Web log ]

Readers may remember my previous rant on the subject of electronic voting, standards for which are being agreed by the Council of Europe as we speak. Some two months ago I wrote to John Prescott to remind him of the problems with electronic voting systems which do not include paper receipts, concluding,

Although electronic voting may be an effective way to increase turnout and make it cheaper to run elections, neither of these aims is as important as making sure that the electoral process properly expresses the will of the people. I urge you to ensure that ODPM does not lend its weight to any proposal for electronic voting which does not require paper receipts. Although the current Council of Europe Recommendation is only a draft, without prompt action the draft may become final, and the Council's proposal -- which in other respects is sensible -- may be adopted for real elections, leaving them open to undetectable fraud.

I hope that I can rely on you to ensure that any electronic voting systems which are adopted in this country always issue paper receipts, so that they can be trusted by voters to accurately record their intentions.

I have now received a response, not from Mr. Prescott, who is presumably too busy supporting the Scottish fishing industry to answer letters from humble Britons such as myself. Instead one Michael Leah from ODPM answered my letter, writing,

I note that you have concerns about the security of these [electronic voting] channels. Issues surrounding the security of e-voting form one of the areas that we are investigating through [pilot projects in remote and on-site electronic voting.]

Recall, from my previous piece, that remote voting cannot be made secure. So the fact that the government are already doing remote voting trials is really bad news.

Nevertheless, Leah assures me that,

At this stage in our programme of pilots, we have found no evidence that would lead us to concluide that electronic voting (whether remote or otherwise) cannot be used without undermining the integrity of the election. In addition, the Electoral Commission has not found any evidence that the pilots have led to an increase in the incidence of electoral offences or malpractice.

Several points here:

Leah continues that,

... at this time, we do not think that it is necessary to insist on the provision of voter-verified receipts to prove that direct recording e-voting kiosks are recording votes correctly. Nevertheless, we remain mindful of this issue and the wider issues surrounding vote verification.

But voter-verifiable receipts are the only way to prove that `direct-recording e-voting kiosks' are recording votes correctly. Leah's statement is equivalent to,

We do not think it important that direct recording e-voting kiosks record votes correctly, but we might change our minds later.

This is very sad.

Leah concludes (emphasis mine),

We are not complacent on this matter and are well aware of the potential threats to the security of e-voting systems, which is why we are still proceeding carefully with our programme of carefully controlled pilots. We will not rush to implement to extend (sic.) innovations to our voting system without considering their implications in detail. However, if we are to put in place a modern electoral system relevant to all our citizens, then we have to explore and develop these new methods of voting.

So all hope is not lost.

(Here I attempt to segue inexpertly from one theme to another.)

We are frequently told that policy should be made on a cost-benefit basis: estimate the costs and the benefits of some course of action, and proceed with it if the ratio of benefits to costs exceeds some threshold -- perhaps 1, so that we follow any policies with a net benefit, or perhaps exceeding the ratio for alternative policies. I don't know whether the electronic voting enthusiasts in ODPM are determining policy on this basis, but let's suppose that they are.

Electronic elections are supposed to be cheaper to run than existing paper elections, and to increase turnout over its current low values. (Political commentators sometimes argue that low turnout is a result of dull politicians, uninspiring policies, and a general apathy to the concerns of our existing elected representatives; but perhaps in fact it is the result of an overcomplicated and inconvenient electoral system which requires electors to leave their sofas and travel sometimes as much as several miles from their homes in order to put a pencil mark on a piece of paper. Well, I'm sure we'll find out.)

These, then, are net benefits which are realised over the life of the system. The costs include the price of the hardware and software, and the expense of training electoral officials and so forth. No doubt (assuming that the electronic voting systems we wind up with work at all and are not humungous Capita-type cockups) interested parties will be able to show that the benefits, in the long run, exceed the costs of the system.

But what about the risk of fraud? It's pretty hard to fit this into the cost-benefit model, because electoral fraud is likely to be fairly rare. Costs and benefits are usually discussed in terms of their expected or mean values. While it's fairly reasonable -- if you think increasing turnout -- to put a value on increased electoral turnout (``We will fund advertising campaigns to reach voters if they will cost less than 1p per additional voter''), it's much harder to do this with the risk of rare electoral fraud. As a society we obviously put a fairly high value on the honesty of the electoral process (you can be gaoled for tampering with an election, and occasionally individual disputed elections are re-run, so clearly the `cost' of electoral fraud exceeds the cost of running an election) but it's pretty hard to price the damage done by an electoral system which left it impossible to determine whether fraud had taken place or not. (As an aside, the Electoral Commission has done some research on attitudes to electoral fraud which might be of interest to some readers.)

The same issue arises in discussion of `carbon taxes' designed to mitigate the effects of climate change. You quite often see the statement that a tax of (say) 20 per tonne of carbon burned should be levied in order (a) to repair damage done by more severe weather, rising sea levels, and whatnot; and (b) to encourage users of energy to be more parsimonious with fossil fuels. Looking at (a) in particular, the idea is to take models which predict the effects of climate change, use them to predict the amount of economic damage done for different amounts of global warming, then linearise them in order to estimate the damage done `by' each tonne of carbon and levy a tax of that amount. This is fine as far as it goes, but unfortunately it's not really a very good model of the effects of climate change.

It's expected that global warming will result in two broad types of damaging effect: gradual, incremental effects such as more frequent severe weather and more serious flooding as a result of slightly higher sea levels; and relatively unlikely but potentially very serious `discontinuity' events, such as having the West Antarctic Ice Shelf melt and sea levels rise by ten meters or more or the thermohaline circulation in the north Atlantic shut down, leaving Europe with the climate of Nova Scotia. The cost of the damage from such events could be very large indeed, but we don't know how likely they would be. To pick random figures out of the air, suppose the probability that global warming will trigger severe cooling in Europe is 1 in 1,000, and that the damage that will be done by this corresponds to 1,000 per tonne of carbon. (The precise figures don't matter: it's just that there is a small probability of a large loss.) That would change the proposed `carbon tax' by only 1, well within the range of estimates proposed; but very few people confronted with the two possibilities, `the weather becomes worse', and `the weather becomes worse plus there's a risk that western Europe might turn into a giant icebox' would regard them as almost equivalent, as the expected-damage model does.

Similarly, it's unlikely that `elections remain as costly as they are now, with low turnout', and `elections get cheaper and more people vote, but there's a risk that catastrophic and undetectable electoral fraud could take place' are alternatives comparable on strictly money terms.

Copyright (c) 2004 Chris Lightfoot; available under a Creative Commons License.