So it’s all over, and we know how it panned out. In a lot of ways, I think the Yes campaign deserved to lose: they were unable to tell a coherent story about why AV was preferable, and were somehow too incompetent to even counter the obvious falsehoods propagated by the No campaign. By dumbing down their own campaign so much, they even ironically played into the No campaigns bizarre claim that AV was too complicated. I was pleased to see that Oxford was one of the 10 regions that voted ‘Yes’—along with Cambridge, inner London, inner Edinburgh, and inner Glasgow. (Pretty striking correlation between expected educational attainment and pro-AV…) I’ve already said quite a bit about AV. But I wanted to make one final observation about AV itself, relating to a distinction that was not drawn during the campaign, and seems to me to lie at the heart of whatever appeal AV has.
Every voting system has two parts: it has a system for collecting certain specified information from individual voters, and it has a system for aggregating that information. These are two separate things. We can easily see this from considerations of the existing system. This system collects one piece of information from voters—their most preferred candidate. And it aggregates this information in the simplest way possible: it adds together the votes from different voters in a given region and whichever candidate scores highest is declared the winner for that region. But one could imagine a completely different aggregation system: one that collects together the votes from a number of regions (say 10), disregards the information about candidates except for their parties, and elects 10 representatives according to the proportions of most preferred votes their parties received. This would be a proportional system but based on the same voter information.
Electoral reformers have generally been concerned with the kinds of outputs produced. In particular, the question of proportionality—should the parliament be composed of party representatives in some sort of proportion to the votes those parties received?—was the central preoccupation of both the Liberal Democrats and the Electoral Reform Society before the referendum. As shown by the example I just gave, proportionality is easily achieved by aggregating votes in a different way to first past the post. And the No campaign was at least right on this: there is no guarantee that AV would be more proportional than first past the post, and may even be less proportional. AV might well have helped the Lib Dems, and given them numbers that were (in fact) more in line with the proportion of votes they received. But that proportionality of representation would not be the aim of AV, and it would be an accident of history (namely, that the third party is ideologically between the two main parties) that would yield this greater proportionality. If the two main parties were both centrist, and there was an extremist third party, AV would probably shut that extremist party out entirely as preferences flowed between the two centrist parties. Perhaps this bias towards the centre is in itself desirable, but it is hardly attaining proportionality in and of itself.
So what is AV for? It is as I conceive of it, primarily a reform to the information collected from voters, and not to the system of aggregation (indeed, both FPTP and PR can be coupled with AV-style ballot papers, with the appropriate aggregation methods). AV recognises that voters have preferences. It doesn’t fully recognise these preferences (since it doesn’t allow for indifference between options [except at the bottom of the ordering], a theoretically fundamental feature of preference orderings). It also doesn’t give those preferences enough structure to define a metric, which would be very useful in many useful systems of aggregation. But the fact remains that any alternative voting system collects information about preferences and also makes non-trivial use of those preferences.
This, it seems to me, is the main selling point for AV, a selling point that was almost entirely absent from the Yes campaign. AV allows you to express your preferences, and has a mechanism that allows the whole preference structure to have an impact on the result. Emphasis on the way that preferences are both noticed and important under AV would have made a much better story about why AV is good. It could have been a story which focussed entirely on the individual voter and their voice, rather than on outcomes. And it was the focus on outcomes which, given the tendency of AV to reward compromise, and people’s unwillingness to most prefer compromise options, ended up giving the No campaign many opportunities to emphasise over and over how AV might end up resulting in the selection of options ‘no one wanted’ (No campaign speak for ‘not most preferred’).
Think instead about ordering from a menu in a restaurant. You want the chicken; but if they’ve run out, as they sometimes do, you’ll have the pasta; and if not that, the fish. You don’t need to announce these preferences when the waiter comes around, since they only matter if indeed the chicken, etc., have run out. But they certainly exist; and in getting your meal, you definitely want them to be considered. The system which asked you to opt for a single dish, and have no say over what you might get if that choice is unavailable, is terrible in the restaurant choice situation, and is one good reason why restaurants, shops, and indeed almost every other situation when one expresses individual preferences allows one to express more about your preference structure than just what comes top. AV can tell a coherent story about giving voice to an individual’s real preferences—why the Yes campaign failed to do this is a mystery to me.
Of course collecting this information doesn’t tell you how to make use of it. The fact that more information is collected opens up many more possible ways to aggregate it too, which doesn’t help matters. But this shouldn’t impact on the decision whether to collect this information about voter preferences in the first place—unless of course one thinks one can’t tell an attractive story about aggregation of such rich data, whereas one can tell an attractive story about how to aggregate less rich data only about first preferences. Indeed, this looks like one area in which PR is more easily accommodated by the current system of balloting, since it is not at all easy to tell an appealing story about how the distribution of second preferences should influence the proportions derived from first preferences. (And one can see why—PR is, in effect, a system that uses voter’s first preferences to establish a proxy preference ordering amongst parties for the electorate as a whole; but that proxy preference of a group is not at all systematically related to the aggregate of individual preferences.) So I suspect that, if you want PR, you should have voted No—just as the group AV2001.co.uk advised. They want single transferable vote, a PR system, and they are refreshingly clear minded about how AV is not a step towards that system.
But I actually don’t care much about proportional representation. It doesn’t seem to me to be intrinsically desirable, nor is it guaranteed to result in the election of workable government. Insofar as government should largely involve compromise, having the information about preferences that AV provides seems to me to provide a better guide to which directions of compromise are desired by the electorate. But the democratic principle that one’s voice should be heard seems to me of paramount importance, and AV respects that principle far more than the current system does. It may not be clear what the sum total of those voices is saying, if indeed there is any sense in which those voices are collectively saying anything. But it is not a virtue of an electoral system that it produces a false impression that there is a coherent view in an electorate, simply because it doesn’t collect the information that would show there was no such coherent view.