If you want a preferential voting system, at least make it preferential

I feel like the kind of pedant that will jump on you for saying ‘less people’, rather than ‘fewer people’; or ‘something I like the sound of’, rather than ‘of which I like the sound’. ‘AV isn’t really a preferential voting system’, I say. Well, yes and no, as it were.

It is the case that, in AV, voters list their candidates ‘in order of preference’. But what does that mean? It doesn’t mean, as you might expect, that if no one’s first preference wins a majority, then everyone’s second preferences will be counted and taken into consideration. Only the second preferences of voters for eliminated candidates are counted, meaning that the second preferences of a majority of voters – i.e. those who voted for the two leading parties – are not even looked at. I don’t know about ‘preferential’; that’s a bit more like giving some voters preferential treatment over others!

So what is needed is a system that treats everyone’s preferences equally. Such a system does exist and is called ‘Bucklin voting‘, which I’ve discussed elsewhere: if there’s no majority on first preferences, every voter’s second preference is counted and added to the candidates’ totals. If there’s still no majority, third preferences are added; and so on till there is a majority for someone (or more than one majority, in which case you take the largest as the winning total) or until the preferences run out and the winner is the candidate with the highest total of votes.

The trouble with Bucklin is that it violates the so-called ‘later no harm’ voting criterion, which says that your lower-preference votes should not be allowed to harm the prospects of your higher-preference votes. For example, if a Tory voter put the Lib Dem candidate down as their second choice, this could help to elect the Lib Dem; whereas if (s)he and other Tory voters hadn’t voted Lib Dem as their second preference, the Tory candidate might have won.

I think you could get over this obstacle by, paradoxically, making it compulsory to list all or at least, say, five candidates in order of preference. Then you could say to voters: ‘Your first preference should be the candidate you most want to win; your second preference should be the candidate you would second-most like to win; and so on until your lowest-ranked candidate is the one you least want to win. The higher you list a candidate in order of preference, the more likely they are to win’. The fact of being forced to select candidates from the most preferred to the least preferred outcome ironically makes it easier to order your preferences ‘sincerely’ without feeling personally responsible for handing the victory to a less preferred candidate. ‘Well, if I have to list five candidates in order of preference, I might as well try and get the best result for myself’.

There would still be tactical voting, but I don’t actually think that’s such a bad thing if tactical voting enables voters to secure a better outcome for themselves. For example, a Tory voter might want to vote for the UKIP candidate as the one they’d second-most want to win. But then, they might think that if they did that, they could let the Labour candidate win (based on the second preferences of Lib Dem and Green voters). So they might decide to put the Lib Dem candidate down as their second choice; and if that candidate won, at least they’d have the satisfaction of having prevented a Labour victory.

It’s a gamble; but almost all single-member voting systems involve an element of that. And at least, this compulsory-ranking version of Bucklin voting allows all voters to put down their actual preferred candidate as No. 1 without fear of wasting their vote; and the tactical vote – if people choose to vote that way – can be reserved for the subsequent preferences. By contrast, it is possible to vote tactically under AV (despite what the Yes camp says), as I’ve discussed elsewhere; but it’s more difficult to work out what to do, and this could paradoxically produce a less satisfactory result for the electorate as a whole.

So the version of the Bucklin system I’m proposing is what AV purports to be – a preferential system – but does it better in that the preferences are all treated equally and so really mean something.

An academic question? Maybe; but like English grammar, I’d rather it was done proper.