To AV or to AV not?

We go to the polls in the UK tomorrow, in a national referendum to decide whether we should change the voting system we use in parliamentary elections. The choice is between our current system, First Past the Post (FPTP), and the Alternative Vote (AV) system. There’s a good, non-partisan explanation of the alternatives here, but briefly, under FPTP each voter gets a single choice, and the candidate with the most votes wins – even if they receive rather a low percentage of all votes cast. Under AV, voters can rank candidates according to preference, and second preferences are counted until one candidate has at least 50% of votes. Proponents of AV system argue that it will result in fewer ‘wasted’ votes – so I could vote, say Green, but have Labour as a second preference to indicate that I would rather they got in than the Conservatives. Opponents balk at the idea that someone can win the first round (i.e., get the most votes), but not get elected. I can see pros and cons in both systems, but the debate has descended into pettiness and misinformation, particularly on the part of the ‘No’ (No to AV, that is) campaign.

In particular, the No campaign have used a range of sporting analogies to suggest that it would be ridiculous in a race for someone to cross the line in first position, yet not be declared the winner. My gut feeling was that this is actually unlikely to happen very often in practice, but I hadn’t seen any data from countries which use AV, to back this up.

The Yes campaign have, however, used Australia as an example of a country which has used AV for years, without any general trend for less stable governments (coalitions, hung parliaments, etc.) than we’ve had in the UK under FPTP. But what I wanted to know was, how often does a candidate finish second or lower based on first preferences, and end up getting elected?

Turns out, there’s a ton of data easily available to look at this. I found information on the 1998 federal election which stated that “99 of the 148 electorates in the House of Representatives required the distribution of preferences. In 7 of these seats… the candidate who led on primary votes lost after the distribution of preferences.” So, in only 5% of cases did the caricature of the No campaign, i.e. a ‘loser’ winning, actually happen.

I had a look at the 2010 election too. The definitive data are available to download here, for someone to do a thorough job of this – if, for example, they were employed on a campaign in favour of AV, say. I have neither time nor inclination to do a proper job, so I relied on the ABC report of the election results.

Looking through each of the 150 seats, I counted 139 (93%) in which the candidate who lead on first preferences, ended up winning the seat. Of those 11 seats where the first preference winner lost on the second preference count, 10 were won by the candidate who came second on first preference. The remaining seat was won by the originally third-placed candidate. Excluding this 11th seat, the candidate who ended up winning was on average about 3.1% behind after the first round (range 0.1-9.5%); including all 11 seats, this changes to 3.7 (0.1-14.5).

So, in general, Australia offers little evidence that candidates who don’t come first based on first preferences will often end up winning seats, except in a (very) few pretty close-run races. Of course, AV may change voter (and campaigning) behaviour in all kinds of ways, some of which may not be desirable. But to base a campaign on such a straw man as ‘losers’ winning, is pretty disingenuous.

In the end, It looks like I will end up making my decision on this issue based on my distaste for a misleading and negative campaign, rather than through any great enthusiasm engendered by a campaign for positive change.