Monday, 28 March 2011

AV – What are the Effects?

There’s been a lot of sound and fury on who’d win and lose from the proposed AV reforms (such as the No2AV claim that effectively Nick Clegg would be the One Man with the One Vote), and it occurred to me that we’ve not had the facts presented in one place.

Whilst some of the arguments over AV have been on the philosophy of the system, much of the undercurrent has been on the party political fallout. With most sides simply guessing.

As it turns out, most of the projections of past elections that have been bandied about over the years aren’t much stronger than educated guesses – Dr Roger Mortimer of MORI has an excellent summary of the AV issues here - which includes much of the facts behind the philosophical discussions as well.

The major issues with projecting these outcomes, he summarises thus:
- Most estimates of how an AV election would turn out assume that voters’ first preferences would be the same as their vote under FPTP. But that is almost certainly not true. The trouble is there is no simple way of finding out or reliably estimating how it will differ at constituency level. How many people are voting tactically (and where) who wouldn’t do so under AV? We know how votes are spread by constituencies at the moment, but we don’t know what the spread of AV first preferences will be.

- Surveys can ask voters what their second (and lower) preferences would be, but we don’t know how this will vary by constituency, which can make all the difference to the election outcome. We certainly can’t assume that, say, Lib Dems in Scotland will split between Labour and the Tories as their second choice precisely the same way as Lib Dems in the South East of England, or that rural and urban are the same.

- Most people don’t really understand alternative electoral systems, so asking them what they would do under a different system may not get accurate answers. Once a new system is in place and they have got the hang of it, they may well vote completely differently to the way they think they would now.

- Parties will behave differently under AV and that may affect the way voters think about them. e.g. If the Lib Dems asked voters in a constituency to give their second preference to the Tories, would that lose them the first preference vote of some left-leaning LibDem voters?
Given those caveats, it’s not surprising to find that some estimates of election results vary between well respected sources. Most are fairly close – I’ve found more than one result for each of the 1983-1997 elections, but all bar the 1997 one are very close to each other. The 1997 one is a pronounced exception, with Conservative seat totals confidently projected at 110, 96 and 70 from three different calculations.

The elections prior to 1983 don’t seem to have had projections made (at least not easily locatable online), but data from the British Election Study (BES) data is now online at the BES Information Site, with data going back to 1964. I’ve run an estimate for the 1979, 1974(October), 1974 (February), 1970, 1966 and 1964 elections.

Enough preamble, what’s the outcome?

Only two elections would have changed their headline result: the October 1974 election would have had Wilson still short of a majority, and 1964 would have seen Home denying Wilson a majority – and the Liberals under Jo Grimond as kingmakers. The reason that the third party has always done better in these projections is that they have (until now) always been preferred to the “other one” of the big two by Con and Lab voters – which may change now they’re in Government.
Who (out of Conservatives and Labour) loses out? It changes – dependant on who is least disliked by Liberal/Lib Dem supporters. This is how it’s happened over the past 46 years:

Net Preference for Con/Lab by Lib/Alliance/LD voters across elections (Source BES 1964-1997, 2001/5 ICM/BBC from Curtice paper, 2010 BES from paper by Sanders et al)

On the above graph, blue corresponds to an advantage for the Tories, red to an advantage for Labour. Current polling suggests that the remaining Lib Dems have a small net pro-Tory preference (that is, the line has crossed back over to the blue side), but 2015 is still a long way off.

Well, that’s what the past looks like – but the big issue with AV is how it will change voting behaviour in the future – with the ability to cast a first preference vote with far less worry about wasting it, will UKIP, the Greens, far left Socialist parties and so on see big boosts in their first preferences? Will we see constituencies where voters have long abandoned hope in their first preference major party suddenly shift sentiment? Subsequent elections will be based on the now-different results that AV will produce and we’ll enter uncharted waters. One thing that does leap out – more constituencies change hands under AV projections than under FPTP ones, so election night may prove rather more anxious for MPs and more entertaining for the rest of us.

Footnote for sources: For this article, I’ve taken the projected totals for the 1983-2010 elections (using John Curtice’s projections at the BBC website here as my starting point. The ERS present different totals for 1983-1997 here, again citing Curtice. The third calculation for the most disputed outcome (1997) is here, by Dunleavy, Margetts, O’Duffy and Weir. For the 2010 election, I’ve used the paper by Sanders, Clarke, Stewart and Whiteley (available here, but an alternative is presented by the ERS and presented by the Guardian here)


RodCrosby said...

It's all very entertaining, but once you retrospectively posit a change to the electoral system, the entire subsequent political history, including the personalities and dates of the elections, might well change. The only election worth forecasting is the next election...

Also, I think Sanders et al are confused regarding their AV variants One and Two. I can't see any practical difference.

My Burning Ears said...

I didn't think this deserved relegation to Channel 2! Good stuff Andy.

One claimed benefit of AV (and logically it sounds reasonable) is that it promotes centrist politics at the expense of extremes. Does this analysis back that up in any way?

[I suppose it does if you accept that Libs/Alliance/LD voters were historically "centrist", and the simulated results seem to have given those voters more say in 2nd-preffing against what is arguably the more "extreme" party of the time, but that's perhaps begging the question. Plenty of left-wingers would say Thatcher was more extremist in the 1980s even than the Bennites were. From the perspective of Alliance supporters, that's obviously not true! But I'm not sure it's fair to let them be the final arbitrator of that.]

Andy Cooke said...

MBE, many thanks. I think it's possibly overly simplistic to claim that AV promotes centrist politics as many do - it depends on which centre we're talking about (radical->reactionary,communist->capitalist, liberal->authoritarian, internationalist->isolationist, tolerant->exclusive, etc) and what of the choices available seems to be the most repulsive option and what the least credible option. There are a number of more extreme potential policies that any of us can think of that would probably gain majority support in the UK today, after all. Overall, there would probably be a tendency towards net "centrist" policies, but certainly not a fixed rule.

It does put the lie to the "Thatcher only won due to a divided opposition" meme that too many have subscribed to for so long.

Of course it's impossible to extrapolate beyond any single election as the following situation would diverge rapidly from the existing history. This is intended to provide a "what if" to each election if that particular one was the first such AV election - as often the question is "if all else were equal, what would have happened in 19xx" - often with the implied subtext that said election would have been altered significantly merely by the incidence of AV (usually hinting that we'd have had PM Foot or Kinnock, or that the Tories could never win under AV, or that the Lib Dems would always be the kingmakers). In reality, the only way that Foot may have become PM is if the Labour Party leadership election of 1976 were held under FPTP rules.