Thursday, 25 July 2013

The Late July Opinion Polls - Testing FPTP To Destruction

One thing I have been doing is analysing the opinion polls produced by YouGov. I have given an outline of how I analyse it, which can be summed up as rather than looking at the headline figures, examine how the vote has shifted between parties and take into account regional variations.

Now, this approach had some problems, and it became reasonable to make an assumption on how many Conservative and Liberal Democrat voters in each constituency had switched from Labour since the June 2001 general election. The modification meant that these (the "switched vote") would be the first people to switch back to Labour, and if necessary, we would then dig in to the rest of the vote (the "retained vote").

Based on recent opinion polls (which I come to in a minute), we get all of the Liberal Democrat switched vote switching back to Labour, along with some of the retained vote (around 20% in Englamd, 24% in Scotland and 6% in Wales). Basically, everyone who was a Labour voter in 2001 and a Liberal Democrat voter at the May 2010 general election is assumed to be voting Labour again, along a significant number of Liberal Democrat 2010 voters who did not vote Labour in 2001.

And in England and Wales, only part of the Conservative switched vote (around 47% in England and 32% in Wales) is switched back to Labour, while in Scotland all of the Conservative switched vote - along with around 1% of the Conservative retained vote - goes to Labour. Basically, in both England and Wales a majority of people who voted Labour in 2001 and Conservative in 2010 are sticking with the Conservatives, while in Scotland eveyone who voted Labour in 2001 and Conservative in 2010 is switching back to Labour.

If I have time, a later modification could be working out approximately how many people per constituency swtiched from the Conservatives or Liberal Democrats in 2001 to one of the other main parties in 2010.

The result of the 2010 general election is:

  • Conservatives - 306 (including 1 Deputy Speaker)
  • Labour - 258 (including 2 Deputy Speakers)
  • Liberal Democrat - 57
  • Northern Ireland parties - 18
  • Scottish National Party - 6
  • Plaid Cymru - 3
  • Greens - 1
  • The Speaker - 1

I will assume that there is no change of Speaker of the House of Commons, so John Bercow would be re-elected in Buckingham as Mr Speaker seeking re-election.

Although I use the YouGov polls, there are 5 of these a week. In order to get a bigger sample size, I have used the past week's:

Date C Lab LD UKIP
Friday 19 July 33% 38% 11% 11%
Sunday 21 July 32% 39% 10% 11%
Tuesday 23 July 35% 38% 11% 10%
Wednesday 24 July 32% 39% 11% 12%
Thursday 25 July 35% 39% 8% 11%
Average 33.4% 38.6% 10.2% 11.0%

Now, much of the focus has been on the Conservatives closing the gap on Labour. But what is also interesting is the Liberal Democrat vote - up till now, the pattern has been for a plurality of the Liberal Democrat 2010 voters to declare their support for Labour. The Tuesday and Wednesday polls implied that this effect has been reversed - but today's poll disagrees.

All this gives us the following:

  • Labour gains 89 seats from the Conservatives, 23 from the Liberal Democrats and 1 from the Greens - a net gain of 113 seats
  • Plaid Cymru gains 1 seat from the Liberal Democrats
  • Kidderminster Hospital & Health Concern gains 1 seat from the Conservatives
  • The Greens lose 1 seat to Labour
  • The Conservatives gain 33 seats from the Liberal Democrats, but lose 1 to Kidderminster Hospital & Health Concern and 89 to Labour - a net loss of 57 seats
  • The Liberal Democrats lose 1 seat to Plaid Cymru, 23 to Labour and 33 to the Liberal Democrats - a net loss of 57 seats

This gives us an overall result of:

  • Labour - 371 (including 2 Deputy Speaker)
  • Conservatives - 249 (including 1 Deputy Speaker)
  • Northern Ireland parties - 18
  • Scottish National Party - 6
  • Plaid Cymru - 4
  • Kidderminster Hospital & Health Concern - 1
  • The Speaker - 1

The thing that stands out is that the Liberal Democrats and Greens do not do well enough to keep any of their seats, and the UK Independence Party fails to make a breakthrough.

The only seat where the Liberal Democrats would achieve over 30% of the vote would be 30.2% in Twickenham, currently held by the Business & Innovation Secretary Vince Cable - a seat that would fall (narrowly) to the Conservatives.

The Greens' best result would be, unsurprisingly, in Brighton Pavilion with 29.7% of the vote. But this is a seat where defeat would be narrow.

There are 6 other seats where the Greens get over 10% of the vote - Edinburgh East (Lahour hold); Norwich South (Labour gain, with the sitting Liberal Democrat MP, Simon Wright, pushed down to fourth place); Glasgow North (Labour hold); Ross, Skye & Lochaber (Labour gain, with sitting Liberal Democrat MP and former leader, Charles Kennedy, pushed into third place behind the Scottish National Party); Herefordshire North (Conservative hold); and Edinburgh North & Leith (Labour hold).

Out of these, the Greens only come third in Edinburgh East (with the Scottish National Party second) and Norwich South (with the Conservatives second). In Glasgow North they are fourth (with the Scottish National Party second and Liberal Democrats third - although there isn't much between the Liberal Democrats and the fifth-placed Conservatives). In Ross, Skye & Lochaber they are fifth (with the Conservatives in fourth place, and even then the battle between the Greens and UKIP for fifth place looks close). In Herefordshire North they are fifth (Labour comes second, UKIP third, and the Liberal Democrats are only just ahead of the Greens). And in Edinburgh North & Leith they are fifth (with the Conservatives second, and the Liberal Democrats only narrowly ahead of the Scottish National Party).

There is another seat where the Greens are third - Aldridge-Brownhills (Conservative hold) - but this is one where they are just ahead of the Liberal Democrats.

Norwich South falls to become the Green's eighth target seat. Brighton Pavilion is first, of course. Second would be Ross, Skye & Lochaber (notwithstanding their fifth place). Third would be Gordon - a fascinating seat which Labour would gain, pushing the sitting Liberal Democrat MP Malcolm Bruce down into fourth place. However, it would be the second closest three-way marginal in the country (a Labour/Scottish National Party/Conservative one) after Aberconwy (Labour gains this from the Conservatives and it becomes a Labour/Conservative/Plaid Cymru marginal).

Despite being their third target, Gordon - like Ross, Skye & Lochaber - would be a seat where the Greens are fifth. Incidentally, Gordon would become the British National Party's top target seat, despite them coming sixth and losing their deposit.

The Greens' fourth target seat is Argyll & Bute, which is the fourth closest three-way mzrginal (a Labour/Conservative/Scottish National Party one) where the sitting Liberal Democrat MP, Alan Reid, is pushed to fourth place, and the Greens come fifth.

The Greens' fifth target seat is Brecon & Radnorshire - a seat the Conservatives gain with the sitting Liberal Democrat MP, Roger Williams, down into fourth place. Here the Greens come sixth, with Labour second, Plaid Cymru third, and UKIP fifth. The 20.1% gap between first and sizth is not the closest - that achievement goes to Ross, Skye & Lochaber, with a 16.6% gap between Labour and UKIP.

The Greens' sixth target seat is interesting, as it is Sheffield Hallam, currently held by the Lord President of the Council and Liberal Democrat leader, Nick Clegg, who is pushed into third place as Labour gain this. The Greens would be fifth, with the Conservatives second and UKIP fourth.

There are only 3 seats where UKIP get over 20% of the vote - Fylde (Conservative hold); Skipton & Ripon (Conservative hold); and Penrith & the Border (Conservative hold). These are among the 325 seats (half the House of Commons) where UKIP come third, with that party coming second in a further 23.

Just below 20% of the vote are a couple of interesting seats - Harrogate & Knaresborough, a Conservative hold where UKIP come second; and Boston & Skegness, which the Conservatives hold with Labour second and UKIP moving up to third.

Boston & Skegness has been in the news today connected with UKIP, as The Times suggests that it is a seat that UKIP leader and Member of the European Parliament for South East England, Nigel Farage, might contest (€*).

[* As is common these days, using a monetary symbol in brackets in a link indicates it is behind a paywall]

If Farage would prefer to contest a seat in his region, then two choices spring to mind. The first is the seat in South East England where UKIP would get the highest share of the vote - New Forest West - or their top target seat in South East England (and seventeenth overall), namely Portsmouth South, currently held by Mike Hancock, who resigned the Liberal Democrat whip last month. This would - despite being UKIP's top target seat in the region - become a three-way Conservative/Labour/Liberal Democrat marginal, with UKIP in fourth place.

UKIP's top target seat would be one of the most interesting seats - Fife North East, currently held by the former Liberal Democrat leader Ming Campbell. Here, like elsewhere, it appears from the polls that the voters will be merciless to the Liberal Democrats, with this seat the third closest three-way marginal and the closest four-way marginal (Labour/Conservative/Scottish National Party/Liberal Democrat) with UKIP obtaining a good fifth place.

Their second target seat is also in Scotland, this time Orkney & Shetland, where Liberal Democrat MP and Government Deputy Chief Whip, Alistair Carmichael, is narrowly defeated by Labour, with the Scottish National Party third, and UKIP in fourth place just ahead of the Conservatives.

UKIP's third target seat is the aforementioned Sheffield Hallam, and fourth target seat is Devon North. Although the Conaervatives win this, sitting Liberal Democrat MP Nick Harvey is the runner-up. UKIP are fourth, just behind Labour.

There are successes for minor parties and others. Independent/Save Our Green Belt would come second in Castle Point (Conservative hold). People's Voice for Blaenau Gwent would come second in Blaenau Gwent (Labour hold). There would be Independents coming third - Khizar Iqbal in Dewsbury (Labour gain from Conservatives, with the Liberal Democrats and British National Party just behind Iqbal); Ian Womersley in Hemsworth (Labour hold); Murdo Murray in Na H-Eileanan An Iar (Scottish National Party hold); and Philip Howe in Rhondda (Labour hold).

The British National Party would come third in Barking (Labour hold); Batley & Spen (Labour hold); Dagenham & Rainham (Labour hold); Jarrow (Labour hold) and South Shields (Labour hold).

Meanwhile, the Liberals would come third in Liverpool West Derby (Labour hold).

Respect - The Unity Coalition come second to Labour in Birmingham Hall Green, and third (behind Labour and the Conservatives) in both Bethnal Green & Bow and neighbouring Poplar & Limehouse. You will notice that I make no mention of Bradford West, which appears as a Labour hold, although Respect's George Galloway won a by-election here.

Not taking by-election results into account is also why Eastleigh doesn't appear high on the list of UKIP targets (it's number 420 on the list) despite UKIP coming second in the by-election. This appears as a Conservative gain, with Labour second, Liberal Democrats third and UKIP losing their deposit in fourth place. I suppose that some analysis of connections between by-election results and the following general election could provide some modifcations to take into account when there has been a by-election.

With losing deposits, you will recall that I mentioned the BNP's top target seat is Gordon, where they come sixth and lose their deposit. How can this be?

One thing to notice how there isn't all that much correlation between share of the vote and how high a seat is on a target list. In the days of two party politics, then there would be a direct link. But now we are in the world of multi-party politics. You may be 5% away from winning a seat where the winner got 25% - but you can be sure that there are other parties which could be closer, or not much far behind you, and also targetting it. You may be 10% away from winning a seat where the winner got 45%. You know that there it's a battle between you and them only.

Especially in Scotland and Wales, where there are nationalist parties, there are seats being won on a low share of the vote. When I saw some of the constituency results for the 2010 election, my thought was that they were random - they could easily have gone in another direction, or indeed a third direction. Small shifts in vote in one constituency can determine which of three (or more) parties win it. And the small (local issues) can influence the big (national destiny).

If a seat can go one of four, or even five or more, ways based on small changes in votes, there is that randomness and unpredictability.

One way around it is switching to the Alternative Vote, which was rejected in May 2011 in a referendum. Late last century, the Jenkins Commission produced a report which called for AV+ - basically a form of the Additional Members System used to elect the Scottish Parliament, National Assembly of Wales and Greater London Assembly, with the constituency MPs elected via AV rather than First Past The Post.

There was a Note of Reservation which suggested that the constituencies in such a model should continue to use FPTP rather than AV. And referring to Winston Churchill's views it states:

He went on to describe AV as containing an element of blind chance and accident which would lower respect for Parliament.

And the Note of Reservation goes on to state:

In addition, as all experts on electoral systems have acknowledged, AV can operate haphazardly depending upon the ranking of candidates on first preference votes.

An example is provided on how the order in which candidates are deleted for coming last can determine the outcome, i.e. if candidate A got only a few more votes, she might have been kept in while candidate B got deleted, and this could have determined whether candidate C or D ultimately won.

Now, this is a serious objection and needs to be looked at. But when we move onto multi-party systems with MPs being elected via FPTP on less than 30% of the vote, and seats which could go one of three or more ways, then it is FPTP that has an element of blind chance and accident which would lower respect for Parliament. And secondly, there are cases where non-monotonicity could happen, but if AV means that a small chnage in lower preferences could make the difference between whether candidate C or D, both capable of commanding an absolute majority, is elected, then FPTP means that a small change in first preferences (and there are no others in FPTP) could make the difference between whether candidate W, X, Y or Z, all with minority support, is elected.

I gave examples of parties which failed to get any MPs, and so far the highest share of the vote for one of these has been 3.10%. We can, just about, live with that.

But what happens when two parties get over 10% of the vote and neither of them get any MPs? Surely even the most vocal supporter of FPTP is going to have questions about that.

If there are no Liberal Democrat, Green or UKIP MPs after the next election, then the case for electoral reform is stronger. And one thought on this - if we went to AV+, couldn't the AV part actually become redundant? By that I note that in Germany, the Free Democrats, Left and Green rarely win constituencies and rely on Land lists to obtain representation in the Bundestag. Is it not possible that the Liberal Democrats, Greens and UKIP would get out of the constituency part (with exceptions, e.g. the Greens in the Brighton area, and the Liberal Democrats across parts of the West Country, Hampshire, south west London and northern Scotland) and focus on winning regional MPs on party lists, leaving the majority of constituencies to be good old-fashioned Conservative/Labour battles?

No comments:

Post a Comment