My post on election campaigns not mattering now seems rather quaint, with the Lib Dems surging in the polls. Of course this is not completely novel, the idea of a 3rd party ‘surge’ was enough ingrained in popular conciousness for Spitting Image to joke about David Steel ‘feeling’ it.

Anyway what is perhaps becoming more clear is that even if the Lib Dems have the same percent of votes as Labour and the Conservatives, they do much worse in terms of numbers of seats. Most of the calculators put a 30% each election as something like Labour on 300 seats, Tories on 200 seats and Lib Dems on 100 seats.

I explained this on Tim Worstall’s site by:

It’s not so much the disposition of the seats but the FPTP system itself, which simply favours geographically concentrated votes. So say the Tories take 60 percent in all southern seats, Libs 30 percent and Lab 10 percent, and the reverse is true in the north. Nationally if equal no. of seats in north and south then vote share is Con 35, Lab 35, Libs 30. But Libs have no seats.

I wasn’t entirely sure if this was right, but I think it is essentially correct – the Lib Dems’ support is too widely spread, and they do reasonably well in most seats, not especially well in enough. Here is a chart of each party’s % share of the vote in the 2005 election, starting with each’s seat where they got the highest share of the vote in % terms, and ending with their lowest. So the 1st point on the chart is not a particular seat, but for each party the share of the vote they receivedin the seat where they received their highest share of the vote.

The box shows the share of the vote – 40% and higher – that typically wins you a seat. Of the 614 seats won by one of the three main parties, only 45 were won with less than 40% of the vote. Similarly in only 32 seats did a party get more than 40% and NOT win.

So taking the 40% line, one can see the Lib Dims get about 60 seats, the Tories 200 and Labour 350 or so – about what happened.

Now let’s assume the vote share – 35.3 Labour, 32.3 Tory and 22.1 Lib Dem in 2005 – becomes 29.9 Labour, 29.9 Tory and 29.9 Lib Dem – not wildly dissimilar to some recent polls.

Now the first thing to note is that the % share of the vote when a candidate will typically win will fall, to something like 37%. This is simply because the Labour and Conservative share has fallen, and we are in a three-way tussle.

But again we can see a good estimate of how many seats each party will get from where their line crosses that 37% line – the Lib Dems about 150, the Conservatives about 210 and Labour about 280. Now this isn’t quite what the uniform polls predict, that is because of a variety factors such as boundary changes since 2005, the particular makeup of some seats whereby Labour can win with a slightly smaller share than the Libs and so on. But it does show us why the Lib Dems fail to match Labour. And basically it’s because their vote is spread reasonably evenly, with still high % shares of the vote in the last 200 constituencies, whereas the Tory and Labour vote has collapsed.

This assumes a uniform national swing (UNS), so the Lib Dems have gained 8% of the vote nationally since 2005 and will gain 8% in every constituency. What they need to form a government is for that extra 8% nationally to be concentrated in seats 200-400, where it would win them the election.

I’m a great supporter of proportional representation, in particularly STV, and not a foul-weather friend like Gordon Brown (my election 2005 first thoughts were “55% of the seats on 36% of the vote though. Can we have PR?”). Thus I was quite surprised last night in Nick Clegg’s interview to see that he didn’t seem to be aware that a party could win most votes but not most seats. It was quite strange – he genuinely didn’t seem to understand Paxman’s point. I guess he might have been pretending not to understand, in order to ignore it, and I suppose it would hardly lose him votes.

Well if this is right, and it is the Telegraph, the policy is as expected, it’s not a support to marriage but a subsidy to not working. And it will be paid for by a tax on successful banks. What’s not to like?

I’d say Lord Carey has been duped.

Last night, Lord Carey of Clifton, the former Archbishop of Canterbury, said: “The recognition of marriage in the tax system is a long overdue restatement of the centrality of this institution to the common good of our society.

But that describes his entire life. A subsidy to marriage would be – e.g. – a £200 cash payment to married couples. This simply penalises the hard working couples.

The election of 2010 really came to life towards the end of the campaign, most observers noted as Labour closed the gap, ‘winning the campaign’. However it was said the whole month was a fascinating battle, with with each party winning on some issues and days and losing on others.

The most accurate pollster YouICMMoriulus had tabulated the ups and downs of each party’s fortunes and the narrative of key events was reasonably clear.

This was especially the case when one examined the Tories’ volatile lead over Labour.

The campaign started with the Tories on 40%, 10% clear of Labour on 30%, with the Liberal Democrats on 20%, and others, mainly the nationalist parties and UKIP, on about 10%. The Tories had the best start, gaining support from business leaders on their policy to scrap an increase in National Insurance, and by day 3 they had taken an 11% lead, easily enough for a Commons majority. This increased to 12% by day 5, as Labour struggled to defend their economic record (the blip on day 4 taking the lead to just 7% seemed attributable to a fleeting reaction to a gaffe by Boris Johnson on plans for a new runway at Heathrow).

Between day 5 and day 14 the main campaign story was a surge higher by the Liberal Democrats on day 11, and a move higher by the BNP (counted in others) on day 12, taking votes from the Lib Dems. The Lib Dems rise, to 24%, was short-lived and sowed the seeds of its own destruction. That day Nick Clegg announced he would work only with certain members of other parties in (what then seemed likely) a hung parliament. This seemed to backfire and the Lib Dems were back down to 19% the following day. The BNP’s rise in the polls was nipped in the bud by a hastily arranged ‘Rock against racism’ concert. Both parties flatlined after that.

By day 14, the Tories had kept their 12% lead and Labour were getting increasingly desperate. Yet that day is what the history books will term “Mandelson Monday”. Labour’s veteran campaigner produced pictures of a member of the Shadow Cabinet in a compromising position with women and drugs. That no-one had heard of the Tory didn’t seem to matter – next day’s polls showed the Tories down at 37% from 42%, Labour up to 34% from 29%, and the lead slashed to just 3% – enough to see Labour home with a majority.

The mood of euphoria didn’t last long, however. On Tuesday (day 15) the Tories showed that the pictures were a crude forgery. Their bounce was immediate and sustained, taking them up to 43% and a 13% lead by day 18, leading to talks of a 1997 or 1983 style majority. This too went down badly with voters, and the Tory lead subsided to 7% by day 22, although a wave of strikes on the railways and at the airports saw the lead back to 13% by day 26, May Day, with the Labour vote share down to just 28%, its worst of the campaign.

That was to be the Tories’ high point, however. It had been little noticed but economic optimism had been rising significantly since the release of initial Q1 2010 GDP data, and the Bank Holiday weekend was the busiest in the nation’s shops on record. Why this suddenly showed up in the polls remains a mystery, but by Tuesday, the Tories were down at 38%, Labour on 32% and the lead slashed to 5%. David Cameron’s advisers told him that Labour might even scrape a majority, especially given the marginals were looking less good.

Cameron consulted former leader Michael Howard, who argued that the Tories needed to put clear blue water between them and Labour. He was thinking about a crackdown on immigration. Cameron’s speech initially went down well, with the Wednesday morning polls showing the gap returning to 9%.

The impact was, however, short-lived. The Tories hastily arranged poster campaign, “Are you thinking what Michael is thinking” reminded voters of what they didn’t like about the old Conservative party, and by polling day the polls were showing just a 3% lead for the Tories, 37% to 34%, with Labour gaining directly and from the smaller parties, whose share fell to 9%. On that basis Labour would have a working majority.

———
Unfortunately for the pollsters the final result was Tories 40%, Labour 30%, Lib Dems 20%, and Others 10% and Cameron got his big majority. Everyone blamed the polls, but it was pointed out that such a volatile race was bound to be difficult. Academics argued over whether Cameron’s immigration appeal had worked, but people were too ashamed to admit it to pollsters. It was agreed the economic news had benefitted Labour, but not enough.

In fact, unbeknown to the pollsters, US search giant Giggle had developed a system which allowed them to know everyone in the country’s voting intention on each day of the campaign. The chart below shows this and their remarkable finding – the campaign made zero difference. Throughout 40% supported the Tories, 30% Labour, 20% the Lib Dems and 10% others. No-one changed their mind at any point.

——–

Apologies for the Andrew Roberts style fantasy, but I expect we’ll get a lot of this kind of ‘explanation’ of poll movements, and yet voting intentions are probably quite rigid (if not as rigid as in this example). The movement in the polls discussed was entirely a function of polling error, I simply ran an opinion poll on a UK electorate whose views were entirely fixed. As
Anthony Wells says:

Party support from a single pollster should randomly vary a couple of points in either direction from poll to poll (the lead will be even more volatile, since you’ve got random variation on two numbers).

This is the margin of error, which typically says something like 95% of the time the figures will be within 2-3% (above or below) of the actual figure (depending on how large they are) where the actual figure is the whole population’s voting intention. So all the movement in the polls was because they are polls, and the final poll was clearly a rogue, i.e outside that range. Here’s the chart of true public opinion and the polls.