Time’s arrow and the pre-election polls
This post originally appeared on Progress Online.
Imagine, for a moment, that there was no Fixed Term Parliaments Act. How would we be assessing the political situation in the late spring of 2014? By past standards, we are very late in the parliament, and everything would be seen in the light of election timing. Would the prime minister go for it in June, October or take a punt on delaying until 2015?
We can be reasonably confident that May 2014 is a year away from a general election, because the term is fixed and the coalition agreement precludes an early end to this parliament. But this is very different from a lot of previous parliaments; it was impossible to tell in May 1978 or June 1986 that there was a year to go until an election. The date of the election was itself partly determined by the state of the polls in the late midterm.
We often tend to assume away this uncertainty and take the election date as given, and work back from it. Leo Barasi has analysed the last years of parliaments. Steve Fisher has produced a statistical model based on pre-election periods, with interesting – and, for Labour, depressing – results, but I wonder how much can be read into them. Let us, for a moment, look at the matter from the other end of the telescope. It will take a better statistician than I am to see how much it will affect election predictions; I suspect it would, but at this stage I am just asking the question.
The following table shows the state of the polls 46 months into a parliament, the point at which many prime ministers will have been pondering whether to have a general election or not, and the point for which we have the most recent full set of polling data in 2014.
|Month 46 voting intention||Con %||Lab %||Lib/ LD %||Opposition lead %||Election timing||Winner||Opposition margin at GE %*||Swing back to Gov’t %|
|May 1949||46||40||11||+6||Feb 1950||Government||-3||+4.5|
|March 1959||46||47||7||+1||Oct 1959||Government||-6||+3.5|
|August 1963||35||49||15||+14||Oct 1964||Opposition||+1||+6.5|
|Jan 1970||48||41||8||+7||June 1970||Opposition||+3||+2|
|August 1978||44||47||6||-2||May 1979||Opposition||+7||-4.5|
|March 1983||41||29||28||-12||June 1983||Government||-15||+1.5|
|April 1987||41||31||26||-10||June 1987||Government||-12||+1|
|April 1991||41||40||15||-1||April 1992||Government||-8||+3.5|
|Feb 1996||27||54||16||+27||May 1997||Opposition||+13||+7|
|March 2001||31||50||13||-19||June 2001||Government||-9||-5|
|April 2005||33||38||21||-6||May 2005||Government||-3||-1.5|
|March 2009||42||30||17||+11||May 2010||Opposition||+7||+2|
Source: Mark Pack’s polling series simply averaging the voting intention polls for that month. For General Elections UK margin before 1974, GB thereafter
There are two clear groups. One is the set of cases where an opposition was plainly struggling at this point during the parliament: 1983, 1987, 2001 and 2005. In each case prime ministers banked the government’s advantage and called an election at around the four-year mark (slightly ahead of it in 2005), which on each occasion they won with healthy majorities. It is not surprising that governments tended to win these voluntary elections at around the four-year mark; if they had not been confident of winning they would have delayed. The result of this is to contaminate a series based on ‘election minus one year’ by selecting a number of the better government recoveries in year four of a parliament, particularly 1982-83 and 1986-87, but also 1969-70 and 1958-59.
The other clear group is when governments were well behind in the polls at this stage and prime ministers would have been suicidally reckless to call an election: 1963, 1996 and 2009. In each case the government had nothing to lose by having another year in power and hoping that their position would improve. They lost in each case, but in each case there was a net improvement in the government’s position by the time the election came round, very dramatically in 1963-64 and – if Labour’s swollen lead in 1996 is for real – 1996-97. The five-year terms in the past will often reflect an unusually bad government performance in year four, which gives more room for a recovery in the year running up to the election.
The middle group is more puzzling, mixed and arguably more relevant. In 1978 and 1991 the polls looked close, with tiny government leads following strong recoveries in support since 1977 and 1990, and the relatively new prime minister decided not to have the election. In 1949, 1959 and 1970 there were small opposition leads and the government went for an election some way into the fifth year – a sudden Labour recovery led to Wilson mistakenly calling the election for June 1970 and losing. The results of these elections are three government wins (1950, 1959 and 1992) and two losses (1970 and 1979). In four of them the government did better in the eventual election than the polls in month 46, in 1979 rather worse thanks to a disastrous series of events in the fifth year of the parliament – the ‘winter of discontent’.
The task for the Conservatives in 2014-15 still looks very hard. The best ever swings back to the government from this point to the general election are around seven per cent, and that would produce a Conservative lead of around 10 percentage points and a small overall majority. But the 1996-97 precedent may be a poor one, because Labour’s lead was unprecedentedly enormous and unrealistic, even if those polls were accurate at the time.
The Tory recovery in 1963-64 is the only really comparable case of a party getting a strong enough recovery from this point in the parliament to a general election, thanks largely to prosperity. Let us also remember what an economic boom looked like in 1964: it involved growth in GDP per head of around 5 per cent which did not involve taking up the spare capacity after a prolonged recession. In November 1963, the chancellor Reggie Maudling quipped, rather cynically, to his economic adviser Alec Cairncross that ‘everything shaping up for a splendid bit of expansion next spring followed by election and then efforts by Callaghan and co to cope with exchange crisis. Nobody would believe we hadn’t planned it’.
While the Conservatives might be able to inflate a consumer spending and property bubble before May 2015, some other elements of the Conservatives’ near-success in 1964 are absent. Fifth years of parliaments are often fag-end sessions in which little legislative work is done; in this parliament the tendency seemed to set in during the third year – another argument for why five years is probably too long between elections. The one exception to the general pattern was 1963-64, when parliament passed 49 Acts (the largest output of the whole 1951-64 Conservative government), including abolition of Resale Price Maintenance, a modernising measure that confronted one of the Tories’ interest groups, small shopkeepers, in the interests of consumers.
The past does not determine the future and one should be careful about reading too much into past poll trends, particularly in a time of coalition. But until this parliament, the captain of the home team had the power to blow the final whistle if their side was ahead in the 70th minute of the match, and that skews the form book a bit. Over to you, proper statistics types…