The Midterm Limbo Dance

How low can you go?

There is a dance the public do with incumbent governments over the course of a Parliament. Incumbents plunge in the polls over the first half, then pull themselves back up again as election day approaches. Fourteen of the nineteen election cycles (74%) since 1945 have featured governments bouncing back substantially from a midterm low.¹ Conservative MPs despairing about their party’s current slump can take some comfort from polling history, which suggests a degree of recovery may be around the corner. But the slump has been awfully large. Like Caribbean limbo dancers approaching the bar, many MPs will now be asking themselves “Can we go this low, and still get back up?”

There are two ways we can measure the midterm limbo downswing and upswing - in terms of governing party vote share, and in terms of opposition lead.² These are not the same, as the share of voters backing third parties - the Liberal Democrats, UKIP, the Greens and so - varies a lot from one election to the next.

In the charts below, I plot where the midterm limbo bar was set in past elections - the lowest monthly average share recorded by the incumbent government. The worst midterm positions are on the left, the best on the right, while the colours of the bars show who was in charge (blue: Conservative, red: Labour) and what happened next (dark bars: incumbent lost, bright bars: incumbent won). As we need to give the government some time to get back up again, the midterm low I use is restricted to polls more than 12 months from election day.³ On average, the worst incumbent polls have come around 20-22 months away from polling day - roughly where we are today if the next election comes as expected late next autumn.

The midterm downswing 1: lowest vote shares for incumbent more than a year before the election

Sources for all charts: Mark Pack polling database (1945-2019), New Statesman/Britain Elects polling averages (2019-24)

The midterm downswing 2: largest opposition leads more than a year before the election

The current government has already notched up one of the worst mid-term lows in post-war history. Only three past governments went lower that the Conservatives’ current low ebb of 26%, and two of those went on to lose the next election, while in the third case, Mrs Thatcher’s first administration, the midterm polls were upended by the emergence of the SDP. The 23 point average Labour lead also recorded in October is the third largest opposition lead at midterm - both of the oppositions who chalked up bigger margins than this went on to win, as did two of the three next best place oppositions.

But there are some glimmers of hope for the government. Three administrations have dipped below 30% at midterm and gone on to win and three administrations have fallen more than 15 points behind the opposition and gone on to win. All four administrations who have done one or both of these things were Tory governments - and 2019 would add a fifth if we had used a six month polling cutoff. By contrast, all three Labour incumbents who fell to 34% or lower and fell more than 20 points behind went on to defeat. Conservatives, it seems, are the better limbo dancers: they can go very low, yet still get back up.

The upswing 1: Difference between worst midterm vote share performance and subsequent general election performance

This pattern is confirmed when we look at the upswing - the difference between an incumbent’s worst midterm polling and their subsequent election performance.⁴ The average Tory bounceback of 9.8 points is more than double the average Labour recovery of 4.3 points, and seven of the eight biggest recoveries were recorded by Conservative governments - including a 17 point upswing in 1983 and a 16 point rebound in 1959. Labour notch up seven of the ten smallest rebounds and among Labour PMs only Harold Wilson in 1970 managed to recover by 10 points or more from his low ebb. He went on to lose anyway.

The upswing 2: Difference between worst midterm poll lead performance and subsequent general election

It is a similar story when we look at poll leads. Eight of the nine biggest bouncebacks have come under Conservative governments, including three 25 point rebounds in 1983, 1992 and 1959. By contrast, six of the seven weakest horse race comebacks have been under Labour. The differences in average performance are once again stark. The average Conservative government narrows the horse race margin by 18 points between the midterm low and election day; Labour government make up half as much ground (9 points) on average.

Rishi Sunak will be cheered to know that two of the largest past comebacks followed changes of leader. The switches from Anthony Eden to Harold Macmillan, and from Margaret Thatcher to John Major were both rewarded by voters with come from behind victories. But there is no historical precedent for the brief and frightening reign of Liz, which drove Conservative poll numbers to their current midterm low last autumn. John Major’s comeback victory would not have been helped if Conservative members had opted to give Norman Tebbit a six week turn at the wheel.

Nor can Sunak bank on other factors which have helped earlier rebounds. A strong economy helped Harold Macmillan recover in 1959, and helped Mrs Thatcher earn a third term in 1987. The optimism of Macmillan’s “You’ve never had it so good” looks unlikely in an economy facing recession, inflation, falling wages and rising bills in 2023-24. Mrs Thatcher managed to come back despite mass unemployment in 1983, but was helped by prosecuting and winning a war in the Falklands, a deeply unpopular opposition leader regarded as unfit to govern, and a major third party intervention from the SDP. Rishi Sunak is unlikely to start and win a war, split the opposition or see Keir Starmer replaced with Jeremy Corbyn before polling day.

It is 1992 which provides Conservatives with most ground for hope, and Labour with most to worry about. A charismatic but divisive election winner imploded, to be replaced in Downing Street by a blander but more reassuring figure. The economy dipped sharply, yet was growing again come polling day. The opposition lead soared, but the Labour leader posted mediocre ratings and doubts lingered about his party’s capacity to govern. While no two elections are exactly alike, each election provides lessons in what is and isn’t possible. Big Conservative comebacks clearly are possible, even when the starting conditions are less than ideal.

The current Labour lead is hefty, but it is not yet sufficient to justify despair on the right or elation on the left. Like Rocky Balboa, Conservative governments can’t be counted out even after taking a pasting. A historically average Conservative rebound would see their poll rating rise 10 points to 36% come election day, and see Labour’s poll lead decline by 18 from its autumn high to just 5 points. Given the current biases in the electoral system, this would certainly be enough to prevent a Labour majority, and perhaps enough even to leave the Conservatives as the largest party in a hung Parliament.

If Sunak were able to repeat John Major’s 1992 performance, he very likely would secure his party another term of office. A 30 point shift in the horse race would transform a 23 point deficit into a 7 point lead, more than enough to secure a comfortable Parliamentary majority.

Such a scenario may look unlikely, but past Conservative governments have recovered from humiliation, crisis and failure⁵ to emerge victorious. Another Conservative comeback may be unlikely, but it is surely not impossible.

1. This total excludes Feb to Oct 1974, which is too short to even have a “midterm”. If we limit ourselves to Parliaments which ran at least three years, the figure increases to 79% (11 out of 14) and the average bounceback gets a bit bigger.

2. When there is a tie, I take the share/lead nearest to the 12 month cutoff.

3. This means excluding the February-October 1974 cycle, when the time between elections was too short. But this government was so shortlived and so unstable that the usual midterm dynamics likely didn’t apply anyway.

4. Yes, part of this is polling error, but that doesn’t change the basic story - no one knows at midterm if the polls will be wrong or by how much. All they have to go on is where the polls are at the time.

5. For example, the foreign policy humiliation of Suez in 1956-7, recession and mass unemployment in 1980-81, and the disastrous introduction of the Poll Tax in 1990.