Are the Tories heading for an electoral wipeout?

Posted in: Data, politics and policy, Political history, Political ideologies, UK politics

Dr Joe Chrisp is a Research Associate in the Institute for Policy Research (IPR) at the University of Bath.

With less than a year until the next general election, politicians and the media are currently finding a multitude of ways to articulate the same basic message: the polls show the Labour Party are edging towards a comfortable majority, even a landslide, but that anything can happen and maybe <Insert obsession of the author> could change things.

YouGov’s latest poll gives Labour a 27 point lead nationally. The Financial Times also has a new general election poll tracker, which shows that even accounting for uncertainty drawn from previous polling inaccuracies and assuming the worst, Labour would still have a commanding lead of over 12 percentage points.

The Telegraph has splashed on the role of Reform in scuppering Conservative election hopes, with a view to encouraging perhaps either a shift in policy to the right or some kind of agreement in the style of the Brexit Party at the 2019 election. The Labour Party is also using every opportunity to brief messages they hope will manage expectations and avoid complacency, including a portfolio of scuppered poll leads.

There are a few commonly rehearsed reasons for Labour pessimism or Conservative optimism. These include the precedent of the 1992 general election (and that this is not 1997), the prospect of don’t knows and Reform voters ‘coming home’ to the Tories, the perceived pro-Tory electoral map and the increased volatility of voters in contemporary democracies. In this blog, we briefly explore these factors after an analysis of the demographic shifts in voting intention.

Age divide

First, one of our key research interests at the IPR in recent years has been mapping out and trying to understand the sizeable and persistent age divide in British politics that first showed signs of growth at the 2015 general election, grew considerably after the EU referendum, and has broadly remained high since. In previous blogs and journal articles, we sought to counter the idea that this is simply due to education or generational social values abstracted from material interests.

The following two graphs show the dynamics of support for the Conservatives and Labour by age group for the period in which the British Election Study Internet Panel has data (February 2014-May 2023; it will continue until the next election), excluding don’t knows and non-voters. They both show that while support for the Conservative Party has collapsed, it has maintained its lead among voters over the age of 65. In absolute terms, the Tories have lost more older voters but the relative size of each age group in their voting coalition has broadly remained stable. Less of the over 65s also appear to have shifted to the Labour Party and have moved to Reform. Thus, despite the dramatic shift in fortunes since the pandemic peak, the age divide remains.

But to what extent is there variation within the broad categories of younger and older voters? In previous analysis (see also this thread on Twitter/X), we divided the electorate into 8 sub-groups distinguished by their age (18-54 or 55+), housing tenure (homeowners or renters) and education (graduates or non-graduates). The 2019 election showed that all these factors were important in driving both voting preferences and turnout, with 18–54-year-old, graduate, renters the stereotypical Labour voter and 55+, non-graduate, homeowners the stereotypical Conservative voter. Labour won a plurality of the vote with any sub-group that featured at least two “Labour-friendly” characteristics (i.e. young graduates, graduate renters and young renters) and likewise the Conservatives won a plurality among those with at least two “Tory-friendly” characteristics (older non-graduates, non-graduate homeowners and older homeowners). The Conservatives won a resounding majority largely because of the relative size of these different sub-groups –there are a lot more non-graduates and homeowners than graduates and renters – and their propensity to vote, with younger renters much less likely to turnout. These factors also explain why Labour lost even in 2017 when it won a plurality of all sub-groups except older homeowners.

One interesting dynamic of the last 3-4 years has been the reverse shift from a 2019-style dynamic in the June 2020 wave of the BES Internet Panel, where each party won a plurality of half of the sub-groups, to a more dramatic version of a 2017-style dynamic in the May 2023 wave where Labour leads with every sub-group except older, non-graduate homeowners. The most dramatic change in those 3 years has been with younger, non-graduate homeowners – a group no doubt affected by the stagnant economy, high inflation and critically the hike in interest rates.

The change is less dramatic if we include voters saying they don’t know who they will vote for or say they will not vote but equally the data do not point to a particularly large growth in undecided voters as compared to when the Conservatives were riding high.

1992 or 1997?

A popular question among journalists in the last couple of years has been whether the Conservatives are heading for a wipeout in the manner of the 1997 election or a surprise victory inspired by the example of John Major in 1992.

As many have argued, the conditions of the 1992 election are simply not here in 2024. The economy shows no serious signs of recovery, the Conservative Party is not more trusted on the economy and polls show Rishi Sunak is not preferred to Keir Starmer the way that John Major was over Neil Kinnock. But equally Starmer’s popularity is not close to that of Tony Blair’s in the run-up to the 1997 election.

Many have also pointed to the low number of ex-Conservative voters switching to the Labour Party and saying they don’t know who they will vote for instead. This may be seen as an indication that that the Labour lead is inflated. The BES Internet Panel suggests the proportion of 2019 Tory voters moving to Labour is about 10% with up to 24% undecided.

However, this is not drastically different from the final flows from the 1992 election to the 1997 election, albeit we are not comparing like with like given one is an election result and the other is a poll one year before in which respondents can state they do not know. Exploring this question in the past, I estimated the voter flows from three different sources, the British Household Panel Survey using respondents that were in the sample in 1992 and 1997 (the old name for Understanding Society), the British Election Study cross-sectional data from 1997 (and using recalled 1992 vote) and the British Election Study panel between 1992 and 1997. Note that I reweighted to account for a large underestimate of non-voters in these surveys. Although the existing BES Internet Panel also undoubtedly underestimates non-voters it also includes a number of people stating they do not know who they will vote for, equally dampening the size of Labour’s vote among Conservative 2019 voters. I include the latter two at the end of the blog but the BHPS data here suggest that roughly 10% of Conservative voters in 1992 switched to the Labour Party in 1997.

Volatile voters and favourable electoral map?

Another reason why neither 1992 or 1997 might be a good guide is that partisan identification has been on a long-term decline, and each younger generation brings with it a lower level of partisanship. As the British Election Study team argue in their book after the last election, voters have never been more volatile and the intention to vote one way now is no guarantee they will continue to do so in a year’s time. This is probably the fact that should most comfort Tories and worry Labour, ignoring that there are no obvious objective reasons to think voters will shift back to the Conservative Party (rather than to say another party on the left). The fact that Labour’s core vote is among the young should also point to the vulnerability it faces.

Voters are not just more volatile from election to election (or poll to poll) but also across constituencies. Uniform National Swing, the measure of average swing in voting Conservative vs. Labour in elections devised by the late David Butler, has become less and less of an accurate measure of specific constituency swing as the variance across the country has grown. The below graph shows the distribution of constituency swing for elections since 1959 (missing Feb 1974 and 1983 when there were constituency boundary changes and no estimates of past election results under new constituency boundaries). Although 2017 was the peak, the long-term trend in volatility has been clear.

The change in variance from election to election can be easier to see if each election is centred by average swing as below.


One way to spin this potential volatility in a positive light for the Conservatives would be perhaps that, they could target their highly superior resources at key marginal constituencies and counter the national trend.

However, I suspect that this volatility across constituencies points to even more danger to the Conservatives. It is likely that Labour will outperform the uniform national swing in the constituencies it needs to win in the so-called Red Wall, Scotland and elsewhere where electoral strategy has been focused, with less impressive performances in areas of existing strength such as London. Indeed, recent polling by YouGov for the Fabian Society suggests this is the case. This combined with tactical anti-Tory voting could mean this volatility means we are understating the extent of Conservative vulnerability. The nature of first past the post is such that favourable electoral geography could easily shift in Labour’s favour and exaggerate its lead as it did in the New Labour years.

All articles posted on this blog give the views of the author(s), and not the position of the IPR, nor of the University of Bath.

Posted in: Data, politics and policy, Political history, Political ideologies, UK politics


  • (we won't publish this)

Write a response