Why it's very hard to know who will win the next election: Analysing pre-election polling fluctuations since 1997

15th February 2023


Which party is going to win the next general election? The latest voting intention poll from Deltapoll has Labour on a 20% lead over the Conservatives. The last time the Conservatives were ahead in any poll published by Deltapoll was in November 2021. Since that time, the Labour lead has been as small as 6% in March 2022 and as large as 32% in October 2023.

Do polling leads for Labour of this size mean that they will secure a majority in the next election? Certainly, it seems difficult to see how the Conservatives can close or even reverse the 20-25 percentage point gap between now and the next election. Yet, if we look at trends in vote intention polling and the final vote share on election day for Labour and the Conservatives in past elections, it is not unprecedented to see large fluctuations throughout the election cycle that ultimately end up as far smaller gaps come the election. Ahead of the 1997 General Election, mid-term polling gaps were as large as Labour polling at 32.4% in February 1995, when the final election result had Labour 12.5% ahead of the Conservatives on election day in May 1997. It should be said, however, that some – indeed much – of the polling work undertaken before the 1997 election was not, in our view, particularly methodologically rigorous, and comparisons with it made today often forget that, especially when polling averages are the comparative method.


Looking at recent election history, how does pre-election vote intention evolve over the course of the election timeline? Building on work by Jennings and Wlezien, I investigate opinion poll fluctuation throughout each general election cycle and how these map onto the results from election day. I assess this for all UK general elections between 1997 and 2019. In the below graph, each panel represents one general election cycle. The dashed lines represent the final election day vote share for the Conservatives in blue and Labour in red. The solid lines represent the monthly average from all vote intention polls between April 1992 and February 2023, collected from PollBase and updated with more recent polling. The x-axis represents the number of months away from the general election the polls were conducted, and the y-axis the vote share of each party.


We need to look back to the late 1990s and early 2000s to find mid-election polling gaps the size currently enjoyed by Labour. However, what is certainly clear is that there is a lot of variation in support for the parties throughout the election timeline. In only three out of eight pre-election periods mapped – 1997, 2001, and 2017 – is the party that ultimately wins the election ahead in essentially all polls throughout the election cycle. For the remaining pre-election periods – 2005, 2010, 2015, 2017, and the current election period – this is not the case. Instead, at different points the gaps between the parties has widened or narrowed. At times, Labour and Conservatives have polled at numbers that are higher than their final share on election day, where at other times they have polled at numbers that are lower.

In the below table, I focus on the final two years before each election. The table shows the percentage point lead between the parties 24 months, 12 months, 9 months, 6 months, 3 months, in the final month preceding the election, and in the final election result, where Labour leads are shaded in red and Conservative leads in blue.


For five out of the seven most recent general elections, the leading party consistently led in the polls in the two years prior to the election and ultimately ended up being the largest party in the election. In the two years ahead of the 2015 General Election, Labour led in the polls up until the month of the election itself, where the Conservatives ultimately ended up winning with a lead of 6.4 percentage points. Although it should be noted that polling ahead of the 2015 General Election was considered some of the most inaccurate since polling began in 1945. Ahead of the 2019 General Election, Labour were leading in the polls 24 months and 6 months ahead of the election, however, the Conservatives led 12, 9, 3, and in the month prior to the election, and ultimately ended up winning with a lead of 11.5 percentage points.

More often than not, the party that leads in the polls in the two years prior to the election ultimately ends up being the largest party come election time. What is clearly common across each of the past elections, however, is that the gap between the parties varies significantly throughout the election cycle. Ahead of the 1997 General Election, the Labour lead two years out from the election was 29.9%, which reduced throughout the cycle to a final lead of 12.5%. Similarly, ahead of the 2001 General Election, the Labour lead reduced from 25.3% two years before the election, to 13.4% one year before, and resulted in a 9% Labour lead on election day. In most cases, the gap between the parties is largest one to two years ahead of the election than it ultimately is come election day. But, again, this is not always the case, as is evident from the 2019 General Election when Labour were ahead in the polls two years ahead and the gap was essentially non-existent a year out, when the Conservatives ultimately ended up with an 11.5% lead.

What can these trends tell us about the positioning of the parties ahead of the next election? Well, ultimately the only common feature across the past seven elections is that things can and do change. To know with any sort of confidence what the polls right now can tell us about the election result, we would need to know both when the next general election will be and what is going to happen to make the public more or less favourable towards the parties ahead of the election. We, of course, know neither of these things.


The latest the next general election can be is 24th January 2025, which is currently 23 months away. Looking at the past seven elections, at no point has a party been polling with a lead of around 20-25 percentage points 23 months out from the election that has not ultimately ended up being the largest party in the election. However, this time around, Labour face an electoral map that is more challenging for them than it has been for any previous contest. Recent estimates suggest that, assuming constituency boundaries stay as they are ahead of the next election, Labour need a swing of 9.9% in order to secure a working majority. Further, they will need to win in the right places, securing 51 seats they lost to the Conservatives in 2019, and taking back a further 47 seats held by the Conservatives since 2010.


Taken together, from examining past election cycle trends, it seems very unlikely that the Conservatives will be able to close or even reverse the current gap between the parties. However, past polling trends also tell us that it seems equally unlikely that the 20-25 percentage point lead currently enjoyed by Labour will be in operation come the next election day. Ultimately, vote intention opinion polls are simply a reflection of public mood towards the parties at the time the poll is conducted. Indeed, public opinion companies do not ask respondents which party they think will win the next election, but instead some variation of “if there were a general election held tomorrow, which party would you vote for”. Polls, therefore, aim to reflect where the public mood is at any given point. The point: polls can, do, and almost certainly will change, and these changes will reflect the performance of each party, the behaviour of their leaders and politicians, their policy proposals, activity in parliament, how they respond to events happening nationally or internationally, among innumerable other factors. Past election timelines make it very clear that public mood can and does shift. How exactly it will shift between now and the next election remains to be seen.


Dr Lotte HargraveAuthor: Dr Lotte Hargrave