THE candidates for the Gorton and Denton Parliamentary by-election include Reform UK’s Matt Goodwin, variously described as an academic, author, broadcaster and pollster.
The polls and betting odds suggest Goodwin has a fair chance of winning with the Greens, followed by Labour, being the other main contenders. The Conservatives don’t seem to be in the running. I have seen one local poll and several projections based on national surveys. While these look good for Reform (and the Green Party) they clearly have their limitations.
Polls ask people which party they would vote for if an election were to take place tomorrow. They also ask how likely they would be to vote. Many collect past voting behaviour to help calibrate and project the data. The survey responses are then weighted to get the demographic balance of the sample right and modelled to allow for voting likelihood. Most polls are based on samples of 1,000 or 2,000 and so have tolerance limits of +/- 1 or 2 per cent. The actual variation in results by polling company is much wider (+/- 3 to 5 per cent). Why is that? And what are the limitations with polls?
1. Elections do not take place tomorrow. They are announced weeks or months ahead. Candidates are selected and they campaign. People consider whether they will vote and for whom they will vote. In the by-election this process will not have taken place yet and has never taken place when polls are conducted mid-term (i.e. when there isn’t going to be an election).
2. People decide whether to vote and whom to vote for based on both parties and candidates. Polls tend to ask about parties and don’t list candidates (and can’t if there isn’t an election).
3. Party loyalty has generally disappeared and so the mix of candidates and their likelihood of winning may determine who people vote for. Tactical voting is important and in this by-election some parties have decided not to stand to help others beat their common enemy (generally Reform or Labour).
4. Turnout (or lack thereof) is a big factor, especially in by-elections. The party who get most people actually to vote for them will win.
5. Factors such as postal voting matter in many contests.
6. The sample quality and assumptions made when calibrating the data explain the significant variations in results by polling company.
In terms of who wins this election the fourth point is most important. The ‘ground’ contest between the Greens and Reform will be important. From the data I’ve seen (at a national level) these two parties are polar opposites both demographically and in terms of priorities. Understanding these factors (the interaction between demographics and voting intentions) does mean that models and projections will have some validity, but the by-election (like all elections) will be decided by who votes and how they vote.
From a professional market research point of view this poses two questions:
What’s the point of polls?
Could pollsters do better?
For most of us polls are a form of infotainment, information which is more entertaining than useful. Actual campaigners on the ground should have more useful information to help them target their activity (i.e. decide what to do). For them polls and data are business information. Central HQ should also have better data. So, as one senior media exec told me a decade or so ago, ‘we (newspapers) don’t pay for polls any more . . . they are a form of promotion for the polling company’. (I have not checked whether this is universally true and I’m happy to be more informed on this.)
As such, therefore, pollsters prefer their questions to be simple (i.e. less time-consuming) rather than complex and reflective of the real voting situation. At this point in the general election cycle we should really be asking about who would you consider voting for and which one party would you be most likely to vote for if there were an election in four weeks. This would be more informative. In an environment where party loyalty has mostly gone this would be a measure of voter stickiness. The same logic would apply to the current by-election.
Predicting whether people will actually vote is more tricky. Only half of those of voting age actually voted in the 2024 general election. I stood as an Independent Network candidate and spoke to about 2,000 people. The most common remark was ‘you politicians are all the same . . . a bunch of [insert derogatory term]’. The Tory majority was slashed from 26,000 to 2,600 partly as a result of the wider field of candidates (only four stood in 2019). I came ninth out of ten, which was reassuringly conclusive. I do not consider myself political; I am an elected representative on a Town Council.
I don’t think pollsters can crack the turnout effect (as the Economist had done) so projections based on polls of polls and different turnout scenarios are helpful. Pollsters who promote their results with an air of authority should realise they are deluded. Given that they are in the market for ‘certainty’ I doubt they will do that. I therefore expect national polls will continue to be all over the place merely providing a general indication of political support at any given point in time. They will still be more entertaining than informative.
Finally, here are my predictions for the by-election and the next general election:
The Greens or Reform could win the by-election; Labour have an outside chance.
Whoever wins the next general election will be a party which most people won’t have voted for.
The general election of 2029 is a long way off. Anything could happen.
