March saw the publication of two of the biggest polls of voting intention for some time. Estimating voting patterns constituency by constituency, they both confirm what conventional national polls have been showing for the last year: that the Conservatives face catastrophic collapse. Both predict a national Labour majority of over 250 MPs. With under 100 seats, the Conservatives face the biggest defeat in the party’s 190-year history.
The polls were very large. Both surveyed over 15,000 people, using MRP methodology to estimate results at the level of individual constituencies. The first was by Electoral Calculus (EC) with Find out Now, carried out over three weeks in January-February. It was followed by Best for Britain (B4B) with Survation, carried out over two weeks in mid-March.
Although there was little dispute about the scale of the Conservative collapse, some commentators doubted the findings at constituency level, especially as they affected the smaller parties. For this reason, it is worth considering their limitations and to look again at what they might mean in East Anglia.
MRP does not mean polling in every constituency
Since the result of a general election is determined by the number of individual MPs elected, rather than the overall national vote share, commentators want to know what might happen in each constituency. MRP polling aims to do this, without the costs of questioning large numbers in every constituency.
To do this, they use a very large national sample to identify how particular kinds of people (by age, sex, ethnicity, previous voting behaviour, etc) intend to vote. The pollster then uses a complex formula to link this with public data on how many people in each of those groups there are in a constituency, to predict the likely result there.
Different MRP polls will produce different results, because each pollster has its own way of balancing the factors in the formula. Since Labour and Liberal Democrat voters generally have more in common with each other than either has with Conservatives, MRP models may have more difficulty distinguishing Labour voters from Liberal Democrats, than distinguishing either from Conservatives. The Green Party is a special case. Because their politics appeals to people on both left and right, and because their numbers are small, the models may have difficulty identifying the profile of a typical Green voter. It could be that this will result in an underestimate of their real strength.
In the 2019 general election, both pollsters claim to have correctly predicted more than 95% of the results. In normal circumstances, that error (of 36 constituencies) could well be the difference between winning and losing a Parliamentary majority, but no polls currently suggest that the margin between Labour and Conservatives is anything like that small.
MRP cannot measure local factors
No poll is perfect, and an important limitation of MRP polls is their inability to recognise local factors. They may miss the real picture in a given constituency because of:
- Local party strength. In a constituency where there is strong opposition to the Conservatives and a majority of people who look like Labour voters, an MRP poll will suggest a Labour win. But if the Liberal Democrats campaign strongly, and the local Labour party is inactive, those people may choose to give their vote to the Liberal Democrat. The Liberal Democrats hope this will win them Chelmsford, despite both polls suggesting a comfortable Labour win there.
- Local government record. If a minority party has been successful in local elections, people may be more likely to see them as a serious option, rather than a wasted vote. The Greens hope that their success in the 2023 local elections in Suffolk, where they took control of Mid Suffolk council, will have this effect in the new Waveney Valley constituency, as well as in Bury St Edmunds, and Suffolk Coastal.
- History. Where a constituency has only recently changed to the Conservatives, a swing against them may return the seat to its previous party, which is seen as the obvious opposition. The Liberal Democrats hope this will work in North Norfolk, a seat which they held for 18 years prior to 2019.
- Candidate recognition. Although many people vote for (or against) a party rather than the candidate, some candidates have a very high profile and local reputation. Whether this benefits the Conservatives in Liz Truss’ South West Norfolk constituency remains to be seen.
- Tactical voting. In some constituencies people may decide to vote for the candidate most likely to overturn the Conservative. And although there are no formal pacts, there is some evidence of the Labour and Liberal Democrat parties quietly reducing their effort in places where the other party stands a better chance.
- Margin of error. MRP polling produces estimates of probability. Survation (B4B) publishes results for each party and constituency, based on a 95% probability. So, while they show Hitchin as a probable Labour win, it could actually be won by the Conservatives.
Do these polls agree in our region?
In East Anglia, there are now 61 constituencies (three more than in 2019, following this year’s boundary review). Both polls predict that Labour will take 26 seats from the Conservatives, and in 48 constituencies they agree on the likely winner.
However, they differ in their overall picture. B4B suggests a predominantly two-horse race, with most constituencies clearly voting either Conservative or Labour. EC, on the other hand, gives a better chance to the three smaller parties, although most of these differences are not big enough to change the results.
The table shows the 13 seats where the polls do disagree about the likely winner. The smaller numbers suggest that B4B is much less confident in its prediction.
Safe and marginal seats
Whatever happens, there will still be some ‘safe seats’, though the polls disagree about how many. EC predicts 30 seats in the East of England region with majorities of over 10%, while B4B predicts only 14.
Both agree that Labour will have at least 12 such seats, and EC puts the number higher, at 24. The highest are Norwich South and Cambridge, where both polls expect a Labour majority of over 35 points.
Both polls show Castle Point and Maldon as safe Conservative seats. B4B also ranks Clacton as safe, but EC disagrees, mainly because they give a much larger figure for Reform UK. Clacton was previously a stronghold of UKIP, the predecessor to Reform, who are taking votes from the Conservatives nationwide.
Will tactical voting matter?
Progressive tactical voting campaigns encourage people to vote for the candidate most likely to defeat the Conservative.
These polls agree that there are well over 40 seats in East Anglia with a clear progressive majority. And in most of those cases, it is obvious which party is the lead challenger to the Conservative.
However, there are a few seats where the three main parties are close together, and the best tactical choice is not obvious. The chart shows the predictions for both polls in the four constituencies where this is most likely. EC sees three constituencies – Ely & East Cambs, Harpenden & Berkhamsted, and St Neots and Mid-Cambs – as three-way contests, with the Liberal Democrats in the lead in two of them. But B4B disagrees, suggesting that in all four, the real battle is probably between Conservative and Labour .
Local Elections may test these polls
On 2 May, 18 councils in East Anglia hold local elections, and commentators will try to use the results to predict the general election outcome. However, there are several reasons why the results may not match.
Local election turnout is likely to be low, so the results will only reflect the view of the most committed. Local factors, including independent groups who are strong in some districts, may have a substantial impact. Reform UK is eating into the Conservative vote in many constituencies, but they will probably not stand in many local elections.
Nevertheless, on 3 May we will have a better idea of the scale of the Conservative disaster, and how far to believe these MRP polls. It they are reliable, we might expect something like the following results:
- Labour win 13 councils. Holding Cambridge, Ipswich and Stevenage; taking Harlow and Thurrock from the Conservatives, Watford from the LibDems, and seven councils from no overall control (noc).
- Conservatives lose two councils. Holding Epping Forest and Basildon & Billericay; and taking Brentwood and Castle Point from noc.
- Liberal Democrats lose one council, but hold St Albans.
Conclusion
As these were both very large polls, the underlying data is probably reliable. The differences between them probably reflect differences in the way the pollsters weight the factors in the calculation of probabilities.
All polls, conventional and MRP, agree that the balance of support between Conservatives and Labour has been fairly steady for months. But this does not make the general election a simple two-horse race. There are two local factors which these polls hint at, which may influence the outcome. The local strength of the Liberal Democrats may swing a few seats from Labour to Liberal Democrat, with no impact on the Conservatives. More significantly, the rise of Reform UK continues to erode the Conservative vote. In probably every case in our region, this benefits Labour.
The MRP findings are interesting, and point us to places to watch, provided we understand their limitations.
