The website tries to help understand the General Election system by translating the support levels for each party into actual seats. The aim is to de-mystify a process which is already well-understood by party insiders, and to make it accessible to anyone who is interested. The site tries to make the least inaccurate prediction possible and is independent of any other organisation.
No. The predictions are based on the law of averages, but any individual seat could be wrongly predicted. Also if the opinion polls are wrong, that will cause more seats to go wrong. Usually about 50 seats are incorrect. Look at the track record for details of elections since 1992.
Yes. The purpose of the predictions is not to tell people how to vote but to predict the election outcome given that people have told the opinion pollsters how they intend to vote. The site lets voters in a particular seat estimate the local strength of the parties from national trends. The ultimate decision in that seat is determined by those voters and no-one else.
Local factors can be very important. These can include a particularly popular (or unpopular) incumbent, a recent by-election, an important local issue, and so on. Unfortunately they cannot be included, as predictions are driven by national levels of support, as measured by national opinion polls. On average local factors must balance out, but the prediction in a particular seat may be incorrect.
Yes, in the user-defined prediction you can now enter tactical voting parameters. The most important of these describes how Liberal Democrat supporters will behave in seats where the Lib Dems are very likely to be in third place, but Labour and the Conservatives are close together. See Tactical Voting model.
Scotland has six major parties (SNP, Con, Lab, LibDem, UKIP, Green), so needs special handling. So the model allows all six parties to move dynamically, as measured by the Scottish opinion polls. But Scottish Westminster voting-intention polls are now quite rare. The prediction for Scottish seats has been relatively static since 2015 and can only be updated when new Scottish opinion polls are published. When available, Scottish polls are shown on the scottish opinion poll page.
The Reform UK and the Green parties are modelled fully in the same way as the other major parties (Con, Lab, Lib and Nationalist parties in Scotland and Wales). Smaller parties such as the BNP are not modelled separately, and are included in the 'Other' category.
The data shown is the snapshot provided by the latest General Election. To keep the consistency of the dataset, it is deliberately not updated to reflect by-elections or changes of party affiliation. This is necessary so that we can predict the next election on the basis of the change in support since the last one. This would make less sense if the base data were measured at different times. Also, by-elections are a famously poor predictor of general elections.
Yes it is. Electoral Calculus no longer uses the Uniform National Swing (UNS) model, since it has serious limitations. You can see details of the current model here (see regression polling). This provides a much better basis for predictions, that allows for regional variation and non-uniform swing. In 2019, we had the most accurate pre-poll predictions: see 2019 case study.
See our articles on regression polling and election prediction.
About once or twice a month, as national opinion polls are published. You can get email notification of updates by joining our mailing list.
All the polls used are opinion polls conducted by the major market research companies, and published in the national media. Occasionally we commission on our polling using one of these companies. History of recent polls.
Due to the lack of Wales-specific polls, it is necessary to keep the PC share of the vote fixed. This is inaccurate, but should not affect too many seats.
This needs regional opinion polls to provide the input data. This can be difficult to obtain. If you have your own regional data, you can use the regional predictor. Sometimes regional data can be derived from aggregating together many polls from the same pollster, though the geographical granularity can be coarse.
Absolutely. The Liberal Democrats vote is modelled to allow it to go up or down at the expense of the other two major parties (and the minor parties). The model is fully multi-party and is not a simplistic one-dimensional swing model. Also modelled are the Reform UK and Green parties, plus the SNP (in Scotland) and Plaid Cymru (in Wales).
The model used has elements of a Markov transitional model (see model details), but has fewer parameters. Markov models break down the current levels of support for each party according to how that person voted in the previous election. That is, it shows for those who voted (say) Labour in 2015, how many still back Labour, and how many back each of the other parties. The model is interesting conceptually, but it has too many parameters to be calibrated using the published national opinion polls. Regression modelling is generally superior to a purely Markovian approach.
The counties used are historical counties, loosely based on the 1974 boundaries. There are some artificial constructs, such as metropolitan urban areas (eg Manchester, Birmingham, Bristol, Newcastle, Glasgow, Edinburgh), and occasional simplifications, such as keeping the Isle of Wight in Hampshire. Each London borough is classified as its own county. See full details of area and county composition with map.
The Battlemap is less powerful than the full multi-party model as it is only two-dimensional. This means that only the three major parties can move, and the minor parties are assumed to stay fixed. It is easier to use the multi-party model to make your own predictions.
Either because you have spotted it before we have, or because it is not considered a high-quality poll from a non-partisan sponsor.
If you do see a poll we haven't used yet (including any for Scotland or Wales) please email us the details.
We are very happy to predict other elections, when commissioned to do so. Please see our Election Prediction Services.
Local parties have permission to republish data extracts from the 'Overview' section of the seat details page, subject to conditions.
Only data from the 'Overview' section can be re-published, including screenshots or re-typeset material. Attribution must be given to 'Electoral Calculus' and should include the date on which the data was observed, such as "Source: Electoral Calculus, retrieved 1 Jan 2024".
Please bear in mind that our predictions update regularly, and that old predictions can become out-of-date quickly. Local parties should not continue to re-publish out-of-date predictions. Other parties may object to such behaviour, and Electoral Calculus will only endorse and validate our current live prediction in the event of a dispute.
For re-publication of other materials, or for any questions, please contact us at .