As the general election campaign comes to a close and 4 July 2024 approaches, polling shows little sign of shifting, with a small drop in Labour support over the campaign. But after 14 years of Conservative rule, Keir Starmer’s Labour has been consistently ahead of the Conservatives in the polls since the start of 2022. The Guardian is tracking latest polling averages, sourced from all major British polling companies, until election day. The Scottish National party (SNP) is not included in the data the Guardian is using in the chart above. In Great Britain-wide polls, the SNP vote sits between 2% and 4% of national vote share. But its geographical concentration in Scotland means it will win many more seats than other small parties with a similar national vote share, such as the Greens. Targeted Scotland-only polls give a much better indication of how well it will do in the next election than the nationwide polls above. Polls only go so far in predicting who will win in the UK’s first-past-the-post electoral system. What matters is the number of seats each party wins in parliament, which is decided by individual races in 650 constituencies. Seat predictions differ, but the one we show above is an average. It takes in three different ways of predicting how voting intention will convert into seats: uniform change projections, multilevel regression and poststratification (MRP) models, and other complex models. Uniform change assumes that national-level swings will apply uniformly across all constituencies. MRP models estimate the connection between characteristics such as age, gender and the area where a person lives, and which party they will vote for, and then use this data to predict votes in each constituency. How accurate are seat projections? In Britain’s first-past-the-post system the numbers in the polls do not correlate cleanly to seats because it depends where votes are located. Describing seat projections from general polling as a “loose yardstick”, Rob Ford, a professor of politics at the University of Manchester, said: “Labour could get a lead of 15 points and not have a majority, a lead of 10 points and have a majority. It depends where those votes are.” If the election is close, the polls become less predictive of the outcome. Other limitations of projecting seat counts from national polling include the fact that the Lib Dem seat count is hard to infer from national polling because, while their national support is much lower than the two main parties, in certain constituencies they have a significant presence. Nor is nationwide polling very informative about what will happen in Scotland, and polls there are more infrequent. Notes on the data The chart shows a rolling 10-day average for the support of each party based on Great Britain-wide polls. This excludes Northern Ireland, which has different political parties. On any given day, the Guardian works out the average support for each party across any poll published in the preceding 10 days. Only polling companies that are members of the British Polling Council are included. Starting 4 June, YouGov changed their polling methodology which they said could lead to smaller Labour leads due to the effects of tactical voting. The seat projections are sourced Professor Stephen Fisher, Dr John Kenny, Paul Furey, and Polina Ryzhuk, Elections Etc and the universities of Oxford and East Anglia. They are based on an average of opinion-poll vote-intention based forecasts including uniform change projections, multilevel regression and poststratification (MRP) models, and other complex models. Forecasts are averaged within those categories before averaging across the categories. They are updated twice a week.