Methodology
The Telegraph relies on a variety of US polls aggregated by FiveThirtyEight. Only head-to-heads between Kamala Harris and Donald Trump are considered.
Exclusively polls of registered voters feed into our national voting intention tracker, which is updated hourly. Our locally estimated scatterplot smoothing (LOESS) model fits multiple regressions over subsets of five per cent of the data around each point.
Polls are also weighted based on the grade assigned to the pollster by FiveThirtyEight – taking into account past empirical accuracy and methodological transparency. Low-rated pollsters are ignored.
Sampling error is unavoidable – there is always a risk that the chosen population’s views are not representative of the public as a whole, and as such an individual poll may not be an exact depiction of reality. This is why we take a poll of polls, and why we depict the 95 per cent confidence interval with shading.
For the running predictor of electoral college votes, we start by taking an average of the last five polls to come out of each state. Due to the lower availability of state-level polling, polls of all population groups – all adults, registered voters and likely voters – are considered.
If either candidate is polling five or more points ahead of the other, the state is deemed to be “solidly” on their side, and the corresponding number of electoral college votes is awarded to them. If their lead is between 0.5 and five points, the state is said to be “leaning” their way. Where margins are tighter, the state is considered a “tossup”. Total votes represent the sum of “solidly” and “leaning” states.
Our swing state tracker charts the latest five individual polls in each 2024 battleground.
Approval ratings are national net proportions of respondents approving, minus those disapproving. Joe Biden’s current presidential approval and Harris’s vice-presidential approval, is compared to Trump’s on the same day of his term in office.
The favourability of vice-presidential picks JD Vance and Tim Walz is the sum of national “Very favourable” and “Somewhat favourable” response shares (and vice-versa for unfavourable views). The net figure is the subtraction of the two.