Editor's note: We have paused our analysis on this page after Kamala Harris replaced Joe Biden as the Democratic nominee. We will resume with new data based on Ms Harris's polling numbers soon. In the meantime, see our presidential election prediction model and poll tracker.
The race between Joe Biden and Donald Trump is very close. Over the first three months of 2024 the candidates were never more than three points apart in our average of national polls, with Mr Trump narrowly ahead for most of that time. That is new for Mr Trump: in his two previous presidential campaigns he never led a general-election polling average for a single day. More worrying still for Mr Biden, Mr Trump is ahead in several of the swing states that he lost in 2020. The outcome in half a dozen states— Arizona, Georgia, Michigan, Nevada, Pennsylvania and Wisconsin—is likely to prove decisive. A small but critical slice of voters who plumped for Mr Biden back then are now telling pollsters that they plan to defect. Who are they?
To find out, we built a statistical model to assess how a hypothetical voter might cast a ballot, based on their demographic traits. Our data come from YouGov, an online pollster, which every week surveys over a thousand people about their demographic profile, voting history and voting intentions. We combined all its survey results since January 2023 to get a detailed portrait of Americans’ voting preferences. Use the drop-down menus below to plug in any combination of attributes—age, sex, religion and more—to construct a hypothetical American and see our estimate of their vote. Or press shuffle to see a voter at random. Our model will continuously update to incorporate each week’s YouGov survey.
Build an American voter
Chance of voting for president*, %
How this group compares to others
Straight white men aged 75+ from a city in California, who are Catholic or Orthodox and have a college degree
*Excluding don't know, won't vote and other parties
The voters propelling Mr Trump’s polling renaissance might come as a surprise. While white voters’ preferences have changed little since 2020, racial minorities—historically the bedrock of Democratic support—have lurched away from Mr Biden. Mr Trump has also sharply cut into his successor’s advantage among young voters, another core Democratic group. Mr Biden will hope these once-loyal Democrats return to the fold once the campaign heats up.
Latina women aged 25-34
Shifted towards Trump
Black people aged 35-54
Became less committed to Biden
Atheists
Remained loyal to Biden
White evangelicals
Remained loyal to Trump
Race is often cited as the central cleavage in American politics, yet the single most powerful predictor of voting intention is religion. A model that knows nothing save for respondents’ religious affiliations can correctly identify their preferred candidate 62% of the time, compared with 59% for race. Of Mormons and evangelical voters, 73% say they support Mr Trump. This compares with just 13% of avowed atheists.
Most likely to make you a Biden voter
Most likely to make you a Trump voter
Rather than the sharp realignment that took place in 2016 and 2020, when Mr Trump attracted working-class white voters while shedding college-educated ones, the voters swinging in either direction this year are more alike: they tend to be young; black or Hispanic; and live in cities. This suggests they have looser party alliances and pay less attention to politics.
So both sides will think they can win as the election approaches. And you can use this tool to explore the type of voter—a 40-year-old high-school-educated black man from rural Georgia, say—who might just swing it.■
Stay up to date on American politics with our new daily update, The US in brief. And explore how British voters may vote in the next election with our UK election trackers.
Methodology
Our model is based on survey data provided by YouGov, which obtains responses from a nationally representative sample of approximately 1,500 Americans each week. We gathered all results since the start of 2023, amounting to nearly 100,000 individual responses. We have removed people who did not say they planned to vote for either Joe Biden or Donald Trump in this year’s presidential election.
To estimate voting intentions based on demographic profiles, we fit a logistic regression model using the LASSO method, a statistical technique that eliminates or reduces the impact of certain variables in order to maximise accuracy on unseen data. Our model accounts not just for the eight demographic features detailed above in isolation, but also for how they interact with each other. For example, switching the listed age group from 75+ to 18-24 sharply increases the chances that a white voter will support Mr Biden, but actually reduces this probability for a black voter. Our model also incorporates the national poll average for the two leading candidates in each week. As a result, if one of them gains or loses ground in the polls overall, the model will automatically shift vote-intention probabilities for each demographic profile in the same direction. We update the model every week to account for additional survey data and new national polling averages.