Anyone who claims they know tomorrow’s result is a fool. But one thing is clear. Whoever wins the US election, the result will tear at the country’s partisan divide.

The vote itself turns on a small number of undecided voters in swing states – disproportionately young, suburban, low attention and nonwhite. They will be deciding based on their views of the candidates, the way conflicting issues tug at their affections, their thoughts on what each party did in the past, and their assessment of what each candidate means for the economy. The degree to which each party’s base is motivated to vote in swing states will also count.

But whoever the thin sliver of marginal voters put in office, we shouldn’t forget the overwhelming bulk whose choice is locked in. They may not matter tomorrow, but in the days and years to follow, they will be the story.

The source of division is primarily cultural. Even people’s views on the economy are heavily shaped by whether they support the Democrats (who have been in charge of the economy for the last four years) or the Republicans.

On the one hand, the economy is the most important concern for people, typically the only issue a majority of voters say is ‘extremely important’ for their vote. As the Gallup chart below shows, this share has fluctuated only modestly over time. At first glance, Bill Clinton is right that ‘It’s the Economy, Stupid’.

However, we need to distinguish what political scientists term valence issues - questions of managerial competence – like maintaining good roads, which matter for all voters and do not predict how a person will vote, from positional issues, which do. Illegal immigration is partly a matter of competence, but is strongly informed by whether people are liberal or conservative on immigration and cultural change. As Mark Cuban argued, the border issue arose because Biden and Harris’ ‘hearts are too big’.

When it comes to the largest partisan divides, which denote positional issues and predict voting, these are mainly cultural. Figure 2 draws on the American National Election Study (ANES) 2024 pilot survey. ANES is the gold standard survey for political scientists, though note that the survey was conducted prior to Kamala becoming leader.

My statistical analysis of the survey shows how the chance of predicting a Trump vote (rather than a Biden vote) increases depending on which information we have about a person.

Much is made of demographics, but these actually don’t tell us much more than a coin flip. Race and education matter more, improving predictive power from 50-50 to almost 60-40. Knowing a person’s race, education, income, age and gender gets us to 62.5 percent, but no higher.

Now consider a person’s warmth toward Black Lives Matter on a 0-100 thermometer. Knowing this information allows us to predict whether someone will vote Trump 78 percent of the time. For comparison, this is only marginally less powerful than knowing someone’s feeling toward the Republican Party (79 percent correctly predicted)! It is vastly more important than all demographics put together.

Likewise, attitudes to immigration tell us a lot more about how a person will vote tomorrow than whether they think inflation or taxes are a problem. The Gallup partisan gap on the importance of taxes is 23 points, compared to 40 points for immigration.

While feelings toward unions and big business also matter in the ANES data, they are considerably less important for predicting the vote than feelings toward BLM or the National Rifle Association. Knowing BLM and NRA sentiment allows us to predict 45 percent of the variation in Trump support while knowing sentiment toward big business and unions only gets us to 17 percent of the variation. All demographics combined predict just 6 percent of the variation.

Source: ANES 2024 Pilot Study.

The figure below illustrates the relative predictive power of BLM support and education. There is an 80-point gap in the likelihood of voting Trump between those who are coldest and warmest toward BLM (see ends of each line) but just a 5-20 point gap between those with and without a college degree (gap between the lines).

Source: ANES 2024 Pilot Survey. N=1,500. Pseudo-R2=.314. No significant interactions.

In political science, Don Green and colleagues hold the view that people form their party attachments early in life based on how well a party meshes with their demographics as men and women, young or old, black or white, college or non-college educated, rich or poor. Once crystallized, attachments are sticky and people ape their party’s stance on the issues. If Trump supports free trade and opposes Russia, so do Republican identifiers. If he changes to oppose free trade and support Russia, they change too. Party identification is emotional and part of a person’s self-concept.

Another view holds that party identification is fluid: people choose their party attachments based on a ‘running tally’ of how many issue positions align with those of a particular party. Party identification is about a rational calculation of which party ticks more of one’s boxes.

Political ideology sits in the middle as something that is partly about deep-rooted emotional attachment to being a liberal or conservative, and partly about a shifting calculus based on a person’s views of the issues.

Finally, party identity strongly predicts voting, but people can still vote against their party or stay home if they like the opposing party leader or dislike their own.

I lean toward the view of Karen Stenner and others that in the US, party identification is more rooted than fluid, more emotional than rational. Its substrate is not so much demography as psychology. Those who prefer national commonality over difference, and tradition over change, tend to be Republican while those who lean the other way are Democrats. The best questions for measuring these psychological dimensions are ‘American culture was better in the past’ and ‘Which is more important for a child, to be considerate or well-behaved.’

Psychology underpins ideology and cultural issue positions, which in turn are close to party identification and voting for most voters – especially whites.

One might argue that people become Republicans first then take their views on BLM and the NRA second, but I am not convinced. If the Republicans - even Trump - applauded BLM and denounced the NRA, some would shift their views but most would not and he’d be in trouble with his base. People’s psychology informs their views on issues and their ideology, which is what underpins their partisanship and vote.

Tomorrow’s result won’t divide people much by race, sex, age, income or education. Instead, expect the vote to split groups, families, friends and even couples along cultural-psychological lines.