The results of the 2020 US presidential election have, once again, shown how dangerous it is to apply broad-brush thinking to the demographics of US politics.
While it’s too soon to declare that the pollsters and forecasters have failed this year (Biden is still favoured to sneak a narrow win, an outcome the New Statesman’s election forecast model suggested as likely), it is clear that anyone who had confidently written Donald Trump’s political obituary made a misjudgement.
The first thing the data tells us is why this was a hard election to predict. Voting patterns so far reveal a complex picture of shifting support across key demographic groups, and these patterns change from state to state.
In previous elections, there have been clear swing patterns across the country. This year, the county-level voting data presents no clear national picture.
So what has helped the incumbent president to, if not win, at least maintain his vote? The New Statesman’s data team has looked at a combination of exit poll data and provisional county-level data to attempt an early analysis of the 2020 race. The figures show an America changing politically in ways people haven’t yet recognised.
Trump has made gains among women of all races
Trump caused a stir in 2016 by outperforming Hillary Clinton by between 41 and 52 per cent among white women, according to exit polling. This year he appears to have extended his support among this group to 55 per cent.
While these on-the-night figures are subject to change, this year’s gender and race breakdowns numbers are striking because Trump appears to have gained ground among every group except white men.
Political commentators – in the UK as well as the US – have tended to assume that Trump’s sexist comments and the repeated allegations of sexual misconduct against him would deter women voters. Similarly, they have predicted that the president’s handling of issues such as immigration or the Black Lives Matter (BLM) protests would make ethnic minority voters more likely to favour his opponent.
This election confirms that, in reality, voters don’t think this way, and that the factors that determined voting decisions are anything but clear cut.
White, working-class Rust Belt voters didn’t desert Trump
Despite a small shift to Biden, the white working class didn’t desert Trump in the droves that pre-election data suggested they might, especially in the hotly contested Rust Belt states.
In the 2016 election Trump outperformed expectations in Michigan, Wisconsin and Pennsylvania. Much of that surge in Republican support came from mobilising, and appealing directly to, America’s white working-class voters – mainly those without a college education.
More than two-thirds of these voters opted for Trump four years ago, and what exit polling data we have now suggests that that grip has been mostly maintained.
Polls throughout the campaign suggested Biden would make huge gains among these voters, reducing Trump’s hold on the group to between 15 and 20 percentage points. This would have represented a sizeable swing on 2016 and turned much, if not all, of the Rust Belt, blue.
What we have seen instead is Trump retaining a lead among them of 30pts. Although this is down by 10pts on 2016, it is still, nonetheless, a demographic that Trump owns.
Hispanic voters can’t be bundled together any more
In the Sun Belt states of Florida, Arizona and Texas, Democrats had seen a growing Hispanic population as a route to a Biden sweep. That didn’t happen – and the data suggests it is because the electorate proved more diverse than strategists assumed.
In Florida, Cuban Latinos, a more socially conservative group than most, spearheaded the shift to Donald Trump. Republicans made a 12pt gain among Latinos in the state, while the Democrats lost 10pts. Overall, Democrats had the support of just 52 per cent of Latino voters in Florida – far less than in other parts of America, and down from 62 per cent in 2016. (But these figures do mean that, even in Florida, a majority of Hispanic voters still supported Biden.)
Crossing west to Texas, there is a different picture. The swing there – albeit still in Trump’s favour – was more muted, with 59 per cent of Latino voters in Texas voting Democrat, down just 2pts on 2016.
Pollsters have long struggled to properly sample Hispanic voters. Historic disengagement and, for some, a language barrier, has often made it difficult to obtain a representative sample. As overall turnout increases, so too does the number of Hispanic voters – and the fact that this brought more of the previously disengaged, socially conservative Latino electorate to the polling booth this year appears to have done direct damage to the Democrats. Hispanic voters nationwide, not just in Florida, have swung in Donald Trump’s favour.
The economy is still the biggest issue
The talk going into the election was that it would be decided on Trump’s handling of the Covid-19 pandemic, with voting also heavily influenced by issues such as the BLM movement. But preliminary figures show that the adage that “It’s the economy, stupid” is as true in 2020 as it was in 1992.
The early exit poll from Edison Research shows that, as in 2016, the economy was the most popular choice when voters were asked to list the issue most pertinent to their vote. As many as 35 per cent identified it as such (down from 52 per cent in 2016), compared to just 17 per cent who said that the Covid-19 pandemic was their priority.
Donald Trump secured an 82 per cent share of those most concerned about the economy. This marks a huge shift compared to 2016, where the majority of voters who said the economy was their prime concern supported Clinton.
This wasn’t the Covid-19 election
Despite Biden leading among those prioritising Covid-19 (with an 82 per cent share), the fact it wasn’t the major issue for as many people as anticipated has harmed his standing in the race.
County-level data shows no statistically significant trend between areas with a higher Covid-19 case rate and a swing towards Biden. Given Trump’s handling of the pandemic, to which almost a quarter of a million American have lost their lives, this is perhaps surprising.
Among the other themes mentioned, the exit poll data suggests Trump won big on the issue of crime and punishment – potentially as a reaction to protests and violence in the wake of the Black Lives Matter movement. However, only 11 per cent said this was their main priority – compared to a fifth who said that racial inequality was the issue that mattered most. Using Crowd Counting Consortium data, we found no relationship between the number or size of BLM protests in a county and a swing towards Trump. All of these factors lead us back to the inescapable conclusion that a lot of political “horse race” discourse is simply not that insightful about how voters decide to vote in elections.
Had Biden won early and comfortably, it would have been easier to gloss over the ways in which political discourse has been lacking in this election cycle. As it is, this awkward set of results should be cause for reflection on how we talk about and analyse voter behaviour.