At a voting centre in San Francisco on primary day in California in March. Covid-19 is likely to be a factor in turnout and the use of absentee ballots in November. Photo / Jim Wilson, New York Times
Could this tilt turnout in Trump's direction and affect the election result?
Identifying likely voters is a challenge for pollsters in every election. This year, the coronavirus, mail voting and a surge in political engagement may make it even harder than usual.
For now, Joe Biden's 9-point lead across thecritical battleground states is so significant that it is essentially invulnerable to assumptions about turnout, according to New York Times/Siena College surveys of the states likeliest to decide the election. But Biden's supporters are far more likely to be concerned about in-person voting during the pandemic, and his wide polling lead among registered voters could narrow if their concerns persist to the election.
Overall, one-quarter of registered voters in the battleground states said they would feel uncomfortable voting in person. Perhaps surprisingly, these voters are fairly representative demographically of the country. The groups hit hardest by the coronavirus so far — older people, Blacks, Latinos, those in densely populated areas — are generally likelier to say they would feel uncomfortable voting, but not vastly so.
Instead, discomfort about voting is mainly a function of political views. People were asked if they would feel uncomfortable voting in person if the election were held during the week they were interviewed in June. About 40 per cent of Biden's supporters said they would feel uncomfortable, compared with just 6 per cent of President Donald Trump's supporters. This political divide transcends demographics. A young Biden supporter in a rural area, for instance, would be likelier to feel uncomfortable voting than an old Trump supporter in a city, even though the health risk is probably quite low for the Biden voter and potentially quite significant for the Trump supporter.
Most of these voters would go to the polls anyway. But about one-quarter of the uncomfortable voters — or about 6 per cent of the overall electorate — said they would feel too uncomfortable to vote in person if the election were held during the week they were interviewed. This includes 8 per cent of all of Biden's supporters in the battleground states, compared with less than 2 per cent of Trump's supporters.
Overall, the voters who said they would be too uncomfortable to vote in person back Biden, 63 per cent to 9 per cent. Biden's lead among registered voters would fall by a net 3 percentage points if these voters stayed home and did not vote by other means.
It is important to emphasise that no-excuse absentee voting, in which any voter can request a mail ballot, is available in all six of the battleground states included in the Times/Siena data. Many of these voters would probably manage to navigate the absentee ballot process and successfully vote by mail, although it is impossible to say just how many.
Yet these voters appear less likely to vote under any circumstances. Only 40% of those who said they would feel too uncomfortable to vote turned out in the 2018 midterm elections, compared with 65 per cent of those who said they would still vote. Similarly, just 35 per cent of those who said they would feel too uncomfortable to vote said they were "almost certain" to turn out, compared with 66 per cent of all other voters.
Their weaker track record of voting indicates that many might not vote anyway, coronavirus or no. And lower political engagement may mean they are less likely to seek out and mail in absentee ballots.
For pollsters, the coronavirus makes the task of modelling the likely electorate even more challenging than usual.
Most political surveys estimate the composition of the electorate in one of two ways: They ask respondents whether they will vote in the coming election or look at whether respondents voted in the past, based on their voter registration file record.
Both approaches have disadvantages. A vote-history-based model might be biased if turnout patterns change considerably from a prior election. Simply asking respondents whether they plan to vote has several problems of its own: Voters tend to overstate their likelihood of voting, and poll respondents tend to be more politically engaged than demographically similar voters who do not take surveys.
At this early stage, neither approach would be likely to account for any eventual coronavirus effect on turnout. A vote history model most certainly would not, although absentee ballot requests and returns could ultimately be helpful closer to the election. It also seems unlikely that voters in June would think much about the coronavirus when evaluating their intention to vote in November, though here again it seems likelier that voters would do so as voting nears.
Either way, Biden would maintain a significant lead among likely voters in the current wave of Times/Siena battleground state surveys using these standard practices. His lead shrinks by about half of a percentage point whether one uses a vote-history-based model or a probabilistic likely-voter screen based on a voter's stated intention to turn out (say, a 90% chance if they are "almost certain" but just a 20% chance if they say they are "not at all likely"). This does not take into account whether respondents said they would feel too uncomfortable to vote in person.
One factor that is not included in most likely-voter screens is enthusiasm. It is certainly possible that the candidate with an enthusiasm edge might hold a turnout edge, but an unenthusiastic vote counts just as much as an enthusiastic one, and most registered voters show up in presidential elections anyway.
Even so, Biden does not appear to be at a meaningful enthusiasm disadvantage: 65% of his supporters said they were "very enthusiastic" about voting this November, compared with 66% of Trump's supporters. Importantly, the survey question asked whether respondents were enthusiastic about voting, not whether they were enthusiastic about the candidate they supported — where Biden appears to be at a more significant disadvantage.
Perhaps more surprising, Biden also enjoys a 9-point lead in a vote-history-based model of the likely electorate, even though this electorate is whiter, older and more Republican than the battleground electorate as a whole, reflecting surprising strength among voters with a robust track record of voting. He also enjoys a 9-point lead using the standard Times/Siena likely voter approach, which blends vote history and self-reported intentions.
The lack of an overall gap between registered and likely voters obscures some modest underlying regional variation. Biden is at a modest turnout disadvantage in the Sun Belt states, where he depends on young and nonwhite voters, which are two typically low-turnout groups. Trump, on the other hand, appears to be at a slight disadvantage among likely voters in the Rust Belt, where he depends on the support of white voters without a college degree, another low-turnout group.
Of course, all of this could change before the election. Biden's position among likely voters may deteriorate if he loses ground among high-turnout older voters or, less likely, high-turnout college-educated voters. And anyone could become more or less likely to vote, even if no one's opinion on a candidate changes. For these reasons, many public pollsters do not report results among likely voters until after Labor Day.
The coronavirus might require even greater caution this cycle.