Democrats on edge: Tight polls fuel election uncertainty

Today, American polls are unable to indicate the winner of the race clearly. Looking at their results, when asked who will win the American election, one can only answer: "it's really impossible to say." This state of uncertainty must cause nervousness in both campaign teams. However, more signs of concern are coming from the Democrats. Does Kamala Harris have reasons to worry?

Not entirely, when looking at the poll numbers alone. In the past week, there hasn't been any reliable poll showing Trump's lead—either nationwide or in key states—that would be significantly higher than the statistical error margin.

As reported by the American "Forbes" on Thursday, out of nine polls from the last week, Harris is leading in six, and Trump in three. In both cases, however, the advantage falls within the statistical error margin. The polling average, calculated ongoing from credible studies by the "New York Times" on Thursday, showed support at 49% for Harris and 48% for Trump.

Of course, elections in the United States are not won by the popular vote. The Electoral College decides, and each state sends a number of electors equal to its congressional delegation, from 3 in states like Vermont and Wyoming to 53 in California. The candidate who wins in a given state, with two exceptions, takes all the electors entitled to them.

In practice, this means that the victory is decided by several states that can vote either way. This year, they include three states from the so-called Rust Belt—Midwestern areas that once formed the industrial heart of the country—Pennsylvania, Michigan, and Wisconsin, and four "Sun Belt" states, those located below the 36th parallel: North Carolina, Georgia, Arizona, and Nevada.

According to the polling averages of the "New York Times," we also have a practical tie in all these states. In the "Rust Belt" states and Nevada, Harris leads, but her advantage over Trump never exceeds one percentage point. In the three "Sun Belt" states, Trump is leading, but only in Arizona does his advantage approach 2 percentage points—still within the margin of error.

Such results could mean anything. Both a victory for Harris and for Trump, an election that is decided by a very small majority in the Electoral College, determined by tens of thousands of votes in several states, as well as a decisive victory for one of the candidates—since current ties could just as easily translate into a clear win with a difference of 2-3 percentage points in each of the 7 decisive states.

Why, then, does this tie cause more concern for the Democrats? Who do the polls underestimate?

Primarily because until now, polls have tended to underestimate Trump. Public opinion centres in 2016 poorly predicted the Republican's victory. In 2020, they correctly predicted Biden's victory, but its scale was smaller than the polls had indicated. The Democrat won the popular vote by 4.5 percentage points, not nearly 8.

There are various theories as to why this happens. One says that Trump voters are ashamed to admit that they support him and hide their preferences from pollsters.

Another theory says that Trump disproportionately mobilises voters characterised by a very low level of trust in all public institutions—including public opinion centres—and low civic participation. For these reasons, they may refuse to participate in surveys at all.

Some analysts argue that the source of problems in correctly estimating Trump's support lies in the fact that the Republican disproportionately mobilises voters with a specific demographic profile—predominantly white men without higher education—and the way public opinion centres select their survey sample doesn't properly account for the strength of these groups' voices.

Besides the fear that Trump might be underestimated again, Democrats may also be concerned about the so-called Bradley effect. The name comes from Tom Bradley, the first black mayor of Los Angeles, who performed worse in the 1982 California governor's election than polls indicated—in some analysts' opinion because voters who didn't want to support him didn't reveal their true preferences to pollsters, fearing they'd be accused of having an issue with a black candidate.

On the other hand, since 2016, public opinion centres have worked hard to correct their models to better measure Trump's support. Some analysts warn that they may have overdone it and could now artificially underestimate Harris's support. As data analyst Nate Silver wrote in the "New York Times" on Thursday, centres might simply be afraid of underestimating Trump's support again and indicate a tie as a precaution to protect their reputation. At the same time, Silver predicts that Trump has a better chance of winning.

Meanwhile, early voting is already underway. About 26 million voters have already voted in person or by post. We don't know who they voted for, but we can observe trends related to turnout and whether more votes are cast by registered Democrats or Republicans. In many states, voters must register as voters of a specific party or as independent—though, of course, registered Republicans can vote for Kamala Harris and Democrats for Trump.

As reported by the British "The Economist" on Thursday, the majority, nearly 48%, of votes cast early come from registered Democrats. The share of votes by registered Republicans is significantly smaller, currently at 32%.

At the same time, it's visible that compared to 2020, the number of early votes cast by Republicans has significantly increased—at this stage of the campaign four years ago, registered Republicans cast about 27% of the votes. The level of votes by Democrats is the same as four years ago.

Does this indicate greater Republican mobilisation this year? Not necessarily. Four years ago, Trump attacked early voting and mail-in voting and cast doubt on their integrity, which certainly discouraged Republican voters. Though in the last four years, Trump claimed it was irregularities in mail-in voting that cost him "victory," during the campaign, Republicans and Trump himself encouraged early voting.

Until all votes are counted, we won't know whether the visible increase in votes for Republicans means that Trump managed to mobilise new voters who vote early or that voters who would have voted for the Republican on 5th November decided to do so before the official election day.

Early voting indicates that turnout will likely be substantial. Traditionally, high turnout favoured Democrats and low turnout favoured Republicans. This year, that pattern may not hold. Trump performs better in polls considering voters who are not entirely decided, or who will even vote. He attracts voter groups—such as young Black and Latino men—characterised by low electoral participation.

In other words, individuals declaring they will vote for Harris are more likely actually to do so than those declaring a vote for Trump.

Nonetheless, more and more election simulations—calculating possible election outcomes based on data considering multiple polls and other data—predict that a Trump victory is more likely.

However, Kamala Harris should not be counted out by any means. Her campaign has raised more money and still has some time to present to American voters why Trump is not the right person for the nation's highest office. Trump has the ability to mobilise fanatical supporters around him but also to mobilise opponents. The Supreme Court ruling against American women's constitutional right to abortion, possibly thanks to the very conservative judges he appointed, weighs heavily on him.

Aside from the sensational success in 2016, Trump and the Republican party closely associated with him have lost all subsequent elections or performed below expectations. Of course, if Trump wins on 5th November, no one will remember the Republican defeats from 2018-22—all that will matter is that against all odds, Trump returns to the White House four years later.