With only a few hours left until Election Day in the United States, pundits and the public alike are scouring opinion polls for early signs of an outcome.

We had months of campaigning and HUNDREDS of surveys. However, there are still few conclusive conclusions other than the unsatisfying slogan of election analysts around the world this year: “it’s too close to call”.

So why does this happen? And what to make of surveys?

The survey is not predictive

Both campaigns have at various times posted favorable — and conflicting — poll numbers. The problem is that no one seems to know which polls to believe.

Sunday, a new Des Moines registry POLL led by survey respected Ann Selzer showed Kamala Harris a surprising three-point lead over Donald Trump in Iowa, giving Harris’ anxious campaign an unexpected boost.

Within hours, a “confidential” Trump campaign memo dismissed Selzer’s numbers. Trump himself he posted on Twitter Favorable poll from AtlasIntel showing him leading in all seven swing states.

In recent cycles, despite glitches in some key states — notably Wisconsin, in 2016 and 2020 – survey averages tended to be relatively accurate estimates of public opinion.

However, there is little to be gleaned from the current swing state margins that are all within the margin of error, other than what we already know: Americans are deeply divided about their choices.

Part of the reason is that polls are not predictive. They are a measure of popular sentiment at the time of the poll from which educated guesses can be made about who might win the next election.

But the margins of error (which are significantly higher than is generally understood), combined with regularly thin final votes in key states and the win-all nature of the Electoral College, limits their ability to predict electoral winners.

The magnitude and direction of the polling errors are unforeseeablemostly because they are often not uniform across the country and historically do not favor one party more than another.

Small misses have a tremendous impact

Methodologically, accurate election polls are hampered in the United States by high non-response rates and nonbinding voting (which requires weighting responses based on anticipated likely voters).

Errors in these assumptions were central to survey errors in 2016 and 2020.

Polls in 2016 underestimated Trump’s support by failing to control for education in their samples.

This meant that they lacked his support white voters without a college education that helped propel him to victory in the Midwest.

But while 2016 is remembered as a catastrophic polling failure for apparently not predicting a Trump victory, the poll averages before Election Day were, in fact, largely accurate.

The national polls were among the most accurate in the 80s, exaggerating Clinton’s popular vote margin by only about a percentage point.

In the ten closest states in the 2016 election, Trump was underestimated by an average of just 1.4%.

Only in a handful of key states such as Wisconsinalthough significant, it had a tremendous impact on the bottom line.

The margins were such that a difference of just a few points of polling error was enough to flip the so-called blue wall states and deliver what most analysts had considered an unthinkable victory for Trump.

However, it was the polls – rather than analytical mistakes and poor media reporting – that bore the brunt of the blame for failing to convey that the unthinkable was actually quite likely, despite the margins of error that showed it was a strong statistical probability of another situation. result.

The polls were much wider in 2020, but avoided the same level of public scrutiny because they correctly “predicted” (if narrowly) a Biden victory.

Have the investigators solved the defects of the past?

Based on the survey AVERAGE in 2024, a uniform margin of error in Trump’s favor of less than 0.8% in the seven key swing states could give him a comfortable 312–226 electoral college victory.

Also, an even swing of less than three percentage points in favor of Harris could give the Democrats a similarly lopsided 319–219 victory.

The critical question is whether pollsters have done enough to change flawed methodologies that underestimated Trump’s support in 2016 and 2020.

Changes to some surveys include an increase in hybrid sampling methods (combining both telephone calls and online interviews) and weighting of past votes.

But such methods could be overcompensating for past and present mistakes underestimating Harris support by sampling them properly.

Until the votes are counted and the winner is declared, there is no sure way to know.

Polls can’t do everything

The polls remain remarkably accurate, given that it is an attempt to divine the opinions of hundreds of millions of people. But it is a mistake to think that polls are predictive or determinative.

The vagaries of the polls—and especially the vagaries of the US electoral system—mean there’s little point in reading the tea leaves beyond the broad public sentiment that the polls capture.

On those numbers, the result could really come down to a handful of votes, or we could see a lopsided electoral college victory instead.

Anyway, don’t blame the polls.