How accurate are the latest Trump and Harris polls?

WWhen Donald Trump left office, the Gallup polling firm reported that he had a job approval rating of 38 percent – a pathetically low number.

A few weeks ago, NBC asked the same question and the answer was a far more credible 48 percent. As fickle as public opinion can be, it can change.

People thought a Joe Biden presidency would offer all the qualities Trump didn't have, like consistency and solidity. When that didn't happen due to Biden's weakness, they were seemingly reminded of Trump's energy. All of his negative aspects – dark conspiracy theories, his flogging, a revolving door in the White House, repeated blatant deception, breaking the law – were forgotten. Suddenly more Americans remembered Trump fondly.

If reason were left to reason, Trump would be nowhere to be found in this election. He carries too much negative baggage with him. But just as Trump defies gravity when it comes to his popularity rating and is able to rewrite history, logic does not determine the outcome of the presidential election. Instead, it depends on all sorts of factors.

The polling institutes are responsible for understanding everything. They say this year's contest is too close to call. When it comes to forecasting, there's been a sort of up-and-down battle all along – one minute one side is rising slightly, the next it's falling.

The latest simulation from the FiveThirtyEight National Polls Tracker (the poll of all polls) puts Trump winning 53 times out of 100 and Harris winning 46 times out of 100. That's how close it is.

But that calculation didn't take into account the local poll in Iowa, which showed Harris with a three-point lead. Yesterday's finding in Iowa threw the balance into disarray, as it is not a swing state but a safe Republican state – and therefore, it was assumed, fixated on Trump. But apparently not if this poll is correct and the reasoning is that many women are moving to Harris, deterred by Trump's support of the anti-abortion law.

The truth is: nobody knows. Even Nate Silver — the doyen of pollsters — isn't so sure anymore. Last week he chose Trump, today he feels a late Harris upswing.

Apparently everyone is expecting it to be close, based on the last two presidential elections where the outcome was determined by 78,000 and 44,000 votes respectively.

But each of the seven critical conflict states has scope or enough scope for the typical poll error, traditionally estimated at 3 percent. It's possible that everyone will go the other way when the count comes, and there's also the possibility that all seven will go one way, so the election won't be close at all.

It may be even more pronounced, as studies suggest that the widely accepted textbook rate of 3 percent may be too low. The actual error rate could be more than double that, at 7 percent, according to an analysis that compared the way people said they would vote with what actually happened.

But because of their inaccuracy, the surveys dominate the coverage – and not only that, since Wall Street and City mutual funds have conducted their own private surveys on which their strategies are based.

Why are we so dependent on them? Because they are all there is. There is no other way to estimate what will happen – nor one that would stand up to scientific scrutiny.

For polling companies, building a reputation for accuracy is their lifeblood. Elections are only a small part of their work, albeit the part that receives the most attention. Your day-to-day job involves surveying consumers about their preferences and views on behalf of paying customers. It is an important industry that new companies are constantly entering as technological advances reduce surveying costs.

The pressure in 2024 will be increased by underestimating support for Trump in the last two presidential elections. This time they really want to do it.

However, that is easier said than done.

In the past, voter forecasts were more reliable because households had landlines and people answered their phones. These days we ignore random, unrecognized calls or text messages, partly out of fear that they might be scammers. This leads to an astonishingly low response rate. One of the most respected ones New York Times/Siena survey has a response rate of just… 1 percent.

This also means that those who respond and say what they mean are definitely in the minority, they are the outsiders. Nobody has any idea who the majority supports.

Pollsters get around this mismatch by weighting responses and using variables like age, gender and race to relate to the broader public. In theory, it allows them to infer how the population will vote from a small sample.

Some surveys this year used up to 12 variables. What they obscure is that the estimate is ultimately little more than an educated guess.

Luckily the clock is ticking. There's only one day left until we know. At the moment, only one thing is certain: that either Harris or Trump will be the next president.