• Race remains too close to call

• Either side could plausibly sweep the swing states

• 2024 electorate may be harder to read than ever

• Follow our 2024 US Election live blog here

• Get the latest data including current prices and volume of bets in our daily update

• View Betfair Exchange politics markets as you have never seen before at Betfair Predicts

Two weeks out, the signals from Betfair markets are increasingly clear. Donald Trump is favourite to regain the presidency. But that cannot be said about the polls.

Trump has definitely had some better numbers of late, but the race remains too close to call. Fivethirtyeight have Harris ahead by 48.2% to 46.4%. That leaves 5% still left to decide, or voting third party. In the seven key states, all deficits are within the margin of error.

Therefore it would take very little to produce a sweep in the seven swing states for either side. Here's five factors that could very easily skew the polls to a small, yet transformative, degree.

1: Shy Trumpers

This theory has been common throughout Trump's nine years in politics. That there is a group of voters who support Trump, who either won't say so publicly or engage with pollsters. It could explain why he outperformed the polling margin in both 2016 and 2020.

I find the second part more plausible. Trump supporters are rarely shy in my experience, but there probably is a vast pool of voters who completely distrust pollsters and are therefore very hard to reach and measure. His 75M vote tally was not something that one could have predicted based on polls.

Another take on those two elections, as explained in this article, is simply that the undecideds broke strongly for Trump. Perhaps they could be described as Conservative-leaners who dislike Trump, but chose him as the lesser of two evils. If so, applying the theory to 2024 is problematic. There are fewer undecideds and it isn't clear which side they will ultimately back. Those 'Trump-sceptic Conservatives' - embodied by Nikki Haley voters in the primary - look just as likely if not more so to end up with Harris.

2: A Red Mirage

If Republicans were underestimated in 2020, the exact opposite was true in 2022. The narrative was swamped with predictions of a 'Red Wave' at the midterm elections and polling averages moved sharply towards them during the campaign. As it turned out, the wave was a mirage. The Democrats retained control of the Senate and, while Republicans did reclaim the House of Representatives, it was by nothing like the projected margin. This despite a turnout advantage, as is standard in midterm elections for the party which doesn't hold the presidency.

Three plausible explanations spring to mind. First, the polling averages were skewed by a disproportionate of lower quality, Republican-backed pollsters. The same is true this cycle, although it isn't clear to what extent it has shifted the averages to Trump. Grant Elliot Morris reckons by 0.3%. But it has impacted the narrative.

More on the huge R effort to push the election narrative to Trump. Red wave 2024 much bigger, more aggressive than the 2022 version. https://t.co/Qy8L9lMfwq

-- Simon Rosenberg (@SimonWDC) October 21, 2024

Is a flood of right-wing polls biasing averages?
We ran the numbers: Even if you use a very expansive list of problematic pollsters, the average change in poll averages is just 0.3 points towards Trump.
The polls are close no matter how you look at ithttps://t.co/Eyfcgb1Uym

-- G Elliott Morris (@gelliottmorris) October 18, 2024

Second, pollsters may have adjusted samples to correct their 2020 misread. A common mistake is to refight the previous election. Third, in a reversal of 2016 and 2020, the late deciders broke for the Democrats, rejecting extreme MAGA candidates, who underperformed across the board.

However, one important warning about comparing 2022 to 2024, is that Trump wasn't on that ballot. Imitators of his MAGA-style simply don't have the same level of cut-through, celebrity or devotion.

3: Female turnout may be underestimated

The biggest single reason for polling misses is differential turnout. Where one side is more motivated than the other. It explains Trump's win in 2016, the Brexit referendum, Theresa May's Tories losing their majority in 2017. Ever since the Dobbs Amendment, reversing the Roe vs Wade ruling on abortion, Democrats have thrived electorally, driven by women voters.

Women usually make up around 52% of the general electorate. If higher - and there is speculation of 54/55%, expect Democrats to overperform polling expectations. Taylor Swift's endorsement of Harris sparked a spike in registration among younger women. Note too that abortion is on the ballot in two of the seven swing states - Arizona and Nevada.

4: Trump's new coalition may be understated

This deserves a different category to 'Shy Trumpers' even if the effect would be similar. As I wrote in Saturday's piece regarding my increasing lack of confidence in Harris winning, Trump's pool of voters is changing, and perhaps expanding.

Politics and discourse has changed dramatically in recent years. From the QaNon conspiracy Cult to Covid-denial/anti-vaxxers to cryptomania, far-Right politics has targeted new segments of the electorate. Younger men in particular - a group that generally hasn't turned out as reliably as others. Observe Trump's campaign almost explicitly rejecting mainstream, legacy media and conventional strategy in favour of podcasts aimed at these audiences. They may be on to something.

Their numbers may well be under-sampled by pollsters. However, this cohort is more evenly spread around the country. It seems that he has fallen back sharply among some core groups - seniors, college educated whites, suburban voters - yet moved forward overall thanks to men, including minorities. The effect may diminish his advantage in the electoral college, meaning Harris only needs to win the popular vote by 1-2%.

5: Herding

One of my sharpest followers on Twitter, @Markgibson2018, raised this matter a few weeks ago and there does seem something slightly fishy about this year's polls. They seem remarkably consistent. Could this be an example of 'herding'?

This is term given to the process where pollsters use results from existing polls to adjust and normalise their own work in an effort to avoid being an outlier and standing out from the crowd.

Perhaps the worst polling miss in living memory came in the 2015 UK General Election. Virtually every poll showed a near statistical dead-heat between the Conservatives and Labour. A hung parliament seemed certain and traded very short in the betting. Yet the Conservatives ended up winning by 6% and an overall majority. When the industry conducted a post-mortem, 'herding' was the conclusion.

If any election were to incentivise such behaviour, it is this one. The narrative is so vicious nowadays, any pollster whose results differed markedly from the crowd would be targeted. Fear of reputational damage may well be a factor. I'm not in a position to judge whether herding is happening this time, but I am not ruling out one side performing 3-4 points better than expected.

Now see the latest US predictions from Betfair predicts