Last-minute polling in the Midwest swing states seem to have given a false impression of the strength of support for Kamala Harris.

Dr Michael Turner – Pollster and strategist

Although they did better than in 2016 and 2020, for a third time in a row, Donald Trump appears to have confounded US pollsters, who, it’s fair to say, are having a hard time reaching Trump supporters.

While polls can only ever present a snapshot in time, never a forecast, it’s hard to stomach the sheer number of late polls that piled in at the death of the US presidential election campaign and pointed to Kamala Harris being highly competitive.

This is particularly problematic given the outcome was so comprehensive for the Republicans, who now control the White House, the Senate, and likely the House of Representatives.

Of the 18 polling organisations that published a nationwide poll in the last week of the presidential election, just three gave a Trump lead. The day before election day, polling aggregators put Harris about one to two points ahead of Trump on the popular vote, and 48 hours later Trump emerged 2.5 per cent ahead of Harris, with a lead of about 4 million votes.

There remains no real explanation for such a consistent divergence between the late polls and the result, other than error. It’s especially concerning when we are reminded that a majority of Americans cast their ballot early (that is, before the late polls were published), and pollsters seem to have underestimated Trump’s vote share consistently for three elections in a row. Not a good look.

The statewide polls were a mixed bag. For the Sun Belt swing states of Arizona, Georgia, Nevada and North Carolina, polling published in the final week pretty much nailed the result. All major polls published in these states put Trump ahead by around two to three points, and broadly speaking, that’s where they landed.

It seems that most of the error, in swing states at least, was confined to several late polls, with small misses that fell on the wrong side of the outcome in all three of the crucial “blue wall” states – Michigan, Pennsylvania and Wisconsin. Then there was the bizarre last-minute release from a reputable pollster in Iowa, who missed the result by a whopping 16 points.

It was the late polls that gave misleading results.

In the 10 most competitive states, the polls overestimated Harris’ support by an average of 2.9 per cent, particularly the late polls.

We saw a similar feature at the recent Queensland state election, where several polls published late in the campaign pointed to a “close result” that was “tightening”, when it really wasn’t.

Then-premier Steven Miles and his Labor team were suckered in by these late polls. In a truly bizarre state of affairs, Miles failed to concede the election on the night and then went on to suggest that his opponents would be unlikely to reach a majority, only for all the major TV networks to call a majority for the LNP just minutes later.

The phone call to Premier-elect David Crisafulli the next morning must have been a tad awkward. Those of us staring at better-quality data on the night could see the result was never in doubt, which made the entire episode even more bewildering.

Like Miles, Kamala Harris also chose not to concede the presidential election on the night. Instead, she sent home thousands of anxious Democrat supporters from her headquarters, while she and her team worked out just what they were going to say. This doesn’t strike me as the actions of a team that had even contemplated losing. Clearly, senior Democrats chose to dismiss polls that didn’t fit their goal and grossly underestimated the risk of losing. An unwise move.

I do feel a little sorry for our American cousins, however. In Australia pollsters have just one job, while in the US they have two. Compulsory elections in Australia mean polls only have to work out the intentions of voters.

In the US, pollsters also have to work out who of them will actually turn up. A very difficult task, given most people don’t always follow up words with actions. (That reminds me, I should probably cancel that gym membership now.)

For a small jurisdiction, Australia is lucky to have the calibre, quality and accuracy of our polling industry. Compulsory voting certainly helps, in this regard, but it also helps that our leading pollsters are taken seriously across the entire media and political landscape.

As the US becomes increasingly polarised, a key feature of the echo chamber has been for partisan commentators and media outlets to start seeing polling firms as politically aligned, or publishing to push a narrative. This is a bad idea, and quite literally the opposite goal of any research organisation worth their salt.

Having worked across the political spectrum in Australia, New Zealand and the UK, my experience has been to publish polling in the media and industry without fear or favour to an alignment or agenda – the truth or be damned, interesting or dull.

After all, polling firms that deviate from objectivity, or get too caught up in the goals of their clients, will eventually fall victim to the cold hard truth of reality.

Dr Michael Turner is a pollster and strategist, and director at Freshwater Strategy which carries out opinion polling for The Australian Financial Review.

This opinion piece was first published in The Australian Financial Review. You can read it here: https://www.afr.com/world/north-america/why-did-polls-get-trump-so-wrong-again-20241105-p5ko2j