Politics: polls apart
By Richard North - November 7, 2024
Now it’s over bar the shouting, one of the more interesting things to emerge from the US presidential election is the performance of the opinion polls.
Despite the polls well into the night last night predicting that the contest was “too close to call”, we emerged into the small hours with indications that Trump was going to sweep to victory.
At the time of writing, the score being shown by the bulk of the media is 294 electoral college votes for Trump against 223 for Kamala Harris, with three states yet to declare. By the time those results are in, the final score will probably be in the order of 311 to Trump and 227 for Harris.
The one thing we can say about that result is that it is anything but close. It represents a clear, decisive win for Trump. Moreover, Trump is the first Republican to win the popular vote, polling 71.85 million votes compared with the 66.98 million polled by Harris.
Currently, the Republicans have also taken the Senate. The House of Representatives hangs in the balance with the Republicans narrowly in the lead taking 205 seats against the Democrat’s 190, 13 short of the winning margin of 218.
If the Republicans do take the House, it will be a clean sweep giving Trump the so-called “trifecta”, a status not enjoyed by a Republican leader since the first term of George W Bush – although Trump in his first term had a trifecta for the first two years as the Democrats only had a technical majority in the Senate – although the kept control of the committees.
There will be no such ambiguity of power if the House falls to the Republicans, giving Trump enormous power to pursue his agenda – a situation that was not being anticipated by the bulk of the pundits at the outset of the voting.
Bloomberg makes the obvious points, stating that, to anyone following the presidential election, the polls made the race out to be a real nail-biter that could take days, if not weeks, to decide.
Yet, the agency says, on the morning after Election Day, Donald Trump has won convincingly, besting Kamala Harris in five swing states and leading her in the other two. His win is part of a broad American shift to the right: Republicans have taken control of the Senate and are poised to retain their majority in the House.
This, it asks; “Did the polls get it wrong?”, answering its own question by saying: “Well, it certainly doesn’t feel like they got it right”. But, it concedes, “it’ll take weeks or months for the industry to do a full autopsy to know what went wrong, and just how wrong things went”.
That hasn’t stopped the Telegraph piling in with an article headed; “Why pollsters got the election wrong… again”, the sub-heading telling us that: “Just two predictions could come close to the results, British firm JL Partners’s election forecast and The Telegraph’s own poll”.
Setting out some of the detail, the paper records that, in North Carolina, when 97 percent of the vote had been counted, Trump had been predicted to have a 1.3-point lead in the final polling averages collected by Real Clear Polling. He exceeded that by three points.
Pennsylvania polling saw him just 0.2 points ahead – hence the “too close to call” mantra. But it appears he has won by around three points with 95 percent of votes in. In Wisconsin and Michigan, he was predicted to lose to Harris, but is currently on track to win both.
On average, Trump performed around three points higher in the seven crucial swing states than polls suggested, these results turning the tide of the election.
Technically, though, the paper says, pollsters will take the outcome as a victory. Three points, plus or minus, lies within the margins of error and they did correctly call the majority of swing states for Trump. Nevertheless, the majority of the polls got the result wrong.
The Economist forecast a narrow victory for Harris and polling accumulator FiveThirtyEight forecast a Trump victory with 276. The British polling firm JL Partners emerged as one of the most bullish in favour of Trump, with a projected 287 electoral college votes.
It is acknowledged that some demographics have been missed in their support for Trump. Latino voters are one group, where their support increased by 13 points for Trump, something that no polls have picked up on. Some groups, such as women and African Americans, were predicted to swing towards Harris but that didn’t happen.
Taking up the same theme, The Times has a slightly more analytical piece with the title: “Why were polls wrong about Donald Trump once again?”, explaining in in the sub-heading that “pollsters have underestimated Trump’s white working-class voter base as their survey methods cater too much to Democractic (sic) voters”.
This view comes from Douglas Rivers, chief scientist for YouGov. In an age where voters are less engaged with political news and ignore phone calls from cold callers, pollsters agreed before election day that reaching an accurate cross-section of the US electorate would be harder than ever.
They did attempt to correct the weighting of their polls in the aftermath of 2020, when the predictions were the least accurate for 40 years. But there were then fears of an over-correction, with suggestions that polls were overestimating Trump’s support. Instead, the opposite was the case. The mix of voters in online panels and phone surveys was off again.
Even the mighty have been humbled by the magnitude of Trump’s victory, we are told. The revered pollster Ann Selzer showed Harris with a four-point lead in Iowa last weekend, sparking feverish excitement among Democrats. In the end, Trump triumphed in Iowa by 13 points with 55.8 percent of the vote, greater than his share in 2016 and 2020.
The Selzer poll is also picked up by the Independent, echoing the other two papers with the heading: “Why the polls got it wrong on Trump — again”, telling us that “An embarrassing night for forecasts raises questions about how to fix broken lines of communication”.
Kamala Harris’s collapse on Tuesday evening, the paper says, was shocking to many in the political press, who had expected a tight election but one that, if anything, looked to be trending in her favour heading into election day.
This paper relies on the founder of J.L. Partners, James Johnson, who argues that understated the Trump voter who is less likely to be engaged politically, and crucially, more likely to be busy, not spending 20 minutes talking to pollsters… people working a pretty common job or, as the case of many Hispanic voters, juggling two or three jobs at a time.”
Whether underestimating Trump support or overestimating Democratic turnout, both may have the same cause. Polling firms are increasingly bad at contacting less politically engaged voters. And unlike the talking heads which make up the political media, most voters are not highly engaged with politics.
These voters are far less likely to be picked up accurately by a pollster. Selzer, for instance, famously refused to contact voters via any method besides a live call – in an age of seemingly nonstop spam calls.
The paper thus argues that some pollsters and, more importantly, the political media needs to get a lot better at talking to low-propensity voters and people who are tuned out from the political news bubble.
Yet, for all that, when we get closer to the action, with the New York Times, the tone changes completely. “Early results”, it headlines, “suggest the polls were notably accurate”.
Collectively, the paper adds, the polls indicated that the presidential race would be close, and it appears they got it right without significantly underestimating Donald Trump’s support.
Of course, the polls weren’t perfect, the paper concedes; they never are. And they sometimes showed Harris with a slight edge when it turned out that Trump had the narrow advantage. But a lead of one point in polling is best interpreted as, effectively, a tie, and the polls’ real value in this election cycle was showing that the race was, in fact, very tight.
While some social media users on Wednesday lamented or mocked the pre-election polls for not predicting Trump’s win more definitively, Peter Enns, a professor of government at Cornell and co-founder of the polling firm Verasight, says that’s neither the intention of polling nor a real possibility in such a closely contested race,
“When we distil it down to, ‘did they get the winner in a relatively close election’, it’s a little bit of an unfair test”, Enns says. Pre-election polls can be very accurate, but their precision is limited because a certain amount of error is baked into the process. Only a sample of the population is polled, and pollsters can make only an educated guess about who is going to cast a ballot.
Thus observes Lakshya Jain, an election modeler with the analytics website Split Ticket, “People want polls to deliver a level of precision that they just can’t”. He concludes that a miss by two or three points is “as good as polls can really get”.
So, we all have it wrong. The polls did a jolly good job, and it’s all our fault for expecting them to tell us who was going to win. After all, that’s why we actually have the voting, when we discover the electorate’s wishes.
