Unless the US polls have messed up for the third consecutive presidential election—and later in this blog we shall discuss why they might have done—three lessons can be learned from the battle between Kamala Harris and Donald Trump.
First, it is extremely close. Polls conducted by more than 20 different companies in the past fortnight show Harris ahead by one percentage point on average. They range from a Trump lead of 3 per cent to a Harris lead of 4 per cent.
Second, the race has tightened in the past six weeks. In late August on the 538 polling analysis website, which combines results from all the main polling companies, Harris’s average lead was 3.5 per cent. Measuring the average irons out the fluctuations to which all polls are prone.
In the British general election four months ago, the polls disagreed on the levels of Labour or Conservative support at any one time, but almost all agreed about the reduction in Labour’s lead towards the end of the campaign. The same has happened in the US.
While that is good news for Trump, the third lesson has better news for Harris. In 2016 and 2020, Trump could win if he kept the Democrats’ national lead to less than three percentage points. He succeeded in 2016, when Hillary Clinton led by 2.1 per cent in the popular vote, but failed in 2020 when Joe Biden led by 4.4 per cent.
This time, the break-even figure may be less than 3 per cent, possibly between 1 per cent and 2 per cent. If Harris wins by the same margin as Clinton, she has a good chance of going one better than her predecessor and winning the electoral college.
There is a simple reason for this. If the latest polls are right, then Democratic support has held up better in the swing states than in the rest of the country. This is particularly noticeable in the Democratic strongholds of New York, California, Massachusetts and Maryland. These four states supplied more than one quarter of Biden’s support four years ago. Harris will win them all—but the polls suggest that her average margins of victory in these states will be six points lower than Biden achieved in 2020. If so, Harris is doing what Labour did in Britain four months ago: losing votes where they matter least, and doing better where they matter most.
If the polls are right… there’s the rub. Most (but not all) of them underestimated Trump support in 2016 and 2020. Might they be doing the same? Alternatively, in their effort to avoid the mistakes of the past, might they now be overestimating his support this time?
Over the weekend, a remarkable poll sent American poll-watchers into a frenzy. It showed Harris winning Iowa by 3 percentage points. Normally, a poll in Iowa would elicit nothing more than a slight shrug. The state has only six electoral college votes. It’s not a swing state. Neither candidate has bothered to go there during the campaign.
The thing is, Trump won Iowa last time by 8 per cent. Two different polls conducted within the past week tell very different stories. One, by Emerson College, put Trump 9 per cent ahead. It used similar methods to those by mainstream pollsters across the country. It weighted its data to a host of demographic factors, such as age and gender—and also education and party registration.
The poll showing Harris ahead was conducted by Selzer Polls, an Iowa company. According to its client, the Des Moines Register, it has weighted its data by age, gender and Congressional district (Iowa has four of them)—but not by education or political affiliation (party registration or past votes).
Normally, I would dismiss such a poll. Too much can go wrong if it has no way of adjusting for these two extra factors. However, Selzer has a remarkable record of getting Iowa right. Its final figures for 2016 and 2020 were significantly closer to the result—that is, more favourable to Trump—than other pollsters. Its figures for the 2014 and 2022 Senate races were also impressively accurate.
Has Selzer done it again? By not adjusting for education or political affiliation, has it achieved a better sample while conventional pollsters, in their attempt to avoid an anti-Trump bias, inserted an anti-Harris bias? If Selzer is right about Iowa, Harris should win comfortable victories in other Rust Belt states, such as Michigan, Pennsylvania and Wisconsin. These would doom Trump’s hopes of returning to the White House.
I know what I want to believe. It is probably what most readers of this blog want to believe. But my heart and head tell me different things. We shall know soon enough.
