The US presidential election will be held on November 5. In analyst Nate Silver's aggregate of national polls, Democrat Kamala Harris leads Republican Donald Trump by 49.3-46.0 - a slight widening of the competition since last Monday, when Harris led Trump by 49.2-46.2.
Author
Adrian Beaumont
Election Analyst (Psephologist) at The Conversation; and Honorary Associate, School of Mathematics and Statistics, The University of Melbourne
President Joe Biden's final position before his withdrawal as Democratic candidate on July 21 was a national poll deficit against Trump of 45.2-41.2.
There will be a debate on Tuesday evening US time between the vice-presidential candidates, Democrat Tim Walz and Republican JD Vance. Vice-presidential debates in previous elections have not had a significant influence on the contest.
The US president isn't elected by the national popular vote, but by the Electoral College, in which each state receives electoral votes equal to its federal House seats (population based) and senators (always two). Almost all states award their electoral votes as winner-takes-all, and it takes 270 electoral votes to win (out of 538 total).
The Electoral College is biased to Trump relative to the national popular vote, with Harris needing at least a two-point popular vote win in Silver's model to be the Electoral College favourite.
In Silver's polling averages, Harris leads Trump by one to two points in Pennsylvania (19 electoral votes), Michigan (15), Wisconsin (ten) and Nevada (six). If Harris wins all these states, she is likely to win the Electoral College by at least a 276-262 margin. Trump is ahead by less than a point in North Carolina (16 electoral votes) and Georgia (16), and if Harris wins both, she wins by 308-230.
In Silver's model, Harris has a 56% chance to win the Electoral College, up from 54% last Monday but down from her peak of 58% two days ago. Earlier this month, there were large differences in win probability between Silver's model and the FiveThirtyEight model, which was more favourable to Harris. But these models have nearly converged, with FiveThirtyEight now giving Harris a 59% win probability.
There are still more than five weeks until election day, so polls could change in either Trump's or Harris' favour by then. Harris' one to two point leads in the key states are tenuous, and this explains why Trump is still rated a good chance to win.
Silver wrote on September 1 that polls in 2020 and 2016 were biased against Trump, but polls in 2012 were biased against Barack Obama. In the last two midterm elections (2022 and 2018), polls have been good. It's plausible there will be a polling error this year, but which candidate such an error would favour can't be predicted.
On Sunday, Silver said if there was a systematic error of three or four points in the polls in either Trump's or Harris' favour, that candidate would sweep all the swing states and easily win the Electoral College. There are other scenarios in which one candidate underperforms the polls with some demographics but overperforms with other demographics.
I wrote about the US election for The Poll Bludger last Thursday, and also covered bleak polls and byelection results in Canada for the governing centre-left Liberals ahead of an election due by October 2025, a dreadful poll for UK Labour Prime Minister Keir Starmer, the new French prime minister, a German state election and a socialist win in Sri Lanka's presidential election.
Upwardly revised economic data
Last Thursday, a revised estimate of June quarter US GDP was released. There was a large upward revision in real disposable personal income compared to the previously reported figures. This has resulted in the personal savings rate being revised up to 4.9% in July from the previously reported 2.9%, and it was 4.8% in August.
With these upward revisions, Silver's economic index that averages six indicators is now at +0.25, up from +0.09. As the incumbent party's candidate, a better economy than was previously believed should help Harris.
Coalition gains narrow lead in Essential
In Australia, a national Essential poll, conducted on September 18-22 from a sample of 1,117 people, gave the Coalition a 48-47 lead (including undecided voters) after a 48-48 tie in early September. It's the Coalition's first lead in the Essential poll since mid-July.
Primary votes were 35% Coalition (steady), 29% Labor (down one), 12% Greens (down one), 8% One Nation (steady), 2% UAP (up one), 9% for all Others (up one) and 5% undecided (steady).
Anthony Albanese's net approval was up five points since August to -5, with 47% disapproving and 42% approving. Peter Dutton's net approval was down one to net zero.
On social media regulations, 48% thought them too weak, 43% about right and 8% too tough. By 67-17, voters supported imposing an age limit for children to access social media (68-15 in July). By 71-12, voters supported making doxing (the public release of personally identifiable data) a criminal offence (62-19 in February).
By 49-18, voters supported Labor's Help to Buy scheme, and by 57-13 they supported the build-to-rent scheme. The questions give detail that few voters would know.
Voters were told the Liberals and Greens had combined to delay Labor's housing policies in the senate. By 48-22, voters thought the Liberals and Greens should pass the policies and argue for their own policies at the next election, rather than block Labor's policies. Greens voters supported passing by 55-21.
Labor keeps narrow lead in Morgan
A national Morgan poll, conducted September 16-22 from a sample of 1,662 people, gave Labor a 50.5-49.5 lead, unchanged from the September 9-15 Morgan poll.
Primary votes were 37.5% Coalition (steady), 32% Labor (up 1.5), 12.5% Greens (steady), 5% One Nation (down 0.5), 9.5% independents (down 0.5) and 3.5% others (down 0.5).
The headline figure is based on respondent preferences. By 2022 election preference flows, Labor led by an unchanged 52-48.
Adrian Beaumont does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
