This is now a Lean Trump race
Our latest model update shows a 62% chance of a Trump victory. Here's why.
Today, the presidential race shifted. The latest update from our J.L. Partners / Daily Mail US presidential model has the race as ‘Lean Trump’ for the first time since July, when Kamala Harris entered the contest.
That means that, based on public polls of the race, we are saying that Donald Trump currently has a 62.4% chance of victory in the Electoral College, compared to a 37.5% chance for Kamala Harris. The graph below shows the current state of play.
The model, designed by data supremo and ex-physicist Callum Hunter - Senior Data Scientist at J.L. Partners - and hosted by the Daily Mail, uses public polls to extrapolate the most likely results in each state.
Generally, the polling has pushed things more in Trump's favour. Over the weekend, the NYT released polls for Arizona and Pennsylvania - the former had Trump on a 5-6 point lead and the latter had Harris on a 3-4 point lead. Redfield & Wilton also released a Pennsylvania poll with Trump on a 2 point lead. Our own PA poll found a dead-heat tie.
We are now saying:
In Arizona, Trump's win probability has increased by a further 2.6 points keeping it firmly in the LIKELY TRUMP camp
Georgia has seen a 1.6 point movement in Trump's favour, putting it on a 72.3% chance of a Trump victor. It remains a LIKELY TRUMP state
Michigan has seen Trump's chances increase by 2.1 points to 52.9%, thus it remains a TOSSUP TRUMP state
Nevada has shifted back towards Trump by 3.1 points and he is now predicted to win 60.5% of the time. The state has shifted from TOSSUP TRUMP to LEAN TRUMP.
North Carolina has moved back to Trump by 2.1 points and he now has a 72.8% chance of winning the state. The state remains a LIKELY TRUMP state.
Pennsylvania has seen a 2 point swing to Trump reaching a new high of 64%. The state remains a LEAN TRUMP state.
Wisconsin has also moved to Trump by 3.3 points and he now stands to win the state 54.4% of the time. The state remains a TOSSUP TRUMP state.
It is important to remember that this model uses correlations as well as national vote shares to obtain each state's average. This means that every state poll affects every other state - so a single poll that is good for Harris is not necessarily going to push things in that state in her favor. The model looks at the WHOLE picture of the country, not single spot polls only.
All together, that means Harris is currently not favored in any of the main swing states, and our central estimate is a Trump Electoral College win by 312 to 226.
You may be looking at this and thinking that we look like an outlier. We are, compared to other models - but we are confident in our projections. There are four things we are utilizing that others are not:
We significantly upweight full-ballot polls and downweight just head-to-head ones. This is because the third party candidates - the most that are standing since 2016 - are important to the dynamic in each states and head-to-head polls simply negate their impact. They are generally working to the detriment of Harris.
We restrict online-only polls "effective" size in the overall model so we are less dependent on them. The reason being they are notoriously more inaccurate and their samples are more biased to the Democrats.
Our prior model does not give Harris an incumbency bonus.
We use voter flows informed by our private polling to direct the full ballots to the main candidates which makes the outcomes more realistic than false head-to-heads.
Nonetheless, other models are now trending towards Trump, even if they are not as decisive as ours.
A 37.5% chance of victory for Harris is still significant. Polling error is of course possible. But this is now a race that is moving to Trump, quickly. The 2024 Election is now a Lean Trump race.
For more on the methodology on our model, please visit: https://jlpartners.com/presidential-race
Does your model publish a list of polls that go into it, a la Silver / 538?