Dominic Cummings

On 2024.

The great Matt Levine, who writes the best column about finance in the entire world, is fond of saying that “everything is securities fraud.” Here at Political Items, we’re fond of saying “everyone’s a political strategist.” In both cases, obviously, it’s mostly untrue.

So when an actual political strategist pops up with some thoughts about the 2024 U.S. presidential election, it’s worth relaying those thoughts, verbatim.

That strategist, Dominic Cummings, is best known as the man who master-minded the Vote Leave campaign in the United Kingdom in 2016. It was a stunning upset that wasn’t really a stunning upset at all. It just happened that Mr. Cummings understood the UK electorate better than anyone at the Remain campaign (or anywhere else for that matter).

What follows are lengthy excerpts from Cummings’ most recent Substack newsletter, which he posted today.

---

After the midterms, a small team did some research to figure out some dynamics for the 2024 Presidential campaign, primaries and general.

We built a model (with many times more data than normal polls) to predict who would win the electoral college in different matchups. We explored attitudes to possible Presidential candidates. We explored crucial issues like the cost of living, health, abortion, crime and Ukraine. We tested some ideas for how a Presidential campaign could best launch — what are the most important waves to ride and avoid, how on crucial issues do you navigate between your core vote and the crucial swing voters in crucial states? We ran focus groups in crucial states including Arizona, Georgia, Michigan, Nevada, Pennsylvania, Wisconsin. These included a) Obama>Trump voters (a fascinating group), b) no vote 2016 > Biden 2020, c) no vote 2016 > Trump 2020 voters, d) hardcore MAGA voters, e) non-MAGA-GOP voters. All data was collected before the Trump-arrest story got going seriously. His position with GOP voters has improved since we collected data. (We also experimented with what we called ‘synthetic focus groups’ and ‘synthetic polling’, i.e running focus groups and polls with ‘synthetic’ voters inside Large Language Models, see below.)

We made more effort than is made in normal polls to ensure we had an accurate sample of low-education, low-trust voters who do not pay much attention to normal news (exploring factors like attitude to vaccination and trust in media). Since 2016 a failure to sample these voters properly has been an important factor in the failure of many pollsters, connected with the failure to appreciate the importance of graduates shifting left on many issues. E.g In 2015-16 in Britain and America many pollsters did not weight by education. An important reason why we and Trump won is that we and Jared Kushner built better data science operations and therefore understood crucial voters (and turnout models) better than the old parties or the old media. With both Brexit and Trump, polling errors reinforced Insiders’ other errors. Despite the stakes, public polling in 2020 in America was also badly wrong but it got less attention than 2016 because Biden won.

---

In our model Trump loses the popular vote by about 2 points but Trump wins the electoral college 294-244.