Congratulations to John Thomas on a Straight Flush
First, a big congratulations to John Thomas and the team at Nestpoint Associates for their outstanding predictions on the outcome of the U.S. elections. It looks like we're on course for a Red sweep, as he predicted. We always push back and debate our partners and consultants during the curation process, regardless of our own views.
Pollsters Nailed It
The pollsters deserve huge praise for their fantastic work this election cycle. If there's one small critique, if we’re nit-picking: since 2016, the polling pendulum has shifted too heavily toward battleground states at the expense of national and stronghold states.
Early indications show that battleground state results are roughly 1% off in Trump’s favor—much less than the 3-5% average margin seen in 2020. It's also worth noting that in traditional stronghold red or blue states, public polling is very sparse, if it exists at all, so it’s not useful for assessing polling errors. However, this doesn't mean that those doing on-the-ground work in focus groups aren’t tracking local trends. This type of information just isn't visible in the public domain.
The Iowa Selzer shock poll, showing Harris +3 and released this weekend, did indeed prove to be a red herring, as John called it. Fortunately, an Emerson survey released the same day showed Trump +9 in Iowa, which aligns with a projected net gain in Trump support in California and national polls. Since 2016, we’ve questioned the popular narrative of the “silent Trump voter” when past survey errors can be explained by sampling and modeling methods. For our purposes, we’re glad we trusted John's assessment and didn’t let the Selzer poll skew our sample.
Poor Election Coverage & Commentary by the Mainstream Media
We’ve always believed that the most dangerous people in markets or politics are those acting on misplaced virtuosity while being precisely wrong. It’s fine to break from scientific orthodoxy if the reasons are valid and it’s done openly, without shutting out alternatives. But the starting point should be well understood before deviating from established rules.
Polls were close in battleground states, with averages slightly favoring Trump by less than 3%. While we maintain that no single poll should be interpreted within a 3% margin of error, this rule is less applicable to polling averages and distributions. So, commentators who disregarded hundreds of state polls 24 hours before the election and declared a “coin flip” were, at minimum, practicing bad science—and worse, perhaps letting their biases masquerade as math and science.
In scenarios where we assumed each battleground state outcome was a coin toss, Trump won in 52% of our first “Shift-F9” 100 randomized (“stochastic”) model simulations, quickly converging to 52.8% by the 300th run. Simplicity and stable convergence are key characteristics of effective models—even in complex ones like the General Circulation Models (GCMs) used to predict the weather. Contrary to common perceptions rooted in the ‘butterfly effect,' weather forecasts don't drastically change when model inputs are slightly changed.
All this is to say that if you’re discarding 100s of state poll data to declare battleground states a coin-flip (50/50) at the last minute, at least call it a 53/47 split to retain an ounce of credibility. This is broadly consistent with the electoral college map’s slight skew toward Republicans.
If Polling Averages Were Done Right, It Was Never a Coin-Flip
We estimated that if polls held, Trump had an 82% chance of winning the electoral college by 291.5/246.5 (+/-10.5 votes) in our central scenario, where Michigan (15) and Nevada (6) were split evenly. Under this model, a 1% systemic poll bias in Trump’s favor shifted the result to 307/231 (+/-5 votes) with Wisconsin as the only split state. A 1% poll bias the other way produced a 255/283 victory for Harris without split states.
We didn’t cover the +2% and -2% systemic poll bias scenarios, which would result in 312/226 (Trump win) and 238/300 (Harris win) with no split states, respectively. We felt this wouldn’t change the essence of our analysis—that Harris would benefit more from polling uncertainties and errors than Trump. With more time, we would have attempted to further reduce the uncertainties by digging into the cross-tabs and raw survey data.
Vibes Don’t Mean Votes
At the end of the day, vibe shifts are worth tracking but shouldn’t influence analysis unless measurable in voter intentions. Both sides spread substantial mis/disinformation this election cycle without foreign assistance.
This isn’t a culture wars statement, or meant to provoke anyone, but consider this post on X:
When was the last time an official survey of 650K Americans showed two clearly identified genders with 100% certainty? If you're going to invent a story, at least stay true to your cause and be inclusive.
The Silent Trump Voter is No Longer
In 2020, the silent Trump voter might have been best represented by a Bay Area resident who avoided sharing non-mainstream views in social circles. These voters were fairly easy to identify at the time with a few questions—assuming they trusted they wouldn’t face backlash. Anecdotally, this phenomenon was less pronounced in New York City, where the general attitude is: do as you wish, just don’t make me late for work.
We suspect that once this election cycle’s results are analyzed in detail, pollsters will reexamine the usefulness of educational attainment as a predictor of voting intentions. This factor was introduced after the 2016 polling issues, mainly to differentiate white college-educated voters from MAGA supporters in survey panels.
Trump Whale Wins $50 million
Let's wrap up our coverage of the US election with an update from notre ami:
How the Trump Whale Correctly Called the Election
The mystery trader who calls himself ‘Théo’ is on track for a payday of nearly $50 million
In his emails and a Zoom conversation with a reporter, Théo repeatedly criticized U.S. opinion polls. He was particularly critical of polls conducted by mainstream-media outlets that, in his view, were biased toward Democrats and tended to produce outlier poll results that favored Harris.“In France this is different!! The pollster credibility is more important: they want to be as close as possible to the actual results. Culture is different on this,” he wrote.
Théo shared a table of numbers he had compiled based on RealClearPolitics polling averages, showing that Trump had overperformed his swing-state polling numbers in 2020. Given the tight polls in swing states in 2024, Théo reasoned that a similar overperformance by Trump would easily push him into the lead.
Polls failed to account for the “shy Trump voter effect,” Théo said. Either Trump backers were reluctant to tell pollsters that they supported the former president, or they didn’t want to participate in polls, Théo wrote.
Based on the WSJ article, he simply took the 2020 polling errors of around 3-5.5% and added them to the 2024 figures. The battleground states were off by at most 2%. So far, Arizona appears to be 1.3% less favorable toward Trump than the polling average we used in our forecasts. As we noted earlier, we would have liked to have a bit more polling data from Arizona closer to the election.
It wasn't that long ago that the second round of the French legislative election produced results far different from what pollsters predicted. It’s all in the mind.
Now, is anyone going to tell “Teo” that he just made $50 million but for the wrong reason? Of course not! Better lucky than clever… and make sure it's big! Welcome to the markets.
Congratulations to him in any case. Double or nothing?
